Merge branch 'develop' into marchcat/b-develop

61 files changed, 3202 insertions, 631 deletions

diff --git a/indra/newview/VIEWER_VERSION.txt b/indra/newview/VIEWER_VERSION.txt
index 346a7e3aa1..e0eaaa0bbc 100644
--- a/indra/newview/VIEWER_VERSION.txt
+++ b/indra/newview/VIEWER_VERSION.txt
@@ -1 +1 @@
-7.1.10
+7.1.11
diff --git a/indra/newview/app_settings/settings.xml b/indra/newview/app_settings/settings.xml
index 1863eaf945..0355bed1bf 100644
--- a/indra/newview/app_settings/settings.xml
+++ b/indra/newview/app_settings/settings.xml
@@ -9821,6 +9821,17 @@
       <key>Value</key>
       <string>00000000-0000-0000-0000-000000000000</string>
     </map>
+    <key>RenderCASSharpness</key>
+    <map>
+        <key>Comment</key>
+        <string>Level of sharpening to apply via Contrast Adaptive Sharpening (0.0(off) - 1.0)</string>
+        <key>Persist</key>
+        <integer>1</integer>
+        <key>Type</key>
+        <string>F32</string>
+        <key>Value</key>
+        <real>0.4</real>
+    </map>
     <key>ReplaySession</key>
     <map>
       <key>Comment</key>
@@ -13158,17 +13169,6 @@
       <key>Value</key>
       <integer>0</integer>
     </map>
-    <key>VoiceEffectExpiryWarningTime</key>
-    <map>
-      <key>Comment</key>
-      <string>How much notice to give of Voice Morph subscriptions expiry, in seconds.</string>
-      <key>Persist</key>
-      <integer>1</integer>
-      <key>Type</key>
-      <string>S32</string>
-      <key>Value</key>
-      <integer>259200</integer>
-    </map>
     <key>VoiceMorphingEnabled</key>
     <map>
       <key>Comment</key>
diff --git a/indra/newview/app_settings/shaders/class1/deferred/CASF.glsl b/indra/newview/app_settings/shaders/class1/deferred/CASF.glsl
new file mode 100644
index 0000000000..96d08058cf
--- /dev/null
+++ b/indra/newview/app_settings/shaders/class1/deferred/CASF.glsl
@@ -0,0 +1,2558 @@
+/**
+ * @file CASF.glsl
+ *
+ * $LicenseInfo:firstyear=2024&license=viewerlgpl$
+ * Second Life Viewer Source Code
+ * Copyright (C) 2024, Linden Research, Inc.
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation;
+ * version 2.1 of the License only.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
+ * $/LicenseInfo$
+ */
+
+/*[EXTRA_CODE_HERE]*/
+
+#ifndef A_CPU
+#define A_GPU
+#define A_GLSL
+#define CAS_BETTER_DIAGONALS
+#define CAS_SLOW
+
+out vec4 frag_color;
+in vec2 vary_fragcoord;
+
+uniform sampler2D diffuseRect;
+uniform vec2 out_screen_res;
+uniform uvec4 cas_param_0;
+uniform uvec4 cas_param_1;
+
+vec3 srgb_to_linear(vec3 cs);
+vec3 linear_to_srgb(vec3 cl);
+#endif
+
+#ifndef SHADER_PORTABILITY
+//==============================================================================================================================
+//
+//                                               [A] SHADER PORTABILITY 1.20210629
+//
+//==============================================================================================================================
+// FidelityFX Super Resolution Sample
+//
+// Copyright (c) 2021 Advanced Micro Devices, Inc. All rights reserved.
+// Permission is hereby granted, free of charge, to any person obtaining a copy
+// of this software and associated documentation files(the "Software"), to deal
+// in the Software without restriction, including without limitation the rights
+// to use, copy, modify, merge, publish, distribute, sublicense, and / or sell
+// copies of the Software, and to permit persons to whom the Software is
+// furnished to do so, subject to the following conditions :
+// The above copyright notice and this permission notice shall be included in
+// all copies or substantial portions of the Software.
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+// THE SOFTWARE.
+//------------------------------------------------------------------------------------------------------------------------------
+// MIT LICENSE
+// ===========
+// Copyright (c) 2014 Michal Drobot (for concepts used in "FLOAT APPROXIMATIONS").
+// -----------
+// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation
+// files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy,
+// modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the
+// Software is furnished to do so, subject to the following conditions:
+// -----------
+// The above copyright notice and this permission notice shall be included in all copies or substantial portions of the
+// Software.
+// -----------
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE
+// WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE AUTHORS OR
+// COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+// ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+//------------------------------------------------------------------------------------------------------------------------------
+#define A_2PI 6.28318530718
+#ifdef A_CPU
+ // Supporting user defined overrides.
+ #ifndef A_RESTRICT
+  #define A_RESTRICT __restrict
+ #endif
+//------------------------------------------------------------------------------------------------------------------------------
+ #ifndef A_STATIC
+  #define A_STATIC static
+ #endif
+//------------------------------------------------------------------------------------------------------------------------------
+ // Same types across CPU and GPU.
+ // Predicate uses 32-bit integer (C friendly bool).
+ typedef uint32_t AP1;
+ typedef float AF1;
+ typedef double AD1;
+ typedef uint8_t AB1;
+ typedef uint16_t AW1;
+ typedef uint32_t AU1;
+ typedef uint64_t AL1;
+ typedef int8_t ASB1;
+ typedef int16_t ASW1;
+ typedef int32_t ASU1;
+ typedef int64_t ASL1;
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AD1_(a) ((AD1)(a))
+ #define AF1_(a) ((AF1)(a))
+ #define AL1_(a) ((AL1)(a))
+ #define AU1_(a) ((AU1)(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define ASL1_(a) ((ASL1)(a))
+ #define ASU1_(a) ((ASU1)(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AU1 AU1_AF1(AF1 a){union{AF1 f;AU1 u;}bits;bits.f=a;return bits.u;}
+//------------------------------------------------------------------------------------------------------------------------------
+ #define A_TRUE 1
+ #define A_FALSE 0
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//                                                       CPU/GPU PORTING
+//
+//------------------------------------------------------------------------------------------------------------------------------
+// Get CPU and GPU to share all setup code, without duplicate code paths.
+// This uses a lower-case prefix for special vector constructs.
+//  - In C restrict pointers are used.
+//  - In the shading language, in/inout/out arguments are used.
+// This depends on the ability to access a vector value in both languages via array syntax (aka color[2]).
+//==============================================================================================================================
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                     VECTOR ARGUMENT/RETURN/INITIALIZATION PORTABILITY
+//==============================================================================================================================
+ #define retAD2 AD1 *A_RESTRICT
+ #define retAD3 AD1 *A_RESTRICT
+ #define retAD4 AD1 *A_RESTRICT
+ #define retAF2 AF1 *A_RESTRICT
+ #define retAF3 AF1 *A_RESTRICT
+ #define retAF4 AF1 *A_RESTRICT
+ #define retAL2 AL1 *A_RESTRICT
+ #define retAL3 AL1 *A_RESTRICT
+ #define retAL4 AL1 *A_RESTRICT
+ #define retAU2 AU1 *A_RESTRICT
+ #define retAU3 AU1 *A_RESTRICT
+ #define retAU4 AU1 *A_RESTRICT
+//------------------------------------------------------------------------------------------------------------------------------
+ #define inAD2 AD1 *A_RESTRICT
+ #define inAD3 AD1 *A_RESTRICT
+ #define inAD4 AD1 *A_RESTRICT
+ #define inAF2 AF1 *A_RESTRICT
+ #define inAF3 AF1 *A_RESTRICT
+ #define inAF4 AF1 *A_RESTRICT
+ #define inAL2 AL1 *A_RESTRICT
+ #define inAL3 AL1 *A_RESTRICT
+ #define inAL4 AL1 *A_RESTRICT
+ #define inAU2 AU1 *A_RESTRICT
+ #define inAU3 AU1 *A_RESTRICT
+ #define inAU4 AU1 *A_RESTRICT
+//------------------------------------------------------------------------------------------------------------------------------
+ #define inoutAD2 AD1 *A_RESTRICT
+ #define inoutAD3 AD1 *A_RESTRICT
+ #define inoutAD4 AD1 *A_RESTRICT
+ #define inoutAF2 AF1 *A_RESTRICT
+ #define inoutAF3 AF1 *A_RESTRICT
+ #define inoutAF4 AF1 *A_RESTRICT
+ #define inoutAL2 AL1 *A_RESTRICT
+ #define inoutAL3 AL1 *A_RESTRICT
+ #define inoutAL4 AL1 *A_RESTRICT
+ #define inoutAU2 AU1 *A_RESTRICT
+ #define inoutAU3 AU1 *A_RESTRICT
+ #define inoutAU4 AU1 *A_RESTRICT
+//------------------------------------------------------------------------------------------------------------------------------
+ #define outAD2 AD1 *A_RESTRICT
+ #define outAD3 AD1 *A_RESTRICT
+ #define outAD4 AD1 *A_RESTRICT
+ #define outAF2 AF1 *A_RESTRICT
+ #define outAF3 AF1 *A_RESTRICT
+ #define outAF4 AF1 *A_RESTRICT
+ #define outAL2 AL1 *A_RESTRICT
+ #define outAL3 AL1 *A_RESTRICT
+ #define outAL4 AL1 *A_RESTRICT
+ #define outAU2 AU1 *A_RESTRICT
+ #define outAU3 AU1 *A_RESTRICT
+ #define outAU4 AU1 *A_RESTRICT
+//------------------------------------------------------------------------------------------------------------------------------
+ #define varAD2(x) AD1 x[2]
+ #define varAD3(x) AD1 x[3]
+ #define varAD4(x) AD1 x[4]
+ #define varAF2(x) AF1 x[2]
+ #define varAF3(x) AF1 x[3]
+ #define varAF4(x) AF1 x[4]
+ #define varAL2(x) AL1 x[2]
+ #define varAL3(x) AL1 x[3]
+ #define varAL4(x) AL1 x[4]
+ #define varAU2(x) AU1 x[2]
+ #define varAU3(x) AU1 x[3]
+ #define varAU4(x) AU1 x[4]
+//------------------------------------------------------------------------------------------------------------------------------
+ #define initAD2(x,y) {x,y}
+ #define initAD3(x,y,z) {x,y,z}
+ #define initAD4(x,y,z,w) {x,y,z,w}
+ #define initAF2(x,y) {x,y}
+ #define initAF3(x,y,z) {x,y,z}
+ #define initAF4(x,y,z,w) {x,y,z,w}
+ #define initAL2(x,y) {x,y}
+ #define initAL3(x,y,z) {x,y,z}
+ #define initAL4(x,y,z,w) {x,y,z,w}
+ #define initAU2(x,y) {x,y}
+ #define initAU3(x,y,z) {x,y,z}
+ #define initAU4(x,y,z,w) {x,y,z,w}
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                     SCALAR RETURN OPS
+//------------------------------------------------------------------------------------------------------------------------------
+// TODO
+// ====
+//  - Replace transcendentals with manual versions.
+//==============================================================================================================================
+ #ifdef A_GCC
+  A_STATIC AD1 AAbsD1(AD1 a){return __builtin_fabs(a);}
+  A_STATIC AF1 AAbsF1(AF1 a){return __builtin_fabsf(a);}
+  A_STATIC AU1 AAbsSU1(AU1 a){return AU1_(__builtin_abs(ASU1_(a)));}
+  A_STATIC AL1 AAbsSL1(AL1 a){return AL1_(__builtin_llabs(ASL1_(a)));}
+ #else
+  A_STATIC AD1 AAbsD1(AD1 a){return fabs(a);}
+  A_STATIC AF1 AAbsF1(AF1 a){return fabsf(a);}
+  A_STATIC AU1 AAbsSU1(AU1 a){return AU1_(abs(ASU1_(a)));}
+  A_STATIC AL1 AAbsSL1(AL1 a){return AL1_(labs((long)ASL1_(a)));}
+ #endif
+//------------------------------------------------------------------------------------------------------------------------------
+ #ifdef A_GCC
+  A_STATIC AD1 ACosD1(AD1 a){return __builtin_cos(a);}
+  A_STATIC AF1 ACosF1(AF1 a){return __builtin_cosf(a);}
+ #else
+  A_STATIC AD1 ACosD1(AD1 a){return cos(a);}
+  A_STATIC AF1 ACosF1(AF1 a){return cosf(a);}
+ #endif
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 ADotD2(inAD2 a,inAD2 b){return a[0]*b[0]+a[1]*b[1];}
+ A_STATIC AD1 ADotD3(inAD3 a,inAD3 b){return a[0]*b[0]+a[1]*b[1]+a[2]*b[2];}
+ A_STATIC AD1 ADotD4(inAD4 a,inAD4 b){return a[0]*b[0]+a[1]*b[1]+a[2]*b[2]+a[3]*b[3];}
+ A_STATIC AF1 ADotF2(inAF2 a,inAF2 b){return a[0]*b[0]+a[1]*b[1];}
+ A_STATIC AF1 ADotF3(inAF3 a,inAF3 b){return a[0]*b[0]+a[1]*b[1]+a[2]*b[2];}
+ A_STATIC AF1 ADotF4(inAF4 a,inAF4 b){return a[0]*b[0]+a[1]*b[1]+a[2]*b[2]+a[3]*b[3];}
+//------------------------------------------------------------------------------------------------------------------------------
+ #ifdef A_GCC
+  A_STATIC AD1 AExp2D1(AD1 a){return __builtin_exp2(a);}
+  A_STATIC AF1 AExp2F1(AF1 a){return __builtin_exp2f(a);}
+ #else
+  A_STATIC AD1 AExp2D1(AD1 a){return exp2(a);}
+  A_STATIC AF1 AExp2F1(AF1 a){return exp2f(a);}
+ #endif
+//------------------------------------------------------------------------------------------------------------------------------
+ #ifdef A_GCC
+  A_STATIC AD1 AFloorD1(AD1 a){return __builtin_floor(a);}
+  A_STATIC AF1 AFloorF1(AF1 a){return __builtin_floorf(a);}
+ #else
+  A_STATIC AD1 AFloorD1(AD1 a){return floor(a);}
+  A_STATIC AF1 AFloorF1(AF1 a){return floorf(a);}
+ #endif
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 ALerpD1(AD1 a,AD1 b,AD1 c){return b*c+(-a*c+a);}
+ A_STATIC AF1 ALerpF1(AF1 a,AF1 b,AF1 c){return b*c+(-a*c+a);}
+//------------------------------------------------------------------------------------------------------------------------------
+ #ifdef A_GCC
+  A_STATIC AD1 ALog2D1(AD1 a){return __builtin_log2(a);}
+  A_STATIC AF1 ALog2F1(AF1 a){return __builtin_log2f(a);}
+ #else
+  A_STATIC AD1 ALog2D1(AD1 a){return log2(a);}
+  A_STATIC AF1 ALog2F1(AF1 a){return log2f(a);}
+ #endif
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 AMaxD1(AD1 a,AD1 b){return a>b?a:b;}
+ A_STATIC AF1 AMaxF1(AF1 a,AF1 b){return a>b?a:b;}
+ A_STATIC AL1 AMaxL1(AL1 a,AL1 b){return a>b?a:b;}
+ A_STATIC AU1 AMaxU1(AU1 a,AU1 b){return a>b?a:b;}
+//------------------------------------------------------------------------------------------------------------------------------
+ // These follow the convention that A integer types don't have signage, until they are operated on.
+ A_STATIC AL1 AMaxSL1(AL1 a,AL1 b){return (ASL1_(a)>ASL1_(b))?a:b;}
+ A_STATIC AU1 AMaxSU1(AU1 a,AU1 b){return (ASU1_(a)>ASU1_(b))?a:b;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 AMinD1(AD1 a,AD1 b){return a<b?a:b;}
+ A_STATIC AF1 AMinF1(AF1 a,AF1 b){return a<b?a:b;}
+ A_STATIC AL1 AMinL1(AL1 a,AL1 b){return a<b?a:b;}
+ A_STATIC AU1 AMinU1(AU1 a,AU1 b){return a<b?a:b;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AL1 AMinSL1(AL1 a,AL1 b){return (ASL1_(a)<ASL1_(b))?a:b;}
+ A_STATIC AU1 AMinSU1(AU1 a,AU1 b){return (ASU1_(a)<ASU1_(b))?a:b;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 ARcpD1(AD1 a){return 1.0/a;}
+ A_STATIC AF1 ARcpF1(AF1 a){return 1.0f/a;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AL1 AShrSL1(AL1 a,AL1 b){return AL1_(ASL1_(a)>>ASL1_(b));}
+ A_STATIC AU1 AShrSU1(AU1 a,AU1 b){return AU1_(ASU1_(a)>>ASU1_(b));}
+//------------------------------------------------------------------------------------------------------------------------------
+ #ifdef A_GCC
+  A_STATIC AD1 ASinD1(AD1 a){return __builtin_sin(a);}
+  A_STATIC AF1 ASinF1(AF1 a){return __builtin_sinf(a);}
+ #else
+  A_STATIC AD1 ASinD1(AD1 a){return sin(a);}
+  A_STATIC AF1 ASinF1(AF1 a){return sinf(a);}
+ #endif
+//------------------------------------------------------------------------------------------------------------------------------
+ #ifdef A_GCC
+  A_STATIC AD1 ASqrtD1(AD1 a){return __builtin_sqrt(a);}
+  A_STATIC AF1 ASqrtF1(AF1 a){return __builtin_sqrtf(a);}
+ #else
+  A_STATIC AD1 ASqrtD1(AD1 a){return sqrt(a);}
+  A_STATIC AF1 ASqrtF1(AF1 a){return sqrtf(a);}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                               SCALAR RETURN OPS - DEPENDENT
+//==============================================================================================================================
+ A_STATIC AD1 AClampD1(AD1 x,AD1 n,AD1 m){return AMaxD1(n,AMinD1(x,m));}
+ A_STATIC AF1 AClampF1(AF1 x,AF1 n,AF1 m){return AMaxF1(n,AMinF1(x,m));}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 AFractD1(AD1 a){return a-AFloorD1(a);}
+ A_STATIC AF1 AFractF1(AF1 a){return a-AFloorF1(a);}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 APowD1(AD1 a,AD1 b){return AExp2D1(b*ALog2D1(a));}
+ A_STATIC AF1 APowF1(AF1 a,AF1 b){return AExp2F1(b*ALog2F1(a));}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 ARsqD1(AD1 a){return ARcpD1(ASqrtD1(a));}
+ A_STATIC AF1 ARsqF1(AF1 a){return ARcpF1(ASqrtF1(a));}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC AD1 ASatD1(AD1 a){return AMinD1(1.0,AMaxD1(0.0,a));}
+ A_STATIC AF1 ASatF1(AF1 a){return AMinF1(1.0f,AMaxF1(0.0f,a));}
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                         VECTOR OPS
+//------------------------------------------------------------------------------------------------------------------------------
+// These are added as needed for production or prototyping, so not necessarily a complete set.
+// They follow a convention of taking in a destination and also returning the destination value to increase utility.
+//==============================================================================================================================
+ A_STATIC retAD2 opAAbsD2(outAD2 d,inAD2 a){d[0]=AAbsD1(a[0]);d[1]=AAbsD1(a[1]);return d;}
+ A_STATIC retAD3 opAAbsD3(outAD3 d,inAD3 a){d[0]=AAbsD1(a[0]);d[1]=AAbsD1(a[1]);d[2]=AAbsD1(a[2]);return d;}
+ A_STATIC retAD4 opAAbsD4(outAD4 d,inAD4 a){d[0]=AAbsD1(a[0]);d[1]=AAbsD1(a[1]);d[2]=AAbsD1(a[2]);d[3]=AAbsD1(a[3]);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opAAbsF2(outAF2 d,inAF2 a){d[0]=AAbsF1(a[0]);d[1]=AAbsF1(a[1]);return d;}
+ A_STATIC retAF3 opAAbsF3(outAF3 d,inAF3 a){d[0]=AAbsF1(a[0]);d[1]=AAbsF1(a[1]);d[2]=AAbsF1(a[2]);return d;}
+ A_STATIC retAF4 opAAbsF4(outAF4 d,inAF4 a){d[0]=AAbsF1(a[0]);d[1]=AAbsF1(a[1]);d[2]=AAbsF1(a[2]);d[3]=AAbsF1(a[3]);return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opAAddD2(outAD2 d,inAD2 a,inAD2 b){d[0]=a[0]+b[0];d[1]=a[1]+b[1];return d;}
+ A_STATIC retAD3 opAAddD3(outAD3 d,inAD3 a,inAD3 b){d[0]=a[0]+b[0];d[1]=a[1]+b[1];d[2]=a[2]+b[2];return d;}
+ A_STATIC retAD4 opAAddD4(outAD4 d,inAD4 a,inAD4 b){d[0]=a[0]+b[0];d[1]=a[1]+b[1];d[2]=a[2]+b[2];d[3]=a[3]+b[3];return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opAAddF2(outAF2 d,inAF2 a,inAF2 b){d[0]=a[0]+b[0];d[1]=a[1]+b[1];return d;}
+ A_STATIC retAF3 opAAddF3(outAF3 d,inAF3 a,inAF3 b){d[0]=a[0]+b[0];d[1]=a[1]+b[1];d[2]=a[2]+b[2];return d;}
+ A_STATIC retAF4 opAAddF4(outAF4 d,inAF4 a,inAF4 b){d[0]=a[0]+b[0];d[1]=a[1]+b[1];d[2]=a[2]+b[2];d[3]=a[3]+b[3];return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opAAddOneD2(outAD2 d,inAD2 a,AD1 b){d[0]=a[0]+b;d[1]=a[1]+b;return d;}
+ A_STATIC retAD3 opAAddOneD3(outAD3 d,inAD3 a,AD1 b){d[0]=a[0]+b;d[1]=a[1]+b;d[2]=a[2]+b;return d;}
+ A_STATIC retAD4 opAAddOneD4(outAD4 d,inAD4 a,AD1 b){d[0]=a[0]+b;d[1]=a[1]+b;d[2]=a[2]+b;d[3]=a[3]+b;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opAAddOneF2(outAF2 d,inAF2 a,AF1 b){d[0]=a[0]+b;d[1]=a[1]+b;return d;}
+ A_STATIC retAF3 opAAddOneF3(outAF3 d,inAF3 a,AF1 b){d[0]=a[0]+b;d[1]=a[1]+b;d[2]=a[2]+b;return d;}
+ A_STATIC retAF4 opAAddOneF4(outAF4 d,inAF4 a,AF1 b){d[0]=a[0]+b;d[1]=a[1]+b;d[2]=a[2]+b;d[3]=a[3]+b;return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opACpyD2(outAD2 d,inAD2 a){d[0]=a[0];d[1]=a[1];return d;}
+ A_STATIC retAD3 opACpyD3(outAD3 d,inAD3 a){d[0]=a[0];d[1]=a[1];d[2]=a[2];return d;}
+ A_STATIC retAD4 opACpyD4(outAD4 d,inAD4 a){d[0]=a[0];d[1]=a[1];d[2]=a[2];d[3]=a[3];return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opACpyF2(outAF2 d,inAF2 a){d[0]=a[0];d[1]=a[1];return d;}
+ A_STATIC retAF3 opACpyF3(outAF3 d,inAF3 a){d[0]=a[0];d[1]=a[1];d[2]=a[2];return d;}
+ A_STATIC retAF4 opACpyF4(outAF4 d,inAF4 a){d[0]=a[0];d[1]=a[1];d[2]=a[2];d[3]=a[3];return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opALerpD2(outAD2 d,inAD2 a,inAD2 b,inAD2 c){d[0]=ALerpD1(a[0],b[0],c[0]);d[1]=ALerpD1(a[1],b[1],c[1]);return d;}
+ A_STATIC retAD3 opALerpD3(outAD3 d,inAD3 a,inAD3 b,inAD3 c){d[0]=ALerpD1(a[0],b[0],c[0]);d[1]=ALerpD1(a[1],b[1],c[1]);d[2]=ALerpD1(a[2],b[2],c[2]);return d;}
+ A_STATIC retAD4 opALerpD4(outAD4 d,inAD4 a,inAD4 b,inAD4 c){d[0]=ALerpD1(a[0],b[0],c[0]);d[1]=ALerpD1(a[1],b[1],c[1]);d[2]=ALerpD1(a[2],b[2],c[2]);d[3]=ALerpD1(a[3],b[3],c[3]);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opALerpF2(outAF2 d,inAF2 a,inAF2 b,inAF2 c){d[0]=ALerpF1(a[0],b[0],c[0]);d[1]=ALerpF1(a[1],b[1],c[1]);return d;}
+ A_STATIC retAF3 opALerpF3(outAF3 d,inAF3 a,inAF3 b,inAF3 c){d[0]=ALerpF1(a[0],b[0],c[0]);d[1]=ALerpF1(a[1],b[1],c[1]);d[2]=ALerpF1(a[2],b[2],c[2]);return d;}
+ A_STATIC retAF4 opALerpF4(outAF4 d,inAF4 a,inAF4 b,inAF4 c){d[0]=ALerpF1(a[0],b[0],c[0]);d[1]=ALerpF1(a[1],b[1],c[1]);d[2]=ALerpF1(a[2],b[2],c[2]);d[3]=ALerpF1(a[3],b[3],c[3]);return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opALerpOneD2(outAD2 d,inAD2 a,inAD2 b,AD1 c){d[0]=ALerpD1(a[0],b[0],c);d[1]=ALerpD1(a[1],b[1],c);return d;}
+ A_STATIC retAD3 opALerpOneD3(outAD3 d,inAD3 a,inAD3 b,AD1 c){d[0]=ALerpD1(a[0],b[0],c);d[1]=ALerpD1(a[1],b[1],c);d[2]=ALerpD1(a[2],b[2],c);return d;}
+ A_STATIC retAD4 opALerpOneD4(outAD4 d,inAD4 a,inAD4 b,AD1 c){d[0]=ALerpD1(a[0],b[0],c);d[1]=ALerpD1(a[1],b[1],c);d[2]=ALerpD1(a[2],b[2],c);d[3]=ALerpD1(a[3],b[3],c);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opALerpOneF2(outAF2 d,inAF2 a,inAF2 b,AF1 c){d[0]=ALerpF1(a[0],b[0],c);d[1]=ALerpF1(a[1],b[1],c);return d;}
+ A_STATIC retAF3 opALerpOneF3(outAF3 d,inAF3 a,inAF3 b,AF1 c){d[0]=ALerpF1(a[0],b[0],c);d[1]=ALerpF1(a[1],b[1],c);d[2]=ALerpF1(a[2],b[2],c);return d;}
+ A_STATIC retAF4 opALerpOneF4(outAF4 d,inAF4 a,inAF4 b,AF1 c){d[0]=ALerpF1(a[0],b[0],c);d[1]=ALerpF1(a[1],b[1],c);d[2]=ALerpF1(a[2],b[2],c);d[3]=ALerpF1(a[3],b[3],c);return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opAMaxD2(outAD2 d,inAD2 a,inAD2 b){d[0]=AMaxD1(a[0],b[0]);d[1]=AMaxD1(a[1],b[1]);return d;}
+ A_STATIC retAD3 opAMaxD3(outAD3 d,inAD3 a,inAD3 b){d[0]=AMaxD1(a[0],b[0]);d[1]=AMaxD1(a[1],b[1]);d[2]=AMaxD1(a[2],b[2]);return d;}
+ A_STATIC retAD4 opAMaxD4(outAD4 d,inAD4 a,inAD4 b){d[0]=AMaxD1(a[0],b[0]);d[1]=AMaxD1(a[1],b[1]);d[2]=AMaxD1(a[2],b[2]);d[3]=AMaxD1(a[3],b[3]);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opAMaxF2(outAF2 d,inAF2 a,inAF2 b){d[0]=AMaxF1(a[0],b[0]);d[1]=AMaxF1(a[1],b[1]);return d;}
+ A_STATIC retAF3 opAMaxF3(outAF3 d,inAF3 a,inAF3 b){d[0]=AMaxF1(a[0],b[0]);d[1]=AMaxF1(a[1],b[1]);d[2]=AMaxF1(a[2],b[2]);return d;}
+ A_STATIC retAF4 opAMaxF4(outAF4 d,inAF4 a,inAF4 b){d[0]=AMaxF1(a[0],b[0]);d[1]=AMaxF1(a[1],b[1]);d[2]=AMaxF1(a[2],b[2]);d[3]=AMaxF1(a[3],b[3]);return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opAMinD2(outAD2 d,inAD2 a,inAD2 b){d[0]=AMinD1(a[0],b[0]);d[1]=AMinD1(a[1],b[1]);return d;}
+ A_STATIC retAD3 opAMinD3(outAD3 d,inAD3 a,inAD3 b){d[0]=AMinD1(a[0],b[0]);d[1]=AMinD1(a[1],b[1]);d[2]=AMinD1(a[2],b[2]);return d;}
+ A_STATIC retAD4 opAMinD4(outAD4 d,inAD4 a,inAD4 b){d[0]=AMinD1(a[0],b[0]);d[1]=AMinD1(a[1],b[1]);d[2]=AMinD1(a[2],b[2]);d[3]=AMinD1(a[3],b[3]);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opAMinF2(outAF2 d,inAF2 a,inAF2 b){d[0]=AMinF1(a[0],b[0]);d[1]=AMinF1(a[1],b[1]);return d;}
+ A_STATIC retAF3 opAMinF3(outAF3 d,inAF3 a,inAF3 b){d[0]=AMinF1(a[0],b[0]);d[1]=AMinF1(a[1],b[1]);d[2]=AMinF1(a[2],b[2]);return d;}
+ A_STATIC retAF4 opAMinF4(outAF4 d,inAF4 a,inAF4 b){d[0]=AMinF1(a[0],b[0]);d[1]=AMinF1(a[1],b[1]);d[2]=AMinF1(a[2],b[2]);d[3]=AMinF1(a[3],b[3]);return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opAMulD2(outAD2 d,inAD2 a,inAD2 b){d[0]=a[0]*b[0];d[1]=a[1]*b[1];return d;}
+ A_STATIC retAD3 opAMulD3(outAD3 d,inAD3 a,inAD3 b){d[0]=a[0]*b[0];d[1]=a[1]*b[1];d[2]=a[2]*b[2];return d;}
+ A_STATIC retAD4 opAMulD4(outAD4 d,inAD4 a,inAD4 b){d[0]=a[0]*b[0];d[1]=a[1]*b[1];d[2]=a[2]*b[2];d[3]=a[3]*b[3];return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opAMulF2(outAF2 d,inAF2 a,inAF2 b){d[0]=a[0]*b[0];d[1]=a[1]*b[1];return d;}
+ A_STATIC retAF3 opAMulF3(outAF3 d,inAF3 a,inAF3 b){d[0]=a[0]*b[0];d[1]=a[1]*b[1];d[2]=a[2]*b[2];return d;}
+ A_STATIC retAF4 opAMulF4(outAF4 d,inAF4 a,inAF4 b){d[0]=a[0]*b[0];d[1]=a[1]*b[1];d[2]=a[2]*b[2];d[3]=a[3]*b[3];return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opAMulOneD2(outAD2 d,inAD2 a,AD1 b){d[0]=a[0]*b;d[1]=a[1]*b;return d;}
+ A_STATIC retAD3 opAMulOneD3(outAD3 d,inAD3 a,AD1 b){d[0]=a[0]*b;d[1]=a[1]*b;d[2]=a[2]*b;return d;}
+ A_STATIC retAD4 opAMulOneD4(outAD4 d,inAD4 a,AD1 b){d[0]=a[0]*b;d[1]=a[1]*b;d[2]=a[2]*b;d[3]=a[3]*b;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opAMulOneF2(outAF2 d,inAF2 a,AF1 b){d[0]=a[0]*b;d[1]=a[1]*b;return d;}
+ A_STATIC retAF3 opAMulOneF3(outAF3 d,inAF3 a,AF1 b){d[0]=a[0]*b;d[1]=a[1]*b;d[2]=a[2]*b;return d;}
+ A_STATIC retAF4 opAMulOneF4(outAF4 d,inAF4 a,AF1 b){d[0]=a[0]*b;d[1]=a[1]*b;d[2]=a[2]*b;d[3]=a[3]*b;return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opANegD2(outAD2 d,inAD2 a){d[0]=-a[0];d[1]=-a[1];return d;}
+ A_STATIC retAD3 opANegD3(outAD3 d,inAD3 a){d[0]=-a[0];d[1]=-a[1];d[2]=-a[2];return d;}
+ A_STATIC retAD4 opANegD4(outAD4 d,inAD4 a){d[0]=-a[0];d[1]=-a[1];d[2]=-a[2];d[3]=-a[3];return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opANegF2(outAF2 d,inAF2 a){d[0]=-a[0];d[1]=-a[1];return d;}
+ A_STATIC retAF3 opANegF3(outAF3 d,inAF3 a){d[0]=-a[0];d[1]=-a[1];d[2]=-a[2];return d;}
+ A_STATIC retAF4 opANegF4(outAF4 d,inAF4 a){d[0]=-a[0];d[1]=-a[1];d[2]=-a[2];d[3]=-a[3];return d;}
+//==============================================================================================================================
+ A_STATIC retAD2 opARcpD2(outAD2 d,inAD2 a){d[0]=ARcpD1(a[0]);d[1]=ARcpD1(a[1]);return d;}
+ A_STATIC retAD3 opARcpD3(outAD3 d,inAD3 a){d[0]=ARcpD1(a[0]);d[1]=ARcpD1(a[1]);d[2]=ARcpD1(a[2]);return d;}
+ A_STATIC retAD4 opARcpD4(outAD4 d,inAD4 a){d[0]=ARcpD1(a[0]);d[1]=ARcpD1(a[1]);d[2]=ARcpD1(a[2]);d[3]=ARcpD1(a[3]);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ A_STATIC retAF2 opARcpF2(outAF2 d,inAF2 a){d[0]=ARcpF1(a[0]);d[1]=ARcpF1(a[1]);return d;}
+ A_STATIC retAF3 opARcpF3(outAF3 d,inAF3 a){d[0]=ARcpF1(a[0]);d[1]=ARcpF1(a[1]);d[2]=ARcpF1(a[2]);return d;}
+ A_STATIC retAF4 opARcpF4(outAF4 d,inAF4 a){d[0]=ARcpF1(a[0]);d[1]=ARcpF1(a[1]);d[2]=ARcpF1(a[2]);d[3]=ARcpF1(a[3]);return d;}
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                     HALF FLOAT PACKING
+//==============================================================================================================================
+ // Convert float to half (in lower 16-bits of output).
+ // Same fast technique as documented here: ftp://ftp.fox-toolkit.org/pub/fasthalffloatconversion.pdf
+ // Supports denormals.
+ // Conversion rules are to make computations possibly "safer" on the GPU,
+ //  -INF & -NaN -> -65504
+ //  +INF & +NaN -> +65504
+ A_STATIC AU1 AU1_AH1_AF1(AF1 f){
+  static AW1 base[512]={
+   0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
+   0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
+   0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
+   0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
+   0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
+   0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
+   0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0001,0x0002,0x0004,0x0008,0x0010,0x0020,0x0040,0x0080,0x0100,
+   0x0200,0x0400,0x0800,0x0c00,0x1000,0x1400,0x1800,0x1c00,0x2000,0x2400,0x2800,0x2c00,0x3000,0x3400,0x3800,0x3c00,
+   0x4000,0x4400,0x4800,0x4c00,0x5000,0x5400,0x5800,0x5c00,0x6000,0x6400,0x6800,0x6c00,0x7000,0x7400,0x7800,0x7bff,
+   0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,
+   0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,
+   0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,
+   0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,
+   0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,
+   0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,
+   0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,0x7bff,
+   0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,
+   0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,
+   0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,
+   0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,
+   0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,
+   0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,
+   0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8001,0x8002,0x8004,0x8008,0x8010,0x8020,0x8040,0x8080,0x8100,
+   0x8200,0x8400,0x8800,0x8c00,0x9000,0x9400,0x9800,0x9c00,0xa000,0xa400,0xa800,0xac00,0xb000,0xb400,0xb800,0xbc00,
+   0xc000,0xc400,0xc800,0xcc00,0xd000,0xd400,0xd800,0xdc00,0xe000,0xe400,0xe800,0xec00,0xf000,0xf400,0xf800,0xfbff,
+   0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,
+   0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,
+   0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,
+   0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,
+   0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,
+   0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,
+   0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff,0xfbff};
+  static AB1 shift[512]={
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x17,0x16,0x15,0x14,0x13,0x12,0x11,0x10,0x0f,
+   0x0e,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,
+   0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x17,0x16,0x15,0x14,0x13,0x12,0x11,0x10,0x0f,
+   0x0e,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,
+   0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x0d,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,
+   0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18};
+  union{AF1 f;AU1 u;}bits;bits.f=f;AU1 u=bits.u;AU1 i=u>>23;return (AU1)(base[i])+((u&0x7fffff)>>shift[i]);}
+//------------------------------------------------------------------------------------------------------------------------------
+ // Used to output packed constant.
+ A_STATIC AU1 AU1_AH2_AF2(inAF2 a){return AU1_AH1_AF1(a[0])+(AU1_AH1_AF1(a[1])<<16);}
+#endif
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                            GLSL
+//==============================================================================================================================
+#if defined(A_GLSL) && defined(A_GPU)
+ #ifndef A_SKIP_EXT
+  #ifdef A_LONG
+   #extension GL_ARB_gpu_shader_int64:require
+   #extension GL_NV_shader_atomic_int64:require
+  #endif
+//------------------------------------------------------------------------------------------------------------------------------
+  #ifdef A_WAVE
+   #extension GL_KHR_shader_subgroup_arithmetic:require
+   #extension GL_KHR_shader_subgroup_ballot:require
+   #extension GL_KHR_shader_subgroup_quad:require
+   #extension GL_KHR_shader_subgroup_shuffle:require
+  #endif
+ #endif
+//==============================================================================================================================
+ #define AP1 bool
+ #define AP2 bvec2
+ #define AP3 bvec3
+ #define AP4 bvec4
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AF1 float
+ #define AF2 vec2
+ #define AF3 vec3
+ #define AF4 vec4
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AU1 uint
+ #define AU2 uvec2
+ #define AU3 uvec3
+ #define AU4 uvec4
+//------------------------------------------------------------------------------------------------------------------------------
+ #define ASU1 int
+ #define ASU2 ivec2
+ #define ASU3 ivec3
+ #define ASU4 ivec4
+//==============================================================================================================================
+ #define AF1_AU1(x) uintBitsToFloat(AU1(x))
+ #define AF2_AU2(x) uintBitsToFloat(AU2(x))
+ #define AF3_AU3(x) uintBitsToFloat(AU3(x))
+ #define AF4_AU4(x) uintBitsToFloat(AU4(x))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AU1_AF1(x) floatBitsToUint(AF1(x))
+ #define AU2_AF2(x) floatBitsToUint(AF2(x))
+ #define AU3_AF3(x) floatBitsToUint(AF3(x))
+ #define AU4_AF4(x) floatBitsToUint(AF4(x))
+//==============================================================================================================================
+ AF1 AF1_x(AF1 a){return AF1(a);}
+ AF2 AF2_x(AF1 a){return AF2(a,a);}
+ AF3 AF3_x(AF1 a){return AF3(a,a,a);}
+ AF4 AF4_x(AF1 a){return AF4(a,a,a,a);}
+ #define AF1_(a) AF1_x(AF1(a))
+ #define AF2_(a) AF2_x(AF1(a))
+ #define AF3_(a) AF3_x(AF1(a))
+ #define AF4_(a) AF4_x(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AU1_x(AU1 a){return AU1(a);}
+ AU2 AU2_x(AU1 a){return AU2(a,a);}
+ AU3 AU3_x(AU1 a){return AU3(a,a,a);}
+ AU4 AU4_x(AU1 a){return AU4(a,a,a,a);}
+ #define AU1_(a) AU1_x(AU1(a))
+ #define AU2_(a) AU2_x(AU1(a))
+ #define AU3_(a) AU3_x(AU1(a))
+ #define AU4_(a) AU4_x(AU1(a))
+//==============================================================================================================================
+ AU1 AAbsSU1(AU1 a){return AU1(abs(ASU1(a)));}
+ AU2 AAbsSU2(AU2 a){return AU2(abs(ASU2(a)));}
+ AU3 AAbsSU3(AU3 a){return AU3(abs(ASU3(a)));}
+ AU4 AAbsSU4(AU4 a){return AU4(abs(ASU4(a)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 ABfe(AU1 src,AU1 off,AU1 bits){return bitfieldExtract(src,ASU1(off),ASU1(bits));}
+ AU1 ABfi(AU1 src,AU1 ins,AU1 mask){return (ins&mask)|(src&(~mask));}
+ // Proxy for V_BFI_B32 where the 'mask' is set as 'bits', 'mask=(1<<bits)-1', and 'bits' needs to be an immediate.
+ AU1 ABfiM(AU1 src,AU1 ins,AU1 bits){return bitfieldInsert(src,ins,0,ASU1(bits));}
+//------------------------------------------------------------------------------------------------------------------------------
+ // V_MED3_F32.
+ AF1 AClampF1(AF1 x,AF1 n,AF1 m){return clamp(x,n,m);}
+ AF2 AClampF2(AF2 x,AF2 n,AF2 m){return clamp(x,n,m);}
+ AF3 AClampF3(AF3 x,AF3 n,AF3 m){return clamp(x,n,m);}
+ AF4 AClampF4(AF4 x,AF4 n,AF4 m){return clamp(x,n,m);}
+//------------------------------------------------------------------------------------------------------------------------------
+ // V_FRACT_F32 (note DX frac() is different).
+ AF1 AFractF1(AF1 x){return fract(x);}
+ AF2 AFractF2(AF2 x){return fract(x);}
+ AF3 AFractF3(AF3 x){return fract(x);}
+ AF4 AFractF4(AF4 x){return fract(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ALerpF1(AF1 x,AF1 y,AF1 a){return mix(x,y,a);}
+ AF2 ALerpF2(AF2 x,AF2 y,AF2 a){return mix(x,y,a);}
+ AF3 ALerpF3(AF3 x,AF3 y,AF3 a){return mix(x,y,a);}
+ AF4 ALerpF4(AF4 x,AF4 y,AF4 a){return mix(x,y,a);}
+//------------------------------------------------------------------------------------------------------------------------------
+ // V_MAX3_F32.
+ AF1 AMax3F1(AF1 x,AF1 y,AF1 z){return max(x,max(y,z));}
+ AF2 AMax3F2(AF2 x,AF2 y,AF2 z){return max(x,max(y,z));}
+ AF3 AMax3F3(AF3 x,AF3 y,AF3 z){return max(x,max(y,z));}
+ AF4 AMax3F4(AF4 x,AF4 y,AF4 z){return max(x,max(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMax3SU1(AU1 x,AU1 y,AU1 z){return AU1(max(ASU1(x),max(ASU1(y),ASU1(z))));}
+ AU2 AMax3SU2(AU2 x,AU2 y,AU2 z){return AU2(max(ASU2(x),max(ASU2(y),ASU2(z))));}
+ AU3 AMax3SU3(AU3 x,AU3 y,AU3 z){return AU3(max(ASU3(x),max(ASU3(y),ASU3(z))));}
+ AU4 AMax3SU4(AU4 x,AU4 y,AU4 z){return AU4(max(ASU4(x),max(ASU4(y),ASU4(z))));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMax3U1(AU1 x,AU1 y,AU1 z){return max(x,max(y,z));}
+ AU2 AMax3U2(AU2 x,AU2 y,AU2 z){return max(x,max(y,z));}
+ AU3 AMax3U3(AU3 x,AU3 y,AU3 z){return max(x,max(y,z));}
+ AU4 AMax3U4(AU4 x,AU4 y,AU4 z){return max(x,max(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMaxSU1(AU1 a,AU1 b){return AU1(max(ASU1(a),ASU1(b)));}
+ AU2 AMaxSU2(AU2 a,AU2 b){return AU2(max(ASU2(a),ASU2(b)));}
+ AU3 AMaxSU3(AU3 a,AU3 b){return AU3(max(ASU3(a),ASU3(b)));}
+ AU4 AMaxSU4(AU4 a,AU4 b){return AU4(max(ASU4(a),ASU4(b)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ // Clamp has an easier pattern match for med3 when some ordering is known.
+ // V_MED3_F32.
+ AF1 AMed3F1(AF1 x,AF1 y,AF1 z){return max(min(x,y),min(max(x,y),z));}
+ AF2 AMed3F2(AF2 x,AF2 y,AF2 z){return max(min(x,y),min(max(x,y),z));}
+ AF3 AMed3F3(AF3 x,AF3 y,AF3 z){return max(min(x,y),min(max(x,y),z));}
+ AF4 AMed3F4(AF4 x,AF4 y,AF4 z){return max(min(x,y),min(max(x,y),z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ // V_MIN3_F32.
+ AF1 AMin3F1(AF1 x,AF1 y,AF1 z){return min(x,min(y,z));}
+ AF2 AMin3F2(AF2 x,AF2 y,AF2 z){return min(x,min(y,z));}
+ AF3 AMin3F3(AF3 x,AF3 y,AF3 z){return min(x,min(y,z));}
+ AF4 AMin3F4(AF4 x,AF4 y,AF4 z){return min(x,min(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMin3SU1(AU1 x,AU1 y,AU1 z){return AU1(min(ASU1(x),min(ASU1(y),ASU1(z))));}
+ AU2 AMin3SU2(AU2 x,AU2 y,AU2 z){return AU2(min(ASU2(x),min(ASU2(y),ASU2(z))));}
+ AU3 AMin3SU3(AU3 x,AU3 y,AU3 z){return AU3(min(ASU3(x),min(ASU3(y),ASU3(z))));}
+ AU4 AMin3SU4(AU4 x,AU4 y,AU4 z){return AU4(min(ASU4(x),min(ASU4(y),ASU4(z))));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMin3U1(AU1 x,AU1 y,AU1 z){return min(x,min(y,z));}
+ AU2 AMin3U2(AU2 x,AU2 y,AU2 z){return min(x,min(y,z));}
+ AU3 AMin3U3(AU3 x,AU3 y,AU3 z){return min(x,min(y,z));}
+ AU4 AMin3U4(AU4 x,AU4 y,AU4 z){return min(x,min(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMinSU1(AU1 a,AU1 b){return AU1(min(ASU1(a),ASU1(b)));}
+ AU2 AMinSU2(AU2 a,AU2 b){return AU2(min(ASU2(a),ASU2(b)));}
+ AU3 AMinSU3(AU3 a,AU3 b){return AU3(min(ASU3(a),ASU3(b)));}
+ AU4 AMinSU4(AU4 a,AU4 b){return AU4(min(ASU4(a),ASU4(b)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ // Normalized trig. Valid input domain is {-256 to +256}. No GLSL compiler intrinsic exists to map to this currently.
+ // V_COS_F32.
+ AF1 ANCosF1(AF1 x){return cos(x*AF1_(A_2PI));}
+ AF2 ANCosF2(AF2 x){return cos(x*AF2_(A_2PI));}
+ AF3 ANCosF3(AF3 x){return cos(x*AF3_(A_2PI));}
+ AF4 ANCosF4(AF4 x){return cos(x*AF4_(A_2PI));}
+//------------------------------------------------------------------------------------------------------------------------------
+ // Normalized trig. Valid input domain is {-256 to +256}. No GLSL compiler intrinsic exists to map to this currently.
+ // V_SIN_F32.
+ AF1 ANSinF1(AF1 x){return sin(x*AF1_(A_2PI));}
+ AF2 ANSinF2(AF2 x){return sin(x*AF2_(A_2PI));}
+ AF3 ANSinF3(AF3 x){return sin(x*AF3_(A_2PI));}
+ AF4 ANSinF4(AF4 x){return sin(x*AF4_(A_2PI));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ARcpF1(AF1 x){return AF1_(1.0)/x;}
+ AF2 ARcpF2(AF2 x){return AF2_(1.0)/x;}
+ AF3 ARcpF3(AF3 x){return AF3_(1.0)/x;}
+ AF4 ARcpF4(AF4 x){return AF4_(1.0)/x;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ARsqF1(AF1 x){return AF1_(1.0)/sqrt(x);}
+ AF2 ARsqF2(AF2 x){return AF2_(1.0)/sqrt(x);}
+ AF3 ARsqF3(AF3 x){return AF3_(1.0)/sqrt(x);}
+ AF4 ARsqF4(AF4 x){return AF4_(1.0)/sqrt(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ASatF1(AF1 x){return clamp(x,AF1_(0.0),AF1_(1.0));}
+ AF2 ASatF2(AF2 x){return clamp(x,AF2_(0.0),AF2_(1.0));}
+ AF3 ASatF3(AF3 x){return clamp(x,AF3_(0.0),AF3_(1.0));}
+ AF4 ASatF4(AF4 x){return clamp(x,AF4_(0.0),AF4_(1.0));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AShrSU1(AU1 a,AU1 b){return AU1(ASU1(a)>>ASU1(b));}
+ AU2 AShrSU2(AU2 a,AU2 b){return AU2(ASU2(a)>>ASU2(b));}
+ AU3 AShrSU3(AU3 a,AU3 b){return AU3(ASU3(a)>>ASU3(b));}
+ AU4 AShrSU4(AU4 a,AU4 b){return AU4(ASU4(a)>>ASU4(b));}
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                          GLSL BYTE
+//==============================================================================================================================
+ #ifdef A_BYTE
+  #define AB1 uint8_t
+  #define AB2 u8vec2
+  #define AB3 u8vec3
+  #define AB4 u8vec4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define ASB1 int8_t
+  #define ASB2 i8vec2
+  #define ASB3 i8vec3
+  #define ASB4 i8vec4
+//------------------------------------------------------------------------------------------------------------------------------
+  AB1 AB1_x(AB1 a){return AB1(a);}
+  AB2 AB2_x(AB1 a){return AB2(a,a);}
+  AB3 AB3_x(AB1 a){return AB3(a,a,a);}
+  AB4 AB4_x(AB1 a){return AB4(a,a,a,a);}
+  #define AB1_(a) AB1_x(AB1(a))
+  #define AB2_(a) AB2_x(AB1(a))
+  #define AB3_(a) AB3_x(AB1(a))
+  #define AB4_(a) AB4_x(AB1(a))
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                         GLSL DOUBLE
+//==============================================================================================================================
+ #ifdef A_DUBL
+  #define AD1 double
+  #define AD2 dvec2
+  #define AD3 dvec3
+  #define AD4 dvec4
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 AD1_x(AD1 a){return AD1(a);}
+  AD2 AD2_x(AD1 a){return AD2(a,a);}
+  AD3 AD3_x(AD1 a){return AD3(a,a,a);}
+  AD4 AD4_x(AD1 a){return AD4(a,a,a,a);}
+  #define AD1_(a) AD1_x(AD1(a))
+  #define AD2_(a) AD2_x(AD1(a))
+  #define AD3_(a) AD3_x(AD1(a))
+  #define AD4_(a) AD4_x(AD1(a))
+//==============================================================================================================================
+  AD1 AFractD1(AD1 x){return fract(x);}
+  AD2 AFractD2(AD2 x){return fract(x);}
+  AD3 AFractD3(AD3 x){return fract(x);}
+  AD4 AFractD4(AD4 x){return fract(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 ALerpD1(AD1 x,AD1 y,AD1 a){return mix(x,y,a);}
+  AD2 ALerpD2(AD2 x,AD2 y,AD2 a){return mix(x,y,a);}
+  AD3 ALerpD3(AD3 x,AD3 y,AD3 a){return mix(x,y,a);}
+  AD4 ALerpD4(AD4 x,AD4 y,AD4 a){return mix(x,y,a);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 ARcpD1(AD1 x){return AD1_(1.0)/x;}
+  AD2 ARcpD2(AD2 x){return AD2_(1.0)/x;}
+  AD3 ARcpD3(AD3 x){return AD3_(1.0)/x;}
+  AD4 ARcpD4(AD4 x){return AD4_(1.0)/x;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 ARsqD1(AD1 x){return AD1_(1.0)/sqrt(x);}
+  AD2 ARsqD2(AD2 x){return AD2_(1.0)/sqrt(x);}
+  AD3 ARsqD3(AD3 x){return AD3_(1.0)/sqrt(x);}
+  AD4 ARsqD4(AD4 x){return AD4_(1.0)/sqrt(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 ASatD1(AD1 x){return clamp(x,AD1_(0.0),AD1_(1.0));}
+  AD2 ASatD2(AD2 x){return clamp(x,AD2_(0.0),AD2_(1.0));}
+  AD3 ASatD3(AD3 x){return clamp(x,AD3_(0.0),AD3_(1.0));}
+  AD4 ASatD4(AD4 x){return clamp(x,AD4_(0.0),AD4_(1.0));}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                         GLSL LONG
+//==============================================================================================================================
+ #ifdef A_LONG
+  #define AL1 uint64_t
+  #define AL2 u64vec2
+  #define AL3 u64vec3
+  #define AL4 u64vec4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define ASL1 int64_t
+  #define ASL2 i64vec2
+  #define ASL3 i64vec3
+  #define ASL4 i64vec4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define AL1_AU2(x) packUint2x32(AU2(x))
+  #define AU2_AL1(x) unpackUint2x32(AL1(x))
+//------------------------------------------------------------------------------------------------------------------------------
+  AL1 AL1_x(AL1 a){return AL1(a);}
+  AL2 AL2_x(AL1 a){return AL2(a,a);}
+  AL3 AL3_x(AL1 a){return AL3(a,a,a);}
+  AL4 AL4_x(AL1 a){return AL4(a,a,a,a);}
+  #define AL1_(a) AL1_x(AL1(a))
+  #define AL2_(a) AL2_x(AL1(a))
+  #define AL3_(a) AL3_x(AL1(a))
+  #define AL4_(a) AL4_x(AL1(a))
+//==============================================================================================================================
+  AL1 AAbsSL1(AL1 a){return AL1(abs(ASL1(a)));}
+  AL2 AAbsSL2(AL2 a){return AL2(abs(ASL2(a)));}
+  AL3 AAbsSL3(AL3 a){return AL3(abs(ASL3(a)));}
+  AL4 AAbsSL4(AL4 a){return AL4(abs(ASL4(a)));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AL1 AMaxSL1(AL1 a,AL1 b){return AL1(max(ASU1(a),ASU1(b)));}
+  AL2 AMaxSL2(AL2 a,AL2 b){return AL2(max(ASU2(a),ASU2(b)));}
+  AL3 AMaxSL3(AL3 a,AL3 b){return AL3(max(ASU3(a),ASU3(b)));}
+  AL4 AMaxSL4(AL4 a,AL4 b){return AL4(max(ASU4(a),ASU4(b)));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AL1 AMinSL1(AL1 a,AL1 b){return AL1(min(ASU1(a),ASU1(b)));}
+  AL2 AMinSL2(AL2 a,AL2 b){return AL2(min(ASU2(a),ASU2(b)));}
+  AL3 AMinSL3(AL3 a,AL3 b){return AL3(min(ASU3(a),ASU3(b)));}
+  AL4 AMinSL4(AL4 a,AL4 b){return AL4(min(ASU4(a),ASU4(b)));}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                      WAVE OPERATIONS
+//==============================================================================================================================
+ #ifdef A_WAVE
+  // Where 'x' must be a compile time literal.
+  AF1 AWaveXorF1(AF1 v,AU1 x){return subgroupShuffleXor(v,x);}
+  AF2 AWaveXorF2(AF2 v,AU1 x){return subgroupShuffleXor(v,x);}
+  AF3 AWaveXorF3(AF3 v,AU1 x){return subgroupShuffleXor(v,x);}
+  AF4 AWaveXorF4(AF4 v,AU1 x){return subgroupShuffleXor(v,x);}
+  AU1 AWaveXorU1(AU1 v,AU1 x){return subgroupShuffleXor(v,x);}
+  AU2 AWaveXorU2(AU2 v,AU1 x){return subgroupShuffleXor(v,x);}
+  AU3 AWaveXorU3(AU3 v,AU1 x){return subgroupShuffleXor(v,x);}
+  AU4 AWaveXorU4(AU4 v,AU1 x){return subgroupShuffleXor(v,x);}
+ #endif
+//==============================================================================================================================
+#endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//
+//                                                            HLSL
+//
+//
+//==============================================================================================================================
+#if defined(A_HLSL) && defined(A_GPU)
+ #ifdef A_HLSL_6_2
+  #define AP1 bool
+  #define AP2 bool2
+  #define AP3 bool3
+  #define AP4 bool4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define AF1 float32_t
+  #define AF2 float32_t2
+  #define AF3 float32_t3
+  #define AF4 float32_t4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define AU1 uint32_t
+  #define AU2 uint32_t2
+  #define AU3 uint32_t3
+  #define AU4 uint32_t4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define ASU1 int32_t
+  #define ASU2 int32_t2
+  #define ASU3 int32_t3
+  #define ASU4 int32_t4
+ #else
+  #define AP1 bool
+  #define AP2 bool2
+  #define AP3 bool3
+  #define AP4 bool4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define AF1 float
+  #define AF2 float2
+  #define AF3 float3
+  #define AF4 float4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define AU1 uint
+  #define AU2 uint2
+  #define AU3 uint3
+  #define AU4 uint4
+//------------------------------------------------------------------------------------------------------------------------------
+  #define ASU1 int
+  #define ASU2 int2
+  #define ASU3 int3
+  #define ASU4 int4
+ #endif
+//==============================================================================================================================
+ #define AF1_AU1(x) asfloat(AU1(x))
+ #define AF2_AU2(x) asfloat(AU2(x))
+ #define AF3_AU3(x) asfloat(AU3(x))
+ #define AF4_AU4(x) asfloat(AU4(x))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AU1_AF1(x) asuint(AF1(x))
+ #define AU2_AF2(x) asuint(AF2(x))
+ #define AU3_AF3(x) asuint(AF3(x))
+ #define AU4_AF4(x) asuint(AF4(x))
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AU1_AH1_AF1_x(AF1 a){return f32tof16(a);}
+ #define AU1_AH1_AF1(a) AU1_AH1_AF1_x(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AU1_AH2_AF2_x(AF2 a){return f32tof16(a.x)|(f32tof16(a.y)<<16);}
+ #define AU1_AH2_AF2(a) AU1_AH2_AF2_x(AF2(a))
+ #define AU1_AB4Unorm_AF4(x) D3DCOLORtoUBYTE4(AF4(x))
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 AF2_AH2_AU1_x(AU1 x){return AF2(f16tof32(x&0xFFFF),f16tof32(x>>16));}
+ #define AF2_AH2_AU1(x) AF2_AH2_AU1_x(AU1(x))
+//==============================================================================================================================
+ AF1 AF1_x(AF1 a){return AF1(a);}
+ AF2 AF2_x(AF1 a){return AF2(a,a);}
+ AF3 AF3_x(AF1 a){return AF3(a,a,a);}
+ AF4 AF4_x(AF1 a){return AF4(a,a,a,a);}
+ #define AF1_(a) AF1_x(AF1(a))
+ #define AF2_(a) AF2_x(AF1(a))
+ #define AF3_(a) AF3_x(AF1(a))
+ #define AF4_(a) AF4_x(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AU1_x(AU1 a){return AU1(a);}
+ AU2 AU2_x(AU1 a){return AU2(a,a);}
+ AU3 AU3_x(AU1 a){return AU3(a,a,a);}
+ AU4 AU4_x(AU1 a){return AU4(a,a,a,a);}
+ #define AU1_(a) AU1_x(AU1(a))
+ #define AU2_(a) AU2_x(AU1(a))
+ #define AU3_(a) AU3_x(AU1(a))
+ #define AU4_(a) AU4_x(AU1(a))
+//==============================================================================================================================
+ AU1 AAbsSU1(AU1 a){return AU1(abs(ASU1(a)));}
+ AU2 AAbsSU2(AU2 a){return AU2(abs(ASU2(a)));}
+ AU3 AAbsSU3(AU3 a){return AU3(abs(ASU3(a)));}
+ AU4 AAbsSU4(AU4 a){return AU4(abs(ASU4(a)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 ABfe(AU1 src,AU1 off,AU1 bits){AU1 mask=(1u<<bits)-1;return (src>>off)&mask;}
+ AU1 ABfi(AU1 src,AU1 ins,AU1 mask){return (ins&mask)|(src&(~mask));}
+ AU1 ABfiM(AU1 src,AU1 ins,AU1 bits){AU1 mask=(1u<<bits)-1;return (ins&mask)|(src&(~mask));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 AClampF1(AF1 x,AF1 n,AF1 m){return max(n,min(x,m));}
+ AF2 AClampF2(AF2 x,AF2 n,AF2 m){return max(n,min(x,m));}
+ AF3 AClampF3(AF3 x,AF3 n,AF3 m){return max(n,min(x,m));}
+ AF4 AClampF4(AF4 x,AF4 n,AF4 m){return max(n,min(x,m));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 AFractF1(AF1 x){return x-floor(x);}
+ AF2 AFractF2(AF2 x){return x-floor(x);}
+ AF3 AFractF3(AF3 x){return x-floor(x);}
+ AF4 AFractF4(AF4 x){return x-floor(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ALerpF1(AF1 x,AF1 y,AF1 a){return lerp(x,y,a);}
+ AF2 ALerpF2(AF2 x,AF2 y,AF2 a){return lerp(x,y,a);}
+ AF3 ALerpF3(AF3 x,AF3 y,AF3 a){return lerp(x,y,a);}
+ AF4 ALerpF4(AF4 x,AF4 y,AF4 a){return lerp(x,y,a);}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 AMax3F1(AF1 x,AF1 y,AF1 z){return max(x,max(y,z));}
+ AF2 AMax3F2(AF2 x,AF2 y,AF2 z){return max(x,max(y,z));}
+ AF3 AMax3F3(AF3 x,AF3 y,AF3 z){return max(x,max(y,z));}
+ AF4 AMax3F4(AF4 x,AF4 y,AF4 z){return max(x,max(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMax3SU1(AU1 x,AU1 y,AU1 z){return AU1(max(ASU1(x),max(ASU1(y),ASU1(z))));}
+ AU2 AMax3SU2(AU2 x,AU2 y,AU2 z){return AU2(max(ASU2(x),max(ASU2(y),ASU2(z))));}
+ AU3 AMax3SU3(AU3 x,AU3 y,AU3 z){return AU3(max(ASU3(x),max(ASU3(y),ASU3(z))));}
+ AU4 AMax3SU4(AU4 x,AU4 y,AU4 z){return AU4(max(ASU4(x),max(ASU4(y),ASU4(z))));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMax3U1(AU1 x,AU1 y,AU1 z){return max(x,max(y,z));}
+ AU2 AMax3U2(AU2 x,AU2 y,AU2 z){return max(x,max(y,z));}
+ AU3 AMax3U3(AU3 x,AU3 y,AU3 z){return max(x,max(y,z));}
+ AU4 AMax3U4(AU4 x,AU4 y,AU4 z){return max(x,max(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMaxSU1(AU1 a,AU1 b){return AU1(max(ASU1(a),ASU1(b)));}
+ AU2 AMaxSU2(AU2 a,AU2 b){return AU2(max(ASU2(a),ASU2(b)));}
+ AU3 AMaxSU3(AU3 a,AU3 b){return AU3(max(ASU3(a),ASU3(b)));}
+ AU4 AMaxSU4(AU4 a,AU4 b){return AU4(max(ASU4(a),ASU4(b)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 AMed3F1(AF1 x,AF1 y,AF1 z){return max(min(x,y),min(max(x,y),z));}
+ AF2 AMed3F2(AF2 x,AF2 y,AF2 z){return max(min(x,y),min(max(x,y),z));}
+ AF3 AMed3F3(AF3 x,AF3 y,AF3 z){return max(min(x,y),min(max(x,y),z));}
+ AF4 AMed3F4(AF4 x,AF4 y,AF4 z){return max(min(x,y),min(max(x,y),z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 AMin3F1(AF1 x,AF1 y,AF1 z){return min(x,min(y,z));}
+ AF2 AMin3F2(AF2 x,AF2 y,AF2 z){return min(x,min(y,z));}
+ AF3 AMin3F3(AF3 x,AF3 y,AF3 z){return min(x,min(y,z));}
+ AF4 AMin3F4(AF4 x,AF4 y,AF4 z){return min(x,min(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMin3SU1(AU1 x,AU1 y,AU1 z){return AU1(min(ASU1(x),min(ASU1(y),ASU1(z))));}
+ AU2 AMin3SU2(AU2 x,AU2 y,AU2 z){return AU2(min(ASU2(x),min(ASU2(y),ASU2(z))));}
+ AU3 AMin3SU3(AU3 x,AU3 y,AU3 z){return AU3(min(ASU3(x),min(ASU3(y),ASU3(z))));}
+ AU4 AMin3SU4(AU4 x,AU4 y,AU4 z){return AU4(min(ASU4(x),min(ASU4(y),ASU4(z))));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMin3U1(AU1 x,AU1 y,AU1 z){return min(x,min(y,z));}
+ AU2 AMin3U2(AU2 x,AU2 y,AU2 z){return min(x,min(y,z));}
+ AU3 AMin3U3(AU3 x,AU3 y,AU3 z){return min(x,min(y,z));}
+ AU4 AMin3U4(AU4 x,AU4 y,AU4 z){return min(x,min(y,z));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AMinSU1(AU1 a,AU1 b){return AU1(min(ASU1(a),ASU1(b)));}
+ AU2 AMinSU2(AU2 a,AU2 b){return AU2(min(ASU2(a),ASU2(b)));}
+ AU3 AMinSU3(AU3 a,AU3 b){return AU3(min(ASU3(a),ASU3(b)));}
+ AU4 AMinSU4(AU4 a,AU4 b){return AU4(min(ASU4(a),ASU4(b)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ANCosF1(AF1 x){return cos(x*AF1_(A_2PI));}
+ AF2 ANCosF2(AF2 x){return cos(x*AF2_(A_2PI));}
+ AF3 ANCosF3(AF3 x){return cos(x*AF3_(A_2PI));}
+ AF4 ANCosF4(AF4 x){return cos(x*AF4_(A_2PI));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ANSinF1(AF1 x){return sin(x*AF1_(A_2PI));}
+ AF2 ANSinF2(AF2 x){return sin(x*AF2_(A_2PI));}
+ AF3 ANSinF3(AF3 x){return sin(x*AF3_(A_2PI));}
+ AF4 ANSinF4(AF4 x){return sin(x*AF4_(A_2PI));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ARcpF1(AF1 x){return rcp(x);}
+ AF2 ARcpF2(AF2 x){return rcp(x);}
+ AF3 ARcpF3(AF3 x){return rcp(x);}
+ AF4 ARcpF4(AF4 x){return rcp(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ARsqF1(AF1 x){return rsqrt(x);}
+ AF2 ARsqF2(AF2 x){return rsqrt(x);}
+ AF3 ARsqF3(AF3 x){return rsqrt(x);}
+ AF4 ARsqF4(AF4 x){return rsqrt(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 ASatF1(AF1 x){return saturate(x);}
+ AF2 ASatF2(AF2 x){return saturate(x);}
+ AF3 ASatF3(AF3 x){return saturate(x);}
+ AF4 ASatF4(AF4 x){return saturate(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+ AU1 AShrSU1(AU1 a,AU1 b){return AU1(ASU1(a)>>ASU1(b));}
+ AU2 AShrSU2(AU2 a,AU2 b){return AU2(ASU2(a)>>ASU2(b));}
+ AU3 AShrSU3(AU3 a,AU3 b){return AU3(ASU3(a)>>ASU3(b));}
+ AU4 AShrSU4(AU4 a,AU4 b){return AU4(ASU4(a)>>ASU4(b));}
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                          HLSL BYTE
+//==============================================================================================================================
+ #ifdef A_BYTE
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                         HLSL DOUBLE
+//==============================================================================================================================
+ #ifdef A_DUBL
+  #ifdef A_HLSL_6_2
+   #define AD1 float64_t
+   #define AD2 float64_t2
+   #define AD3 float64_t3
+   #define AD4 float64_t4
+  #else
+   #define AD1 double
+   #define AD2 double2
+   #define AD3 double3
+   #define AD4 double4
+  #endif
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 AD1_x(AD1 a){return AD1(a);}
+  AD2 AD2_x(AD1 a){return AD2(a,a);}
+  AD3 AD3_x(AD1 a){return AD3(a,a,a);}
+  AD4 AD4_x(AD1 a){return AD4(a,a,a,a);}
+  #define AD1_(a) AD1_x(AD1(a))
+  #define AD2_(a) AD2_x(AD1(a))
+  #define AD3_(a) AD3_x(AD1(a))
+  #define AD4_(a) AD4_x(AD1(a))
+//==============================================================================================================================
+  AD1 AFractD1(AD1 a){return a-floor(a);}
+  AD2 AFractD2(AD2 a){return a-floor(a);}
+  AD3 AFractD3(AD3 a){return a-floor(a);}
+  AD4 AFractD4(AD4 a){return a-floor(a);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 ALerpD1(AD1 x,AD1 y,AD1 a){return lerp(x,y,a);}
+  AD2 ALerpD2(AD2 x,AD2 y,AD2 a){return lerp(x,y,a);}
+  AD3 ALerpD3(AD3 x,AD3 y,AD3 a){return lerp(x,y,a);}
+  AD4 ALerpD4(AD4 x,AD4 y,AD4 a){return lerp(x,y,a);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 ARcpD1(AD1 x){return rcp(x);}
+  AD2 ARcpD2(AD2 x){return rcp(x);}
+  AD3 ARcpD3(AD3 x){return rcp(x);}
+  AD4 ARcpD4(AD4 x){return rcp(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 ARsqD1(AD1 x){return rsqrt(x);}
+  AD2 ARsqD2(AD2 x){return rsqrt(x);}
+  AD3 ARsqD3(AD3 x){return rsqrt(x);}
+  AD4 ARsqD4(AD4 x){return rsqrt(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD1 ASatD1(AD1 x){return saturate(x);}
+  AD2 ASatD2(AD2 x){return saturate(x);}
+  AD3 ASatD3(AD3 x){return saturate(x);}
+  AD4 ASatD4(AD4 x){return saturate(x);}
+ #endif
+//==============================================================================================================================
+//                                                         HLSL WAVE
+//==============================================================================================================================
+ #ifdef A_WAVE
+  // Where 'x' must be a compile time literal.
+  AF1 AWaveXorF1(AF1 v,AU1 x){return WaveReadLaneAt(v,WaveGetLaneIndex()^x);}
+  AF2 AWaveXorF2(AF2 v,AU1 x){return WaveReadLaneAt(v,WaveGetLaneIndex()^x);}
+  AF3 AWaveXorF3(AF3 v,AU1 x){return WaveReadLaneAt(v,WaveGetLaneIndex()^x);}
+  AF4 AWaveXorF4(AF4 v,AU1 x){return WaveReadLaneAt(v,WaveGetLaneIndex()^x);}
+  AU1 AWaveXorU1(AU1 v,AU1 x){return WaveReadLaneAt(v,WaveGetLaneIndex()^x);}
+  AU2 AWaveXorU1(AU2 v,AU1 x){return WaveReadLaneAt(v,WaveGetLaneIndex()^x);}
+  AU3 AWaveXorU1(AU3 v,AU1 x){return WaveReadLaneAt(v,WaveGetLaneIndex()^x);}
+  AU4 AWaveXorU1(AU4 v,AU1 x){return WaveReadLaneAt(v,WaveGetLaneIndex()^x);}
+ #endif
+//==============================================================================================================================
+#endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//
+//                                                          GPU COMMON
+//
+//
+//==============================================================================================================================
+#ifdef A_GPU
+ // Negative and positive infinity.
+ #define A_INFP_F AF1_AU1(0x7f800000u)
+ #define A_INFN_F AF1_AU1(0xff800000u)
+//------------------------------------------------------------------------------------------------------------------------------
+ // Copy sign from 's' to positive 'd'.
+ AF1 ACpySgnF1(AF1 d,AF1 s){return AF1_AU1(AU1_AF1(d)|(AU1_AF1(s)&AU1_(0x80000000u)));}
+ AF2 ACpySgnF2(AF2 d,AF2 s){return AF2_AU2(AU2_AF2(d)|(AU2_AF2(s)&AU2_(0x80000000u)));}
+ AF3 ACpySgnF3(AF3 d,AF3 s){return AF3_AU3(AU3_AF3(d)|(AU3_AF3(s)&AU3_(0x80000000u)));}
+ AF4 ACpySgnF4(AF4 d,AF4 s){return AF4_AU4(AU4_AF4(d)|(AU4_AF4(s)&AU4_(0x80000000u)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ // Single operation to return (useful to create a mask to use in lerp for branch free logic),
+ //  m=NaN := 0
+ //  m>=0  := 0
+ //  m<0   := 1
+ // Uses the following useful floating point logic,
+ //  saturate(+a*(-INF)==-INF) := 0
+ //  saturate( 0*(-INF)== NaN) := 0
+ //  saturate(-a*(-INF)==+INF) := 1
+ AF1 ASignedF1(AF1 m){return ASatF1(m*AF1_(A_INFN_F));}
+ AF2 ASignedF2(AF2 m){return ASatF2(m*AF2_(A_INFN_F));}
+ AF3 ASignedF3(AF3 m){return ASatF3(m*AF3_(A_INFN_F));}
+ AF4 ASignedF4(AF4 m){return ASatF4(m*AF4_(A_INFN_F));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF1 AGtZeroF1(AF1 m){return ASatF1(m*AF1_(A_INFP_F));}
+ AF2 AGtZeroF2(AF2 m){return ASatF2(m*AF2_(A_INFP_F));}
+ AF3 AGtZeroF3(AF3 m){return ASatF3(m*AF3_(A_INFP_F));}
+ AF4 AGtZeroF4(AF4 m){return ASatF4(m*AF4_(A_INFP_F));}
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                [FIS] FLOAT INTEGER SORTABLE
+//------------------------------------------------------------------------------------------------------------------------------
+// Float to integer sortable.
+//  - If sign bit=0, flip the sign bit (positives).
+//  - If sign bit=1, flip all bits     (negatives).
+// Integer sortable to float.
+//  - If sign bit=1, flip the sign bit (positives).
+//  - If sign bit=0, flip all bits     (negatives).
+// Has nice side effects.
+//  - Larger integers are more positive values.
+//  - Float zero is mapped to center of integers (so clear to integer zero is a nice default for atomic max usage).
+// Burns 3 ops for conversion {shift,or,xor}.
+//==============================================================================================================================
+ AU1 AFisToU1(AU1 x){return x^(( AShrSU1(x,AU1_(31)))|AU1_(0x80000000));}
+ AU1 AFisFromU1(AU1 x){return x^((~AShrSU1(x,AU1_(31)))|AU1_(0x80000000));}
+//------------------------------------------------------------------------------------------------------------------------------
+ // Just adjust high 16-bit value (useful when upper part of 32-bit word is a 16-bit float value).
+ AU1 AFisToHiU1(AU1 x){return x^(( AShrSU1(x,AU1_(15)))|AU1_(0x80000000));}
+ AU1 AFisFromHiU1(AU1 x){return x^((~AShrSU1(x,AU1_(15)))|AU1_(0x80000000));}
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                               [BUC] BYTE UNSIGNED CONVERSION
+//------------------------------------------------------------------------------------------------------------------------------
+// Designed to use the optimal conversion, enables the scaling to possibly be factored into other computation.
+// Works on a range of {0 to A_BUC_<32,16>}, for <32-bit, and 16-bit> respectively.
+//------------------------------------------------------------------------------------------------------------------------------
+// OPCODE NOTES
+// ============
+// GCN does not do UNORM or SNORM for bytes in opcodes.
+//  - V_CVT_F32_UBYTE{0,1,2,3} - Unsigned byte to float.
+//  - V_CVT_PKACC_U8_F32 - Float to unsigned byte (does bit-field insert into 32-bit integer).
+// V_PERM_B32 does byte packing with ability to zero fill bytes as well.
+//  - Can pull out byte values from two sources, and zero fill upper 8-bits of packed hi and lo.
+//------------------------------------------------------------------------------------------------------------------------------
+// BYTE : FLOAT - ABuc{0,1,2,3}{To,From}U1() - Designed for V_CVT_F32_UBYTE* and V_CVT_PKACCUM_U8_F32 ops.
+// ====   =====
+//    0 : 0
+//    1 : 1
+//     ...
+//  255 : 255
+//      : 256 (just outside the encoding range)
+//------------------------------------------------------------------------------------------------------------------------------
+// BYTE : FLOAT - ABuc{0,1,2,3}{To,From}U2() - Designed for 16-bit denormal tricks and V_PERM_B32.
+// ====   =====
+//    0 : 0
+//    1 : 1/512
+//    2 : 1/256
+//     ...
+//   64 : 1/8
+//  128 : 1/4
+//  255 : 255/512
+//      : 1/2 (just outside the encoding range)
+//------------------------------------------------------------------------------------------------------------------------------
+// OPTIMAL IMPLEMENTATIONS ON AMD ARCHITECTURES
+// ============================================
+// r=ABuc0FromU1(i)
+//   V_CVT_F32_UBYTE0 r,i
+// --------------------------------------------
+// r=ABuc0ToU1(d,i)
+//   V_CVT_PKACCUM_U8_F32 r,i,0,d
+// --------------------------------------------
+// d=ABuc0FromU2(i)
+//   Where 'k0' is an SGPR with 0x0E0A
+//   Where 'k1' is an SGPR with {32768.0} packed into the lower 16-bits
+//   V_PERM_B32 d,i.x,i.y,k0
+//   V_PK_FMA_F16 d,d,k1.x,0
+// --------------------------------------------
+// r=ABuc0ToU2(d,i)
+//   Where 'k0' is an SGPR with {1.0/32768.0} packed into the lower 16-bits
+//   Where 'k1' is an SGPR with 0x????
+//   Where 'k2' is an SGPR with 0x????
+//   V_PK_FMA_F16 i,i,k0.x,0
+//   V_PERM_B32 r.x,i,i,k1
+//   V_PERM_B32 r.y,i,i,k2
+//==============================================================================================================================
+ // Peak range for 32-bit and 16-bit operations.
+ #define A_BUC_32 (255.0)
+ #define A_BUC_16 (255.0/512.0)
+//==============================================================================================================================
+ #if 1
+  // Designed to be one V_CVT_PKACCUM_U8_F32.
+  // The extra min is required to pattern match to V_CVT_PKACCUM_U8_F32.
+  AU1 ABuc0ToU1(AU1 d,AF1 i){return (d&0xffffff00u)|((min(AU1(i),255u)    )&(0x000000ffu));}
+  AU1 ABuc1ToU1(AU1 d,AF1 i){return (d&0xffff00ffu)|((min(AU1(i),255u)<< 8)&(0x0000ff00u));}
+  AU1 ABuc2ToU1(AU1 d,AF1 i){return (d&0xff00ffffu)|((min(AU1(i),255u)<<16)&(0x00ff0000u));}
+  AU1 ABuc3ToU1(AU1 d,AF1 i){return (d&0x00ffffffu)|((min(AU1(i),255u)<<24)&(0xff000000u));}
+//------------------------------------------------------------------------------------------------------------------------------
+  // Designed to be one V_CVT_F32_UBYTE*.
+  AF1 ABuc0FromU1(AU1 i){return AF1((i    )&255u);}
+  AF1 ABuc1FromU1(AU1 i){return AF1((i>> 8)&255u);}
+  AF1 ABuc2FromU1(AU1 i){return AF1((i>>16)&255u);}
+  AF1 ABuc3FromU1(AU1 i){return AF1((i>>24)&255u);}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                 [BSC] BYTE SIGNED CONVERSION
+//------------------------------------------------------------------------------------------------------------------------------
+// Similar to [BUC].
+// Works on a range of {-/+ A_BSC_<32,16>}, for <32-bit, and 16-bit> respectively.
+//------------------------------------------------------------------------------------------------------------------------------
+// ENCODING (without zero-based encoding)
+// ========
+//   0 = unused (can be used to mean something else)
+//   1 = lowest value
+// 128 = exact zero center (zero based encoding
+// 255 = highest value
+//------------------------------------------------------------------------------------------------------------------------------
+// Zero-based [Zb] flips the MSB bit of the byte (making 128 "exact zero" actually zero).
+// This is useful if there is a desire for cleared values to decode as zero.
+//------------------------------------------------------------------------------------------------------------------------------
+// BYTE : FLOAT - ABsc{0,1,2,3}{To,From}U2() - Designed for 16-bit denormal tricks and V_PERM_B32.
+// ====   =====
+//    0 : -127/512 (unused)
+//    1 : -126/512
+//    2 : -125/512
+//     ...
+//  128 : 0
+//     ...
+//  255 : 127/512
+//      : 1/4 (just outside the encoding range)
+//==============================================================================================================================
+ // Peak range for 32-bit and 16-bit operations.
+ #define A_BSC_32 (127.0)
+ #define A_BSC_16 (127.0/512.0)
+//==============================================================================================================================
+ #if 1
+  AU1 ABsc0ToU1(AU1 d,AF1 i){return (d&0xffffff00u)|((min(AU1(i+128.0),255u)    )&(0x000000ffu));}
+  AU1 ABsc1ToU1(AU1 d,AF1 i){return (d&0xffff00ffu)|((min(AU1(i+128.0),255u)<< 8)&(0x0000ff00u));}
+  AU1 ABsc2ToU1(AU1 d,AF1 i){return (d&0xff00ffffu)|((min(AU1(i+128.0),255u)<<16)&(0x00ff0000u));}
+  AU1 ABsc3ToU1(AU1 d,AF1 i){return (d&0x00ffffffu)|((min(AU1(i+128.0),255u)<<24)&(0xff000000u));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AU1 ABsc0ToZbU1(AU1 d,AF1 i){return ((d&0xffffff00u)|((min(AU1(trunc(i)+128.0),255u)    )&(0x000000ffu)))^0x00000080u;}
+  AU1 ABsc1ToZbU1(AU1 d,AF1 i){return ((d&0xffff00ffu)|((min(AU1(trunc(i)+128.0),255u)<< 8)&(0x0000ff00u)))^0x00008000u;}
+  AU1 ABsc2ToZbU1(AU1 d,AF1 i){return ((d&0xff00ffffu)|((min(AU1(trunc(i)+128.0),255u)<<16)&(0x00ff0000u)))^0x00800000u;}
+  AU1 ABsc3ToZbU1(AU1 d,AF1 i){return ((d&0x00ffffffu)|((min(AU1(trunc(i)+128.0),255u)<<24)&(0xff000000u)))^0x80000000u;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 ABsc0FromU1(AU1 i){return AF1((i    )&255u)-128.0;}
+  AF1 ABsc1FromU1(AU1 i){return AF1((i>> 8)&255u)-128.0;}
+  AF1 ABsc2FromU1(AU1 i){return AF1((i>>16)&255u)-128.0;}
+  AF1 ABsc3FromU1(AU1 i){return AF1((i>>24)&255u)-128.0;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 ABsc0FromZbU1(AU1 i){return AF1(((i    )&255u)^0x80u)-128.0;}
+  AF1 ABsc1FromZbU1(AU1 i){return AF1(((i>> 8)&255u)^0x80u)-128.0;}
+  AF1 ABsc2FromZbU1(AU1 i){return AF1(((i>>16)&255u)^0x80u)-128.0;}
+  AF1 ABsc3FromZbU1(AU1 i){return AF1(((i>>24)&255u)^0x80u)-128.0;}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                    FLOAT APPROXIMATIONS
+//------------------------------------------------------------------------------------------------------------------------------
+// Michal Drobot has an excellent presentation on these: "Low Level Optimizations For GCN",
+//  - Idea dates back to SGI, then to Quake 3, etc.
+//  - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf
+//     - sqrt(x)=rsqrt(x)*x
+//     - rcp(x)=rsqrt(x)*rsqrt(x) for positive x
+//  - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h
+//------------------------------------------------------------------------------------------------------------------------------
+// These below are from perhaps less complete searching for optimal.
+// Used FP16 normal range for testing with +4096 32-bit step size for sampling error.
+// So these match up well with the half approximations.
+//==============================================================================================================================
+ AF1 APrxLoSqrtF1(AF1 a){return AF1_AU1((AU1_AF1(a)>>AU1_(1))+AU1_(0x1fbc4639));}
+ AF1 APrxLoRcpF1(AF1 a){return AF1_AU1(AU1_(0x7ef07ebb)-AU1_AF1(a));}
+ AF1 APrxMedRcpF1(AF1 a){AF1 b=AF1_AU1(AU1_(0x7ef19fff)-AU1_AF1(a));return b*(-b*a+AF1_(2.0));}
+ AF1 APrxLoRsqF1(AF1 a){return AF1_AU1(AU1_(0x5f347d74)-(AU1_AF1(a)>>AU1_(1)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 APrxLoSqrtF2(AF2 a){return AF2_AU2((AU2_AF2(a)>>AU2_(1))+AU2_(0x1fbc4639));}
+ AF2 APrxLoRcpF2(AF2 a){return AF2_AU2(AU2_(0x7ef07ebb)-AU2_AF2(a));}
+ AF2 APrxMedRcpF2(AF2 a){AF2 b=AF2_AU2(AU2_(0x7ef19fff)-AU2_AF2(a));return b*(-b*a+AF2_(2.0));}
+ AF2 APrxLoRsqF2(AF2 a){return AF2_AU2(AU2_(0x5f347d74)-(AU2_AF2(a)>>AU2_(1)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF3 APrxLoSqrtF3(AF3 a){return AF3_AU3((AU3_AF3(a)>>AU3_(1))+AU3_(0x1fbc4639));}
+ AF3 APrxLoRcpF3(AF3 a){return AF3_AU3(AU3_(0x7ef07ebb)-AU3_AF3(a));}
+ AF3 APrxMedRcpF3(AF3 a){AF3 b=AF3_AU3(AU3_(0x7ef19fff)-AU3_AF3(a));return b*(-b*a+AF3_(2.0));}
+ AF3 APrxLoRsqF3(AF3 a){return AF3_AU3(AU3_(0x5f347d74)-(AU3_AF3(a)>>AU3_(1)));}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF4 APrxLoSqrtF4(AF4 a){return AF4_AU4((AU4_AF4(a)>>AU4_(1))+AU4_(0x1fbc4639));}
+ AF4 APrxLoRcpF4(AF4 a){return AF4_AU4(AU4_(0x7ef07ebb)-AU4_AF4(a));}
+ AF4 APrxMedRcpF4(AF4 a){AF4 b=AF4_AU4(AU4_(0x7ef19fff)-AU4_AF4(a));return b*(-b*a+AF4_(2.0));}
+ AF4 APrxLoRsqF4(AF4 a){return AF4_AU4(AU4_(0x5f347d74)-(AU4_AF4(a)>>AU4_(1)));}
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                    PQ APPROXIMATIONS
+//------------------------------------------------------------------------------------------------------------------------------
+// PQ is very close to x^(1/8). The functions below Use the fast float approximation method to do
+// PQ<~>Gamma2 (4th power and fast 4th root) and PQ<~>Linear (8th power and fast 8th root). Maximum error is ~0.2%.
+//==============================================================================================================================
+// Helpers
+ AF1 Quart(AF1 a) { a = a * a; return a * a;}
+ AF1 Oct(AF1 a) { a = a * a; a = a * a; return a * a; }
+ AF2 Quart(AF2 a) { a = a * a; return a * a; }
+ AF2 Oct(AF2 a) { a = a * a; a = a * a; return a * a; }
+ AF3 Quart(AF3 a) { a = a * a; return a * a; }
+ AF3 Oct(AF3 a) { a = a * a; a = a * a; return a * a; }
+ AF4 Quart(AF4 a) { a = a * a; return a * a; }
+ AF4 Oct(AF4 a) { a = a * a; a = a * a; return a * a; }
+ //------------------------------------------------------------------------------------------------------------------------------
+ AF1 APrxPQToGamma2(AF1 a) { return Quart(a); }
+ AF1 APrxPQToLinear(AF1 a) { return Oct(a); }
+ AF1 APrxLoGamma2ToPQ(AF1 a) { return AF1_AU1((AU1_AF1(a) >> AU1_(2)) + AU1_(0x2F9A4E46)); }
+ AF1 APrxMedGamma2ToPQ(AF1 a) { AF1 b = AF1_AU1((AU1_AF1(a) >> AU1_(2)) + AU1_(0x2F9A4E46)); AF1 b4 = Quart(b); return b - b * (b4 - a) / (AF1_(4.0) * b4); }
+ AF1 APrxHighGamma2ToPQ(AF1 a) { return sqrt(sqrt(a)); }
+ AF1 APrxLoLinearToPQ(AF1 a) { return AF1_AU1((AU1_AF1(a) >> AU1_(3)) + AU1_(0x378D8723)); }
+ AF1 APrxMedLinearToPQ(AF1 a) { AF1 b = AF1_AU1((AU1_AF1(a) >> AU1_(3)) + AU1_(0x378D8723)); AF1 b8 = Oct(b); return b - b * (b8 - a) / (AF1_(8.0) * b8); }
+ AF1 APrxHighLinearToPQ(AF1 a) { return sqrt(sqrt(sqrt(a))); }
+ //------------------------------------------------------------------------------------------------------------------------------
+ AF2 APrxPQToGamma2(AF2 a) { return Quart(a); }
+ AF2 APrxPQToLinear(AF2 a) { return Oct(a); }
+ AF2 APrxLoGamma2ToPQ(AF2 a) { return AF2_AU2((AU2_AF2(a) >> AU2_(2)) + AU2_(0x2F9A4E46)); }
+ AF2 APrxMedGamma2ToPQ(AF2 a) { AF2 b = AF2_AU2((AU2_AF2(a) >> AU2_(2)) + AU2_(0x2F9A4E46)); AF2 b4 = Quart(b); return b - b * (b4 - a) / (AF1_(4.0) * b4); }
+ AF2 APrxHighGamma2ToPQ(AF2 a) { return sqrt(sqrt(a)); }
+ AF2 APrxLoLinearToPQ(AF2 a) { return AF2_AU2((AU2_AF2(a) >> AU2_(3)) + AU2_(0x378D8723)); }
+ AF2 APrxMedLinearToPQ(AF2 a) { AF2 b = AF2_AU2((AU2_AF2(a) >> AU2_(3)) + AU2_(0x378D8723)); AF2 b8 = Oct(b); return b - b * (b8 - a) / (AF1_(8.0) * b8); }
+ AF2 APrxHighLinearToPQ(AF2 a) { return sqrt(sqrt(sqrt(a))); }
+ //------------------------------------------------------------------------------------------------------------------------------
+ AF3 APrxPQToGamma2(AF3 a) { return Quart(a); }
+ AF3 APrxPQToLinear(AF3 a) { return Oct(a); }
+ AF3 APrxLoGamma2ToPQ(AF3 a) { return AF3_AU3((AU3_AF3(a) >> AU3_(2)) + AU3_(0x2F9A4E46)); }
+ AF3 APrxMedGamma2ToPQ(AF3 a) { AF3 b = AF3_AU3((AU3_AF3(a) >> AU3_(2)) + AU3_(0x2F9A4E46)); AF3 b4 = Quart(b); return b - b * (b4 - a) / (AF1_(4.0) * b4); }
+ AF3 APrxHighGamma2ToPQ(AF3 a) { return sqrt(sqrt(a)); }
+ AF3 APrxLoLinearToPQ(AF3 a) { return AF3_AU3((AU3_AF3(a) >> AU3_(3)) + AU3_(0x378D8723)); }
+ AF3 APrxMedLinearToPQ(AF3 a) { AF3 b = AF3_AU3((AU3_AF3(a) >> AU3_(3)) + AU3_(0x378D8723)); AF3 b8 = Oct(b); return b - b * (b8 - a) / (AF1_(8.0) * b8); }
+ AF3 APrxHighLinearToPQ(AF3 a) { return sqrt(sqrt(sqrt(a))); }
+ //------------------------------------------------------------------------------------------------------------------------------
+ AF4 APrxPQToGamma2(AF4 a) { return Quart(a); }
+ AF4 APrxPQToLinear(AF4 a) { return Oct(a); }
+ AF4 APrxLoGamma2ToPQ(AF4 a) { return AF4_AU4((AU4_AF4(a) >> AU4_(2)) + AU4_(0x2F9A4E46)); }
+ AF4 APrxMedGamma2ToPQ(AF4 a) { AF4 b = AF4_AU4((AU4_AF4(a) >> AU4_(2)) + AU4_(0x2F9A4E46)); AF4 b4 = Quart(b); return b - b * (b4 - a) / (AF1_(4.0) * b4); }
+ AF4 APrxHighGamma2ToPQ(AF4 a) { return sqrt(sqrt(a)); }
+ AF4 APrxLoLinearToPQ(AF4 a) { return AF4_AU4((AU4_AF4(a) >> AU4_(3)) + AU4_(0x378D8723)); }
+ AF4 APrxMedLinearToPQ(AF4 a) { AF4 b = AF4_AU4((AU4_AF4(a) >> AU4_(3)) + AU4_(0x378D8723)); AF4 b8 = Oct(b); return b - b * (b8 - a) / (AF1_(8.0) * b8); }
+ AF4 APrxHighLinearToPQ(AF4 a) { return sqrt(sqrt(sqrt(a))); }
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                    PARABOLIC SIN & COS
+//------------------------------------------------------------------------------------------------------------------------------
+// Approximate answers to transcendental questions.
+//------------------------------------------------------------------------------------------------------------------------------
+//==============================================================================================================================
+ #if 1
+  // Valid input range is {-1 to 1} representing {0 to 2 pi}.
+  // Output range is {-1/4 to 1/4} representing {-1 to 1}.
+  AF1 APSinF1(AF1 x){return x*abs(x)-x;} // MAD.
+  AF2 APSinF2(AF2 x){return x*abs(x)-x;}
+  AF1 APCosF1(AF1 x){x=AFractF1(x*AF1_(0.5)+AF1_(0.75));x=x*AF1_(2.0)-AF1_(1.0);return APSinF1(x);} // 3x MAD, FRACT
+  AF2 APCosF2(AF2 x){x=AFractF2(x*AF2_(0.5)+AF2_(0.75));x=x*AF2_(2.0)-AF2_(1.0);return APSinF2(x);}
+  AF2 APSinCosF1(AF1 x){AF1 y=AFractF1(x*AF1_(0.5)+AF1_(0.75));y=y*AF1_(2.0)-AF1_(1.0);return APSinF2(AF2(x,y));}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                     [ZOL] ZERO ONE LOGIC
+//------------------------------------------------------------------------------------------------------------------------------
+// Conditional free logic designed for easy 16-bit packing, and backwards porting to 32-bit.
+//------------------------------------------------------------------------------------------------------------------------------
+// 0 := false
+// 1 := true
+//------------------------------------------------------------------------------------------------------------------------------
+// AndNot(x,y)   -> !(x&y) .... One op.
+// AndOr(x,y,z)  -> (x&y)|z ... One op.
+// GtZero(x)     -> x>0.0 ..... One op.
+// Sel(x,y,z)    -> x?y:z ..... Two ops, has no precision loss.
+// Signed(x)     -> x<0.0 ..... One op.
+// ZeroPass(x,y) -> x?0:y ..... Two ops, 'y' is a pass through safe for aliasing as integer.
+//------------------------------------------------------------------------------------------------------------------------------
+// OPTIMIZATION NOTES
+// ==================
+// - On Vega to use 2 constants in a packed op, pass in as one AW2 or one AH2 'k.xy' and use as 'k.xx' and 'k.yy'.
+//   For example 'a.xy*k.xx+k.yy'.
+//==============================================================================================================================
+ #if 1
+  AU1 AZolAndU1(AU1 x,AU1 y){return min(x,y);}
+  AU2 AZolAndU2(AU2 x,AU2 y){return min(x,y);}
+  AU3 AZolAndU3(AU3 x,AU3 y){return min(x,y);}
+  AU4 AZolAndU4(AU4 x,AU4 y){return min(x,y);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AU1 AZolNotU1(AU1 x){return x^AU1_(1);}
+  AU2 AZolNotU2(AU2 x){return x^AU2_(1);}
+  AU3 AZolNotU3(AU3 x){return x^AU3_(1);}
+  AU4 AZolNotU4(AU4 x){return x^AU4_(1);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AU1 AZolOrU1(AU1 x,AU1 y){return max(x,y);}
+  AU2 AZolOrU2(AU2 x,AU2 y){return max(x,y);}
+  AU3 AZolOrU3(AU3 x,AU3 y){return max(x,y);}
+  AU4 AZolOrU4(AU4 x,AU4 y){return max(x,y);}
+//==============================================================================================================================
+  AU1 AZolF1ToU1(AF1 x){return AU1(x);}
+  AU2 AZolF2ToU2(AF2 x){return AU2(x);}
+  AU3 AZolF3ToU3(AF3 x){return AU3(x);}
+  AU4 AZolF4ToU4(AF4 x){return AU4(x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  // 2 ops, denormals don't work in 32-bit on PC (and if they are enabled, OMOD is disabled).
+  AU1 AZolNotF1ToU1(AF1 x){return AU1(AF1_(1.0)-x);}
+  AU2 AZolNotF2ToU2(AF2 x){return AU2(AF2_(1.0)-x);}
+  AU3 AZolNotF3ToU3(AF3 x){return AU3(AF3_(1.0)-x);}
+  AU4 AZolNotF4ToU4(AF4 x){return AU4(AF4_(1.0)-x);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AZolU1ToF1(AU1 x){return AF1(x);}
+  AF2 AZolU2ToF2(AU2 x){return AF2(x);}
+  AF3 AZolU3ToF3(AU3 x){return AF3(x);}
+  AF4 AZolU4ToF4(AU4 x){return AF4(x);}
+//==============================================================================================================================
+  AF1 AZolAndF1(AF1 x,AF1 y){return min(x,y);}
+  AF2 AZolAndF2(AF2 x,AF2 y){return min(x,y);}
+  AF3 AZolAndF3(AF3 x,AF3 y){return min(x,y);}
+  AF4 AZolAndF4(AF4 x,AF4 y){return min(x,y);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 ASolAndNotF1(AF1 x,AF1 y){return (-x)*y+AF1_(1.0);}
+  AF2 ASolAndNotF2(AF2 x,AF2 y){return (-x)*y+AF2_(1.0);}
+  AF3 ASolAndNotF3(AF3 x,AF3 y){return (-x)*y+AF3_(1.0);}
+  AF4 ASolAndNotF4(AF4 x,AF4 y){return (-x)*y+AF4_(1.0);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AZolAndOrF1(AF1 x,AF1 y,AF1 z){return ASatF1(x*y+z);}
+  AF2 AZolAndOrF2(AF2 x,AF2 y,AF2 z){return ASatF2(x*y+z);}
+  AF3 AZolAndOrF3(AF3 x,AF3 y,AF3 z){return ASatF3(x*y+z);}
+  AF4 AZolAndOrF4(AF4 x,AF4 y,AF4 z){return ASatF4(x*y+z);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AZolGtZeroF1(AF1 x){return ASatF1(x*AF1_(A_INFP_F));}
+  AF2 AZolGtZeroF2(AF2 x){return ASatF2(x*AF2_(A_INFP_F));}
+  AF3 AZolGtZeroF3(AF3 x){return ASatF3(x*AF3_(A_INFP_F));}
+  AF4 AZolGtZeroF4(AF4 x){return ASatF4(x*AF4_(A_INFP_F));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AZolNotF1(AF1 x){return AF1_(1.0)-x;}
+  AF2 AZolNotF2(AF2 x){return AF2_(1.0)-x;}
+  AF3 AZolNotF3(AF3 x){return AF3_(1.0)-x;}
+  AF4 AZolNotF4(AF4 x){return AF4_(1.0)-x;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AZolOrF1(AF1 x,AF1 y){return max(x,y);}
+  AF2 AZolOrF2(AF2 x,AF2 y){return max(x,y);}
+  AF3 AZolOrF3(AF3 x,AF3 y){return max(x,y);}
+  AF4 AZolOrF4(AF4 x,AF4 y){return max(x,y);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AZolSelF1(AF1 x,AF1 y,AF1 z){AF1 r=(-x)*z+z;return x*y+r;}
+  AF2 AZolSelF2(AF2 x,AF2 y,AF2 z){AF2 r=(-x)*z+z;return x*y+r;}
+  AF3 AZolSelF3(AF3 x,AF3 y,AF3 z){AF3 r=(-x)*z+z;return x*y+r;}
+  AF4 AZolSelF4(AF4 x,AF4 y,AF4 z){AF4 r=(-x)*z+z;return x*y+r;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AZolSignedF1(AF1 x){return ASatF1(x*AF1_(A_INFN_F));}
+  AF2 AZolSignedF2(AF2 x){return ASatF2(x*AF2_(A_INFN_F));}
+  AF3 AZolSignedF3(AF3 x){return ASatF3(x*AF3_(A_INFN_F));}
+  AF4 AZolSignedF4(AF4 x){return ASatF4(x*AF4_(A_INFN_F));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AZolZeroPassF1(AF1 x,AF1 y){return AF1_AU1((AU1_AF1(x)!=AU1_(0))?AU1_(0):AU1_AF1(y));}
+  AF2 AZolZeroPassF2(AF2 x,AF2 y){return AF2_AU2((AU2_AF2(x)!=AU2_(0))?AU2_(0):AU2_AF2(y));}
+  AF3 AZolZeroPassF3(AF3 x,AF3 y){return AF3_AU3((AU3_AF3(x)!=AU3_(0))?AU3_(0):AU3_AF3(y));}
+  AF4 AZolZeroPassF4(AF4 x,AF4 y){return AF4_AU4((AU4_AF4(x)!=AU4_(0))?AU4_(0):AU4_AF4(y));}
+ #endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                      COLOR CONVERSIONS
+//------------------------------------------------------------------------------------------------------------------------------
+// These are all linear to/from some other space (where 'linear' has been shortened out of the function name).
+// So 'ToGamma' is 'LinearToGamma', and 'FromGamma' is 'LinearFromGamma'.
+// These are branch free implementations.
+// The AToSrgbF1() function is useful for stores for compute shaders for GPUs without hardware linear->sRGB store conversion.
+//------------------------------------------------------------------------------------------------------------------------------
+// TRANSFER FUNCTIONS
+// ==================
+// 709 ..... Rec709 used for some HDTVs
+// Gamma ... Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native
+// Pq ...... PQ native for HDR10
+// Srgb .... The sRGB output, typical of PC displays, useful for 10-bit output, or storing to 8-bit UNORM without SRGB type
+// Two ..... Gamma 2.0, fastest conversion (useful for intermediate pass approximations)
+// Three ... Gamma 3.0, less fast, but good for HDR.
+//------------------------------------------------------------------------------------------------------------------------------
+// KEEPING TO SPEC
+// ===============
+// Both Rec.709 and sRGB have a linear segment which as spec'ed would intersect the curved segment 2 times.
+//  (a.) For 8-bit sRGB, steps {0 to 10.3} are in the linear region (4% of the encoding range).
+//  (b.) For 8-bit  709, steps {0 to 20.7} are in the linear region (8% of the encoding range).
+// Also there is a slight step in the transition regions.
+// Precision of the coefficients in the spec being the likely cause.
+// Main usage case of the sRGB code is to do the linear->sRGB converstion in a compute shader before store.
+// This is to work around lack of hardware (typically only ROP does the conversion for free).
+// To "correct" the linear segment, would be to introduce error, because hardware decode of sRGB->linear is fixed (and free).
+// So this header keeps with the spec.
+// For linear->sRGB transforms, the linear segment in some respects reduces error, because rounding in that region is linear.
+// Rounding in the curved region in hardware (and fast software code) introduces error due to rounding in non-linear.
+//------------------------------------------------------------------------------------------------------------------------------
+// FOR PQ
+// ======
+// Both input and output is {0.0-1.0}, and where output 1.0 represents 10000.0 cd/m^2.
+// All constants are only specified to FP32 precision.
+// External PQ source reference,
+//  - https://github.com/ampas/aces-dev/blob/master/transforms/ctl/utilities/ACESlib.Utilities_Color.a1.0.1.ctl
+//------------------------------------------------------------------------------------------------------------------------------
+// PACKED VERSIONS
+// ===============
+// These are the A*H2() functions.
+// There is no PQ functions as FP16 seemed to not have enough precision for the conversion.
+// The remaining functions are "good enough" for 8-bit, and maybe 10-bit if not concerned about a few 1-bit errors.
+// Precision is lowest in the 709 conversion, higher in sRGB, higher still in Two and Gamma (when using 2.2 at least).
+//------------------------------------------------------------------------------------------------------------------------------
+// NOTES
+// =====
+// Could be faster for PQ conversions to be in ALU or a texture lookup depending on usage case.
+//==============================================================================================================================
+ #if 1
+  AF1 ATo709F1(AF1 c){AF3 j=AF3(0.018*4.5,4.5,0.45);AF2 k=AF2(1.099,-0.099);
+   return clamp(j.x  ,c*j.y  ,pow(c,j.z  )*k.x  +k.y  );}
+  AF2 ATo709F2(AF2 c){AF3 j=AF3(0.018*4.5,4.5,0.45);AF2 k=AF2(1.099,-0.099);
+   return clamp(j.xx ,c*j.yy ,pow(c,j.zz )*k.xx +k.yy );}
+  AF3 ATo709F3(AF3 c){AF3 j=AF3(0.018*4.5,4.5,0.45);AF2 k=AF2(1.099,-0.099);
+   return clamp(j.xxx,c*j.yyy,pow(c,j.zzz)*k.xxx+k.yyy);}
+//------------------------------------------------------------------------------------------------------------------------------
+  // Note 'rcpX' is '1/x', where the 'x' is what would be used in AFromGamma().
+  AF1 AToGammaF1(AF1 c,AF1 rcpX){return pow(c,AF1_(rcpX));}
+  AF2 AToGammaF2(AF2 c,AF1 rcpX){return pow(c,AF2_(rcpX));}
+  AF3 AToGammaF3(AF3 c,AF1 rcpX){return pow(c,AF3_(rcpX));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AToPqF1(AF1 x){AF1 p=pow(x,AF1_(0.159302));
+   return pow((AF1_(0.835938)+AF1_(18.8516)*p)/(AF1_(1.0)+AF1_(18.6875)*p),AF1_(78.8438));}
+  AF2 AToPqF1(AF2 x){AF2 p=pow(x,AF2_(0.159302));
+   return pow((AF2_(0.835938)+AF2_(18.8516)*p)/(AF2_(1.0)+AF2_(18.6875)*p),AF2_(78.8438));}
+  AF3 AToPqF1(AF3 x){AF3 p=pow(x,AF3_(0.159302));
+   return pow((AF3_(0.835938)+AF3_(18.8516)*p)/(AF3_(1.0)+AF3_(18.6875)*p),AF3_(78.8438));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AToSrgbF1(AF1 c){AF3 j=AF3(0.0031308*12.92,12.92,1.0/2.4);AF2 k=AF2(1.055,-0.055);
+   return clamp(j.x  ,c*j.y  ,pow(c,j.z  )*k.x  +k.y  );}
+  AF2 AToSrgbF2(AF2 c){AF3 j=AF3(0.0031308*12.92,12.92,1.0/2.4);AF2 k=AF2(1.055,-0.055);
+   return clamp(j.xx ,c*j.yy ,pow(c,j.zz )*k.xx +k.yy );}
+  AF3 AToSrgbF3(AF3 c){AF3 j=AF3(0.0031308*12.92,12.92,1.0/2.4);AF2 k=AF2(1.055,-0.055);
+   return clamp(j.xxx,c*j.yyy,pow(c,j.zzz)*k.xxx+k.yyy);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AToTwoF1(AF1 c){return sqrt(c);}
+  AF2 AToTwoF2(AF2 c){return sqrt(c);}
+  AF3 AToTwoF3(AF3 c){return sqrt(c);}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AToThreeF1(AF1 c){return pow(c,AF1_(1.0/3.0));}
+  AF2 AToThreeF2(AF2 c){return pow(c,AF2_(1.0/3.0));}
+  AF3 AToThreeF3(AF3 c){return pow(c,AF3_(1.0/3.0));}
+ #endif
+//==============================================================================================================================
+ #if 1
+  // Unfortunately median won't work here.
+  AF1 AFrom709F1(AF1 c){AF3 j=AF3(0.081/4.5,1.0/4.5,1.0/0.45);AF2 k=AF2(1.0/1.099,0.099/1.099);
+   return AZolSelF1(AZolSignedF1(c-j.x  ),c*j.y  ,pow(c*k.x  +k.y  ,j.z  ));}
+  AF2 AFrom709F2(AF2 c){AF3 j=AF3(0.081/4.5,1.0/4.5,1.0/0.45);AF2 k=AF2(1.0/1.099,0.099/1.099);
+   return AZolSelF2(AZolSignedF2(c-j.xx ),c*j.yy ,pow(c*k.xx +k.yy ,j.zz ));}
+  AF3 AFrom709F3(AF3 c){AF3 j=AF3(0.081/4.5,1.0/4.5,1.0/0.45);AF2 k=AF2(1.0/1.099,0.099/1.099);
+   return AZolSelF3(AZolSignedF3(c-j.xxx),c*j.yyy,pow(c*k.xxx+k.yyy,j.zzz));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AFromGammaF1(AF1 c,AF1 x){return pow(c,AF1_(x));}
+  AF2 AFromGammaF2(AF2 c,AF1 x){return pow(c,AF2_(x));}
+  AF3 AFromGammaF3(AF3 c,AF1 x){return pow(c,AF3_(x));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AFromPqF1(AF1 x){AF1 p=pow(x,AF1_(0.0126833));
+   return pow(ASatF1(p-AF1_(0.835938))/(AF1_(18.8516)-AF1_(18.6875)*p),AF1_(6.27739));}
+  AF2 AFromPqF1(AF2 x){AF2 p=pow(x,AF2_(0.0126833));
+   return pow(ASatF2(p-AF2_(0.835938))/(AF2_(18.8516)-AF2_(18.6875)*p),AF2_(6.27739));}
+  AF3 AFromPqF1(AF3 x){AF3 p=pow(x,AF3_(0.0126833));
+   return pow(ASatF3(p-AF3_(0.835938))/(AF3_(18.8516)-AF3_(18.6875)*p),AF3_(6.27739));}
+//------------------------------------------------------------------------------------------------------------------------------
+  // Unfortunately median won't work here.
+  AF1 AFromSrgbF1(AF1 c){AF3 j=AF3(0.04045/12.92,1.0/12.92,2.4);AF2 k=AF2(1.0/1.055,0.055/1.055);
+   return AZolSelF1(AZolSignedF1(c-j.x  ),c*j.y  ,pow(c*k.x  +k.y  ,j.z  ));}
+  AF2 AFromSrgbF2(AF2 c){AF3 j=AF3(0.04045/12.92,1.0/12.92,2.4);AF2 k=AF2(1.0/1.055,0.055/1.055);
+   return AZolSelF2(AZolSignedF2(c-j.xx ),c*j.yy ,pow(c*k.xx +k.yy ,j.zz ));}
+  AF3 AFromSrgbF3(AF3 c){AF3 j=AF3(0.04045/12.92,1.0/12.92,2.4);AF2 k=AF2(1.0/1.055,0.055/1.055);
+   return AZolSelF3(AZolSignedF3(c-j.xxx),c*j.yyy,pow(c*k.xxx+k.yyy,j.zzz));}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AFromTwoF1(AF1 c){return c*c;}
+  AF2 AFromTwoF2(AF2 c){return c*c;}
+  AF3 AFromTwoF3(AF3 c){return c*c;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AF1 AFromThreeF1(AF1 c){return c*c*c;}
+  AF2 AFromThreeF2(AF2 c){return c*c*c;}
+  AF3 AFromThreeF3(AF3 c){return c*c*c;}
+ #endif
+//==============================================================================================================================
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                          CS REMAP
+//==============================================================================================================================
+ // Simple remap 64x1 to 8x8 with rotated 2x2 pixel quads in quad linear.
+ //  543210
+ //  ======
+ //  ..xxx.
+ //  yy...y
+ AU2 ARmp8x8(AU1 a){return AU2(ABfe(a,1u,3u),ABfiM(ABfe(a,3u,3u),a,1u));}
+//==============================================================================================================================
+ // More complex remap 64x1 to 8x8 which is necessary for 2D wave reductions.
+ //  543210
+ //  ======
+ //  .xx..x
+ //  y..yy.
+ // Details,
+ //  LANE TO 8x8 MAPPING
+ //  ===================
+ //  00 01 08 09 10 11 18 19
+ //  02 03 0a 0b 12 13 1a 1b
+ //  04 05 0c 0d 14 15 1c 1d
+ //  06 07 0e 0f 16 17 1e 1f
+ //  20 21 28 29 30 31 38 39
+ //  22 23 2a 2b 32 33 3a 3b
+ //  24 25 2c 2d 34 35 3c 3d
+ //  26 27 2e 2f 36 37 3e 3f
+ AU2 ARmpRed8x8(AU1 a){return AU2(ABfiM(ABfe(a,2u,3u),a,1u),ABfiM(ABfe(a,3u,3u),ABfe(a,1u,2u),2u));}
+//==============================================================================================================================
+#endif
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//                                                          REFERENCE
+//
+//------------------------------------------------------------------------------------------------------------------------------
+// IEEE FLOAT RULES
+// ================
+//  - saturate(NaN)=0, saturate(-INF)=0, saturate(+INF)=1
+//  - {+/-}0 * {+/-}INF = NaN
+//  - -INF + (+INF) = NaN
+//  - {+/-}0 / {+/-}0 = NaN
+//  - {+/-}INF / {+/-}INF = NaN
+//  - a<(-0) := sqrt(a) = NaN (a=-0.0 won't NaN)
+//  - 0 == -0
+//  - 4/0 = +INF
+//  - 4/-0 = -INF
+//  - 4+INF = +INF
+//  - 4-INF = -INF
+//  - 4*(+INF) = +INF
+//  - 4*(-INF) = -INF
+//  - -4*(+INF) = -INF
+//  - sqrt(+INF) = +INF
+//------------------------------------------------------------------------------------------------------------------------------
+// FP16 ENCODING
+// =============
+// fedcba9876543210
+// ----------------
+// ......mmmmmmmmmm  10-bit mantissa (encodes 11-bit 0.5 to 1.0 except for denormals)
+// .eeeee..........  5-bit exponent
+// .00000..........  denormals
+// .00001..........  -14 exponent
+// .11110..........   15 exponent
+// .111110000000000  infinity
+// .11111nnnnnnnnnn  NaN with n!=0
+// s...............  sign
+//------------------------------------------------------------------------------------------------------------------------------
+// FP16/INT16 ALIASING DENORMAL
+// ============================
+// 11-bit unsigned integers alias with half float denormal/normal values,
+//     1 = 2^(-24) = 1/16777216 ....................... first denormal value
+//     2 = 2^(-23)
+//   ...
+//  1023 = 2^(-14)*(1-2^(-10)) = 2^(-14)*(1-1/1024) ... last denormal value
+//  1024 = 2^(-14) = 1/16384 .......................... first normal value that still maps to integers
+//  2047 .............................................. last normal value that still maps to integers
+// Scaling limits,
+//  2^15 = 32768 ...................................... largest power of 2 scaling
+// Largest pow2 conversion mapping is at *32768,
+//     1 : 2^(-9) = 1/512
+//     2 : 1/256
+//     4 : 1/128
+//     8 : 1/64
+//    16 : 1/32
+//    32 : 1/16
+//    64 : 1/8
+//   128 : 1/4
+//   256 : 1/2
+//   512 : 1
+//  1024 : 2
+//  2047 : a little less than 4
+//==============================================================================================================================
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//
+//                                                     GPU/CPU PORTABILITY
+//
+//
+//------------------------------------------------------------------------------------------------------------------------------
+// This is the GPU implementation.
+// See the CPU implementation for docs.
+//==============================================================================================================================
+#ifdef A_GPU
+ #define A_TRUE true
+ #define A_FALSE false
+ #define A_STATIC
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                     VECTOR ARGUMENT/RETURN/INITIALIZATION PORTABILITY
+//==============================================================================================================================
+ #define retAD2 AD2
+ #define retAD3 AD3
+ #define retAD4 AD4
+ #define retAF2 AF2
+ #define retAF3 AF3
+ #define retAF4 AF4
+ #define retAL2 AL2
+ #define retAL3 AL3
+ #define retAL4 AL4
+ #define retAU2 AU2
+ #define retAU3 AU3
+ #define retAU4 AU4
+//------------------------------------------------------------------------------------------------------------------------------
+ #define inAD2 in AD2
+ #define inAD3 in AD3
+ #define inAD4 in AD4
+ #define inAF2 in AF2
+ #define inAF3 in AF3
+ #define inAF4 in AF4
+ #define inAL2 in AL2
+ #define inAL3 in AL3
+ #define inAL4 in AL4
+ #define inAU2 in AU2
+ #define inAU3 in AU3
+ #define inAU4 in AU4
+//------------------------------------------------------------------------------------------------------------------------------
+ #define inoutAD2 inout AD2
+ #define inoutAD3 inout AD3
+ #define inoutAD4 inout AD4
+ #define inoutAF2 inout AF2
+ #define inoutAF3 inout AF3
+ #define inoutAF4 inout AF4
+ #define inoutAL2 inout AL2
+ #define inoutAL3 inout AL3
+ #define inoutAL4 inout AL4
+ #define inoutAU2 inout AU2
+ #define inoutAU3 inout AU3
+ #define inoutAU4 inout AU4
+//------------------------------------------------------------------------------------------------------------------------------
+ #define outAD2 out AD2
+ #define outAD3 out AD3
+ #define outAD4 out AD4
+ #define outAF2 out AF2
+ #define outAF3 out AF3
+ #define outAF4 out AF4
+ #define outAL2 out AL2
+ #define outAL3 out AL3
+ #define outAL4 out AL4
+ #define outAU2 out AU2
+ #define outAU3 out AU3
+ #define outAU4 out AU4
+//------------------------------------------------------------------------------------------------------------------------------
+ #define varAD2(x) AD2 x
+ #define varAD3(x) AD3 x
+ #define varAD4(x) AD4 x
+ #define varAF2(x) AF2 x
+ #define varAF3(x) AF3 x
+ #define varAF4(x) AF4 x
+ #define varAL2(x) AL2 x
+ #define varAL3(x) AL3 x
+ #define varAL4(x) AL4 x
+ #define varAU2(x) AU2 x
+ #define varAU3(x) AU3 x
+ #define varAU4(x) AU4 x
+//------------------------------------------------------------------------------------------------------------------------------
+ #define initAD2(x,y) AD2(x,y)
+ #define initAD3(x,y,z) AD3(x,y,z)
+ #define initAD4(x,y,z,w) AD4(x,y,z,w)
+ #define initAF2(x,y) AF2(x,y)
+ #define initAF3(x,y,z) AF3(x,y,z)
+ #define initAF4(x,y,z,w) AF4(x,y,z,w)
+ #define initAL2(x,y) AL2(x,y)
+ #define initAL3(x,y,z) AL3(x,y,z)
+ #define initAL4(x,y,z,w) AL4(x,y,z,w)
+ #define initAU2(x,y) AU2(x,y)
+ #define initAU3(x,y,z) AU3(x,y,z)
+ #define initAU4(x,y,z,w) AU4(x,y,z,w)
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                     SCALAR RETURN OPS
+//==============================================================================================================================
+ #define AAbsD1(a) abs(AD1(a))
+ #define AAbsF1(a) abs(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define ACosD1(a) cos(AD1(a))
+ #define ACosF1(a) cos(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define ADotD2(a,b) dot(AD2(a),AD2(b))
+ #define ADotD3(a,b) dot(AD3(a),AD3(b))
+ #define ADotD4(a,b) dot(AD4(a),AD4(b))
+ #define ADotF2(a,b) dot(AF2(a),AF2(b))
+ #define ADotF3(a,b) dot(AF3(a),AF3(b))
+ #define ADotF4(a,b) dot(AF4(a),AF4(b))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AExp2D1(a) exp2(AD1(a))
+ #define AExp2F1(a) exp2(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AFloorD1(a) floor(AD1(a))
+ #define AFloorF1(a) floor(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define ALog2D1(a) log2(AD1(a))
+ #define ALog2F1(a) log2(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AMaxD1(a,b) max(a,b)
+ #define AMaxF1(a,b) max(a,b)
+ #define AMaxL1(a,b) max(a,b)
+ #define AMaxU1(a,b) max(a,b)
+//------------------------------------------------------------------------------------------------------------------------------
+ #define AMinD1(a,b) min(a,b)
+ #define AMinF1(a,b) min(a,b)
+ #define AMinL1(a,b) min(a,b)
+ #define AMinU1(a,b) min(a,b)
+//------------------------------------------------------------------------------------------------------------------------------
+ #define ASinD1(a) sin(AD1(a))
+ #define ASinF1(a) sin(AF1(a))
+//------------------------------------------------------------------------------------------------------------------------------
+ #define ASqrtD1(a) sqrt(AD1(a))
+ #define ASqrtF1(a) sqrt(AF1(a))
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                               SCALAR RETURN OPS - DEPENDENT
+//==============================================================================================================================
+ #define APowD1(a,b) pow(AD1(a),AF1(b))
+ #define APowF1(a,b) pow(AF1(a),AF1(b))
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                         VECTOR OPS
+//------------------------------------------------------------------------------------------------------------------------------
+// These are added as needed for production or prototyping, so not necessarily a complete set.
+// They follow a convention of taking in a destination and also returning the destination value to increase utility.
+//==============================================================================================================================
+ #ifdef A_DUBL
+  AD2 opAAbsD2(outAD2 d,inAD2 a){d=abs(a);return d;}
+  AD3 opAAbsD3(outAD3 d,inAD3 a){d=abs(a);return d;}
+  AD4 opAAbsD4(outAD4 d,inAD4 a){d=abs(a);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opAAddD2(outAD2 d,inAD2 a,inAD2 b){d=a+b;return d;}
+  AD3 opAAddD3(outAD3 d,inAD3 a,inAD3 b){d=a+b;return d;}
+  AD4 opAAddD4(outAD4 d,inAD4 a,inAD4 b){d=a+b;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opAAddOneD2(outAD2 d,inAD2 a,AD1 b){d=a+AD2_(b);return d;}
+  AD3 opAAddOneD3(outAD3 d,inAD3 a,AD1 b){d=a+AD3_(b);return d;}
+  AD4 opAAddOneD4(outAD4 d,inAD4 a,AD1 b){d=a+AD4_(b);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opACpyD2(outAD2 d,inAD2 a){d=a;return d;}
+  AD3 opACpyD3(outAD3 d,inAD3 a){d=a;return d;}
+  AD4 opACpyD4(outAD4 d,inAD4 a){d=a;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opALerpD2(outAD2 d,inAD2 a,inAD2 b,inAD2 c){d=ALerpD2(a,b,c);return d;}
+  AD3 opALerpD3(outAD3 d,inAD3 a,inAD3 b,inAD3 c){d=ALerpD3(a,b,c);return d;}
+  AD4 opALerpD4(outAD4 d,inAD4 a,inAD4 b,inAD4 c){d=ALerpD4(a,b,c);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opALerpOneD2(outAD2 d,inAD2 a,inAD2 b,AD1 c){d=ALerpD2(a,b,AD2_(c));return d;}
+  AD3 opALerpOneD3(outAD3 d,inAD3 a,inAD3 b,AD1 c){d=ALerpD3(a,b,AD3_(c));return d;}
+  AD4 opALerpOneD4(outAD4 d,inAD4 a,inAD4 b,AD1 c){d=ALerpD4(a,b,AD4_(c));return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opAMaxD2(outAD2 d,inAD2 a,inAD2 b){d=max(a,b);return d;}
+  AD3 opAMaxD3(outAD3 d,inAD3 a,inAD3 b){d=max(a,b);return d;}
+  AD4 opAMaxD4(outAD4 d,inAD4 a,inAD4 b){d=max(a,b);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opAMinD2(outAD2 d,inAD2 a,inAD2 b){d=min(a,b);return d;}
+  AD3 opAMinD3(outAD3 d,inAD3 a,inAD3 b){d=min(a,b);return d;}
+  AD4 opAMinD4(outAD4 d,inAD4 a,inAD4 b){d=min(a,b);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opAMulD2(outAD2 d,inAD2 a,inAD2 b){d=a*b;return d;}
+  AD3 opAMulD3(outAD3 d,inAD3 a,inAD3 b){d=a*b;return d;}
+  AD4 opAMulD4(outAD4 d,inAD4 a,inAD4 b){d=a*b;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opAMulOneD2(outAD2 d,inAD2 a,AD1 b){d=a*AD2_(b);return d;}
+  AD3 opAMulOneD3(outAD3 d,inAD3 a,AD1 b){d=a*AD3_(b);return d;}
+  AD4 opAMulOneD4(outAD4 d,inAD4 a,AD1 b){d=a*AD4_(b);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opANegD2(outAD2 d,inAD2 a){d=-a;return d;}
+  AD3 opANegD3(outAD3 d,inAD3 a){d=-a;return d;}
+  AD4 opANegD4(outAD4 d,inAD4 a){d=-a;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+  AD2 opARcpD2(outAD2 d,inAD2 a){d=ARcpD2(a);return d;}
+  AD3 opARcpD3(outAD3 d,inAD3 a){d=ARcpD3(a);return d;}
+  AD4 opARcpD4(outAD4 d,inAD4 a){d=ARcpD4(a);return d;}
+ #endif
+//==============================================================================================================================
+ AF2 opAAbsF2(outAF2 d,inAF2 a){d=abs(a);return d;}
+ AF3 opAAbsF3(outAF3 d,inAF3 a){d=abs(a);return d;}
+ AF4 opAAbsF4(outAF4 d,inAF4 a){d=abs(a);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opAAddF2(outAF2 d,inAF2 a,inAF2 b){d=a+b;return d;}
+ AF3 opAAddF3(outAF3 d,inAF3 a,inAF3 b){d=a+b;return d;}
+ AF4 opAAddF4(outAF4 d,inAF4 a,inAF4 b){d=a+b;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opAAddOneF2(outAF2 d,inAF2 a,AF1 b){d=a+AF2_(b);return d;}
+ AF3 opAAddOneF3(outAF3 d,inAF3 a,AF1 b){d=a+AF3_(b);return d;}
+ AF4 opAAddOneF4(outAF4 d,inAF4 a,AF1 b){d=a+AF4_(b);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opACpyF2(outAF2 d,inAF2 a){d=a;return d;}
+ AF3 opACpyF3(outAF3 d,inAF3 a){d=a;return d;}
+ AF4 opACpyF4(outAF4 d,inAF4 a){d=a;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opALerpF2(outAF2 d,inAF2 a,inAF2 b,inAF2 c){d=ALerpF2(a,b,c);return d;}
+ AF3 opALerpF3(outAF3 d,inAF3 a,inAF3 b,inAF3 c){d=ALerpF3(a,b,c);return d;}
+ AF4 opALerpF4(outAF4 d,inAF4 a,inAF4 b,inAF4 c){d=ALerpF4(a,b,c);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opALerpOneF2(outAF2 d,inAF2 a,inAF2 b,AF1 c){d=ALerpF2(a,b,AF2_(c));return d;}
+ AF3 opALerpOneF3(outAF3 d,inAF3 a,inAF3 b,AF1 c){d=ALerpF3(a,b,AF3_(c));return d;}
+ AF4 opALerpOneF4(outAF4 d,inAF4 a,inAF4 b,AF1 c){d=ALerpF4(a,b,AF4_(c));return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opAMaxF2(outAF2 d,inAF2 a,inAF2 b){d=max(a,b);return d;}
+ AF3 opAMaxF3(outAF3 d,inAF3 a,inAF3 b){d=max(a,b);return d;}
+ AF4 opAMaxF4(outAF4 d,inAF4 a,inAF4 b){d=max(a,b);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opAMinF2(outAF2 d,inAF2 a,inAF2 b){d=min(a,b);return d;}
+ AF3 opAMinF3(outAF3 d,inAF3 a,inAF3 b){d=min(a,b);return d;}
+ AF4 opAMinF4(outAF4 d,inAF4 a,inAF4 b){d=min(a,b);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opAMulF2(outAF2 d,inAF2 a,inAF2 b){d=a*b;return d;}
+ AF3 opAMulF3(outAF3 d,inAF3 a,inAF3 b){d=a*b;return d;}
+ AF4 opAMulF4(outAF4 d,inAF4 a,inAF4 b){d=a*b;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opAMulOneF2(outAF2 d,inAF2 a,AF1 b){d=a*AF2_(b);return d;}
+ AF3 opAMulOneF3(outAF3 d,inAF3 a,AF1 b){d=a*AF3_(b);return d;}
+ AF4 opAMulOneF4(outAF4 d,inAF4 a,AF1 b){d=a*AF4_(b);return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opANegF2(outAF2 d,inAF2 a){d=-a;return d;}
+ AF3 opANegF3(outAF3 d,inAF3 a){d=-a;return d;}
+ AF4 opANegF4(outAF4 d,inAF4 a){d=-a;return d;}
+//------------------------------------------------------------------------------------------------------------------------------
+ AF2 opARcpF2(outAF2 d,inAF2 a){d=ARcpF2(a);return d;}
+ AF3 opARcpF3(outAF3 d,inAF3 a){d=ARcpF3(a);return d;}
+ AF4 opARcpF4(outAF4 d,inAF4 a){d=ARcpF4(a);return d;}
+#endif
+
+#endif
+
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//
+//                                 [CAS] FIDELITY FX - CONSTRAST ADAPTIVE SHARPENING 1.20190610
+//
+//==============================================================================================================================
+// LICENSE
+// =======
+// Copyright (c) 2017-2019 Advanced Micro Devices, Inc. All rights reserved.
+// -------
+// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation
+// files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy,
+// modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the
+// Software is furnished to do so, subject to the following conditions:
+// -------
+// The above copyright notice and this permission notice shall be included in all copies or substantial portions of the
+// Software.
+// -------
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE
+// WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE AUTHORS OR
+// COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+// ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+//------------------------------------------------------------------------------------------------------------------------------
+#define CAS_AREA_LIMIT 4.0
+//------------------------------------------------------------------------------------------------------------------------------
+// Pass in output and input resolution in pixels.
+// This returns true if CAS supports scaling in the given configuration.
+AP1 CasSupportScaling(AF1 outX,AF1 outY,AF1 inX,AF1 inY){return ((outX*outY)*ARcpF1(inX*inY))<=CAS_AREA_LIMIT;}
+//==============================================================================================================================
+// Call to setup required constant values (works on CPU or GPU).
+#ifndef A_GPU
+A_STATIC void CasSetup(
+ outAU4 const0,
+ outAU4 const1,
+ AF1 sharpness, // 0 := default (lower ringing), 1 := maximum (higest ringing)
+ AF1 inputSizeInPixelsX,
+ AF1 inputSizeInPixelsY,
+ AF1 outputSizeInPixelsX,
+ AF1 outputSizeInPixelsY){
+  // Scaling terms.
+  const0[0]=AU1_AF1(inputSizeInPixelsX*ARcpF1(outputSizeInPixelsX));
+  const0[1]=AU1_AF1(inputSizeInPixelsY*ARcpF1(outputSizeInPixelsY));
+  const0[2]=AU1_AF1(AF1_(0.5)*inputSizeInPixelsX*ARcpF1(outputSizeInPixelsX)-AF1_(0.5));
+  const0[3]=AU1_AF1(AF1_(0.5)*inputSizeInPixelsY*ARcpF1(outputSizeInPixelsY)-AF1_(0.5));
+  // Sharpness value.
+  AF1 sharp=-ARcpF1(ALerpF1(8.0,5.0,ASatF1(sharpness)));
+  varAF2(hSharp)=initAF2(sharp,0.0);
+  const1[0]=AU1_AF1(sharp);
+  const1[1]=AU1_AH2_AF2(hSharp);
+  const1[2]=AU1_AF1(AF1_(8.0)*inputSizeInPixelsX*ARcpF1(outputSizeInPixelsX));
+  const1[3]=0;}
+#endif
+
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+//_____________________________________________________________/\_______________________________________________________________
+//==============================================================================================================================
+//                                                     NON-PACKED VERSION
+//==============================================================================================================================
+#ifdef A_GPU
+ AF3 CasLoad(ASU2 p) { return texelFetch(diffuseRect, p, 0).rgb; }
+ void CasInput(inout AF1 r,inout AF1 g,inout AF1 b)
+ {
+    r = AFromSrgbF1(r);
+    g = AFromSrgbF1(g);
+    b = AFromSrgbF1(b);
+ }
+
+//------------------------------------------------------------------------------------------------------------------------------
+ void CasFilter(
+ out AF1 pixR, // Output values, non-vector so port between CasFilter() and CasFilterH() is easy.
+ out AF1 pixG,
+ out AF1 pixB,
+ AU2 ip, // Integer pixel position in output.
+ AU4 const0, // Constants generated by CasSetup().
+ AU4 const1,
+ AP1 noScaling){ // Must be a compile-time literal value, true = sharpen only (no resize).
+//------------------------------------------------------------------------------------------------------------------------------
+  // Debug a checker pattern of on/off tiles for visual inspection.
+  #ifdef CAS_DEBUG_CHECKER
+   if((((ip.x^ip.y)>>8u)&1u)==0u){AF3 pix0=CasLoad(ASU2(ip));
+    pixR=pix0.r;pixG=pix0.g;pixB=pix0.b;CasInput(pixR,pixG,pixB);return;}
+  #endif
+//------------------------------------------------------------------------------------------------------------------------------
+  // No scaling algorithm uses minimal 3x3 pixel neighborhood.
+  if(noScaling){
+   // a b c
+   // d e f
+   // g h i
+   ASU2 sp=ASU2(ip);
+   AF3 a=CasLoad(sp+ASU2(-1,-1));
+   AF3 b=CasLoad(sp+ASU2( 0,-1));
+   AF3 c=CasLoad(sp+ASU2( 1,-1));
+   AF3 d=CasLoad(sp+ASU2(-1, 0));
+   AF3 e=CasLoad(sp);
+   AF3 f=CasLoad(sp+ASU2( 1, 0));
+   AF3 g=CasLoad(sp+ASU2(-1, 1));
+   AF3 h=CasLoad(sp+ASU2( 0, 1));
+   AF3 i=CasLoad(sp+ASU2( 1, 1));
+   // Run optional input transform.
+   CasInput(a.r,a.g,a.b);
+   CasInput(b.r,b.g,b.b);
+   CasInput(c.r,c.g,c.b);
+   CasInput(d.r,d.g,d.b);
+   CasInput(e.r,e.g,e.b);
+   CasInput(f.r,f.g,f.b);
+   CasInput(g.r,g.g,g.b);
+   CasInput(h.r,h.g,h.b);
+   CasInput(i.r,i.g,i.b);
+   // Soft min and max.
+   //  a b c             b
+   //  d e f * 0.5  +  d e f * 0.5
+   //  g h i             h
+   // These are 2.0x bigger (factored out the extra multiply).
+   AF1 mnR=AMin3F1(AMin3F1(d.r,e.r,f.r),b.r,h.r);
+   AF1 mnG=AMin3F1(AMin3F1(d.g,e.g,f.g),b.g,h.g);
+   AF1 mnB=AMin3F1(AMin3F1(d.b,e.b,f.b),b.b,h.b);
+   #ifdef CAS_BETTER_DIAGONALS
+    AF1 mnR2=AMin3F1(AMin3F1(mnR,a.r,c.r),g.r,i.r);
+    AF1 mnG2=AMin3F1(AMin3F1(mnG,a.g,c.g),g.g,i.g);
+    AF1 mnB2=AMin3F1(AMin3F1(mnB,a.b,c.b),g.b,i.b);
+    mnR=mnR+mnR2;
+    mnG=mnG+mnG2;
+    mnB=mnB+mnB2;
+   #endif
+   AF1 mxR=AMax3F1(AMax3F1(d.r,e.r,f.r),b.r,h.r);
+   AF1 mxG=AMax3F1(AMax3F1(d.g,e.g,f.g),b.g,h.g);
+   AF1 mxB=AMax3F1(AMax3F1(d.b,e.b,f.b),b.b,h.b);
+   #ifdef CAS_BETTER_DIAGONALS
+    AF1 mxR2=AMax3F1(AMax3F1(mxR,a.r,c.r),g.r,i.r);
+    AF1 mxG2=AMax3F1(AMax3F1(mxG,a.g,c.g),g.g,i.g);
+    AF1 mxB2=AMax3F1(AMax3F1(mxB,a.b,c.b),g.b,i.b);
+    mxR=mxR+mxR2;
+    mxG=mxG+mxG2;
+    mxB=mxB+mxB2;
+   #endif
+   // Smooth minimum distance to signal limit divided by smooth max.
+   #ifdef CAS_GO_SLOWER
+    AF1 rcpMR=ARcpF1(mxR);
+    AF1 rcpMG=ARcpF1(mxG);
+    AF1 rcpMB=ARcpF1(mxB);
+   #else
+    AF1 rcpMR=APrxLoRcpF1(mxR);
+    AF1 rcpMG=APrxLoRcpF1(mxG);
+    AF1 rcpMB=APrxLoRcpF1(mxB);
+   #endif
+   #ifdef CAS_BETTER_DIAGONALS
+    AF1 ampR=ASatF1(min(mnR,AF1_(2.0)-mxR)*rcpMR);
+    AF1 ampG=ASatF1(min(mnG,AF1_(2.0)-mxG)*rcpMG);
+    AF1 ampB=ASatF1(min(mnB,AF1_(2.0)-mxB)*rcpMB);
+   #else
+    AF1 ampR=ASatF1(min(mnR,AF1_(1.0)-mxR)*rcpMR);
+    AF1 ampG=ASatF1(min(mnG,AF1_(1.0)-mxG)*rcpMG);
+    AF1 ampB=ASatF1(min(mnB,AF1_(1.0)-mxB)*rcpMB);
+   #endif
+   // Shaping amount of sharpening.
+   #ifdef CAS_GO_SLOWER
+    ampR=sqrt(ampR);
+    ampG=sqrt(ampG);
+    ampB=sqrt(ampB);
+   #else
+    ampR=APrxLoSqrtF1(ampR);
+    ampG=APrxLoSqrtF1(ampG);
+    ampB=APrxLoSqrtF1(ampB);
+   #endif
+   // Filter shape.
+   //  0 w 0
+   //  w 1 w
+   //  0 w 0
+   AF1 peak=AF1_AU1(const1.x);
+   AF1 wR=ampR*peak;
+   AF1 wG=ampG*peak;
+   AF1 wB=ampB*peak;
+   // Filter.
+   #ifndef CAS_SLOW
+    // Using green coef only, depending on dead code removal to strip out the extra overhead.
+    #ifdef CAS_GO_SLOWER
+     AF1 rcpWeight=ARcpF1(AF1_(1.0)+AF1_(4.0)*wG);
+    #else
+     AF1 rcpWeight=APrxMedRcpF1(AF1_(1.0)+AF1_(4.0)*wG);
+    #endif
+    pixR=ASatF1((b.r*wG+d.r*wG+f.r*wG+h.r*wG+e.r)*rcpWeight);
+    pixG=ASatF1((b.g*wG+d.g*wG+f.g*wG+h.g*wG+e.g)*rcpWeight);
+    pixB=ASatF1((b.b*wG+d.b*wG+f.b*wG+h.b*wG+e.b)*rcpWeight);
+   #else
+    #ifdef CAS_GO_SLOWER
+     AF1 rcpWeightR=ARcpF1(AF1_(1.0)+AF1_(4.0)*wR);
+     AF1 rcpWeightG=ARcpF1(AF1_(1.0)+AF1_(4.0)*wG);
+     AF1 rcpWeightB=ARcpF1(AF1_(1.0)+AF1_(4.0)*wB);
+    #else
+     AF1 rcpWeightR=APrxMedRcpF1(AF1_(1.0)+AF1_(4.0)*wR);
+     AF1 rcpWeightG=APrxMedRcpF1(AF1_(1.0)+AF1_(4.0)*wG);
+     AF1 rcpWeightB=APrxMedRcpF1(AF1_(1.0)+AF1_(4.0)*wB);
+    #endif
+    pixR=ASatF1((b.r*wR+d.r*wR+f.r*wR+h.r*wR+e.r)*rcpWeightR);
+    pixG=ASatF1((b.g*wG+d.g*wG+f.g*wG+h.g*wG+e.g)*rcpWeightG);
+    pixB=ASatF1((b.b*wB+d.b*wB+f.b*wB+h.b*wB+e.b)*rcpWeightB);
+   #endif
+   return;}
+//------------------------------------------------------------------------------------------------------------------------------
+  // Scaling algorithm adaptively interpolates between nearest 4 results of the non-scaling algorithm.
+  //  a b c d
+  //  e f g h
+  //  i j k l
+  //  m n o p
+  // Working these 4 results.
+  //  +-----+-----+
+  //  |     |     |
+  //  |  f..|..g  |
+  //  |  .  |  .  |
+  //  +-----+-----+
+  //  |  .  |  .  |
+  //  |  j..|..k  |
+  //  |     |     |
+  //  +-----+-----+
+  AF2 pp=AF2(ip)*AF2_AU2(const0.xy)+AF2_AU2(const0.zw);
+  AF2 fp=floor(pp);
+  pp-=fp;
+  ASU2 sp=ASU2(fp);
+  AF3 a=CasLoad(sp+ASU2(-1,-1));
+  AF3 b=CasLoad(sp+ASU2( 0,-1));
+  AF3 e=CasLoad(sp+ASU2(-1, 0));
+  AF3 f=CasLoad(sp);
+  AF3 c=CasLoad(sp+ASU2( 1,-1));
+  AF3 d=CasLoad(sp+ASU2( 2,-1));
+  AF3 g=CasLoad(sp+ASU2( 1, 0));
+  AF3 h=CasLoad(sp+ASU2( 2, 0));
+  AF3 i=CasLoad(sp+ASU2(-1, 1));
+  AF3 j=CasLoad(sp+ASU2( 0, 1));
+  AF3 m=CasLoad(sp+ASU2(-1, 2));
+  AF3 n=CasLoad(sp+ASU2( 0, 2));
+  AF3 k=CasLoad(sp+ASU2( 1, 1));
+  AF3 l=CasLoad(sp+ASU2( 2, 1));
+  AF3 o=CasLoad(sp+ASU2( 1, 2));
+  AF3 p=CasLoad(sp+ASU2( 2, 2));
+  // Run optional input transform.
+  CasInput(a.r,a.g,a.b);
+  CasInput(b.r,b.g,b.b);
+  CasInput(c.r,c.g,c.b);
+  CasInput(d.r,d.g,d.b);
+  CasInput(e.r,e.g,e.b);
+  CasInput(f.r,f.g,f.b);
+  CasInput(g.r,g.g,g.b);
+  CasInput(h.r,h.g,h.b);
+  CasInput(i.r,i.g,i.b);
+  CasInput(j.r,j.g,j.b);
+  CasInput(k.r,k.g,k.b);
+  CasInput(l.r,l.g,l.b);
+  CasInput(m.r,m.g,m.b);
+  CasInput(n.r,n.g,n.b);
+  CasInput(o.r,o.g,o.b);
+  CasInput(p.r,p.g,p.b);
+  // Soft min and max.
+  // These are 2.0x bigger (factored out the extra multiply).
+  //  a b c             b
+  //  e f g * 0.5  +  e f g * 0.5  [F]
+  //  i j k             j
+  AF1 mnfR=AMin3F1(AMin3F1(b.r,e.r,f.r),g.r,j.r);
+  AF1 mnfG=AMin3F1(AMin3F1(b.g,e.g,f.g),g.g,j.g);
+  AF1 mnfB=AMin3F1(AMin3F1(b.b,e.b,f.b),g.b,j.b);
+  #ifdef CAS_BETTER_DIAGONALS
+   AF1 mnfR2=AMin3F1(AMin3F1(mnfR,a.r,c.r),i.r,k.r);
+   AF1 mnfG2=AMin3F1(AMin3F1(mnfG,a.g,c.g),i.g,k.g);
+   AF1 mnfB2=AMin3F1(AMin3F1(mnfB,a.b,c.b),i.b,k.b);
+   mnfR=mnfR+mnfR2;
+   mnfG=mnfG+mnfG2;
+   mnfB=mnfB+mnfB2;
+  #endif
+  AF1 mxfR=AMax3F1(AMax3F1(b.r,e.r,f.r),g.r,j.r);
+  AF1 mxfG=AMax3F1(AMax3F1(b.g,e.g,f.g),g.g,j.g);
+  AF1 mxfB=AMax3F1(AMax3F1(b.b,e.b,f.b),g.b,j.b);
+  #ifdef CAS_BETTER_DIAGONALS
+   AF1 mxfR2=AMax3F1(AMax3F1(mxfR,a.r,c.r),i.r,k.r);
+   AF1 mxfG2=AMax3F1(AMax3F1(mxfG,a.g,c.g),i.g,k.g);
+   AF1 mxfB2=AMax3F1(AMax3F1(mxfB,a.b,c.b),i.b,k.b);
+   mxfR=mxfR+mxfR2;
+   mxfG=mxfG+mxfG2;
+   mxfB=mxfB+mxfB2;
+  #endif
+  //  b c d             c
+  //  f g h * 0.5  +  f g h * 0.5  [G]
+  //  j k l             k
+  AF1 mngR=AMin3F1(AMin3F1(c.r,f.r,g.r),h.r,k.r);
+  AF1 mngG=AMin3F1(AMin3F1(c.g,f.g,g.g),h.g,k.g);
+  AF1 mngB=AMin3F1(AMin3F1(c.b,f.b,g.b),h.b,k.b);
+  #ifdef CAS_BETTER_DIAGONALS
+   AF1 mngR2=AMin3F1(AMin3F1(mngR,b.r,d.r),j.r,l.r);
+   AF1 mngG2=AMin3F1(AMin3F1(mngG,b.g,d.g),j.g,l.g);
+   AF1 mngB2=AMin3F1(AMin3F1(mngB,b.b,d.b),j.b,l.b);
+   mngR=mngR+mngR2;
+   mngG=mngG+mngG2;
+   mngB=mngB+mngB2;
+  #endif
+  AF1 mxgR=AMax3F1(AMax3F1(c.r,f.r,g.r),h.r,k.r);
+  AF1 mxgG=AMax3F1(AMax3F1(c.g,f.g,g.g),h.g,k.g);
+  AF1 mxgB=AMax3F1(AMax3F1(c.b,f.b,g.b),h.b,k.b);
+  #ifdef CAS_BETTER_DIAGONALS
+   AF1 mxgR2=AMax3F1(AMax3F1(mxgR,b.r,d.r),j.r,l.r);
+   AF1 mxgG2=AMax3F1(AMax3F1(mxgG,b.g,d.g),j.g,l.g);
+   AF1 mxgB2=AMax3F1(AMax3F1(mxgB,b.b,d.b),j.b,l.b);
+   mxgR=mxgR+mxgR2;
+   mxgG=mxgG+mxgG2;
+   mxgB=mxgB+mxgB2;
+  #endif
+  //  e f g             f
+  //  i j k * 0.5  +  i j k * 0.5  [J]
+  //  m n o             n
+  AF1 mnjR=AMin3F1(AMin3F1(f.r,i.r,j.r),k.r,n.r);
+  AF1 mnjG=AMin3F1(AMin3F1(f.g,i.g,j.g),k.g,n.g);
+  AF1 mnjB=AMin3F1(AMin3F1(f.b,i.b,j.b),k.b,n.b);
+  #ifdef CAS_BETTER_DIAGONALS
+   AF1 mnjR2=AMin3F1(AMin3F1(mnjR,e.r,g.r),m.r,o.r);
+   AF1 mnjG2=AMin3F1(AMin3F1(mnjG,e.g,g.g),m.g,o.g);
+   AF1 mnjB2=AMin3F1(AMin3F1(mnjB,e.b,g.b),m.b,o.b);
+   mnjR=mnjR+mnjR2;
+   mnjG=mnjG+mnjG2;
+   mnjB=mnjB+mnjB2;
+  #endif
+  AF1 mxjR=AMax3F1(AMax3F1(f.r,i.r,j.r),k.r,n.r);
+  AF1 mxjG=AMax3F1(AMax3F1(f.g,i.g,j.g),k.g,n.g);
+  AF1 mxjB=AMax3F1(AMax3F1(f.b,i.b,j.b),k.b,n.b);
+  #ifdef CAS_BETTER_DIAGONALS
+   AF1 mxjR2=AMax3F1(AMax3F1(mxjR,e.r,g.r),m.r,o.r);
+   AF1 mxjG2=AMax3F1(AMax3F1(mxjG,e.g,g.g),m.g,o.g);
+   AF1 mxjB2=AMax3F1(AMax3F1(mxjB,e.b,g.b),m.b,o.b);
+   mxjR=mxjR+mxjR2;
+   mxjG=mxjG+mxjG2;
+   mxjB=mxjB+mxjB2;
+  #endif
+  //  f g h             g
+  //  j k l * 0.5  +  j k l * 0.5  [K]
+  //  n o p             o
+  AF1 mnkR=AMin3F1(AMin3F1(g.r,j.r,k.r),l.r,o.r);
+  AF1 mnkG=AMin3F1(AMin3F1(g.g,j.g,k.g),l.g,o.g);
+  AF1 mnkB=AMin3F1(AMin3F1(g.b,j.b,k.b),l.b,o.b);
+  #ifdef CAS_BETTER_DIAGONALS
+   AF1 mnkR2=AMin3F1(AMin3F1(mnkR,f.r,h.r),n.r,p.r);
+   AF1 mnkG2=AMin3F1(AMin3F1(mnkG,f.g,h.g),n.g,p.g);
+   AF1 mnkB2=AMin3F1(AMin3F1(mnkB,f.b,h.b),n.b,p.b);
+   mnkR=mnkR+mnkR2;
+   mnkG=mnkG+mnkG2;
+   mnkB=mnkB+mnkB2;
+  #endif
+  AF1 mxkR=AMax3F1(AMax3F1(g.r,j.r,k.r),l.r,o.r);
+  AF1 mxkG=AMax3F1(AMax3F1(g.g,j.g,k.g),l.g,o.g);
+  AF1 mxkB=AMax3F1(AMax3F1(g.b,j.b,k.b),l.b,o.b);
+  #ifdef CAS_BETTER_DIAGONALS
+   AF1 mxkR2=AMax3F1(AMax3F1(mxkR,f.r,h.r),n.r,p.r);
+   AF1 mxkG2=AMax3F1(AMax3F1(mxkG,f.g,h.g),n.g,p.g);
+   AF1 mxkB2=AMax3F1(AMax3F1(mxkB,f.b,h.b),n.b,p.b);
+   mxkR=mxkR+mxkR2;
+   mxkG=mxkG+mxkG2;
+   mxkB=mxkB+mxkB2;
+  #endif
+  // Smooth minimum distance to signal limit divided by smooth max.
+  #ifdef CAS_GO_SLOWER
+   AF1 rcpMfR=ARcpF1(mxfR);
+   AF1 rcpMfG=ARcpF1(mxfG);
+   AF1 rcpMfB=ARcpF1(mxfB);
+   AF1 rcpMgR=ARcpF1(mxgR);
+   AF1 rcpMgG=ARcpF1(mxgG);
+   AF1 rcpMgB=ARcpF1(mxgB);
+   AF1 rcpMjR=ARcpF1(mxjR);
+   AF1 rcpMjG=ARcpF1(mxjG);
+   AF1 rcpMjB=ARcpF1(mxjB);
+   AF1 rcpMkR=ARcpF1(mxkR);
+   AF1 rcpMkG=ARcpF1(mxkG);
+   AF1 rcpMkB=ARcpF1(mxkB);
+  #else
+   AF1 rcpMfR=APrxLoRcpF1(mxfR);
+   AF1 rcpMfG=APrxLoRcpF1(mxfG);
+   AF1 rcpMfB=APrxLoRcpF1(mxfB);
+   AF1 rcpMgR=APrxLoRcpF1(mxgR);
+   AF1 rcpMgG=APrxLoRcpF1(mxgG);
+   AF1 rcpMgB=APrxLoRcpF1(mxgB);
+   AF1 rcpMjR=APrxLoRcpF1(mxjR);
+   AF1 rcpMjG=APrxLoRcpF1(mxjG);
+   AF1 rcpMjB=APrxLoRcpF1(mxjB);
+   AF1 rcpMkR=APrxLoRcpF1(mxkR);
+   AF1 rcpMkG=APrxLoRcpF1(mxkG);
+   AF1 rcpMkB=APrxLoRcpF1(mxkB);
+  #endif
+  #ifdef CAS_BETTER_DIAGONALS
+   AF1 ampfR=ASatF1(min(mnfR,AF1_(2.0)-mxfR)*rcpMfR);
+   AF1 ampfG=ASatF1(min(mnfG,AF1_(2.0)-mxfG)*rcpMfG);
+   AF1 ampfB=ASatF1(min(mnfB,AF1_(2.0)-mxfB)*rcpMfB);
+   AF1 ampgR=ASatF1(min(mngR,AF1_(2.0)-mxgR)*rcpMgR);
+   AF1 ampgG=ASatF1(min(mngG,AF1_(2.0)-mxgG)*rcpMgG);
+   AF1 ampgB=ASatF1(min(mngB,AF1_(2.0)-mxgB)*rcpMgB);
+   AF1 ampjR=ASatF1(min(mnjR,AF1_(2.0)-mxjR)*rcpMjR);
+   AF1 ampjG=ASatF1(min(mnjG,AF1_(2.0)-mxjG)*rcpMjG);
+   AF1 ampjB=ASatF1(min(mnjB,AF1_(2.0)-mxjB)*rcpMjB);
+   AF1 ampkR=ASatF1(min(mnkR,AF1_(2.0)-mxkR)*rcpMkR);
+   AF1 ampkG=ASatF1(min(mnkG,AF1_(2.0)-mxkG)*rcpMkG);
+   AF1 ampkB=ASatF1(min(mnkB,AF1_(2.0)-mxkB)*rcpMkB);
+  #else
+   AF1 ampfR=ASatF1(min(mnfR,AF1_(1.0)-mxfR)*rcpMfR);
+   AF1 ampfG=ASatF1(min(mnfG,AF1_(1.0)-mxfG)*rcpMfG);
+   AF1 ampfB=ASatF1(min(mnfB,AF1_(1.0)-mxfB)*rcpMfB);
+   AF1 ampgR=ASatF1(min(mngR,AF1_(1.0)-mxgR)*rcpMgR);
+   AF1 ampgG=ASatF1(min(mngG,AF1_(1.0)-mxgG)*rcpMgG);
+   AF1 ampgB=ASatF1(min(mngB,AF1_(1.0)-mxgB)*rcpMgB);
+   AF1 ampjR=ASatF1(min(mnjR,AF1_(1.0)-mxjR)*rcpMjR);
+   AF1 ampjG=ASatF1(min(mnjG,AF1_(1.0)-mxjG)*rcpMjG);
+   AF1 ampjB=ASatF1(min(mnjB,AF1_(1.0)-mxjB)*rcpMjB);
+   AF1 ampkR=ASatF1(min(mnkR,AF1_(1.0)-mxkR)*rcpMkR);
+   AF1 ampkG=ASatF1(min(mnkG,AF1_(1.0)-mxkG)*rcpMkG);
+   AF1 ampkB=ASatF1(min(mnkB,AF1_(1.0)-mxkB)*rcpMkB);
+  #endif
+  // Shaping amount of sharpening.
+  #ifdef CAS_GO_SLOWER
+   ampfR=sqrt(ampfR);
+   ampfG=sqrt(ampfG);
+   ampfB=sqrt(ampfB);
+   ampgR=sqrt(ampgR);
+   ampgG=sqrt(ampgG);
+   ampgB=sqrt(ampgB);
+   ampjR=sqrt(ampjR);
+   ampjG=sqrt(ampjG);
+   ampjB=sqrt(ampjB);
+   ampkR=sqrt(ampkR);
+   ampkG=sqrt(ampkG);
+   ampkB=sqrt(ampkB);
+  #else
+   ampfR=APrxLoSqrtF1(ampfR);
+   ampfG=APrxLoSqrtF1(ampfG);
+   ampfB=APrxLoSqrtF1(ampfB);
+   ampgR=APrxLoSqrtF1(ampgR);
+   ampgG=APrxLoSqrtF1(ampgG);
+   ampgB=APrxLoSqrtF1(ampgB);
+   ampjR=APrxLoSqrtF1(ampjR);
+   ampjG=APrxLoSqrtF1(ampjG);
+   ampjB=APrxLoSqrtF1(ampjB);
+   ampkR=APrxLoSqrtF1(ampkR);
+   ampkG=APrxLoSqrtF1(ampkG);
+   ampkB=APrxLoSqrtF1(ampkB);
+  #endif
+  // Filter shape.
+  //  0 w 0
+  //  w 1 w
+  //  0 w 0
+  AF1 peak=AF1_AU1(const1.x);
+  AF1 wfR=ampfR*peak;
+  AF1 wfG=ampfG*peak;
+  AF1 wfB=ampfB*peak;
+  AF1 wgR=ampgR*peak;
+  AF1 wgG=ampgG*peak;
+  AF1 wgB=ampgB*peak;
+  AF1 wjR=ampjR*peak;
+  AF1 wjG=ampjG*peak;
+  AF1 wjB=ampjB*peak;
+  AF1 wkR=ampkR*peak;
+  AF1 wkG=ampkG*peak;
+  AF1 wkB=ampkB*peak;
+  // Blend between 4 results.
+  //  s t
+  //  u v
+  AF1 s=(AF1_(1.0)-pp.x)*(AF1_(1.0)-pp.y);
+  AF1 t=           pp.x *(AF1_(1.0)-pp.y);
+  AF1 u=(AF1_(1.0)-pp.x)*           pp.y ;
+  AF1 v=           pp.x *           pp.y ;
+  // Thin edges to hide bilinear interpolation (helps diagonals).
+  AF1 thinB=1.0/32.0;
+  #ifdef CAS_GO_SLOWER
+   s*=ARcpF1(thinB+(mxfG-mnfG));
+   t*=ARcpF1(thinB+(mxgG-mngG));
+   u*=ARcpF1(thinB+(mxjG-mnjG));
+   v*=ARcpF1(thinB+(mxkG-mnkG));
+  #else
+   s*=APrxLoRcpF1(thinB+(mxfG-mnfG));
+   t*=APrxLoRcpF1(thinB+(mxgG-mngG));
+   u*=APrxLoRcpF1(thinB+(mxjG-mnjG));
+   v*=APrxLoRcpF1(thinB+(mxkG-mnkG));
+  #endif
+  // Final weighting.
+  //    b c
+  //  e f g h
+  //  i j k l
+  //    n o
+  //  _____  _____  _____  _____
+  //         fs        gt
+  //
+  //  _____  _____  _____  _____
+  //  fs      s gt  fs  t     gt
+  //         ju        kv
+  //  _____  _____  _____  _____
+  //         fs        gt
+  //  ju      u kv  ju  v     kv
+  //  _____  _____  _____  _____
+  //
+  //         ju        kv
+  AF1 qbeR=wfR*s;
+  AF1 qbeG=wfG*s;
+  AF1 qbeB=wfB*s;
+  AF1 qchR=wgR*t;
+  AF1 qchG=wgG*t;
+  AF1 qchB=wgB*t;
+  AF1 qfR=wgR*t+wjR*u+s;
+  AF1 qfG=wgG*t+wjG*u+s;
+  AF1 qfB=wgB*t+wjB*u+s;
+  AF1 qgR=wfR*s+wkR*v+t;
+  AF1 qgG=wfG*s+wkG*v+t;
+  AF1 qgB=wfB*s+wkB*v+t;
+  AF1 qjR=wfR*s+wkR*v+u;
+  AF1 qjG=wfG*s+wkG*v+u;
+  AF1 qjB=wfB*s+wkB*v+u;
+  AF1 qkR=wgR*t+wjR*u+v;
+  AF1 qkG=wgG*t+wjG*u+v;
+  AF1 qkB=wgB*t+wjB*u+v;
+  AF1 qinR=wjR*u;
+  AF1 qinG=wjG*u;
+  AF1 qinB=wjB*u;
+  AF1 qloR=wkR*v;
+  AF1 qloG=wkG*v;
+  AF1 qloB=wkB*v;
+  // Filter.
+  #ifndef CAS_SLOW
+   // Using green coef only, depending on dead code removal to strip out the extra overhead.
+   #ifdef CAS_GO_SLOWER
+    AF1 rcpWG=ARcpF1(AF1_(2.0)*qbeG+AF1_(2.0)*qchG+AF1_(2.0)*qinG+AF1_(2.0)*qloG+qfG+qgG+qjG+qkG);
+   #else
+    AF1 rcpWG=APrxMedRcpF1(AF1_(2.0)*qbeG+AF1_(2.0)*qchG+AF1_(2.0)*qinG+AF1_(2.0)*qloG+qfG+qgG+qjG+qkG);
+   #endif
+   pixR=ASatF1((b.r*qbeG+e.r*qbeG+c.r*qchG+h.r*qchG+i.r*qinG+n.r*qinG+l.r*qloG+o.r*qloG+f.r*qfG+g.r*qgG+j.r*qjG+k.r*qkG)*rcpWG);
+   pixG=ASatF1((b.g*qbeG+e.g*qbeG+c.g*qchG+h.g*qchG+i.g*qinG+n.g*qinG+l.g*qloG+o.g*qloG+f.g*qfG+g.g*qgG+j.g*qjG+k.g*qkG)*rcpWG);
+   pixB=ASatF1((b.b*qbeG+e.b*qbeG+c.b*qchG+h.b*qchG+i.b*qinG+n.b*qinG+l.b*qloG+o.b*qloG+f.b*qfG+g.b*qgG+j.b*qjG+k.b*qkG)*rcpWG);
+  #else
+   #ifdef CAS_GO_SLOWER
+    AF1 rcpWR=ARcpF1(AF1_(2.0)*qbeR+AF1_(2.0)*qchR+AF1_(2.0)*qinR+AF1_(2.0)*qloR+qfR+qgR+qjR+qkR);
+    AF1 rcpWG=ARcpF1(AF1_(2.0)*qbeG+AF1_(2.0)*qchG+AF1_(2.0)*qinG+AF1_(2.0)*qloG+qfG+qgG+qjG+qkG);
+    AF1 rcpWB=ARcpF1(AF1_(2.0)*qbeB+AF1_(2.0)*qchB+AF1_(2.0)*qinB+AF1_(2.0)*qloB+qfB+qgB+qjB+qkB);
+   #else
+    AF1 rcpWR=APrxMedRcpF1(AF1_(2.0)*qbeR+AF1_(2.0)*qchR+AF1_(2.0)*qinR+AF1_(2.0)*qloR+qfR+qgR+qjR+qkR);
+    AF1 rcpWG=APrxMedRcpF1(AF1_(2.0)*qbeG+AF1_(2.0)*qchG+AF1_(2.0)*qinG+AF1_(2.0)*qloG+qfG+qgG+qjG+qkG);
+    AF1 rcpWB=APrxMedRcpF1(AF1_(2.0)*qbeB+AF1_(2.0)*qchB+AF1_(2.0)*qinB+AF1_(2.0)*qloB+qfB+qgB+qjB+qkB);
+   #endif
+   pixR=ASatF1((b.r*qbeR+e.r*qbeR+c.r*qchR+h.r*qchR+i.r*qinR+n.r*qinR+l.r*qloR+o.r*qloR+f.r*qfR+g.r*qgR+j.r*qjR+k.r*qkR)*rcpWR);
+   pixG=ASatF1((b.g*qbeG+e.g*qbeG+c.g*qchG+h.g*qchG+i.g*qinG+n.g*qinG+l.g*qloG+o.g*qloG+f.g*qfG+g.g*qgG+j.g*qjG+k.g*qkG)*rcpWG);
+   pixB=ASatF1((b.b*qbeB+e.b*qbeB+c.b*qchB+h.b*qchB+i.b*qinB+n.b*qinB+l.b*qloB+o.b*qloB+f.b*qfB+g.b*qgB+j.b*qjB+k.b*qkB)*rcpWB);
+  #endif
+ }
+#endif
+
+#ifdef A_GPU
+void main()
+{
+    vec4 diff = vec4(0.f);
+    uvec2 point = uvec2(vary_fragcoord * out_screen_res.xy);
+    CasFilter(diff.r, diff.g, diff.b, point, cas_param_0, cas_param_1, true);
+    frag_color = vec4(linear_to_srgb(diff.rgb), 1.0);
+}
+#endif
diff --git a/indra/newview/app_settings/shaders/class1/deferred/fxaaF.glsl b/indra/newview/app_settings/shaders/class1/deferred/fxaaF.glsl
index 94dac7e5a9..16e23a3da7 100644
--- a/indra/newview/app_settings/shaders/class1/deferred/fxaaF.glsl
+++ b/indra/newview/app_settings/shaders/class1/deferred/fxaaF.glsl
@@ -256,6 +256,10 @@ A. Or use FXAA_GREEN_AS_LUMA.
     #define FXAA_GLSL_130 0
 #endif
 /*--------------------------------------------------------------------------*/
+#ifndef FXAA_GLSL_400
+    #define FXAA_GLSL_400 0
+#endif
+/*--------------------------------------------------------------------------*/
 #ifndef FXAA_HLSL_3
     #define FXAA_HLSL_3 0
 #endif
@@ -342,8 +346,8 @@ A. Or use FXAA_GREEN_AS_LUMA.
     // 1 = API supports gather4 on alpha channel.
     // 0 = API does not support gather4 on alpha channel.
     //
-    #if (FXAA_GLSL_130 == 0)
-        #define FXAA_GATHER4_ALPHA 0
+    #if (FXAA_GLSL_400 == 1)
+        #define FXAA_GATHER4_ALPHA 1
     #endif
     #if (FXAA_HLSL_5 == 1)
         #define FXAA_GATHER4_ALPHA 1
@@ -652,7 +656,7 @@ NOTE the other tuning knobs are now in the shader function inputs!
                                 API PORTING
 
 ============================================================================*/
-#if (FXAA_GLSL_120 == 1) || (FXAA_GLSL_130 == 1)
+#if (FXAA_GLSL_120 == 1) || (FXAA_GLSL_130 == 1) || (FXAA_GLSL_400 == 1)
     #define FxaaBool bool
     #define FxaaDiscard discard
     #define FxaaFloat float
@@ -714,6 +718,16 @@ NOTE the other tuning knobs are now in the shader function inputs!
     #endif
 #endif
 /*--------------------------------------------------------------------------*/
+#if (FXAA_GLSL_400 == 1)
+    // Requires "#version 400" or better
+    #define FxaaTexTop(t, p) textureLod(t, p, 0.0)
+    #define FxaaTexOff(t, p, o, r) textureLodOffset(t, p, 0.0, o)
+    #define FxaaTexAlpha4(t, p) textureGather(t, p, 3)
+    #define FxaaTexOffAlpha4(t, p, o) textureGatherOffset(t, p, o, 3)
+    #define FxaaTexGreen4(t, p) textureGather(t, p, 1)
+    #define FxaaTexOffGreen4(t, p, o) textureGatherOffset(t, p, o, 1)
+#endif
+/*--------------------------------------------------------------------------*/
 #if (FXAA_HLSL_3 == 1) || (FXAA_360 == 1) || (FXAA_PS3 == 1)
     #define FxaaInt2 float2
     #define FxaaTex sampler2D
diff --git a/indra/newview/app_settings/shaders/class1/windlight/atmosphericsFuncs.glsl b/indra/newview/app_settings/shaders/class1/windlight/atmosphericsFuncs.glsl
index 55daa83750..359bfe8253 100644
--- a/indra/newview/app_settings/shaders/class1/windlight/atmosphericsFuncs.glsl
+++ b/indra/newview/app_settings/shaders/class1/windlight/atmosphericsFuncs.glsl
@@ -124,6 +124,9 @@ void calcAtmosphericVars(vec3 inPositionEye, vec3 light_dir, float ambFactor, ou
     amblit = tmpAmbient;
 
     additive *= vec3(1.0 - combined_haze);
+
+    // sanity clamp haze contribution
+    additive = min(additive, vec3(10));
 }
 
 vec3 srgb_to_linear(vec3 col);
diff --git a/indra/newview/featuretable.txt b/indra/newview/featuretable.txt
index 213f0ab845..9be7254a7a 100644
--- a/indra/newview/featuretable.txt
+++ b/indra/newview/featuretable.txt
@@ -79,6 +79,7 @@ RenderHeroProbeDistance		1	16
 RenderHeroProbeUpdateRate	1	4
 RenderHeroProbeConservativeUpdateMultiplier 1 16
 RenderDownScaleMethod       1   1
+RenderCASSharpness          1   1
 
 
 //
@@ -115,6 +116,7 @@ RenderHeroProbeResolution	1	256
 RenderHeroProbeDistance		1	4
 RenderHeroProbeUpdateRate	1	6
 RenderHeroProbeConservativeUpdateMultiplier 1 16
+RenderCASSharpness          1   0
 
 //
 // Medium Low Graphics Settings
@@ -150,6 +152,7 @@ RenderHeroProbeResolution	1	256
 RenderHeroProbeDistance		1	6
 RenderHeroProbeUpdateRate	1	3
 RenderHeroProbeConservativeUpdateMultiplier 1 16
+RenderCASSharpness          1   0
 
 //
 // Medium Graphics Settings (standard)
@@ -185,6 +188,7 @@ RenderHeroProbeResolution	1	512
 RenderHeroProbeDistance		1	6
 RenderHeroProbeUpdateRate	1	3
 RenderHeroProbeConservativeUpdateMultiplier 1 16
+RenderCASSharpness          1   0
 
 //
 // Medium High Graphics Settings
@@ -220,6 +224,7 @@ RenderHeroProbeResolution	1	512
 RenderHeroProbeDistance		1	6
 RenderHeroProbeUpdateRate	1	2
 RenderHeroProbeConservativeUpdateMultiplier 1 8
+RenderCASSharpness          1   0
 
 //
 // High Graphics Settings (SSAO + sun shadows)
@@ -255,6 +260,7 @@ RenderHeroProbeResolution	1	512
 RenderHeroProbeDistance		1	8
 RenderHeroProbeUpdateRate	1	2
 RenderHeroProbeConservativeUpdateMultiplier 1 8
+RenderCASSharpness          1   0.4
 
 //
 // High Ultra Graphics Settings (deferred + SSAO + all shadows)
@@ -290,6 +296,7 @@ RenderHeroProbeResolution	1	1024
 RenderHeroProbeDistance		1	16
 RenderHeroProbeUpdateRate	1	1
 RenderHeroProbeConservativeUpdateMultiplier 1 4
+RenderCASSharpness          1   0.4
 
 //
 // Ultra graphics (REALLY PURTY!)
@@ -325,6 +332,7 @@ RenderHeroProbeResolution	1	2048
 RenderHeroProbeDistance		1	16
 RenderHeroProbeUpdateRate	1	1
 RenderHeroProbeConservativeUpdateMultiplier 1 4
+RenderCASSharpness          1   0.4
 
 //
 // Class Unknown Hardware (unknown)
diff --git a/indra/newview/featuretable_linux.txt b/indra/newview/featuretable_linux.txt
index 4bcefc1546..6f7f553cec 100644
--- a/indra/newview/featuretable_linux.txt
+++ b/indra/newview/featuretable_linux.txt
@@ -1,4 +1,4 @@
-version 28
+version 29
 // The version number above should be incremented IF AND ONLY IF some
 // change has been made that is sufficiently important to justify
 // resetting the graphics preferences of all users to the recommended
@@ -67,6 +67,7 @@ RenderUseAdvancedAtmospherics 1 0
 RenderShadowDetail			1	2
 RenderFSAASamples			1	16
 RenderMaxTextureIndex		1	16
+RenderMirrors				1	1
 
 //
 // Low Graphics Settings (fixed function)
@@ -97,6 +98,12 @@ RenderUseAdvancedAtmospherics 1 0
 RenderShadowDetail			1	0
 WLSkyDetail					1	48
 RenderFSAASamples			1	0
+RenderMirrors				1	0
+RenderHeroProbeResolution	1	256
+RenderHeroProbeDistance		1	4
+RenderHeroProbeUpdateRate	1   6
+RenderHeroProbeConservativeUpdateMultiplier 1 16
+
 
 //
 // Low Graphics Settings
@@ -127,6 +134,11 @@ RenderUseAdvancedAtmospherics 1 0
 RenderShadowDetail			1	0
 WLSkyDetail					1	48
 RenderFSAASamples			1	0
+RenderMirrors				1	0
+RenderHeroProbeResolution	1	256
+RenderHeroProbeDistance		1	4
+RenderHeroProbeUpdateRate	1   6
+RenderHeroProbeConservativeUpdateMultiplier 1 16
 
 //
 // Medium Low Graphics Settings
@@ -156,6 +168,11 @@ RenderUseAdvancedAtmospherics 1 0
 RenderShadowDetail			1	0
 WLSkyDetail					1	48
 RenderFSAASamples			1	0
+RenderMirrors				1	0
+RenderHeroProbeResolution	1	256
+RenderHeroProbeDistance		1	6
+RenderHeroProbeUpdateRate	1	3
+RenderHeroProbeConservativeUpdateMultiplier 1 16
 
 //
 // Medium Graphics Settings (standard)
@@ -185,6 +202,11 @@ RenderUseAdvancedAtmospherics 1 0
 RenderShadowDetail			1	0
 WLSkyDetail					1	48
 RenderFSAASamples			1	2
+RenderMirrors				1	0
+RenderHeroProbeResolution	1	512
+RenderHeroProbeDistance		1	6
+RenderHeroProbeUpdateRate	1	3
+RenderHeroProbeConservativeUpdateMultiplier 1 16
 
 //
 // Medium High Graphics Settings (deferred enabled)
@@ -214,6 +236,11 @@ RenderDeferredSSAO			1	0
 RenderShadowDetail			1	0
 WLSkyDetail					1	48
 RenderFSAASamples			1	2
+RenderMirrors				1	0
+RenderHeroProbeResolution	1	512
+RenderHeroProbeDistance		1	6
+RenderHeroProbeUpdateRate	1	2
+RenderHeroProbeConservativeUpdateMultiplier 1 8
 
 //
 // High Graphics Settings (deferred + SSAO)
@@ -243,6 +270,11 @@ RenderUseAdvancedAtmospherics 1 0
 RenderShadowDetail			1	0
 WLSkyDetail					1	48
 RenderFSAASamples			1	2
+RenderMirrors				1	0
+RenderHeroProbeResolution	1	512
+RenderHeroProbeDistance		1	8
+RenderHeroProbeUpdateRate	1	2
+RenderHeroProbeConservativeUpdateMultiplier 1 8
 
 //
 // High Ultra Graphics Settings (deferred + SSAO + shadows)
@@ -272,6 +304,11 @@ RenderUseAdvancedAtmospherics 1 0
 RenderShadowDetail			1	2
 WLSkyDetail					1	48
 RenderFSAASamples			1	2
+RenderMirrors				1	0
+RenderHeroProbeResolution	1	512
+RenderHeroProbeDistance		1	16
+RenderHeroProbeUpdateRate	1	1
+RenderHeroProbeConservativeUpdateMultiplier 1 4
 
 //
 // Ultra graphics (REALLY PURTY!)
@@ -300,6 +337,11 @@ RenderDeferredSSAO			1	1
 RenderUseAdvancedAtmospherics 1 0
 RenderShadowDetail			1	2
 RenderFSAASamples			1	2
+RenderMirrors				1	0
+RenderHeroProbeResolution	1	1024
+RenderHeroProbeDistance		1	16
+RenderHeroProbeUpdateRate	1	1
+RenderHeroProbeConservativeUpdateMultiplier 1 4
 
 //
 // Class Unknown Hardware (unknown)
diff --git a/indra/newview/featuretable_mac.txt b/indra/newview/featuretable_mac.txt
index 4f361a8dfe..8eaf5894ab 100644
--- a/indra/newview/featuretable_mac.txt
+++ b/indra/newview/featuretable_mac.txt
@@ -1,4 +1,4 @@
-version 58
+version 59
 // The version number above should be incremented IF AND ONLY IF some
 // change has been made that is sufficiently important to justify
 // resetting the graphics preferences of all users to the recommended
@@ -77,6 +77,7 @@ RenderHeroProbeResolution	1	2048
 RenderHeroProbeDistance		1	16
 RenderHeroProbeUpdateRate	1	4
 RenderHeroProbeConservativeUpdateMultiplier 1 16
+RenderCASSharpness          1   1
 
 //
 // Low Graphics Settings
@@ -112,6 +113,7 @@ RenderHeroProbeResolution	1	256
 RenderHeroProbeDistance		1	4
 RenderHeroProbeUpdateRate	1   6
 RenderHeroProbeConservativeUpdateMultiplier 1 16
+RenderCASSharpness          1   0
 
 //
 // Medium Low Graphics Settings
@@ -147,6 +149,7 @@ RenderHeroProbeResolution	1	256
 RenderHeroProbeDistance		1	6
 RenderHeroProbeUpdateRate	1	3
 RenderHeroProbeConservativeUpdateMultiplier 1 16
+RenderCASSharpness          1   0
 
 //
 // Medium Graphics Settings (standard)
@@ -182,6 +185,7 @@ RenderHeroProbeResolution	1	512
 RenderHeroProbeDistance		1	6
 RenderHeroProbeUpdateRate	1	3
 RenderHeroProbeConservativeUpdateMultiplier 1 16
+RenderCASSharpness          1   0
 
 //
 // Medium High Graphics Settings
@@ -217,6 +221,7 @@ RenderHeroProbeResolution	1	512
 RenderHeroProbeDistance		1	6
 RenderHeroProbeUpdateRate	1	2
 RenderHeroProbeConservativeUpdateMultiplier 1 8
+RenderCASSharpness          1   0
 
 //
 // High Graphics Settings (SSAO + sun shadows)
@@ -247,11 +252,12 @@ RenderReflectionsEnabled    1   1
 RenderReflectionProbeDetail	1	1
 RenderScreenSpaceReflections 1  0
 RenderReflectionProbeLevel  1   1
-RenderMirrors				1	1
+RenderMirrors				1	0
 RenderHeroProbeResolution	1	512
 RenderHeroProbeDistance		1	8
 RenderHeroProbeUpdateRate	1	2
 RenderHeroProbeConservativeUpdateMultiplier 1 8
+RenderCASSharpness          1   0
 
 //
 // High Ultra Graphics Settings (SSAO + all shadows)
@@ -287,6 +293,7 @@ RenderHeroProbeResolution	1	512
 RenderHeroProbeDistance		1	16
 RenderHeroProbeUpdateRate	1	1
 RenderHeroProbeConservativeUpdateMultiplier 1 4
+RenderCASSharpness          1   0.4
 
 //
 // Ultra graphics (REALLY PURTY!)
@@ -322,6 +329,7 @@ RenderHeroProbeResolution	1	1024
 RenderHeroProbeDistance		1	16
 RenderHeroProbeUpdateRate	1	1
 RenderHeroProbeConservativeUpdateMultiplier 1 4
+RenderCASSharpness          1   0.4
 
 //
 // Class Unknown Hardware (unknown)
diff --git a/indra/newview/lldrawpool.cpp b/indra/newview/lldrawpool.cpp
index 739975eab4..7bd5206453 100644
--- a/indra/newview/lldrawpool.cpp
+++ b/indra/newview/lldrawpool.cpp
@@ -407,22 +407,19 @@ void LLRenderPass::renderRiggedGroup(LLSpatialGroup* group, U32 type, bool textu
 {
     LL_PROFILE_ZONE_SCOPED_CATEGORY_DRAWPOOL;
     LLSpatialGroup::drawmap_elem_t& draw_info = group->mDrawMap[type];
-    LLVOAvatar* lastAvatar = nullptr;
+    const LLVOAvatar* lastAvatar = nullptr;
     U64 lastMeshId = 0;
+    bool skipLastSkin = false;
 
     for (LLSpatialGroup::drawmap_elem_t::iterator k = draw_info.begin(); k != draw_info.end(); ++k)
     {
         LLDrawInfo* pparams = *k;
         if (pparams)
         {
-            if (lastAvatar != pparams->mAvatar || lastMeshId != pparams->mSkinInfo->mHash)
+            if (uploadMatrixPalette(pparams->mAvatar, pparams->mSkinInfo, lastAvatar, lastMeshId, skipLastSkin))
             {
-                uploadMatrixPalette(*pparams);
-                lastAvatar = pparams->mAvatar;
-                lastMeshId = pparams->mSkinInfo->mHash;
+                pushBatch(*pparams, texture);
             }
-
-            pushBatch(*pparams, texture);
         }
     }
 }
@@ -468,8 +465,9 @@ void LLRenderPass::pushRiggedBatches(U32 type, bool texture, bool batch_textures
 
     if (texture)
     {
-        LLVOAvatar* lastAvatar = nullptr;
+        const LLVOAvatar* lastAvatar = nullptr;
         U64 lastMeshId = 0;
+        bool skipLastSkin = false;
         auto* begin = gPipeline.beginRenderMap(type);
         auto* end = gPipeline.endRenderMap(type);
         for (LLCullResult::drawinfo_iterator i = begin; i != end; )
@@ -477,14 +475,10 @@ void LLRenderPass::pushRiggedBatches(U32 type, bool texture, bool batch_textures
             LLDrawInfo* pparams = *i;
             LLCullResult::increment_iterator(i, end);
 
-            if (pparams->mAvatar.notNull() && (lastAvatar != pparams->mAvatar || lastMeshId != pparams->mSkinInfo->mHash))
+            if (uploadMatrixPalette(pparams->mAvatar, pparams->mSkinInfo, lastAvatar, lastMeshId, skipLastSkin))
             {
-                uploadMatrixPalette(*pparams);
-                lastAvatar = pparams->mAvatar;
-                lastMeshId = pparams->mSkinInfo->mHash;
+                pushBatch(*pparams, texture, batch_textures);
             }
-
-            pushBatch(*pparams, texture, batch_textures);
         }
     }
     else
@@ -496,8 +490,9 @@ void LLRenderPass::pushRiggedBatches(U32 type, bool texture, bool batch_textures
 void LLRenderPass::pushUntexturedRiggedBatches(U32 type)
 {
     LL_PROFILE_ZONE_SCOPED_CATEGORY_DRAWPOOL;
-    LLVOAvatar* lastAvatar = nullptr;
+    const LLVOAvatar* lastAvatar = nullptr;
     U64 lastMeshId = 0;
+    bool skipLastSkin = false;
     auto* begin = gPipeline.beginRenderMap(type);
     auto* end = gPipeline.endRenderMap(type);
     for (LLCullResult::drawinfo_iterator i = begin; i != end; )
@@ -505,14 +500,10 @@ void LLRenderPass::pushUntexturedRiggedBatches(U32 type)
         LLDrawInfo* pparams = *i;
         LLCullResult::increment_iterator(i, end);
 
-        if (pparams->mAvatar.notNull() && (lastAvatar != pparams->mAvatar || lastMeshId != pparams->mSkinInfo->mHash))
+        if (uploadMatrixPalette(pparams->mAvatar, pparams->mSkinInfo, lastAvatar, lastMeshId, skipLastSkin))
         {
-            uploadMatrixPalette(*pparams);
-            lastAvatar = pparams->mAvatar;
-            lastMeshId = pparams->mSkinInfo->mHash;
+            pushUntexturedBatch(*pparams);
         }
-
-        pushUntexturedBatch(*pparams);
     }
 }
 
@@ -533,8 +524,9 @@ void LLRenderPass::pushMaskBatches(U32 type, bool texture, bool batch_textures)
 void LLRenderPass::pushRiggedMaskBatches(U32 type, bool texture, bool batch_textures)
 {
     LL_PROFILE_ZONE_SCOPED_CATEGORY_DRAWPOOL;
-    LLVOAvatar* lastAvatar = nullptr;
+    const LLVOAvatar* lastAvatar = nullptr;
     U64 lastMeshId = 0;
+    bool skipLastSkin = false;
     auto* begin = gPipeline.beginRenderMap(type);
     auto* end = gPipeline.endRenderMap(type);
     for (LLCullResult::drawinfo_iterator i = begin; i != end; )
@@ -545,23 +537,12 @@ void LLRenderPass::pushRiggedMaskBatches(U32 type, bool texture, bool batch_text
 
         llassert(pparams);
 
-        if (LLGLSLShader::sCurBoundShaderPtr)
-        {
-            LLGLSLShader::sCurBoundShaderPtr->setMinimumAlpha(pparams->mAlphaMaskCutoff);
-        }
-        else
-        {
-            gGL.flush();
-        }
+        LLGLSLShader::sCurBoundShaderPtr->setMinimumAlpha(pparams->mAlphaMaskCutoff);
 
-        if (lastAvatar != pparams->mAvatar || lastMeshId != pparams->mSkinInfo->mHash)
+        if (uploadMatrixPalette(pparams->mAvatar, pparams->mSkinInfo, lastAvatar, lastMeshId, skipLastSkin))
         {
-            uploadMatrixPalette(*pparams);
-            lastAvatar = pparams->mAvatar;
-            lastMeshId = pparams->mSkinInfo->mHash;
+            pushBatch(*pparams, texture, batch_textures);
         }
-
-        pushBatch(*pparams, texture, batch_textures);
     }
 }
 
@@ -667,6 +648,8 @@ bool LLRenderPass::uploadMatrixPalette(LLDrawInfo& params)
 //static
 bool LLRenderPass::uploadMatrixPalette(LLVOAvatar* avatar, LLMeshSkinInfo* skinInfo)
 {
+    LL_PROFILE_ZONE_SCOPED_CATEGORY_AVATAR;
+
     if (!avatar)
     {
         return false;
@@ -688,6 +671,81 @@ bool LLRenderPass::uploadMatrixPalette(LLVOAvatar* avatar, LLMeshSkinInfo* skinI
     return true;
 }
 
+// Returns true if rendering should proceed
+//static
+bool LLRenderPass::uploadMatrixPalette(LLVOAvatar* avatar, LLMeshSkinInfo* skinInfo, const LLVOAvatar*& lastAvatar, U64& lastMeshId, bool& skipLastSkin)
+{
+    LL_PROFILE_ZONE_SCOPED_CATEGORY_AVATAR;
+
+    llassert(skinInfo);
+    llassert(LLGLSLShader::sCurBoundShaderPtr);
+
+    if (!avatar)
+    {
+        return false;
+    }
+
+    if (avatar == lastAvatar && skinInfo->mHash == lastMeshId)
+    {
+        return !skipLastSkin;
+    }
+
+    const LLVOAvatar::MatrixPaletteCache& mpc = avatar->updateSkinInfoMatrixPalette(skinInfo);
+    U32 count = static_cast<U32>(mpc.mMatrixPalette.size());
+    // skipLastSkin -> skin info not loaded yet, don't render
+    skipLastSkin = !bool(count);
+    lastAvatar = avatar;
+    lastMeshId = skinInfo->mHash;
+
+    if (!skipLastSkin)
+    {
+        LLGLSLShader::sCurBoundShaderPtr->uniformMatrix3x4fv(LLViewerShaderMgr::AVATAR_MATRIX,
+            count,
+            false,
+            (GLfloat*)&(mpc.mGLMp[0]));
+    }
+
+    return !skipLastSkin;
+}
+
+// Returns true if rendering should proceed
+//static
+bool LLRenderPass::uploadMatrixPalette(LLVOAvatar* avatar, LLMeshSkinInfo* skinInfo, const LLVOAvatar*& lastAvatar, U64& lastMeshId, const LLGLSLShader*& lastAvatarShader, bool& skipLastSkin)
+{
+    LL_PROFILE_ZONE_SCOPED_CATEGORY_AVATAR;
+
+    llassert(skinInfo);
+    llassert(LLGLSLShader::sCurBoundShaderPtr);
+
+    if (!avatar)
+    {
+        return false;
+    }
+
+    if (avatar == lastAvatar && skinInfo->mHash == lastMeshId && lastAvatarShader == LLGLSLShader::sCurBoundShaderPtr)
+    {
+        return !skipLastSkin;
+    }
+
+    const LLVOAvatar::MatrixPaletteCache& mpc = avatar->updateSkinInfoMatrixPalette(skinInfo);
+    U32 count = static_cast<U32>(mpc.mMatrixPalette.size());
+    // skipLastSkin -> skin info not loaded yet, don't render
+    skipLastSkin = !bool(count);
+    lastAvatar = avatar;
+    lastMeshId = skinInfo->mHash;
+    lastAvatarShader = LLGLSLShader::sCurBoundShaderPtr;
+
+    if (!skipLastSkin)
+    {
+        LLGLSLShader::sCurBoundShaderPtr->uniformMatrix3x4fv(LLViewerShaderMgr::AVATAR_MATRIX,
+            count,
+            false,
+            (GLfloat*)&(mpc.mGLMp[0]));
+    }
+
+    return !skipLastSkin;
+}
+
 void setup_texture_matrix(LLDrawInfo& params)
 {
     if (params.mTextureMatrix)
@@ -717,7 +775,7 @@ void LLRenderPass::pushGLTFBatches(U32 type, bool textured)
     }
     else
     {
-        pushRiggedGLTFBatches(type);
+        pushUntexturedGLTFBatches(type);
     }
 }
 
@@ -801,8 +859,9 @@ void LLRenderPass::pushRiggedGLTFBatches(U32 type, bool textured)
 void LLRenderPass::pushRiggedGLTFBatches(U32 type)
 {
     LL_PROFILE_ZONE_SCOPED_CATEGORY_DRAWPOOL;
-    LLVOAvatar* lastAvatar = nullptr;
+    const LLVOAvatar* lastAvatar = nullptr;
     U64 lastMeshId = 0;
+    bool skipLastSkin = false;
 
     auto* begin = gPipeline.beginRenderMap(type);
     auto* end = gPipeline.endRenderMap(type);
@@ -812,15 +871,16 @@ void LLRenderPass::pushRiggedGLTFBatches(U32 type)
         LLDrawInfo& params = **i;
         LLCullResult::increment_iterator(i, end);
 
-        pushRiggedGLTFBatch(params, lastAvatar, lastMeshId);
+        pushRiggedGLTFBatch(params, lastAvatar, lastMeshId, skipLastSkin);
     }
 }
 
 void LLRenderPass::pushUntexturedRiggedGLTFBatches(U32 type)
 {
     LL_PROFILE_ZONE_SCOPED_CATEGORY_DRAWPOOL;
-    LLVOAvatar* lastAvatar = nullptr;
+    const LLVOAvatar* lastAvatar = nullptr;
     U64 lastMeshId = 0;
+    bool skipLastSkin = false;
 
     auto* begin = gPipeline.beginRenderMap(type);
     auto* end = gPipeline.endRenderMap(type);
@@ -830,34 +890,26 @@ void LLRenderPass::pushUntexturedRiggedGLTFBatches(U32 type)
         LLDrawInfo& params = **i;
         LLCullResult::increment_iterator(i, end);
 
-        pushUntexturedRiggedGLTFBatch(params, lastAvatar, lastMeshId);
+        pushUntexturedRiggedGLTFBatch(params, lastAvatar, lastMeshId, skipLastSkin);
     }
 }
 
 
 // static
-void LLRenderPass::pushRiggedGLTFBatch(LLDrawInfo& params, LLVOAvatar*& lastAvatar, U64& lastMeshId)
+void LLRenderPass::pushRiggedGLTFBatch(LLDrawInfo& params, const LLVOAvatar*& lastAvatar, U64& lastMeshId, bool& skipLastSkin)
 {
-    if (params.mAvatar.notNull() && (lastAvatar != params.mAvatar || lastMeshId != params.mSkinInfo->mHash))
+    if (uploadMatrixPalette(params.mAvatar, params.mSkinInfo, lastAvatar, lastMeshId, skipLastSkin))
     {
-        uploadMatrixPalette(params);
-        lastAvatar = params.mAvatar;
-        lastMeshId = params.mSkinInfo->mHash;
+        pushGLTFBatch(params);
     }
-
-    pushGLTFBatch(params);
 }
 
 // static
-void LLRenderPass::pushUntexturedRiggedGLTFBatch(LLDrawInfo& params, LLVOAvatar*& lastAvatar, U64& lastMeshId)
+void LLRenderPass::pushUntexturedRiggedGLTFBatch(LLDrawInfo& params, const LLVOAvatar*& lastAvatar, U64& lastMeshId, bool& skipLastSkin)
 {
-    if (params.mAvatar.notNull() && (lastAvatar != params.mAvatar || lastMeshId != params.mSkinInfo->mHash))
+    if (uploadMatrixPalette(params.mAvatar, params.mSkinInfo, lastAvatar, lastMeshId, skipLastSkin))
     {
-        uploadMatrixPalette(params);
-        lastAvatar = params.mAvatar;
-        lastMeshId = params.mSkinInfo->mHash;
+        pushUntexturedGLTFBatch(params);
     }
-
-    pushUntexturedGLTFBatch(params);
 }
 
diff --git a/indra/newview/lldrawpool.h b/indra/newview/lldrawpool.h
index 01a29f71a0..bc412214c7 100644
--- a/indra/newview/lldrawpool.h
+++ b/indra/newview/lldrawpool.h
@@ -38,6 +38,7 @@ class LLViewerFetchedTexture;
 class LLSpatialGroup;
 class LLDrawInfo;
 class LLVOAvatar;
+class LLGLSLShader;
 class LLMeshSkinInfo;
 
 class LLDrawPool
@@ -375,9 +376,9 @@ public:
 
     // push a single GLTF draw call
     static void pushGLTFBatch(LLDrawInfo& params);
-    static void pushRiggedGLTFBatch(LLDrawInfo& params, LLVOAvatar*& lastAvatar, U64& lastMeshId);
+    static void pushRiggedGLTFBatch(LLDrawInfo& params, const LLVOAvatar*& lastAvatar, U64& lastMeshId, bool& skipLastSkin);
     static void pushUntexturedGLTFBatch(LLDrawInfo& params);
-    static void pushUntexturedRiggedGLTFBatch(LLDrawInfo& params, LLVOAvatar*& lastAvatar, U64& lastMeshId);
+    static void pushUntexturedRiggedGLTFBatch(LLDrawInfo& params, const LLVOAvatar*& lastAvatar, U64& lastMeshId, bool& skipLastSkin);
 
     void pushMaskBatches(U32 type, bool texture = true, bool batch_textures = false);
     void pushRiggedMaskBatches(U32 type, bool texture = true, bool batch_textures = false);
@@ -386,6 +387,8 @@ public:
     void pushBumpBatch(LLDrawInfo& params, bool texture, bool batch_textures = false);
     static bool uploadMatrixPalette(LLDrawInfo& params);
     static bool uploadMatrixPalette(LLVOAvatar* avatar, LLMeshSkinInfo* skinInfo);
+    static bool uploadMatrixPalette(LLVOAvatar* avatar, LLMeshSkinInfo* skinInfo, const LLVOAvatar*& lastAvatar, U64& lastMeshId, bool& skipLastSkin);
+    static bool uploadMatrixPalette(LLVOAvatar* avatar, LLMeshSkinInfo* skinInfo, const LLVOAvatar*& lastAvatar, U64& lastMeshId, const LLGLSLShader*& lastAvatarShader, bool& skipLastSkin);
     virtual void renderGroup(LLSpatialGroup* group, U32 type, bool texture = true);
     virtual void renderRiggedGroup(LLSpatialGroup* group, U32 type, bool texture = true);
 };
diff --git a/indra/newview/lldrawpoolalpha.cpp b/indra/newview/lldrawpoolalpha.cpp
index cb1439b403..b4d14e22f3 100644
--- a/indra/newview/lldrawpoolalpha.cpp
+++ b/indra/newview/lldrawpoolalpha.cpp
@@ -327,8 +327,9 @@ void LLDrawPoolAlpha::renderAlphaHighlight()
 {
     for (int pass = 0; pass < 2; ++pass)
     { //two passes, one rigged and one not
-        LLVOAvatar* lastAvatar = nullptr;
+        const LLVOAvatar* lastAvatar = nullptr;
         U64 lastMeshId = 0;
+        bool skipLastSkin = false;
 
         LLCullResult::sg_iterator begin = pass == 0 ? gPipeline.beginAlphaGroups() : gPipeline.beginRiggedAlphaGroups();
         LLCullResult::sg_iterator end = pass == 0 ? gPipeline.endAlphaGroups() : gPipeline.endRiggedAlphaGroups();
@@ -347,22 +348,16 @@ void LLDrawPoolAlpha::renderAlphaHighlight()
 
                     bool rigged = (params.mAvatar != nullptr);
                     gHighlightProgram.bind(rigged);
-                    gGL.diffuseColor4f(1, 0, 0, 1);
 
                     if (rigged)
                     {
-                        if (lastAvatar != params.mAvatar ||
-                            lastMeshId != params.mSkinInfo->mHash)
-                        {
-                            if (!uploadMatrixPalette(params))
-                            {
-                                continue;
-                            }
-                            lastAvatar = params.mAvatar;
-                            lastMeshId = params.mSkinInfo->mHash;
+                        if (!uploadMatrixPalette(params.mAvatar, params.mSkinInfo, lastAvatar, lastMeshId, skipLastSkin))
+                        { // failed to upload matrix palette, skip rendering
+                            continue;
                         }
                     }
 
+                    gGL.diffuseColor4f(1, 0, 0, 1);
                     LLRenderPass::applyModelMatrix(params);
                     params.mVertexBuffer->setBuffer();
                     params.mVertexBuffer->drawRange(LLRender::TRIANGLES, params.mStart, params.mEnd, params.mCount, params.mOffset);
@@ -526,25 +521,20 @@ void LLDrawPoolAlpha::renderRiggedEmissives(std::vector<LLDrawInfo*>& emissives)
     shader->bind();
     shader->uniform1f(LLShaderMgr::EMISSIVE_BRIGHTNESS, 1.f);
 
-    LLVOAvatar* lastAvatar = nullptr;
+    const LLVOAvatar* lastAvatar = nullptr;
     U64 lastMeshId = 0;
+    bool skipLastSkin = false;
 
     for (LLDrawInfo* draw : emissives)
     {
         LL_PROFILE_ZONE_NAMED_CATEGORY_DRAWPOOL("Emissives");
 
-        bool tex_setup = TexSetup(draw, false);
-        if (lastAvatar != draw->mAvatar || lastMeshId != draw->mSkinInfo->mHash)
+        if (uploadMatrixPalette(draw->mAvatar, draw->mSkinInfo, lastAvatar, lastMeshId, skipLastSkin))
         {
-            if (!uploadMatrixPalette(*draw))
-            { // failed to upload matrix palette, skip rendering
-                continue;
-            }
-            lastAvatar = draw->mAvatar;
-            lastMeshId = draw->mSkinInfo->mHash;
+            bool tex_setup = TexSetup(draw, false);
+            drawEmissive(draw);
+            RestoreTexSetup(tex_setup);
         }
-        drawEmissive(draw);
-        RestoreTexSetup(tex_setup);
     }
 }
 
@@ -553,19 +543,15 @@ void LLDrawPoolAlpha::renderRiggedPbrEmissives(std::vector<LLDrawInfo*>& emissiv
     LLGLDepthTest depth(GL_TRUE, GL_FALSE); //disable depth writes since "emissive" is additive so sorting doesn't matter
     pbr_emissive_shader->bind(true);
 
-    LLVOAvatar* lastAvatar = nullptr;
+    const LLVOAvatar* lastAvatar = nullptr;
     U64 lastMeshId = 0;
+    bool skipLastSkin = false;
 
     for (LLDrawInfo* draw : emissives)
     {
-        if (lastAvatar != draw->mAvatar || lastMeshId != draw->mSkinInfo->mHash)
-        {
-            if (!uploadMatrixPalette(*draw))
-            { // failed to upload matrix palette, skip rendering
-                continue;
-            }
-            lastAvatar = draw->mAvatar;
-            lastMeshId = draw->mSkinInfo->mHash;
+        if (!uploadMatrixPalette(draw->mAvatar, draw->mSkinInfo, lastAvatar, lastMeshId, skipLastSkin))
+        { // failed to upload matrix palette, skip rendering
+            continue;
         }
 
         LLGLDisable cull_face(draw->mGLTFMaterial->mDoubleSided ? GL_CULL_FACE : 0);
@@ -581,9 +567,10 @@ void LLDrawPoolAlpha::renderAlpha(U32 mask, bool depth_only, bool rigged)
     bool initialized_lighting = false;
     bool light_enabled = true;
 
-    LLVOAvatar* lastAvatar = nullptr;
+    const LLVOAvatar* lastAvatar = nullptr;
     U64 lastMeshId = 0;
-    LLGLSLShader* lastAvatarShader = nullptr;
+    const LLGLSLShader* lastAvatarShader = nullptr;
+    bool skipLastSkin = false;
 
     LLCullResult::sg_iterator begin;
     LLCullResult::sg_iterator end;
@@ -776,20 +763,9 @@ void LLDrawPoolAlpha::renderAlpha(U32 mask, bool depth_only, bool rigged)
                     }
                 }
 
-                if (params.mAvatar != nullptr)
+                if (params.mAvatar && !uploadMatrixPalette(params.mAvatar, params.mSkinInfo, lastAvatar, lastMeshId, lastAvatarShader, skipLastSkin))
                 {
-                    if (lastAvatar != params.mAvatar ||
-                        lastMeshId != params.mSkinInfo->mHash ||
-                        lastAvatarShader != LLGLSLShader::sCurBoundShaderPtr)
-                    {
-                        if (!uploadMatrixPalette(params))
-                        {
-                            continue;
-                        }
-                        lastAvatar = params.mAvatar;
-                        lastMeshId = params.mSkinInfo->mHash;
-                        lastAvatarShader = LLGLSLShader::sCurBoundShaderPtr;
-                    }
+                    continue;
                 }
 
                 bool tex_setup = TexSetup(&params, (mat != nullptr));
@@ -912,26 +888,3 @@ void LLDrawPoolAlpha::renderAlpha(U32 mask, bool depth_only, bool rigged)
         gPipeline.enableLightsDynamic();
     }
 }
-
-bool LLDrawPoolAlpha::uploadMatrixPalette(const LLDrawInfo& params)
-{
-    if (params.mAvatar.isNull())
-    {
-        return false;
-    }
-    const LLVOAvatar::MatrixPaletteCache& mpc = params.mAvatar.get()->updateSkinInfoMatrixPalette(params.mSkinInfo);
-    U32 count = static_cast<U32>(mpc.mMatrixPalette.size());
-
-    if (count == 0)
-    {
-        //skin info not loaded yet, don't render
-        return false;
-    }
-
-    LLGLSLShader::sCurBoundShaderPtr->uniformMatrix3x4fv(LLViewerShaderMgr::AVATAR_MATRIX,
-        count,
-        false,
-        (GLfloat*)&(mpc.mGLMp[0]));
-
-    return true;
-}
diff --git a/indra/newview/lldrawpoolalpha.h b/indra/newview/lldrawpoolalpha.h
index 0abe001714..25044beda0 100644
--- a/indra/newview/lldrawpoolalpha.h
+++ b/indra/newview/lldrawpoolalpha.h
@@ -66,7 +66,6 @@ public:
     void renderGroupAlpha(LLSpatialGroup* group, U32 type, U32 mask, bool texture = true);
     void renderAlpha(U32 mask, bool depth_only = false, bool rigged = false);
     void renderAlphaHighlight();
-    bool uploadMatrixPalette(const LLDrawInfo& params);
 
     static bool sShowDebugAlpha;
 
diff --git a/indra/newview/lldrawpoolbump.cpp b/indra/newview/lldrawpoolbump.cpp
index 7289e95b6e..bf593bff07 100644
--- a/indra/newview/lldrawpoolbump.cpp
+++ b/indra/newview/lldrawpoolbump.cpp
@@ -558,8 +558,9 @@ void LLDrawPoolBump::renderDeferred(S32 pass)
         LLCullResult::drawinfo_iterator begin = gPipeline.beginRenderMap(type);
         LLCullResult::drawinfo_iterator end = gPipeline.endRenderMap(type);
 
-        LLVOAvatar* avatar = nullptr;
-        U64 skin = 0;
+        const LLVOAvatar* lastAvatar = nullptr;
+        U64 lastMeshId = 0;
+        bool skipLastSkin = false;
 
         for (LLCullResult::drawinfo_iterator i = begin; i != end; )
         {
@@ -572,13 +573,10 @@ void LLDrawPoolBump::renderDeferred(S32 pass)
 
             if (rigged)
             {
-                if (avatar != params.mAvatar || skin != params.mSkinInfo->mHash)
+                if (uploadMatrixPalette(params.mAvatar, params.mSkinInfo, lastAvatar, lastMeshId, skipLastSkin))
                 {
-                    uploadMatrixPalette(params);
-                    avatar = params.mAvatar;
-                    skin = params.mSkinInfo->mHash;
+                    pushBumpBatch(params, true, false);
                 }
-                pushBumpBatch(params, true, false);
             }
             else
             {
@@ -974,8 +972,9 @@ void LLBumpImageList::onSourceUpdated(LLViewerTexture* src, EBumpEffect bump_cod
 
 void LLDrawPoolBump::pushBumpBatches(U32 type)
 {
-    LLVOAvatar* avatar = nullptr;
-    U64 skin = 0;
+    const LLVOAvatar* lastAvatar = nullptr;
+    U64 lastMeshId = 0;
+    bool skipLastSkin = false;
 
     if (mRigged)
     { // nudge type enum and include skinweights for rigged pass
@@ -993,17 +992,9 @@ void LLDrawPoolBump::pushBumpBatches(U32 type)
         {
             if (mRigged)
             {
-                if (avatar != params.mAvatar || skin != params.mSkinInfo->mHash)
+                if (!uploadMatrixPalette(params.mAvatar, params.mSkinInfo, lastAvatar, lastMeshId, skipLastSkin))
                 {
-                    if (uploadMatrixPalette(params))
-                    {
-                        avatar = params.mAvatar;
-                        skin = params.mSkinInfo->mHash;
-                    }
-                    else
-                    {
-                        continue;
-                    }
+                    continue;
                 }
             }
             pushBumpBatch(params, false);
diff --git a/indra/newview/lldrawpoolmaterials.cpp b/indra/newview/lldrawpoolmaterials.cpp
index bbc93c2af9..e7ec2022d2 100644
--- a/indra/newview/lldrawpoolmaterials.cpp
+++ b/indra/newview/lldrawpoolmaterials.cpp
@@ -183,7 +183,9 @@ void LLDrawPoolMaterials::renderDeferred(S32 pass)
         glUniform4fv(specular, 1, lastSpecular.mV);
     }
 
-    LLVOAvatar* lastAvatar = nullptr;
+    const LLVOAvatar* lastAvatar = nullptr;
+    U64 lastMeshId = 0;
+    bool skipLastSkin = false;
 
     for (LLCullResult::drawinfo_iterator i = begin; i != end; )
     {
@@ -245,23 +247,11 @@ void LLDrawPoolMaterials::renderDeferred(S32 pass)
         }
 
         // upload matrix palette to shader
-        if (rigged && params.mAvatar.notNull())
+        if (rigged)
         {
-            if (params.mAvatar != lastAvatar)
+            if (!uploadMatrixPalette(params.mAvatar, params.mSkinInfo, lastAvatar, lastMeshId, skipLastSkin))
             {
-                const LLVOAvatar::MatrixPaletteCache& mpc = params.mAvatar->updateSkinInfoMatrixPalette(params.mSkinInfo);
-                U32 count = static_cast<U32>(mpc.mMatrixPalette.size());
-
-                if (count == 0)
-                {
-                    //skin info not loaded yet, don't render
-                    return;
-                }
-
-                mShader->uniformMatrix3x4fv(LLViewerShaderMgr::AVATAR_MATRIX,
-                    count,
-                    false,
-                    (GLfloat*)&(mpc.mGLMp[0]));
+                continue;
             }
         }
 
diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp
index 82a0bfb9bb..d92a5be8b6 100644
--- a/indra/newview/llface.cpp
+++ b/indra/newview/llface.cpp
@@ -822,11 +822,6 @@ bool LLFace::genVolumeBBoxes(const LLVolume &volume, S32 f,
 
         const LLVolumeFace &face = volume.getVolumeFace(f);
 
-        LL_DEBUGS("RiggedBox") << "updating extents for face " << f
-                               << " starting extents " << mExtents[0] << ", " << mExtents[1]
-                               << " starting vf extents " << face.mExtents[0] << ", " << face.mExtents[1]
-                               << " num verts " << face.mNumVertices << LL_ENDL;
-
         // MAINT-8264 - stray vertices, especially in low LODs, cause bounding box errors.
         if (face.mNumVertices < 3)
         {
@@ -844,21 +839,14 @@ bool LLFace::genVolumeBBoxes(const LLVolume &volume, S32 f,
 
         matMulBoundBox(mat_vert, face.mExtents, mExtents);
 
-        LL_DEBUGS("RiggedBox") << "updated extents for face " << f
-                               << " bbox gave extents " << mExtents[0] << ", " << mExtents[1] << LL_ENDL;
-
         if (!mDrawablep->isActive())
         {   // Shift position for region
             LLVector4a offset;
             offset.load3(mDrawablep->getRegion()->getOriginAgent().mV);
             mExtents[0].add(offset);
             mExtents[1].add(offset);
-            LL_DEBUGS("RiggedBox") << "updating extents for face " << f
-                                   << " not active, added offset " << offset << LL_ENDL;
         }
 
-        LL_DEBUGS("RiggedBox") << "updated extents for face " << f
-                               << " to " << mExtents[0] << ", " << mExtents[1] << LL_ENDL;
         LLVector4a t;
         t.setAdd(mExtents[0],mExtents[1]);
         t.mul(0.5f);
@@ -2148,7 +2136,7 @@ F32 LLFace::getTextureVirtualSize()
         face_area =  mPixelArea / llclamp(texel_area, 0.015625f, 128.f);
     }
 
-    face_area = LLFace::adjustPixelArea(mImportanceToCamera, face_area) ;
+    face_area = LLFace::adjustPixelArea(mImportanceToCamera, face_area);
     if(face_area > LLViewerTexture::sMinLargeImageSize) //if is large image, shrink face_area by considering the partial overlapping.
     {
         if(mImportanceToCamera > LEAST_IMPORTANCE_FOR_LARGE_IMAGE && mTexture[LLRender::DIFFUSE_MAP].notNull() && mTexture[LLRender::DIFFUSE_MAP]->isLargeImage())
@@ -2166,7 +2154,6 @@ bool LLFace::calcPixelArea(F32& cos_angle_to_view_dir, F32& radius)
 {
     LL_PROFILE_ZONE_SCOPED_CATEGORY_FACE;
 
-    //VECTORIZE THIS
     //get area of circle around face
 
     LLVector4a center;
@@ -2205,13 +2192,6 @@ bool LLFace::calcPixelArea(F32& cos_angle_to_view_dir, F32& radius)
 
     F32 dist = lookAt.getLength3().getF32();
     dist = llmax(dist-size.getLength3().getF32(), 0.001f);
-    //ramp down distance for nearby objects
-    if (dist < 16.f)
-    {
-        dist /= 16.f;
-        dist *= dist;
-        dist *= 16.f;
-    }
 
     lookAt.normalize3fast() ;
 
@@ -2298,6 +2278,7 @@ const F32 FACE_IMPORTANCE_TO_CAMERA_OVER_ANGLE[FACE_IMPORTANCE_LEVEL][2] =    //
 //static
 F32 LLFace::calcImportanceToCamera(F32 cos_angle_to_view_dir, F32 dist)
 {
+    LL_PROFILE_ZONE_SCOPED_CATEGORY_FACE;
     F32 importance = 0.f ;
 
     if(cos_angle_to_view_dir > LLViewerCamera::getInstance()->getCosHalfFov() &&
diff --git a/indra/newview/llface.h b/indra/newview/llface.h
index 7cf256f731..99642016f7 100644
--- a/indra/newview/llface.h
+++ b/indra/newview/llface.h
@@ -209,7 +209,6 @@ public:
     void        setDrawInfo(LLDrawInfo* draw_info);
 
     F32         getTextureVirtualSize() ;
-    F32         getImportanceToCamera()const {return mImportanceToCamera ;}
     void        resetVirtualSize();
 
     void        setHasMedia(bool has_media)  { mHasMedia = has_media ;}
@@ -265,6 +264,11 @@ public:
     // return mSkinInfo->mHash or 0 if mSkinInfo is null
     U64 getSkinHash();
 
+    // true if face was recently in the main camera frustum according to LLViewerTextureList updates
+    bool mInFrustum = false;
+    // value of gFrameCount the last time the face was touched by LLViewerTextureList::updateImageDecodePriority
+    U32 mLastTextureUpdate = 0;
+
 private:
     LLPointer<LLVertexBuffer> mVertexBuffer;
     LLPointer<LLVertexBuffer> mVertexBufferGLTF;
diff --git a/indra/newview/llfloaterimsession.cpp b/indra/newview/llfloaterimsession.cpp
index 557b3f27c5..97e0d01b52 100644
--- a/indra/newview/llfloaterimsession.cpp
+++ b/indra/newview/llfloaterimsession.cpp
@@ -368,8 +368,6 @@ bool LLFloaterIMSession::postBuild()
     add_btn->setEnabled(isInviteAllowed());
     add_btn->setClickedCallback(boost::bind(&LLFloaterIMSession::onAddButtonClicked, this));
 
-    childSetAction("voice_call_btn", boost::bind(&LLFloaterIMSession::onCallButtonClicked, this));
-
     LLVoiceClient::addObserver(this);
 
     //*TODO if session is not initialized yet, add some sort of a warning message like "starting session...blablabla"
@@ -551,23 +549,6 @@ void LLFloaterIMSession::boundVoiceChannel()
     }
 }
 
-void LLFloaterIMSession::onCallButtonClicked()
-{
-    LLVoiceChannel* voice_channel = LLIMModel::getInstance()->getVoiceChannel(mSessionID);
-    if (voice_channel)
-    {
-        bool is_call_active = voice_channel->getState() >= LLVoiceChannel::STATE_CALL_STARTED;
-        if (is_call_active)
-        {
-            gIMMgr->endCall(mSessionID);
-        }
-        else
-        {
-            gIMMgr->startCall(mSessionID);
-        }
-    }
-}
-
 void LLFloaterIMSession::onChange(EStatusType status, const LLSD& channelInfo, bool proximal)
 {
     if(status != STATUS_JOINING && status != STATUS_LEFT_CHANNEL)
diff --git a/indra/newview/llfloaterimsession.h b/indra/newview/llfloaterimsession.h
index 7bc54cf3da..a442099ad1 100644
--- a/indra/newview/llfloaterimsession.h
+++ b/indra/newview/llfloaterimsession.h
@@ -159,8 +159,6 @@ private:
     void sendParticipantsAddedNotification(const uuid_vec_t& uuids);
     bool canAddSelectedToChat(const uuid_vec_t& uuids);
 
-    void onCallButtonClicked();
-
     void onVoiceChannelChanged(const LLUUID &session_id);
 
     void boundVoiceChannel();
diff --git a/indra/newview/llfloaterimsessiontab.cpp b/indra/newview/llfloaterimsessiontab.cpp
index 2621ce576c..00e7dae657 100644
--- a/indra/newview/llfloaterimsessiontab.cpp
+++ b/indra/newview/llfloaterimsessiontab.cpp
@@ -302,6 +302,7 @@ bool LLFloaterIMSessionTab::postBuild()
     mGearBtn = getChild<LLButton>("gear_btn");
     mAddBtn = getChild<LLButton>("add_btn");
     mVoiceButton = getChild<LLButton>("voice_call_btn");
+    mVoiceButton->setClickedCallback([this](LLUICtrl*, const LLSD&) { onCallButtonClicked(); });
 
     mParticipantListPanel = getChild<LLLayoutPanel>("speakers_list_panel");
     mRightPartPanel = getChild<LLLayoutPanel>("right_part_holder");
@@ -434,16 +435,34 @@ void LLFloaterIMSessionTab::draw()
 
 void LLFloaterIMSessionTab::enableDisableCallBtn()
 {
-    if (LLVoiceClient::instanceExists() && mVoiceButton)
+    if (!mVoiceButton)
+        return;
+
+    bool enable = false;
+
+    if (mSessionID.notNull() && mSession && mSession->mSessionInitialized && mSession->mCallBackEnabled)
     {
-        mVoiceButton->setEnabled(
-            mSessionID.notNull()
-            && mSession
-            && mSession->mSessionInitialized
-            && LLVoiceClient::getInstance()->voiceEnabled()
-            && LLVoiceClient::getInstance()->isVoiceWorking()
-            && mSession->mCallBackEnabled);
+        if (mVoiceButtonHangUpMode)
+        {
+            // We allow to hang up from any state
+            enable = true;
+        }
+        else
+        {
+            // We allow to start call from this state only
+            if (mSession->mVoiceChannel->getState() == LLVoiceChannel::STATE_NO_CHANNEL_INFO &&
+                LLVoiceClient::instanceExists())
+            {
+                LLVoiceClient* client = LLVoiceClient::getInstance();
+                if (client->voiceEnabled() && client->isVoiceWorking())
+                {
+                    enable = true;
+                }
+            }
+        }
     }
+
+    mVoiceButton->setEnabled(enable);
 }
 
 // virtual
@@ -466,6 +485,25 @@ void LLFloaterIMSessionTab::onFocusLost()
     super::onFocusLost();
 }
 
+void LLFloaterIMSessionTab::onCallButtonClicked()
+{
+    if (mVoiceButtonHangUpMode)
+    {
+        // We allow to hang up from any state
+        gIMMgr->endCall(mSessionID);
+    }
+    else
+    {
+        LLVoiceChannel::EState channel_state = mSession && mSession->mVoiceChannel ?
+            mSession->mVoiceChannel->getState() : LLVoiceChannel::STATE_NO_CHANNEL_INFO;
+        // We allow to start call from this state only
+        if (channel_state == LLVoiceChannel::STATE_NO_CHANNEL_INFO)
+        {
+            gIMMgr->startCall(mSessionID);
+        }
+    }
+}
+
 void LLFloaterIMSessionTab::onInputEditorClicked()
 {
     LLFloaterIMContainer* im_box = LLFloaterIMContainer::findInstance();
@@ -1040,6 +1078,7 @@ void LLFloaterIMSessionTab::updateCallBtnState(bool callIsActive)
 {
     mVoiceButton->setImageOverlay(callIsActive? getString("call_btn_stop") : getString("call_btn_start"));
     mVoiceButton->setToolTip(callIsActive? getString("end_call_button_tooltip") : getString("start_call_button_tooltip"));
+    mVoiceButtonHangUpMode = callIsActive;
 
     enableDisableCallBtn();
 }
diff --git a/indra/newview/llfloaterimsessiontab.h b/indra/newview/llfloaterimsessiontab.h
index 6dd8e62482..bee5c8c2c4 100644
--- a/indra/newview/llfloaterimsessiontab.h
+++ b/indra/newview/llfloaterimsessiontab.h
@@ -198,6 +198,11 @@ protected:
     LLButton* mAddBtn;
     LLButton* mVoiceButton;
 
+    // Since mVoiceButton can work in one of two modes, "Start call" or "Hang up",
+    // (with different images and tooltips depending on the currently chosen mode)
+    // we should track the mode we're currently using to react on click accordingly
+    bool mVoiceButtonHangUpMode { false };
+
 private:
     // Handling selection and contextual menu
     void doToSelected(const LLSD& userdata);
@@ -216,6 +221,8 @@ private:
      */
     void reshapeChatLayoutPanel();
 
+    void onCallButtonClicked();
+
     void onInputEditorClicked();
 
     void onEmojiRecentPanelToggleBtnClicked();
diff --git a/indra/newview/llhudnametag.cpp b/indra/newview/llhudnametag.cpp
index 205089c662..11f049564a 100644
--- a/indra/newview/llhudnametag.cpp
+++ b/indra/newview/llhudnametag.cpp
@@ -229,35 +229,20 @@ void LLHUDNameTag::render()
     if (sDisplayText)
     {
         LLGLDepthTest gls_depth(GL_TRUE, GL_FALSE);
-        //LLGLDisable gls_stencil(GL_STENCIL_TEST);
-        renderText(false);
+        renderText();
     }
 }
 
-void LLHUDNameTag::renderText(bool for_select)
+void LLHUDNameTag::renderText()
 {
     if (!mVisible || mHidden)
     {
         return;
     }
 
-    // don't pick text that isn't bound to a viewerobject
-    if (for_select &&
-        (!mSourceObject || mSourceObject->mDrawable.isNull()))
-    {
-        return;
-    }
-
-    if (for_select)
-    {
-        gGL.getTexUnit(0)->disable();
-    }
-    else
-    {
-        gGL.getTexUnit(0)->enable(LLTexUnit::TT_TEXTURE);
-    }
+    LL_PROFILE_ZONE_SCOPED_CATEGORY_UI;
 
-    LLGLState gls_blend(GL_BLEND, !for_select);
+    gGL.getTexUnit(0)->enable(LLTexUnit::TT_TEXTURE);
 
     LLColor4 shadow_color(0.f, 0.f, 0.f, 1.f);
     F32 alpha_factor = 1.f;
@@ -393,10 +378,6 @@ void LLHUDNameTag::renderText(bool for_select)
     }
     /// Reset the default color to white.  The renderer expects this to be the default.
     gGL.color4f(1.0f, 1.0f, 1.0f, 1.0f);
-    if (for_select)
-    {
-        gGL.getTexUnit(0)->enable(LLTexUnit::TT_TEXTURE);
-    }
 }
 
 void LLHUDNameTag::setString(const std::string &text_utf8)
diff --git a/indra/newview/llhudnametag.h b/indra/newview/llhudnametag.h
index 9abd8f33cc..5cb7fa877e 100644
--- a/indra/newview/llhudnametag.h
+++ b/indra/newview/llhudnametag.h
@@ -145,7 +145,7 @@ protected:
     LLHUDNameTag(const U8 type);
 
     /*virtual*/ void render();
-    void renderText(bool for_select);
+    void renderText();
     static void updateAll();
     void setLOD(S32 lod);
     S32 getMaxLines();
diff --git a/indra/newview/lllogininstance.cpp b/indra/newview/lllogininstance.cpp
index c54d2c080f..ad04c11cc6 100644
--- a/indra/newview/lllogininstance.cpp
+++ b/indra/newview/lllogininstance.cpp
@@ -602,13 +602,14 @@ std::string construct_start_string()
         {
             // a startup URL was specified
             LLVector3 position = start_slurl.getPosition();
-            std::string unescaped_start =
+            // NOTE - do not xml escape here, will get escaped properly later by LLSD::asXMLRPCValue()
+            // see secondlife/viewer#2395
+            start =
             STRINGIZE(  "uri:"
                       << start_slurl.getRegion() << "&"
                         << position[VX] << "&"
                         << position[VY] << "&"
                         << position[VZ]);
-            start = LLStringFn::xml_encode(unescaped_start, true);
             break;
         }
         case LLSLURL::HOME_LOCATION:
diff --git a/indra/newview/llpanelenvironment.cpp b/indra/newview/llpanelenvironment.cpp
index 51e2c05070..2a4323d3e6 100644
--- a/indra/newview/llpanelenvironment.cpp
+++ b/indra/newview/llpanelenvironment.cpp
@@ -296,7 +296,7 @@ void LLPanelEnvironmentInfo::refresh()
     F32Hours dayoffset(mCurrentEnvironment->mDayOffset);
 
     if (dayoffset.value() > 12.0f)
-        dayoffset -= F32Hours(24.0);
+        dayoffset -= daylength;
 
     mSliderDayLength->setValue(daylength.value());
     mSliderDayOffset->setValue(dayoffset.value());
@@ -723,6 +723,11 @@ void LLPanelEnvironmentInfo::onSldDayLengthChanged(F32 value)
         F32Hours daylength(value);
 
         mCurrentEnvironment->mDayLength = daylength;
+        F32 offset = (F32)mSliderDayOffset->getValue().asReal();
+        if (offset <= 0.0f)
+        {
+            onSldDayOffsetChanged(offset);
+        }
         setDirtyFlag(DIRTY_FLAG_DAYLENGTH);
 
         udpateApparentTimeOfDay();
@@ -736,7 +741,8 @@ void LLPanelEnvironmentInfo::onSldDayOffsetChanged(F32 value)
         F32Hours dayoffset(value);
 
         if (dayoffset.value() <= 0.0f)
-            dayoffset += F32Hours(24.0);
+            // if day cycle is 5 hours long, we want -1h offset to result in 4h
+            dayoffset += mCurrentEnvironment->mDayLength;
 
         mCurrentEnvironment->mDayOffset = dayoffset;
         setDirtyFlag(DIRTY_FLAG_DAYOFFSET);
@@ -929,7 +935,7 @@ void LLPanelEnvironmentInfo::udpateApparentTimeOfDay()
 {
     static const F32 SECONDSINDAY(24.0 * 60.0 * 60.0);
 
-    if ((!mCurrentEnvironment) || (mCurrentEnvironment->mDayLength.value() < 1.0) || (mCurrentEnvironment->mDayOffset.value() < 1.0))
+    if ((!mCurrentEnvironment) || (mCurrentEnvironment->mDayLength.value() < 1.0))
     {
         mLabelApparentTime->setVisible(false);
         return;
diff --git a/indra/newview/llspatialpartition.cpp b/indra/newview/llspatialpartition.cpp
index b1f80e48af..e517d009f5 100644
--- a/indra/newview/llspatialpartition.cpp
+++ b/indra/newview/llspatialpartition.cpp
@@ -29,7 +29,6 @@
 #include "llspatialpartition.h"
 
 #include "llappviewer.h"
-#include "llcallstack.h"
 #include "lltexturecache.h"
 #include "lltexturefetch.h"
 #include "llimageworker.h"
@@ -755,14 +754,6 @@ bool LLSpatialGroup::changeLOD()
 
         if (fabsf(ratio) >= getSpatialPartition()->mSlopRatio)
         {
-            LL_DEBUGS("RiggedBox") << "changeLOD true because of ratio compare "
-                                   << fabsf(ratio) << " " << getSpatialPartition()->mSlopRatio << LL_ENDL;
-            LL_DEBUGS("RiggedBox") << "sg " << this << "\nmDistance " << mDistance
-                                   << " mLastUpdateDistance " << mLastUpdateDistance
-                                   << " mRadius " << mRadius
-                                   << " fab ratio " << fabsf(ratio)
-                                   << " slop " << getSpatialPartition()->mSlopRatio << LL_ENDL;
-
             return true;
         }
     }
@@ -1676,8 +1667,9 @@ void renderOctree(LLSpatialGroup* group)
             glLineWidth(1.f);
             gGL.flush();
 
-            LLVOAvatar* lastAvatar = nullptr;
+            const LLVOAvatar* lastAvatar = nullptr;
             U64 lastMeshId = 0;
+            bool skipLastSkin = false;
 
             for (LLSpatialGroup::element_iter i = group->getDataBegin(); i != group->getDataEnd(); ++i)
             {
@@ -1706,15 +1698,9 @@ void renderOctree(LLSpatialGroup* group)
                 {
                     gGL.pushMatrix();
                     gGL.loadMatrix(gGLModelView);
-                    if (lastAvatar != face->mAvatar ||
-                        lastMeshId != face->mSkinInfo->mHash)
+                    if (!LLRenderPass::uploadMatrixPalette(face->mAvatar, face->mSkinInfo, lastAvatar, lastMeshId, skipLastSkin))
                     {
-                        if (!LLRenderPass::uploadMatrixPalette(face->mAvatar, face->mSkinInfo))
-                        {
-                            continue;
-                        }
-                        lastAvatar = face->mAvatar;
-                        lastMeshId = face->mSkinInfo->mHash;
+                        continue;
                     }
                 }
                 for (S32 j = 0; j < drawable->getNumFaces(); j++)
diff --git a/indra/newview/llstartup.cpp b/indra/newview/llstartup.cpp
index 7cefa21125..b32b80331a 100644
--- a/indra/newview/llstartup.cpp
+++ b/indra/newview/llstartup.cpp
@@ -28,7 +28,6 @@
 
 #include "llappviewer.h"
 #include "llstartup.h"
-#include "llcallstack.h"
 
 #if LL_WINDOWS
 #   include <process.h>     // _spawnl()
diff --git a/indra/newview/llviewerdisplay.cpp b/indra/newview/llviewerdisplay.cpp
index 8c6a38876a..9bd0973cc0 100644
--- a/indra/newview/llviewerdisplay.cpp
+++ b/indra/newview/llviewerdisplay.cpp
@@ -1502,6 +1502,7 @@ void draw_axes()
 
 void render_ui_3d()
 {
+    LL_PROFILE_ZONE_SCOPED_CATEGORY_UI;
     LLGLSPipeline gls_pipeline;
 
     //////////////////////////////////////
@@ -1550,6 +1551,7 @@ void render_ui_3d()
 
 void render_ui_2d()
 {
+    LL_PROFILE_ZONE_SCOPED_CATEGORY_UI;
     LLGLSUIDefault gls_ui;
 
     /////////////////////////////////////////////////////////////
diff --git a/indra/newview/llviewerobject.cpp b/indra/newview/llviewerobject.cpp
index 1131ec7239..bc6e13ff8e 100644
--- a/indra/newview/llviewerobject.cpp
+++ b/indra/newview/llviewerobject.cpp
@@ -103,7 +103,6 @@
 #include "llfloaterperms.h"
 #include "llvocache.h"
 #include "llcleanup.h"
-#include "llcallstack.h"
 #include "llmeshrepository.h"
 #include "llgltfmateriallist.h"
 #include "llgl.h"
@@ -151,7 +150,6 @@ LLViewerObject *LLViewerObject::createObject(const LLUUID &id, const LLPCode pco
 {
     LL_PROFILE_ZONE_SCOPED;
     LL_DEBUGS("ObjectUpdate") << "creating " << id << LL_ENDL;
-    dumpStack("ObjectUpdateStack");
 
     LLViewerObject *res = NULL;
 
@@ -1165,7 +1163,6 @@ U32 LLViewerObject::processUpdateMessage(LLMessageSystem *mesgsys,
     LL_DEBUGS_ONCE("SceneLoadTiming") << "Received viewer object data" << LL_ENDL;
 
     LL_DEBUGS("ObjectUpdate") << " mesgsys " << mesgsys << " dp " << dp << " id " << getID() << " update_type " << (S32) update_type << LL_ENDL;
-    dumpStack("ObjectUpdateStack");
 
     // The new OBJECTDATA_FIELD_SIZE_124, OBJECTDATA_FIELD_SIZE_140, OBJECTDATA_FIELD_SIZE_80
     // and OBJECTDATA_FIELD_SIZE_64 lengths should be supported in the existing cases below.
@@ -6161,6 +6158,7 @@ bool LLViewerObject::isParticleSource() const
 
 void LLViewerObject::setParticleSource(const LLPartSysData& particle_parameters, const LLUUID& owner_id)
 {
+    LL_PROFILE_ZONE_SCOPED_CATEGORY_VIEWER;
     if (mPartSourcep)
     {
         deleteParticleSource();
@@ -6192,6 +6190,7 @@ void LLViewerObject::setParticleSource(const LLPartSysData& particle_parameters,
 
 void LLViewerObject::unpackParticleSource(const S32 block_num, const LLUUID& owner_id)
 {
+    LL_PROFILE_ZONE_SCOPED_CATEGORY_VIEWER;
     if (!mPartSourcep.isNull() && mPartSourcep->isDead())
     {
         mPartSourcep = NULL;
@@ -6227,7 +6226,7 @@ void LLViewerObject::unpackParticleSource(const S32 block_num, const LLUUID& own
             LLViewerTexture* image;
             if (mPartSourcep->mPartSysData.mPartImageID == LLUUID::null)
             {
-                image = LLViewerTextureManager::getFetchedTextureFromFile("pixiesmall.j2c");
+                image = LLViewerFetchedTexture::sDefaultParticleImagep;
             }
             else
             {
@@ -6240,6 +6239,7 @@ void LLViewerObject::unpackParticleSource(const S32 block_num, const LLUUID& own
 
 void LLViewerObject::unpackParticleSource(LLDataPacker &dp, const LLUUID& owner_id, bool legacy)
 {
+    LL_PROFILE_ZONE_SCOPED_CATEGORY_VIEWER;
     if (!mPartSourcep.isNull() && mPartSourcep->isDead())
     {
         mPartSourcep = NULL;
@@ -6274,7 +6274,7 @@ void LLViewerObject::unpackParticleSource(LLDataPacker &dp, const LLUUID& owner_
             LLViewerTexture* image;
             if (mPartSourcep->mPartSysData.mPartImageID == LLUUID::null)
             {
-                image = LLViewerTextureManager::getFetchedTextureFromFile("pixiesmall.j2c");
+                image = LLViewerFetchedTexture::sDefaultParticleImagep;
             }
             else
             {
diff --git a/indra/newview/llviewerobjectlist.cpp b/indra/newview/llviewerobjectlist.cpp
index 435bd60917..9e1d86faac 100644
--- a/indra/newview/llviewerobjectlist.cpp
+++ b/indra/newview/llviewerobjectlist.cpp
@@ -68,7 +68,6 @@
 #include "u64.h"
 #include "llviewertexturelist.h"
 #include "lldatapacker.h"
-#include "llcallstack.h"
 #ifdef LL_USESYSTEMLIBS
 #include <zlib.h>
 #else
@@ -245,7 +244,6 @@ void LLViewerObjectList::processUpdateCore(LLViewerObject* objectp,
     // ignore returned flags
     LL_DEBUGS("ObjectUpdate") << "uuid " << objectp->mID << " calling processUpdateMessage "
                               << objectp << " just_created " << just_created << " from_cache " << from_cache << " msg " << msg << LL_ENDL;
-    dumpStack("ObjectUpdateStack");
 
     objectp->processUpdateMessage(msg, user_data, i, update_type, dpp);
 
@@ -362,7 +360,6 @@ LLViewerObject* LLViewerObjectList::processObjectUpdateFromCache(LLVOCacheEntry*
         objectp = createObjectFromCache(pcode, regionp, fullid, entry->getLocalID());
 
         LL_DEBUGS("ObjectUpdate") << "uuid " << fullid << " created objectp " << objectp << LL_ENDL;
-        dumpStack("ObjectUpdateStack");
 
         if (!objectp)
         {
@@ -557,7 +554,6 @@ void LLViewerObjectList::processObjectUpdate(LLMessageSystem *mesgsys,
         LL_DEBUGS("ObjectUpdate") << "uuid " << fullid << " objectp " << objectp
                                      << " update_cache " << (S32) update_cache << " compressed " << compressed
                                      << " update_type "  << update_type << LL_ENDL;
-        dumpStack("ObjectUpdateStack");
 
         if(update_cache)
         {
@@ -635,7 +631,6 @@ void LLViewerObjectList::processObjectUpdate(LLMessageSystem *mesgsys,
             objectp = createObject(pcode, regionp, fullid, local_id, gMessageSystem->getSender());
 
             LL_DEBUGS("ObjectUpdate") << "creating object " << fullid << " result " << objectp << LL_ENDL;
-            dumpStack("ObjectUpdateStack");
 
             if (!objectp)
             {
@@ -729,7 +724,6 @@ void LLViewerObjectList::processCachedObjectUpdate(LLMessageSystem *mesgsys,
         mesgsys->getU32Fast(_PREHASH_ObjectData, _PREHASH_UpdateFlags, flags, i);
 
         LL_DEBUGS("ObjectUpdate") << "got probe for id " << id << " crc " << crc << LL_ENDL;
-        dumpStack("ObjectUpdateStack");
 
         // Lookup data packer and add this id to cache miss lists if necessary.
         U8 cache_miss_type = LLViewerRegion::CACHE_MISS_TYPE_NONE;
@@ -1305,7 +1299,6 @@ void LLViewerObjectList::cleanupReferences(LLViewerObject *objectp)
     // Remove from object map so noone can look it up.
 
     LL_DEBUGS("ObjectUpdate") << " dereferencing id " << objectp->mID << LL_ENDL;
-    dumpStack("ObjectUpdateStack");
 
     mUUIDObjectMap.erase(objectp->mID);
 
@@ -1857,7 +1850,6 @@ LLViewerObject *LLViewerObjectList::createObjectFromCache(const LLPCode pcode, L
     llassert_always(uuid.notNull());
 
     LL_DEBUGS("ObjectUpdate") << "creating " << uuid << " local_id " << local_id << LL_ENDL;
-    dumpStack("ObjectUpdateStack");
 
     LLViewerObject *objectp = LLViewerObject::createObject(uuid, pcode, regionp);
     if (!objectp)
@@ -1893,7 +1885,6 @@ LLViewerObject *LLViewerObjectList::createObject(const LLPCode pcode, LLViewerRe
     }
 
     LL_DEBUGS("ObjectUpdate") << "createObject creating " << fullid << LL_ENDL;
-    dumpStack("ObjectUpdateStack");
 
     LLViewerObject *objectp = LLViewerObject::createObject(fullid, pcode, regionp);
     if (!objectp)
diff --git a/indra/newview/llviewerpartsource.cpp b/indra/newview/llviewerpartsource.cpp
index dd6a404836..54e0470604 100644
--- a/indra/newview/llviewerpartsource.cpp
+++ b/indra/newview/llviewerpartsource.cpp
@@ -114,7 +114,7 @@ LLViewerPartSourceScript::LLViewerPartSourceScript(LLViewerObject *source_objp)
     llassert(source_objp);
     mSourceObjectp = source_objp;
     mPosAgent = mSourceObjectp->getPositionAgent();
-    mImagep = LLViewerTextureManager::getFetchedTextureFromFile("pixiesmall.j2c");
+    mImagep = LLViewerFetchedTexture::sDefaultParticleImagep;
 
     mImagep->setAddressMode(LLTexUnit::TAM_CLAMP);
 }
@@ -619,7 +619,7 @@ void LLViewerPartSourceSpiral::update(const F32 dt)
 {
     if (!mImagep)
     {
-        mImagep = LLViewerTextureManager::getFetchedTextureFromFile("pixiesmall.j2c");
+        mImagep = LLViewerFetchedTexture::sDefaultParticleImagep;
     }
 
     const F32 RATE = 0.025f;
@@ -789,7 +789,7 @@ void LLViewerPartSourceBeam::update(const F32 dt)
 
         if (!mImagep)
         {
-            mImagep = LLViewerTextureManager::getFetchedTextureFromFile("pixiesmall.j2c");
+            mImagep = LLViewerFetchedTexture::sDefaultParticleImagep;
         }
 
         LLViewerPart* part = new LLViewerPart();
@@ -876,7 +876,7 @@ void LLViewerPartSourceChat::update(const F32 dt)
 {
     if (!mImagep)
     {
-        mImagep = LLViewerTextureManager::getFetchedTextureFromFile("pixiesmall.j2c");
+        mImagep = LLViewerFetchedTexture::sDefaultParticleImagep;
     }
 
 
diff --git a/indra/newview/llviewerregion.cpp b/indra/newview/llviewerregion.cpp
index 57ed0da06b..72e066eb74 100755
--- a/indra/newview/llviewerregion.cpp
+++ b/indra/newview/llviewerregion.cpp
@@ -81,7 +81,6 @@
 #include "llcoros.h"
 #include "lleventcoro.h"
 #include "llcorehttputil.h"
-#include "llcallstack.h"
 #include "llsettingsdaycycle.h"
 
 #include <boost/regex.hpp>
@@ -2716,7 +2715,6 @@ LLViewerRegion::eCacheUpdateResult LLViewerRegion::cacheFullUpdate(LLDataPackerB
         if (entry->getCRC() == crc)
         {
             LL_DEBUGS("AnimatedObjects") << " got dupe for local_id " << local_id << LL_ENDL;
-            dumpStack("AnimatedObjectsStack");
 
             // Record a hit
             entry->recordDupe();
@@ -2725,7 +2723,6 @@ LLViewerRegion::eCacheUpdateResult LLViewerRegion::cacheFullUpdate(LLDataPackerB
         else //CRC changed
         {
             LL_DEBUGS("AnimatedObjects") << " got update for local_id " << local_id << LL_ENDL;
-            dumpStack("AnimatedObjectsStack");
 
             // Update the cache entry
             entry->updateEntry(crc, dp);
@@ -2738,7 +2735,6 @@ LLViewerRegion::eCacheUpdateResult LLViewerRegion::cacheFullUpdate(LLDataPackerB
     else
     {
         LL_DEBUGS("AnimatedObjects") << " got first notification for local_id " << local_id << LL_ENDL;
-        dumpStack("AnimatedObjectsStack");
 
         // we haven't seen this object before
         // Create new entry and add to map
diff --git a/indra/newview/llviewershadermgr.cpp b/indra/newview/llviewershadermgr.cpp
index 073a1787d5..d1dea16bc0 100644
--- a/indra/newview/llviewershadermgr.cpp
+++ b/indra/newview/llviewershadermgr.cpp
@@ -158,6 +158,7 @@ LLGLSLShader            gDeferredMultiLightProgram[16];
 LLGLSLShader            gDeferredSpotLightProgram;
 LLGLSLShader            gDeferredMultiSpotLightProgram;
 LLGLSLShader            gDeferredSunProgram;
+LLGLSLShader            gDeferredSunProbeProgram;
 LLGLSLShader            gHazeProgram;
 LLGLSLShader            gHazeWaterProgram;
 LLGLSLShader            gDeferredBlurLightProgram;
@@ -200,6 +201,7 @@ LLGLSLShader            gExposureProgram;
 LLGLSLShader            gExposureProgramNoFade;
 LLGLSLShader            gLuminanceProgram;
 LLGLSLShader            gFXAAProgram;
+LLGLSLShader            gCASProgram;
 LLGLSLShader            gDeferredPostNoDoFProgram;
 LLGLSLShader            gDeferredWLSkyProgram;
 LLGLSLShader            gEnvironmentMapProgram;
@@ -400,6 +402,7 @@ void LLViewerShaderMgr::finalizeShaderList()
     mShaderList.push_back(&gObjectAlphaMaskNoColorProgram);
     mShaderList.push_back(&gUnderWaterProgram);
     mShaderList.push_back(&gDeferredSunProgram);
+    mShaderList.push_back(&gDeferredSunProbeProgram);
     mShaderList.push_back(&gHazeProgram);
     mShaderList.push_back(&gHazeWaterProgram);
     mShaderList.push_back(&gDeferredSoftenProgram);
@@ -1600,10 +1603,7 @@ bool LLViewerShaderMgr::loadShadersDeferred()
     if (success)
     {
         std::string fragment;
-        std::string vertex = "deferred/sunLightV.glsl";
-
         bool use_ao = gSavedSettings.getBOOL("RenderDeferredSSAO");
-
         if (use_ao)
         {
             fragment = "deferred/sunLightSSAOF.glsl";
@@ -1611,10 +1611,6 @@ bool LLViewerShaderMgr::loadShadersDeferred()
         else
         {
             fragment = "deferred/sunLightF.glsl";
-            if (mShaderLevel[SHADER_DEFERRED] == 1)
-            { //no shadows, no SSAO, no frag coord
-                vertex = "deferred/sunLightNoFragCoordV.glsl";
-            }
         }
 
         gDeferredSunProgram.mName = "Deferred Sun Shader";
@@ -1623,7 +1619,7 @@ bool LLViewerShaderMgr::loadShadersDeferred()
         gDeferredSunProgram.mFeatures.hasAmbientOcclusion = use_ao;
 
         gDeferredSunProgram.mShaderFiles.clear();
-        gDeferredSunProgram.mShaderFiles.push_back(make_pair(vertex, GL_VERTEX_SHADER));
+        gDeferredSunProgram.mShaderFiles.push_back(make_pair("deferred/sunLightV.glsl", GL_VERTEX_SHADER));
         gDeferredSunProgram.mShaderFiles.push_back(make_pair(fragment, GL_FRAGMENT_SHADER));
         gDeferredSunProgram.mShaderLevel = mShaderLevel[SHADER_DEFERRED];
 
@@ -1633,6 +1629,21 @@ bool LLViewerShaderMgr::loadShadersDeferred()
 
     if (success)
     {
+        gDeferredSunProbeProgram.mName = "Deferred Sun Probe Shader";
+        gDeferredSunProbeProgram.mFeatures.isDeferred = true;
+        gDeferredSunProbeProgram.mFeatures.hasShadows = true;
+
+        gDeferredSunProbeProgram.mShaderFiles.clear();
+        gDeferredSunProbeProgram.mShaderFiles.push_back(make_pair("deferred/sunLightV.glsl", GL_VERTEX_SHADER));
+        gDeferredSunProbeProgram.mShaderFiles.push_back(make_pair("deferred/sunLightF.glsl", GL_FRAGMENT_SHADER));
+        gDeferredSunProbeProgram.mShaderLevel = mShaderLevel[SHADER_DEFERRED];
+
+        success = gDeferredSunProbeProgram.createShader();
+        llassert(success);
+    }
+
+    if (success)
+    {
         gDeferredBlurLightProgram.mName = "Deferred Blur Light Shader";
         gDeferredBlurLightProgram.mFeatures.isDeferred = true;
 
@@ -2334,6 +2345,16 @@ bool LLViewerShaderMgr::loadShadersDeferred()
         gFXAAProgram.mShaderFiles.clear();
         gFXAAProgram.mShaderFiles.push_back(make_pair("deferred/postDeferredV.glsl", GL_VERTEX_SHADER));
         gFXAAProgram.mShaderFiles.push_back(make_pair("deferred/fxaaF.glsl", GL_FRAGMENT_SHADER));
+
+        if (gGLManager.mGLVersion > 3.9)
+        {
+            gFXAAProgram.addPermutation("FXAA_GLSL_400", "1");
+        }
+        else
+        {
+            gFXAAProgram.addPermutation("FXAA_GLSL_130", "1");
+        }
+
         gFXAAProgram.mShaderLevel = mShaderLevel[SHADER_DEFERRED];
         success = gFXAAProgram.createShader();
         llassert(success);
@@ -2341,6 +2362,17 @@ bool LLViewerShaderMgr::loadShadersDeferred()
 
     if (success)
     {
+        gCASProgram.mName = "Contrast Adaptive Sharpening Shader";
+        gCASProgram.mFeatures.hasSrgb = true;
+        gCASProgram.mShaderFiles.clear();
+        gCASProgram.mShaderFiles.push_back(make_pair("deferred/postDeferredNoTCV.glsl", GL_VERTEX_SHADER));
+        gCASProgram.mShaderFiles.push_back(make_pair("deferred/CASF.glsl", GL_FRAGMENT_SHADER));
+        gCASProgram.mShaderLevel = mShaderLevel[SHADER_DEFERRED];
+        gCASProgram.createShader();
+    }
+
+    if (success)
+    {
         gDeferredPostProgram.mName = "Deferred Post Shader";
         gDeferredPostProgram.mFeatures.isDeferred = true;
         gDeferredPostProgram.mShaderFiles.clear();
diff --git a/indra/newview/llviewershadermgr.h b/indra/newview/llviewershadermgr.h
index af47014a43..b858530c3e 100644
--- a/indra/newview/llviewershadermgr.h
+++ b/indra/newview/llviewershadermgr.h
@@ -230,6 +230,7 @@ extern LLGLSLShader         gDeferredMultiLightProgram[LL_DEFERRED_MULTI_LIGHT_C
 extern LLGLSLShader         gDeferredSpotLightProgram;
 extern LLGLSLShader         gDeferredMultiSpotLightProgram;
 extern LLGLSLShader         gDeferredSunProgram;
+extern LLGLSLShader         gDeferredSunProbeProgram;
 extern LLGLSLShader         gHazeProgram;
 extern LLGLSLShader         gHazeWaterProgram;
 extern LLGLSLShader         gDeferredBlurLightProgram;
@@ -245,6 +246,7 @@ extern LLGLSLShader         gDeferredPostProgram;
 extern LLGLSLShader         gDeferredCoFProgram;
 extern LLGLSLShader         gDeferredDoFCombineProgram;
 extern LLGLSLShader         gFXAAProgram;
+extern LLGLSLShader         gCASProgram;
 extern LLGLSLShader         gDeferredPostNoDoFProgram;
 extern LLGLSLShader         gDeferredPostGammaCorrectProgram;
 extern LLGLSLShader         gNoPostGammaCorrectProgram;
diff --git a/indra/newview/llviewerstats.cpp b/indra/newview/llviewerstats.cpp
index b351f3e12f..0cc2eb8b7c 100644
--- a/indra/newview/llviewerstats.cpp
+++ b/indra/newview/llviewerstats.cpp
@@ -137,7 +137,6 @@ LLTrace::CountStatHandle<>  FPS("FPS", "Frames rendered"),
                             UPLOAD_TEXTURE("uploadtexture", "Textures uploaded"),
                             EDIT_TEXTURE("edittexture", "Changes to textures on objects"),
                             KILLED("killed", "Number of times killed"),
-                            FRAMETIME_DOUBLED("frametimedoubled", "Ratio of frames 2x longer than previous"),
                             TEX_BAKES("texbakes", "Number of times avatar textures have been baked"),
                             TEX_REBAKES("texrebakes", "Number of times avatar textures have been forced to rebake"),
                             NUM_NEW_OBJECTS("numnewobjectsstat", "Number of objects in scene that were not previously in cache");
@@ -157,11 +156,6 @@ LLTrace::CountStatHandle<F64Kilobytes >
                             MESSAGE_SYSTEM_DATA_IN("messagedatain", "Incoming message system network data"),
                             MESSAGE_SYSTEM_DATA_OUT("messagedataout", "Outgoing message system network data");
 
-LLTrace::CountStatHandle<F64Seconds >
-                            SIM_20_FPS_TIME("sim20fpstime", "Seconds with sim FPS below 20"),
-                            SIM_PHYSICS_20_FPS_TIME("simphysics20fpstime", "Seconds with physics FPS below 20"),
-                            LOSS_5_PERCENT_TIME("loss5percenttime", "Seconds with packet loss > 5%");
-
 SimMeasurement<>            SIM_TIME_DILATION("simtimedilation", "Simulator time scale", LL_SIM_STAT_TIME_DILATION),
                             SIM_FPS("simfps", "Simulator framerate", LL_SIM_STAT_FPS),
                             SIM_PHYSICS_FPS("simphysicsfps", "Simulator physics framerate", LL_SIM_STAT_PHYSFPS),
@@ -205,9 +199,6 @@ static LLTrace::SampleStatHandle<bool>
                             CHAT_BUBBLES("chatbubbles", "Chat Bubbles Enabled");
 
 LLTrace::SampleStatHandle<F64Megabytes > FORMATTED_MEM("formattedmemstat");
-LLTrace::SampleStatHandle<F64Kilobytes >    DELTA_BANDWIDTH("deltabandwidth", "Increase/Decrease in bandwidth based on packet loss"),
-                                                            MAX_BANDWIDTH("maxbandwidth", "Max bandwidth setting");
-
 
 SimMeasurement<F64Milliseconds >    SIM_FRAME_TIME("simframemsec", "", LL_SIM_STAT_FRAMEMS),
                                                     SIM_NET_TIME("simnetmsec", "", LL_SIM_STAT_NETMS),
@@ -228,7 +219,6 @@ SimMeasurement<F64Kilobytes >   SIM_UNACKED_BYTES("simtotalunackedbytes", "", LL
 SimMeasurement<F64Megabytes >   SIM_PHYSICS_MEM("physicsmemoryallocated", "", LL_SIM_STAT_SIMPHYSICSMEMORY);
 
 LLTrace::SampleStatHandle<F64Milliseconds > FRAMETIME_JITTER("frametimejitter", "Average delta between successive frame times"),
-                                            FRAMETIME_SLEW("frametimeslew", "Average delta between frame time and mean"),
                                             FRAMETIME("frametime", "Measured frame time"),
                                             SIM_PING("simpingstat");
 
@@ -246,10 +236,7 @@ LLTrace::EventStatHandle<F64Milliseconds >  REGION_CROSSING_TIME("regioncrossing
 
 LLTrace::EventStatHandle<F64Seconds >   AVATAR_EDIT_TIME("avataredittime", "Seconds in Edit Appearance"),
                                                             TOOLBOX_TIME("toolboxtime", "Seconds using Toolbox"),
-                                                            MOUSELOOK_TIME("mouselooktime", "Seconds in Mouselook"),
-                                                            FPS_10_TIME("fps10time", "Seconds below 10 FPS"),
-                                                            FPS_8_TIME("fps8time", "Seconds below 8 FPS"),
-                                                            FPS_2_TIME("fps2time", "Seconds below 2 FPS");
+                                                            MOUSELOOK_TIME("mouselooktime", "Seconds in Mouselook");
 
 LLTrace::EventStatHandle<LLUnit<F32, LLUnits::Percent> > OBJECT_CACHE_HIT_RATE("object_cache_hits");
 
@@ -279,55 +266,12 @@ void LLViewerStats::resetStats()
 
 void LLViewerStats::updateFrameStats(const F64Seconds time_diff)
 {
-    if (getRecording().getLastValue(LLStatViewer::PACKETS_LOST_PERCENT) > F32Percent(5.0))
-    {
-        add(LLStatViewer::LOSS_5_PERCENT_TIME, time_diff);
-    }
-
-    F32 sim_fps = (F32)getRecording().getLastValue(LLStatViewer::SIM_FPS);
-    if (0.f < sim_fps && sim_fps < 20.f)
-    {
-        add(LLStatViewer::SIM_20_FPS_TIME, time_diff);
-    }
-
-    F32 sim_physics_fps = (F32)getRecording().getLastValue(LLStatViewer::SIM_PHYSICS_FPS);
-
-    if (0.f < sim_physics_fps && sim_physics_fps < 20.f)
-    {
-        add(LLStatViewer::SIM_PHYSICS_20_FPS_TIME, time_diff);
-    }
-
-    if (time_diff >= (F64Seconds)0.5)
-    {
-        record(LLStatViewer::FPS_2_TIME, time_diff);
-    }
-    if (time_diff >= (F64Seconds)0.125)
-    {
-        record(LLStatViewer::FPS_8_TIME, time_diff);
-    }
-    if (time_diff >= (F64Seconds)0.1)
-    {
-        record(LLStatViewer::FPS_10_TIME, time_diff);
-    }
-
     if (gFrameCount && mLastTimeDiff > (F64Seconds)0.0)
     {
-        // new "stutter" meter
-        add(LLStatViewer::FRAMETIME_DOUBLED, time_diff >= 2.0 * mLastTimeDiff ? 1 : 0);
-
         sample(LLStatViewer::FRAMETIME, time_diff);
-
         // old stats that were never really used
-        F64Seconds jit = (F64Seconds) std::fabs((mLastTimeDiff - time_diff));
+        F64Seconds jit = (F64Seconds)std::fabs((mLastTimeDiff - time_diff));
         sample(LLStatViewer::FRAMETIME_JITTER, jit);
-
-        F32Seconds average_frametime = gRenderStartTime.getElapsedTimeF32() / (F32)gFrameCount;
-        sample(LLStatViewer::FRAMETIME_SLEW, F64Milliseconds (average_frametime - time_diff));
-
-        F32 max_bandwidth = gViewerThrottle.getMaxBandwidth();
-        F32 delta_bandwidth = gViewerThrottle.getCurrentBandwidth() - max_bandwidth;
-        sample(LLStatViewer::DELTA_BANDWIDTH, F64Bits(delta_bandwidth));
-        sample(LLStatViewer::MAX_BANDWIDTH, F64Bits(max_bandwidth));
     }
 
     mLastTimeDiff = time_diff;
diff --git a/indra/newview/llviewerstats.h b/indra/newview/llviewerstats.h
index a857af53cc..8aed1c537e 100644
--- a/indra/newview/llviewerstats.h
+++ b/indra/newview/llviewerstats.h
@@ -135,7 +135,6 @@ extern LLTrace::CountStatHandle<>           FPS,
                                             UPLOAD_TEXTURE,
                                             EDIT_TEXTURE,
                                             KILLED,
-                                            FRAMETIME_DOUBLED,
                                             TEX_BAKES,
                                             TEX_REBAKES,
                                             NUM_NEW_OBJECTS;
@@ -150,10 +149,6 @@ extern LLTrace::CountStatHandle<F64Kilobytes >  ACTIVE_MESSAGE_DATA_RECEIVED,
                                                                     MESSAGE_SYSTEM_DATA_IN,
                                                                     MESSAGE_SYSTEM_DATA_OUT;
 
-extern LLTrace::CountStatHandle<F64Seconds >        SIM_20_FPS_TIME,
-                                                                    SIM_PHYSICS_20_FPS_TIME,
-                                                                    LOSS_5_PERCENT_TIME;
-
 extern SimMeasurement<>                     SIM_TIME_DILATION,
                                             SIM_FPS,
                                             SIM_PHYSICS_FPS,
@@ -194,8 +189,6 @@ extern LLTrace::SampleStatHandle<LLUnit<F32, LLUnits::Percent> > PACKETS_LOST_PE
 
 extern LLTrace::SampleStatHandle<F64Megabytes > FORMATTED_MEM;
 
-extern LLTrace::SampleStatHandle<F64Kilobytes > DELTA_BANDWIDTH,
-                                                                    MAX_BANDWIDTH;
 extern SimMeasurement<F64Milliseconds > SIM_FRAME_TIME,
                                                             SIM_NET_TIME,
                                                             SIM_OTHER_TIME,
@@ -216,7 +209,6 @@ extern SimMeasurement<F64Megabytes >    SIM_PHYSICS_MEM;
 
 
 extern LLTrace::SampleStatHandle<F64Milliseconds >  FRAMETIME_JITTER,
-                                                    FRAMETIME_SLEW,
                                                     SIM_PING;
 
 extern LLTrace::EventStatHandle<LLUnit<F64, LLUnits::Meters> > AGENT_POSITION_SNAP;
@@ -233,10 +225,7 @@ extern LLTrace::EventStatHandle<F64Milliseconds >   REGION_CROSSING_TIME,
 
 extern LLTrace::EventStatHandle<F64Seconds >    AVATAR_EDIT_TIME,
                                                                 TOOLBOX_TIME,
-                                                                MOUSELOOK_TIME,
-                                                                FPS_10_TIME,
-                                                                FPS_8_TIME,
-                                                                FPS_2_TIME;
+                                                                MOUSELOOK_TIME;
 
 extern LLTrace::EventStatHandle<LLUnit<F32, LLUnits::Percent> > OBJECT_CACHE_HIT_RATE;
 
diff --git a/indra/newview/llviewertexture.cpp b/indra/newview/llviewertexture.cpp
index 4922df6738..bb16fec32a 100644
--- a/indra/newview/llviewertexture.cpp
+++ b/indra/newview/llviewertexture.cpp
@@ -70,6 +70,7 @@ LLPointer<LLViewerTexture>        LLViewerTexture::sBlackImagep = nullptr;
 LLPointer<LLViewerTexture>        LLViewerTexture::sCheckerBoardImagep = nullptr;
 LLPointer<LLViewerFetchedTexture> LLViewerFetchedTexture::sMissingAssetImagep = nullptr;
 LLPointer<LLViewerFetchedTexture> LLViewerFetchedTexture::sWhiteImagep = nullptr;
+LLPointer<LLViewerFetchedTexture> LLViewerFetchedTexture::sDefaultParticleImagep = nullptr;
 LLPointer<LLViewerFetchedTexture> LLViewerFetchedTexture::sDefaultImagep = nullptr;
 LLPointer<LLViewerFetchedTexture> LLViewerFetchedTexture::sSmokeImagep = nullptr;
 LLPointer<LLViewerFetchedTexture> LLViewerFetchedTexture::sFlatNormalImagep = nullptr;
@@ -547,7 +548,37 @@ void LLViewerTexture::updateClass()
 
     was_low = is_low;
 
-    sDesiredDiscardBias = llclamp(sDesiredDiscardBias, 1.f, 3.f);
+
+    // set to max discard bias if the window has been backgrounded for a while
+    static bool was_backgrounded = false;
+    static LLFrameTimer backgrounded_timer;
+
+    bool in_background = (gViewerWindow && !gViewerWindow->getWindow()->getVisible()) || !gFocusMgr.getAppHasFocus();
+
+    if (in_background)
+    {
+        if (backgrounded_timer.getElapsedTimeF32() > 10.f)
+        {
+            if (!was_backgrounded)
+            {
+                LL_INFOS() << "Viewer is backgrounded, freeing up video memory." << LL_ENDL;
+            }
+            was_backgrounded = true;
+            sDesiredDiscardBias = 4.f;
+        }
+    }
+    else
+    {
+        backgrounded_timer.reset();
+        if (was_backgrounded)
+        { // if the viewer was backgrounded
+            LL_INFOS() << "Viewer is no longer backgrounded, resuming normal texture usage." << LL_ENDL;
+            was_backgrounded = false;
+            sDesiredDiscardBias = 1.f;
+        }
+    }
+
+    sDesiredDiscardBias = llclamp(sDesiredDiscardBias, 1.f, 4.f);
 
     LLViewerTexture::sFreezeImageUpdates = false;
 }
diff --git a/indra/newview/llviewertexture.h b/indra/newview/llviewertexture.h
index 9046daeab2..65fa633f81 100644
--- a/indra/newview/llviewertexture.h
+++ b/indra/newview/llviewertexture.h
@@ -146,6 +146,8 @@ public:
 
     virtual F32  getMaxVirtualSize() ;
 
+    LLFrameTimer* getLastReferencedTimer() { return &mLastReferencedTimer; }
+
     S32 getFullWidth() const { return mFullWidth; }
     S32 getFullHeight() const { return mFullHeight; }
     /*virtual*/ void setKnownDrawSize(S32 width, S32 height);
@@ -195,6 +197,7 @@ protected:
     mutable F32 mMaxVirtualSize = 0.f;  // The largest virtual size of the image, in pixels - how much data to we need?
     mutable S32  mMaxVirtualSizeResetCounter;
     mutable S32  mMaxVirtualSizeResetInterval;
+    LLFrameTimer mLastReferencedTimer;
 
     ll_face_list_t    mFaceList[LLRender::NUM_TEXTURE_CHANNELS]; //reverse pointer pointing to the faces using this image as texture
     U32               mNumFaces[LLRender::NUM_TEXTURE_CHANNELS];
@@ -504,6 +507,7 @@ public:
     static LLPointer<LLViewerFetchedTexture> sDefaultImagep; // "Default" texture for error cases, the only case of fetched texture which is generated in local.
     static LLPointer<LLViewerFetchedTexture> sFlatNormalImagep; // Flat normal map denoting no bumpiness on a surface
     static LLPointer<LLViewerFetchedTexture> sDefaultIrradiancePBRp; // PBR: irradiance
+    static LLPointer<LLViewerFetchedTexture> sDefaultParticleImagep; // Default particle texture
 
     // not sure why, but something is iffy about the loading of this particular texture, use the accessor instead of accessing directly
     static LLPointer<LLViewerFetchedTexture> sSmokeImagep; // Old "Default" translucent texture
@@ -571,10 +575,7 @@ public:
     /*virtual*/ void addFace(U32 ch, LLFace* facep) ;
     /*virtual*/ void removeFace(U32 ch, LLFace* facep) ;
 
-    /*virtual*/ F32  getMaxVirtualSize() ;
-
-    // get the timer that tracks the last time reinit was called
-    LLFrameTimer* getLastReferencedTimer() { return &mLastReferencedTimer; }
+    /*virtual*/ F32  getMaxVirtualSize();
 
 private:
     void switchTexture(U32 ch, LLFace* facep) ;
@@ -596,9 +597,6 @@ private:
     bool mIsPlaying ;
     U32  mUpdateVirtualSizeTime ;
 
-    // tracks last time reinit was called
-    LLFrameTimer mLastReferencedTimer;
-
 public:
     static void updateClass() ;
     static void cleanUpClass() ;
diff --git a/indra/newview/llviewertexturelist.cpp b/indra/newview/llviewertexturelist.cpp
index 1ac97a20cf..2df60dbf6c 100644
--- a/indra/newview/llviewertexturelist.cpp
+++ b/indra/newview/llviewertexturelist.cpp
@@ -123,6 +123,9 @@ void LLViewerTextureList::doPreloadImages()
     LLTexUnit::sWhiteTexture = LLViewerFetchedTexture::sWhiteImagep->getTexName();
     LLUIImageList* image_list = LLUIImageList::getInstance();
 
+    // Set default particle texture
+    LLViewerFetchedTexture::sDefaultParticleImagep = LLViewerTextureManager::getFetchedTextureFromFile("pixiesmall.j2c");
+
     // Set the default flat normal map
     // BLANK_OBJECT_NORMAL has a version on dataserver, but it has compression artifacts
     LLViewerFetchedTexture::sFlatNormalImagep =
@@ -403,6 +406,7 @@ LLViewerFetchedTexture* LLViewerTextureList::getImageFromFile(const std::string&
                                                    const LLUUID& force_id)
 {
     LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE;
+    LL_PROFILE_ZONE_TEXT(filename.c_str(), filename.size());
     if(!mInitialized)
     {
         return NULL ;
@@ -885,66 +889,101 @@ void LLViewerTextureList::updateImageDecodePriority(LLViewerFetchedTexture* imag
 {
     llassert(!gCubeSnapshot);
 
-    static LLCachedControl<F32> bias_distance_scale(gSavedSettings, "TextureBiasDistanceScale", 1.f);
-    static LLCachedControl<F32> texture_scale_min(gSavedSettings, "TextureScaleMinAreaFactor", 0.04f);
-    static LLCachedControl<F32> texture_scale_max(gSavedSettings, "TextureScaleMaxAreaFactor", 25.f);
+    if (imagep->getBoostLevel() < LLViewerFetchedTexture::BOOST_HIGH)  // don't bother checking face list for boosted textures
+    {
+        static LLCachedControl<F32> bias_distance_scale(gSavedSettings, "TextureBiasDistanceScale", 1.f);
+        static LLCachedControl<F32> texture_scale_min(gSavedSettings, "TextureScaleMinAreaFactor", 0.04f);
+        static LLCachedControl<F32> texture_scale_max(gSavedSettings, "TextureScaleMaxAreaFactor", 25.f);
 
-    if (imagep->getType() == LLViewerTexture::LOD_TEXTURE && imagep->getBoostLevel() == LLViewerTexture::BOOST_NONE)
-    { // reset max virtual size for unboosted LOD_TEXTURES
-        // this is an alternative to decaying mMaxVirtualSize over time
-        // that keeps textures from continously downrezzing and uprezzing in the background
-        imagep->mMaxVirtualSize = 0.f;
-    }
+        F32 max_vsize = 0.f;
+        bool on_screen = false;
 
-    LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE;
-    for (U32 i = 0; i < LLRender::NUM_TEXTURE_CHANNELS; ++i)
-    {
-        for (S32 fi = 0; fi < imagep->getNumFaces(i); ++fi)
-        {
-            LLFace* face = (*(imagep->getFaceList(i)))[fi];
+        U32 face_count = 0;
 
-            if (face && face->getViewerObject())
+        F32 bias = (F32) llroundf(powf(4, LLViewerTexture::sDesiredDiscardBias - 1.f));
+
+        LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE;
+        for (U32 i = 0; i < LLRender::NUM_TEXTURE_CHANNELS; ++i)
+        {
+            for (S32 fi = 0; fi < imagep->getNumFaces(i); ++fi)
             {
-                F32 radius;
-                F32 cos_angle_to_view_dir;
-                static LLCachedControl<F32> bias_unimportant_threshold(gSavedSettings, "TextureBiasUnimportantFactor", 0.25f);
-                F32 vsize = face->getPixelArea();
-                bool in_frustum = face->calcPixelArea(cos_angle_to_view_dir, radius);
-
-                // Scale desired texture resolution higher or lower depending on texture scale
-                //
-                // Minimum usage examples: a 1024x1024 texture with aplhabet, runing string
-                // shows one letter at a time
-                //
-                // Maximum usage examples: huge chunk of terrain repeats texture
-                S32 te_offset = face->getTEOffset();  // offset is -1 if not inited
-                LLViewerObject* objp = face->getViewerObject();
-                const LLTextureEntry* te = (te_offset < 0 || te_offset >= objp->getNumTEs()) ? nullptr : objp->getTE(te_offset);
-                F32 min_scale = te ? llmin(fabsf(te->getScaleS()), fabsf(te->getScaleT())) : 1.f;
-                min_scale = llclamp(min_scale * min_scale, texture_scale_min(), texture_scale_max());
-                vsize /= min_scale;
-
-                // if bias is > 2, apply to on-screen textures as well
-                bool apply_bias = LLViewerTexture::sDesiredDiscardBias > 2.f;
-
-                // apply bias to off screen objects or objects that are small on screen all the time
-                if (!in_frustum || !face->getDrawable()->isVisible() || face->getImportanceToCamera() < bias_unimportant_threshold)
-                { // further reduce by discard bias when off screen or occluded
-                    apply_bias = true;
-                }
+                LLFace* face = (*(imagep->getFaceList(i)))[fi];
 
-                if (apply_bias)
+                if (face && face->getViewerObject())
                 {
-                    F32 bias = powf(4, LLViewerTexture::sDesiredDiscardBias - 1.f);
-                    bias = (F32)llround(bias);
-                    vsize /= bias;
+                    ++face_count;
+                    F32 radius;
+                    F32 cos_angle_to_view_dir;
+                    static LLCachedControl<F32> bias_unimportant_threshold(gSavedSettings, "TextureBiasUnimportantFactor", 0.25f);
+
+                    if ((gFrameCount - face->mLastTextureUpdate) > 10)
+                    { // only call calcPixelArea at most once every 10 frames for a given face
+                        // this helps eliminate redundant calls to calcPixelArea for faces that have multiple textures
+                        // assigned to them, such as is the case with GLTF materials or Blinn-Phong materials
+                        face->mInFrustum = face->calcPixelArea(cos_angle_to_view_dir, radius);
+                        face->mLastTextureUpdate = gFrameCount;
+                    }
+
+                    F32 vsize = face->getPixelArea();
+
+                    on_screen = face->mInFrustum;
+
+                    // Scale desired texture resolution higher or lower depending on texture scale
+                    //
+                    // Minimum usage examples: a 1024x1024 texture with aplhabet, runing string
+                    // shows one letter at a time
+                    //
+                    // Maximum usage examples: huge chunk of terrain repeats texture
+                    // TODO: make this work with the GLTF texture transforms
+                    S32 te_offset = face->getTEOffset();  // offset is -1 if not inited
+                    LLViewerObject* objp = face->getViewerObject();
+                    const LLTextureEntry* te = (te_offset < 0 || te_offset >= objp->getNumTEs()) ? nullptr : objp->getTE(te_offset);
+                    F32 min_scale = te ? llmin(fabsf(te->getScaleS()), fabsf(te->getScaleT())) : 1.f;
+                    min_scale = llclamp(min_scale * min_scale, texture_scale_min(), texture_scale_max());
+                    vsize /= min_scale;
+
+                    // apply bias to offscreen faces all the time, but only to onscreen faces when bias is large
+                    if (!face->mInFrustum || LLViewerTexture::sDesiredDiscardBias > 2.f)
+                    {
+                        vsize /= bias;
+                    }
+
+                    max_vsize = llmax(max_vsize, vsize);
                 }
+            }
+        }
 
-                imagep->addTextureStats(vsize);
+        if (face_count > 1024)
+        { // this texture is used in so many places we should just boost it and not bother checking its vsize
+            // this is especially important because the above is not time sliced and can hit multiple ms for a single texture
+            imagep->setBoostLevel(LLViewerFetchedTexture::BOOST_HIGH);
+        }
+
+        if (imagep->getType() == LLViewerTexture::LOD_TEXTURE && imagep->getBoostLevel() == LLViewerTexture::BOOST_NONE)
+        { // conditionally reset max virtual size for unboosted LOD_TEXTURES
+          // this is an alternative to decaying mMaxVirtualSize over time
+          // that keeps textures from continously downrezzing and uprezzing in the background
+
+            if (LLViewerTexture::sDesiredDiscardBias > 2.f ||
+                (!on_screen && LLViewerTexture::sDesiredDiscardBias > 1.f))
+            {
+                imagep->mMaxVirtualSize = 0.f;
             }
         }
+
+        imagep->addTextureStats(max_vsize);
     }
 
+#if 0
+    imagep->setDebugText(llformat("%d/%d - %d/%d -- %d/%d",
+        (S32)sqrtf(max_vsize),
+        (S32)sqrtf(imagep->mMaxVirtualSize),
+        imagep->getDiscardLevel(),
+        imagep->getDesiredDiscardLevel(),
+        imagep->getWidth(),
+        imagep->getFullWidth()));
+#endif
+
     // make sure to addTextureStats for any spotlights that are using this texture
     for (S32 vi = 0; vi < imagep->getNumVolumes(LLRender::LIGHT_TEX); ++vi)
     {
@@ -955,18 +994,26 @@ void LLViewerTextureList::updateImageDecodePriority(LLViewerFetchedTexture* imag
     F32 max_inactive_time = 20.f; // inactive time before deleting saved raw image
     S32 min_refs = 3; // 1 for mImageList, 1 for mUUIDMap, and 1 for "entries" in updateImagesFetchTextures
 
+    F32 lazy_flush_timeout = 30.f; // delete unused images after 30 seconds
+
     //
     // Flush formatted images using a lazy flush
     //
     S32 num_refs = imagep->getNumRefs();
     if (num_refs <= min_refs && flush_images)
     {
-        // Remove the unused image from the image list
-        deleteImage(imagep);
-        return;
+        if (imagep->getLastReferencedTimer()->getElapsedTimeF32() > lazy_flush_timeout)
+        {
+            // Remove the unused image from the image list
+            deleteImage(imagep);
+            return;
+        }
     }
     else
     {
+        // still referenced outside of image list, reset timer
+        imagep->getLastReferencedTimer()->reset();
+
         if (imagep->hasSavedRawImage())
         {
             if (imagep->getElapsedLastReferencedSavedRawImageTime() > max_inactive_time)
diff --git a/indra/newview/llvoavatar.cpp b/indra/newview/llvoavatar.cpp
index 09375a6930..e638e5b8c3 100644
--- a/indra/newview/llvoavatar.cpp
+++ b/indra/newview/llvoavatar.cpp
@@ -109,7 +109,6 @@
 #include "llsdutil.h"
 #include "llscenemonitor.h"
 #include "llsdserialize.h"
-#include "llcallstack.h"
 #include "llrendersphere.h"
 #include "llskinningutil.h"
 
@@ -2666,8 +2665,6 @@ void LLVOAvatar::idleUpdate(LLAgent &agent, const F64 &time)
         mNeedsExtentUpdate = ((thisFrame + mID.mData[0]) % upd_freq == 0);
     }
 
-    LLScopedContextString str("avatar_idle_update " + getFullname());
-
     checkTextureLoading() ;
 
     // force immediate pixel area update on avatars using last frames data (before drawable or camera updates)
@@ -4689,10 +4686,6 @@ bool LLVOAvatar::updateCharacter(LLAgent &agent)
         is_attachment = cav && cav->mRootVolp && cav->mRootVolp->isAttachment(); // For attached animated objects
     }
 
-    LLScopedContextString str("updateCharacter " + getFullname() + " is_control_avatar "
-                              + boost::lexical_cast<std::string>(is_control_avatar)
-                              + " is_attachment " + boost::lexical_cast<std::string>(is_attachment));
-
     // For fading out the names above heads, only let the timer
     // run if we're visible.
     if (mDrawable.notNull() && !visible)
@@ -6358,8 +6351,6 @@ bool LLVOAvatar::jointIsRiggedTo(const LLJoint *joint) const
 
 void LLVOAvatar::clearAttachmentOverrides()
 {
-    LLScopedContextString str("clearAttachmentOverrides " + getFullname());
-
     for (S32 i=0; i<LL_CHARACTER_MAX_ANIMATED_JOINTS; i++)
     {
         LLJoint *pJoint = getJoint(i);
@@ -6390,10 +6381,7 @@ void LLVOAvatar::clearAttachmentOverrides()
 //-----------------------------------------------------------------------------
 void LLVOAvatar::rebuildAttachmentOverrides()
 {
-    LLScopedContextString str("rebuildAttachmentOverrides " + getFullname());
-
     LL_DEBUGS("AnimatedObjects") << "rebuilding" << LL_ENDL;
-    dumpStack("AnimatedObjectsStack");
 
     clearAttachmentOverrides();
 
@@ -6441,10 +6429,7 @@ void LLVOAvatar::rebuildAttachmentOverrides()
 // -----------------------------------------------------------------------------
 void LLVOAvatar::updateAttachmentOverrides()
 {
-    LLScopedContextString str("updateAttachmentOverrides " + getFullname());
-
     LL_DEBUGS("AnimatedObjects") << "updating" << LL_ENDL;
-    dumpStack("AnimatedObjectsStack");
 
     std::set<LLUUID> meshes_seen;
 
@@ -6573,15 +6558,12 @@ void LLVOAvatar::addAttachmentOverridesForObject(LLViewerObject *vo, std::set<LL
         return;
     }
 
-    LLScopedContextString str("addAttachmentOverridesForObject " + getFullname());
-
     if (getOverallAppearance() != AOA_NORMAL)
     {
         return;
     }
 
     LL_DEBUGS("AnimatedObjects") << "adding" << LL_ENDL;
-    dumpStack("AnimatedObjectsStack");
 
     // Process all children
     if (recursive)
diff --git a/indra/newview/llvoavatarself.cpp b/indra/newview/llvoavatarself.cpp
index 00a7325db2..49859bb585 100644
--- a/indra/newview/llvoavatarself.cpp
+++ b/indra/newview/llvoavatarself.cpp
@@ -59,7 +59,6 @@
 #include "llsdutil.h"
 #include "llstartup.h"
 #include "llsdserialize.h"
-#include "llcallstack.h"
 #include "llcorehttputil.h"
 #include "lluiusage.h"
 
diff --git a/indra/newview/llvoicechannel.cpp b/indra/newview/llvoicechannel.cpp
index eb1cd00940..55769f567b 100644
--- a/indra/newview/llvoicechannel.cpp
+++ b/indra/newview/llvoicechannel.cpp
@@ -328,6 +328,16 @@ void LLVoiceChannel::setState(EState state)
 
 void LLVoiceChannel::doSetState(const EState& new_state)
 {
+    LL_DEBUGS("Voice") << "session '" << mSessionName << "' state " << mState << ", new_state " << new_state << ": "
+        << (new_state == STATE_ERROR ? "ERROR" :
+            new_state == STATE_HUNG_UP ? "HUNG_UP" :
+            new_state == STATE_READY ? "READY" :
+            new_state == STATE_CALL_STARTED ? "CALL_STARTED" :
+            new_state == STATE_RINGING ? "RINGING" :
+            new_state == STATE_CONNECTED ? "CONNECTED" :
+            "NO_INFO")
+        << LL_ENDL;
+
     EState old_state = mState;
     mState = new_state;
 
diff --git a/indra/newview/llvoicevisualizer.cpp b/indra/newview/llvoicevisualizer.cpp
index 305fd77126..9412136272 100644
--- a/indra/newview/llvoicevisualizer.cpp
+++ b/indra/newview/llvoicevisualizer.cpp
@@ -342,6 +342,8 @@ void LLVoiceVisualizer::render()
         return;
     }
 
+    LL_PROFILE_ZONE_SCOPED_CATEGORY_UI;
+
     if ( mSoundSymbol.mActive )
     {
         mPreviousTime = mCurrentTime;
diff --git a/indra/newview/llvoicevivox.cpp b/indra/newview/llvoicevivox.cpp
index ac54445786..f3b624e968 100644
--- a/indra/newview/llvoicevivox.cpp
+++ b/indra/newview/llvoicevivox.cpp
@@ -6460,7 +6460,6 @@ LLVivoxVoiceClient::voiceFontEntry::voiceFontEntry(LLUUID& id) :
     mIsNew(false)
 {
     mExpiryTimer.stop();
-    mExpiryWarningTimer.stop();
 }
 
 LLVivoxVoiceClient::voiceFontEntry::~voiceFontEntry()
@@ -6571,20 +6570,6 @@ void LLVivoxVoiceClient::addVoiceFont(const S32 font_index,
             font->mExpiryTimer.start();
             font->mExpiryTimer.setExpiryAt(expiration_date.secondsSinceEpoch() - VOICE_FONT_EXPIRY_INTERVAL);
 
-            // Set the warning timer to some interval before actual expiry.
-            S32 warning_time = gSavedSettings.getS32("VoiceEffectExpiryWarningTime");
-            if (warning_time != 0)
-            {
-                font->mExpiryWarningTimer.start();
-                F64 expiry_time = (expiration_date.secondsSinceEpoch() - (F64)warning_time);
-                font->mExpiryWarningTimer.setExpiryAt(expiry_time - VOICE_FONT_EXPIRY_INTERVAL);
-            }
-            else
-            {
-                // Disable the warning timer.
-                font->mExpiryWarningTimer.stop();
-            }
-
              // Only flag new session fonts after the first time we have fetched the list.
             if (mVoiceFontsReceived)
             {
@@ -6626,7 +6611,6 @@ void LLVivoxVoiceClient::expireVoiceFonts()
     // than checking each font individually.
 
     bool have_expired = false;
-    bool will_expire = false;
     bool expired_in_use = false;
 
     LLUUID current_effect = LLVoiceClient::instance().getVoiceEffectDefault();
@@ -6636,7 +6620,6 @@ void LLVivoxVoiceClient::expireVoiceFonts()
     {
         voiceFontEntry* voice_font = iter->second;
         LLFrameTimer& expiry_timer  = voice_font->mExpiryTimer;
-        LLFrameTimer& warning_timer = voice_font->mExpiryWarningTimer;
 
         // Check for expired voice fonts
         if (expiry_timer.getStarted() && expiry_timer.hasExpired())
@@ -6653,14 +6636,6 @@ void LLVivoxVoiceClient::expireVoiceFonts()
             deleteVoiceFont(voice_font->mID);
             have_expired = true;
         }
-
-        // Check for voice fonts that will expire in less that the warning time
-        if (warning_timer.getStarted() && warning_timer.hasExpired())
-        {
-            LL_DEBUGS("VoiceFont") << "Voice Font " << voice_font->mName << " will expire soon." << LL_ENDL;
-            will_expire = true;
-            warning_timer.stop();
-        }
     }
 
     LLSD args;
@@ -6682,15 +6657,6 @@ void LLVivoxVoiceClient::expireVoiceFonts()
         // Refresh voice font lists in the UI.
         notifyVoiceFontObservers();
     }
-
-    // Give a warning notification if any voice fonts are due to expire.
-    if (will_expire)
-    {
-        S32Seconds seconds(gSavedSettings.getS32("VoiceEffectExpiryWarningTime"));
-        args["INTERVAL"] = llformat("%d", LLUnit<S32, LLUnits::Days>(seconds).value());
-
-        LLNotificationsUtil::add("VoiceEffectsWillExpire", args);
-    }
 }
 
 void LLVivoxVoiceClient::deleteVoiceFont(const LLUUID& id)
diff --git a/indra/newview/llvoicevivox.h b/indra/newview/llvoicevivox.h
index 7862e492b2..3167705528 100644
--- a/indra/newview/llvoicevivox.h
+++ b/indra/newview/llvoicevivox.h
@@ -880,7 +880,6 @@ private:
         bool        mIsNew;
 
         LLFrameTimer    mExpiryTimer;
-        LLFrameTimer    mExpiryWarningTimer;
     };
 
     bool mVoiceFontsReceived;
diff --git a/indra/newview/llvoicewebrtc.cpp b/indra/newview/llvoicewebrtc.cpp
index af93bdcaf0..fdb7b3c3b9 100644
--- a/indra/newview/llvoicewebrtc.cpp
+++ b/indra/newview/llvoicewebrtc.cpp
@@ -548,13 +548,19 @@ void LLWebRTCVoiceClient::voiceConnectionCoro()
                     updatePosition();
                 }
             }
-
-            sessionState::processSessionStates();
-            if (mProcessChannels && voiceEnabled && !mHidden)
-            {
-                sendPositionUpdate(false);
-                updateOwnVolume();
-            }
+            LL::WorkQueue::postMaybe(mMainQueue,
+                [=] {
+                    if  (sShuttingDown)
+                    {
+                        return;
+                    }
+                    sessionState::processSessionStates();
+                    if (mProcessChannels && voiceEnabled && !mHidden)
+                    {
+                        sendPositionUpdate(false);
+                        updateOwnVolume();
+                    }
+            });
         }
     }
     catch (const LLCoros::Stop&)
@@ -2220,6 +2226,7 @@ void LLVoiceWebRTCConnection::OnIceCandidate(const llwebrtc::LLWebRTCIceCandidat
 void LLVoiceWebRTCConnection::processIceUpdates()
 {
     mOutstandingRequests++;
+
     LLCoros::getInstance()->launch("LLVoiceWebRTCConnection::processIceUpdatesCoro",
                                    boost::bind(&LLVoiceWebRTCConnection::processIceUpdatesCoro, this->shared_from_this()));
 }
diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp
index 813a1d286c..f3e8c9ab3c 100644
--- a/indra/newview/llvovolume.cpp
+++ b/indra/newview/llvovolume.cpp
@@ -85,7 +85,6 @@
 #include "llanimationstates.h"
 #include "llinventorytype.h"
 #include "llviewerinventory.h"
-#include "llcallstack.h"
 #include "llsculptidsize.h"
 #include "llavatarappearancedefines.h"
 #include "llgltfmateriallist.h"
@@ -357,7 +356,6 @@ U32 LLVOVolume::processUpdateMessage(LLMessageSystem *mesgsys,
         sculpt_type = sculpt_params->getSculptType();
 
         LL_DEBUGS("ObjectUpdate") << "uuid " << mID << " set sculpt_id " << sculpt_id << LL_ENDL;
-        dumpStack("ObjectUpdateStack");
     }
 
     if (!dp)
@@ -1448,7 +1446,6 @@ bool LLVOVolume::calcLOD()
             const LLVector3* box = avatar->getLastAnimExtents();
             LLVector3 diag = box[1] - box[0];
             radius = diag.magVec() * 0.5f;
-            LL_DEBUGS("DynamicBox") << avatar->getFullname() << " diag " << diag << " radius " << radius << LL_ENDL;
         }
         else
         {
@@ -1459,11 +1456,9 @@ bool LLVOVolume::calcLOD()
             const LLVector3* box = avatar->getLastAnimExtents();
             LLVector3 diag = box[1] - box[0];
             radius = diag.magVec(); // preserve old BinRadius behavior - 2x off
-            LL_DEBUGS("DynamicBox") << avatar->getFullname() << " diag " << diag << " radius " << radius << LL_ENDL;
         }
         if (distance <= 0.f || radius <= 0.f)
         {
-            LL_DEBUGS("DynamicBox","CalcLOD") << "avatar distance/radius uninitialized, skipping" << LL_ENDL;
             return false;
         }
     }
@@ -1473,7 +1468,6 @@ bool LLVOVolume::calcLOD()
         radius = getVolume() ? getVolume()->mLODScaleBias.scaledVec(getScale()).length() : getScale().length();
         if (distance <= 0.f || radius <= 0.f)
         {
-            LL_DEBUGS("DynamicBox","CalcLOD") << "non-avatar distance/radius uninitialized, skipping" << LL_ENDL;
             return false;
         }
     }
@@ -1554,13 +1548,6 @@ bool LLVOVolume::calcLOD()
 
     if (cur_detail != mLOD)
     {
-        LL_DEBUGS("DynamicBox","CalcLOD") << "new LOD " << cur_detail << " change from " << mLOD
-                             << " distance " << distance << " radius " << radius << " rampDist " << rampDist
-                             << " drawable rigged? " << (mDrawable ? (S32) mDrawable->isState(LLDrawable::RIGGED) : (S32) -1)
-                             << " mRiggedVolume " << (void*)getRiggedVolume()
-                             << " distanceWRTCamera " << (mDrawable ? mDrawable->mDistanceWRTCamera : -1.f)
-                             << LL_ENDL;
-
         mAppAngle = ll_round((F32) atan2( mDrawable->getRadius(), mDrawable->mDistanceWRTCamera) * RAD_TO_DEG, 0.01f);
         mLOD = cur_detail;
 
@@ -1760,11 +1747,6 @@ bool LLVOVolume::genBBoxes(bool force_global, bool should_update_octree_bounds)
 
     bool any_valid_boxes = false;
 
-    if (getRiggedVolume())
-    {
-        LL_DEBUGS("RiggedBox") << "rebuilding box, volume face count " << getVolume()->getNumVolumeFaces() << " drawable face count " << mDrawable->getNumFaces() << LL_ENDL;
-    }
-
     // There's no guarantee that getVolume()->getNumFaces() == mDrawable->getNumFaces()
     for (S32 i = 0;
         i < getVolume()->getNumVolumeFaces() && i < mDrawable->getNumFaces() && i < getNumTEs();
@@ -1788,10 +1770,6 @@ bool LLVOVolume::genBBoxes(bool force_global, bool should_update_octree_bounds)
         }
         if (rebuild)
         {
-            if (getRiggedVolume())
-            {
-                LL_DEBUGS("RiggedBox") << "rebuilding box, face " << i << " extents " << face->mExtents[0] << ", " << face->mExtents[1] << LL_ENDL;
-            }
             if (!any_valid_boxes)
             {
                 min = face->mExtents[0];
@@ -3241,6 +3219,7 @@ void LLVOVolume::updateSpotLightPriority()
     {
         return;
     }
+    LL_PROFILE_ZONE_SCOPED_CATEGORY_VOLUME;
 
     F32 r = getLightRadius();
     LLVector3 pos = mDrawable->getPositionAgent();
@@ -5655,8 +5634,6 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group)
             // apply any pending material overrides
             gGLTFMaterialList.applyQueuedOverrides(vobj);
 
-            std::string vobj_name = llformat("Vol%p", vobj);
-
             bool is_mesh = vobj->isMesh();
             if (is_mesh)
             {
@@ -5681,24 +5658,8 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group)
                 group->mSurfaceArea += volume->getSurfaceArea() * llmax(llmax(scale.mV[0], scale.mV[1]), scale.mV[2]);
             }
 
-
-            F32 est_tris = vobj->getEstTrianglesMax();
-
             vobj->updateControlAvatar();
 
-            LL_DEBUGS("AnimatedObjectsLinkset") << vobj_name << " rebuilding, isAttachment: " << (U32) vobj->isAttachment()
-                                                << " is_mesh " << is_mesh
-                                                << " est_tris " << est_tris
-                                                << " is_animated " << vobj->isAnimatedObject()
-                                                << " can_animate " << vobj->canBeAnimatedObject()
-                                                << " cav " << vobj->getControlAvatar()
-                                                << " lod " << vobj->getLOD()
-                                                << " drawable rigged " << (drawablep->isState(LLDrawable::RIGGED))
-                                                << " drawable state " << drawablep->getState()
-                                                << " playing " << (U32) (vobj->getControlAvatar() ? vobj->getControlAvatar()->mPlaying : false)
-                                                << " frame " << LLFrameTimer::getFrameCount()
-                                                << LL_ENDL;
-
             llassert_always(vobj);
             vobj->updateTextureVirtualSize(true);
             vobj->preRebuild();
diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp
index 98fdb68222..eaf54356a2 100644
--- a/indra/newview/pipeline.cpp
+++ b/indra/newview/pipeline.cpp
@@ -117,6 +117,17 @@
 #include "llenvironment.h"
 #include "llsettingsvo.h"
 
+#ifndef LL_WINDOWS
+#define A_GCC 1
+#pragma GCC diagnostic ignored "-Wunused-function"
+#pragma GCC diagnostic ignored "-Wunused-variable"
+#if LL_LINUX
+#pragma GCC diagnostic ignored "-Wrestrict"
+#endif
+#endif
+#define A_CPU 1
+#include "app_settings/shaders/class1/deferred/CASF.glsl" // This is also C++
+
 extern bool gSnapshot;
 bool gShiftFrame = false;
 
@@ -6624,8 +6635,14 @@ void LLPipeline::renderAlphaObjects(bool rigged)
     S32 sun_up = LLEnvironment::instance().getIsSunUp() ? 1 : 0;
     U32 target_width = LLRenderTarget::sCurResX;
     U32 type = LLRenderPass::PASS_ALPHA;
-    LLVOAvatar* lastAvatar = nullptr;
+    // for gDeferredShadowAlphaMaskProgram
+    const LLVOAvatar* lastAvatar = nullptr;
     U64 lastMeshId = 0;
+    bool skipLastSkin;
+    // for gDeferredShadowGLTFAlphaBlendProgram
+    const LLVOAvatar* lastAvatarGLTF = nullptr;
+    U64 lastMeshIdGLTF = 0;
+    bool skipLastSkinGLTF;
     auto* begin = gPipeline.beginRenderMap(type);
     auto* end = gPipeline.endRenderMap(type);
 
@@ -6649,7 +6666,7 @@ void LLPipeline::renderAlphaObjects(bool rigged)
                 LLGLSLShader::sCurBoundShaderPtr->uniform1i(LLShaderMgr::SUN_UP_FACTOR, sun_up);
                 LLGLSLShader::sCurBoundShaderPtr->uniform1f(LLShaderMgr::DEFERRED_SHADOW_TARGET_WIDTH, (float)target_width);
                 LLGLSLShader::sCurBoundShaderPtr->setMinimumAlpha(ALPHA_BLEND_CUTOFF);
-                LLRenderPass::pushRiggedGLTFBatch(*pparams, lastAvatar, lastMeshId);
+                LLRenderPass::pushRiggedGLTFBatch(*pparams, lastAvatarGLTF, lastMeshIdGLTF, skipLastSkinGLTF);
             }
             else
             {
@@ -6657,14 +6674,10 @@ void LLPipeline::renderAlphaObjects(bool rigged)
                 LLGLSLShader::sCurBoundShaderPtr->uniform1i(LLShaderMgr::SUN_UP_FACTOR, sun_up);
                 LLGLSLShader::sCurBoundShaderPtr->uniform1f(LLShaderMgr::DEFERRED_SHADOW_TARGET_WIDTH, (float)target_width);
                 LLGLSLShader::sCurBoundShaderPtr->setMinimumAlpha(ALPHA_BLEND_CUTOFF);
-                if (lastAvatar != pparams->mAvatar || lastMeshId != pparams->mSkinInfo->mHash)
+                if (mSimplePool->uploadMatrixPalette(pparams->mAvatar, pparams->mSkinInfo, lastAvatar, lastMeshId, skipLastSkin))
                 {
-                    mSimplePool->uploadMatrixPalette(*pparams);
-                    lastAvatar = pparams->mAvatar;
-                    lastMeshId = pparams->mSkinInfo->mHash;
+                    mSimplePool->pushBatch(*pparams, true, true);
                 }
-
-                mSimplePool->pushBatch(*pparams, true, true);
             }
         }
         else
@@ -7131,6 +7144,51 @@ void LLPipeline::generateGlow(LLRenderTarget* src)
     }
 }
 
+void LLPipeline::applyCAS(LLRenderTarget* src, LLRenderTarget* dst)
+{
+    static LLCachedControl<F32> cas_sharpness(gSavedSettings, "RenderCASSharpness", 0.4f);
+    if (cas_sharpness == 0.0f)
+    {
+        gPipeline.copyRenderTarget(src, dst);
+        return;
+    }
+
+    LLGLSLShader* sharpen_shader = &gCASProgram;
+
+    // Bind setup:
+    dst->bindTarget();
+
+    sharpen_shader->bind();
+
+    {
+        static LLStaticHashedString cas_param_0("cas_param_0");
+        static LLStaticHashedString cas_param_1("cas_param_1");
+        static LLStaticHashedString out_screen_res("out_screen_res");
+
+        varAU4(const0);
+        varAU4(const1);
+        CasSetup(const0, const1,
+            cas_sharpness(),             // Sharpness tuning knob (0.0 to 1.0).
+            (AF1)src->getWidth(), (AF1)src->getHeight(),  // Input size.
+            (AF1)dst->getWidth(), (AF1)dst->getHeight()); // Output size.
+
+        sharpen_shader->uniform4uiv(cas_param_0, 1, const0);
+        sharpen_shader->uniform4uiv(cas_param_1, 1, const1);
+
+        sharpen_shader->uniform2f(out_screen_res, (AF1)dst->getWidth(), (AF1)dst->getHeight());
+    }
+
+    sharpen_shader->bindTexture(LLShaderMgr::DEFERRED_DIFFUSE, src, false, LLTexUnit::TFO_POINT);
+
+    // Draw
+    gPipeline.mScreenTriangleVB->setBuffer();
+    gPipeline.mScreenTriangleVB->drawArrays(LLRender::TRIANGLES, 0, 3);
+
+    sharpen_shader->unbind();
+
+    dst->flush();
+}
+
 void LLPipeline::applyFXAA(LLRenderTarget* src, LLRenderTarget* dst)
 {
     {
@@ -7500,13 +7558,15 @@ void LLPipeline::renderFinalize()
     gGLViewport[3] = gViewerWindow->getWorldViewRectRaw().getHeight();
     glViewport(gGLViewport[0], gGLViewport[1], gGLViewport[2], gGLViewport[3]);
 
-    renderDoF(&mRT->screen, &mPostMap);
+    applyCAS(&mRT->screen, &mPostMap);
+
+    renderDoF(&mPostMap, &mRT->screen);
 
-    applyFXAA(&mPostMap, &mRT->screen);
-    LLRenderTarget* finalBuffer = &mRT->screen;
+    applyFXAA(&mRT->screen, &mPostMap);
+    LLRenderTarget* finalBuffer = &mPostMap;
     if (RenderBufferVisualization > -1)
     {
-        finalBuffer = &mPostMap;
+        finalBuffer = &mRT->screen;
         switch (RenderBufferVisualization)
         {
         case 0:
@@ -7928,13 +7988,15 @@ void LLPipeline::renderDeferredLighting()
         mat.mult_matrix_vec(tc_moon);
         mTransformedMoonDir.set(tc_moon.v);
 
-        if (RenderDeferredSSAO || RenderShadowDetail > 0)
+        if ((RenderDeferredSSAO && !gCubeSnapshot) || RenderShadowDetail > 0)
         {
             LL_PROFILE_GPU_ZONE("sun program");
             deferred_light_target->bindTarget();
             {  // paint shadow/SSAO light map (direct lighting lightmap)
                 LL_PROFILE_ZONE_NAMED_CATEGORY_PIPELINE("renderDeferredLighting - sun shadow");
-                bindDeferredShader(gDeferredSunProgram, deferred_light_target);
+
+                LLGLSLShader& sun_shader = gCubeSnapshot ? gDeferredSunProbeProgram : gDeferredSunProgram;
+                bindDeferredShader(sun_shader, deferred_light_target);
                 mScreenTriangleVB->setBuffer();
                 glClearColor(1, 1, 1, 1);
                 deferred_light_target->clear(GL_COLOR_BUFFER_BIT);
@@ -7959,8 +8021,8 @@ void LLPipeline::renderDeferredLighting()
                     }
                 }
 
-                gDeferredSunProgram.uniform3fv(sOffset, slice, offset);
-                gDeferredSunProgram.uniform2f(LLShaderMgr::DEFERRED_SCREEN_RES,
+                sun_shader.uniform3fv(sOffset, slice, offset);
+                sun_shader.uniform2f(LLShaderMgr::DEFERRED_SCREEN_RES,
                                               (GLfloat)deferred_light_target->getWidth(),
                                               (GLfloat)deferred_light_target->getHeight());
 
@@ -7970,12 +8032,12 @@ void LLPipeline::renderDeferredLighting()
                     mScreenTriangleVB->drawArrays(LLRender::TRIANGLES, 0, 3);
                 }
 
-                unbindDeferredShader(gDeferredSunProgram);
+                unbindDeferredShader(sun_shader);
             }
             deferred_light_target->flush();
         }
 
-        if (RenderDeferredSSAO)
+        if (RenderDeferredSSAO && !gCubeSnapshot)
         {
             // soften direct lighting lightmap
             LL_PROFILE_ZONE_NAMED_CATEGORY_PIPELINE("renderDeferredLighting - soften shadow");
diff --git a/indra/newview/pipeline.h b/indra/newview/pipeline.h
index 024303d5f0..1c87b71d06 100644
--- a/indra/newview/pipeline.h
+++ b/indra/newview/pipeline.h
@@ -157,6 +157,7 @@ public:
     void generateExposure(LLRenderTarget* src, LLRenderTarget* dst, bool use_history = true);
     void gammaCorrect(LLRenderTarget* src, LLRenderTarget* dst);
     void generateGlow(LLRenderTarget* src);
+    void applyCAS(LLRenderTarget* src, LLRenderTarget* dst);
     void applyFXAA(LLRenderTarget* src, LLRenderTarget* dst);
     void renderDoF(LLRenderTarget* src, LLRenderTarget* dst);
     void copyRenderTarget(LLRenderTarget* src, LLRenderTarget* dst);
diff --git a/indra/newview/skins/default/xui/da/notifications.xml b/indra/newview/skins/default/xui/da/notifications.xml
index 283a7b2a43..4a4b7269dc 100644
--- a/indra/newview/skins/default/xui/da/notifications.xml
+++ b/indra/newview/skins/default/xui/da/notifications.xml
@@ -1574,10 +1574,6 @@ Klik på Acceptér for at deltage eller Afvis for at afvise invitationen. Klik p
 		Den aktive stemme &quot;morph&quot; er udløbet og din normale stemme opsætning er genaktiveret.
 [[URL] Click here] for at forny dit abbonnement.
 	</notification>
-	<notification name="VoiceEffectsWillExpire">
-		En eller flere af dine stemme &quot;morphs&quot; vil udløbe om mindre end [INTERVAL] dage.
-[[URL] Click here] for at forny dit abbonnement.
-	</notification>
 	<notification name="VoiceEffectsNew">
 		Nye stemme &quot;morphs&quot; er tilgængelige!
 	</notification>
diff --git a/indra/newview/skins/default/xui/de/notifications.xml b/indra/newview/skins/default/xui/de/notifications.xml
index 6ad71e0ad1..76bebedeec 100644
--- a/indra/newview/skins/default/xui/de/notifications.xml
+++ b/indra/newview/skins/default/xui/de/notifications.xml
@@ -2466,10 +2466,6 @@ Wenn Sie Premium-Mitglied sind, [[PREMIUM_URL] klicken Sie hier], um Ihren Voice
 [[URL] Klicken Sie hier], um Ihr Abo zu erneuern. 
 
 Wenn Sie Premium-Mitglied sind, [[PREMIUM_URL] klicken Sie hier], um Ihren Voice-Morphing-Vorteil zu nutzen.</notification>
-	<notification name="VoiceEffectsWillExpire">Ein oder mehrere Ihrer Voice-Morph-Abos laufen in weniger als [INTERVAL] Tagen ab. 
-[[URL] Klicken Sie hier], um Ihr Abo zu erneuern. 
-
-Wenn Sie Premium-Mitglied sind, [[PREMIUM_URL] klicken Sie hier], um Ihren Voice-Morphing-Vorteil zu nutzen.</notification>
 	<notification name="VoiceEffectsNew">Neue Voice-Morph-Effekte sind erhältlich!</notification>
 	<notification name="Cannot enter parcel: not a group member">Nur Mitglieder einer bestimmten Gruppe dürfen diesen Bereich betreten.</notification>
 	<notification name="Cannot enter parcel: banned">Zugang zur Parzelle verweigert. Sie wurden verbannt.</notification>
diff --git a/indra/newview/skins/default/xui/en/floater_preferences_graphics_advanced.xml b/indra/newview/skins/default/xui/en/floater_preferences_graphics_advanced.xml
index 8c4019a668..8ed25e22e4 100644
--- a/indra/newview/skins/default/xui/en/floater_preferences_graphics_advanced.xml
+++ b/indra/newview/skins/default/xui/en/floater_preferences_graphics_advanced.xml
@@ -881,7 +881,26 @@
       value="1"/>
   </combo_box>
   <!-- End of mirror settings -->
-
+  <!-- Sharpening Settings -->
+  <slider
+    control_name="RenderCASSharpness"
+    decimal_digits="1"
+    follows="left|top"
+    height="16"
+    increment="0.1"
+    initial_value="160"
+    label="Sharpening:"
+    label_width="145"
+    layout="topleft"
+    left="420"
+    min_val="0.0"
+    max_val="1.0"
+    name="RenderSharpness"
+    show_text="true"
+    top_delta="24"
+    width="260">
+  </slider>
+  <!-- End of Sharpening Settings-->
   <!-- End of Advanced Settings block -->
 	<view_border
       bevel_style="in"
diff --git a/indra/newview/skins/default/xui/en/notifications.xml b/indra/newview/skins/default/xui/en/notifications.xml
index e539388735..4a9dd62666 100644
--- a/indra/newview/skins/default/xui/en/notifications.xml
+++ b/indra/newview/skins/default/xui/en/notifications.xml
@@ -8860,21 +8860,6 @@ If you are a Premium Member, [[PREMIUM_URL] click here] to receive your voice mo
 
   <notification
    icon="notify.tga"
-   name="VoiceEffectsWillExpire"
-   sound="UISndAlert"
-   persist="true"
-   type="notify">
-    <unique/>    
-One or more of your Voice Morphs will expire in less than [INTERVAL] days.
-[[URL] Click here] to renew your subscription.
-
-If you are a Premium Member, [[PREMIUM_URL] click here] to receive your voice morphing perk.
-  <tag>fail</tag>
-    <tag>voice</tag>
-  </notification>
-
-  <notification
-   icon="notify.tga"
    name="VoiceEffectsNew"
    sound="UISndAlert"
    persist="true"
diff --git a/indra/newview/skins/default/xui/es/notifications.xml b/indra/newview/skins/default/xui/es/notifications.xml
index 739391b965..bf55e2c443 100644
--- a/indra/newview/skins/default/xui/es/notifications.xml
+++ b/indra/newview/skins/default/xui/es/notifications.xml
@@ -2452,10 +2452,6 @@ Si eres un miembro Premium [[PREMIUM_URL] pulsa aquí] para recibir tu beneficio
 [[URL] Pulsa aquí] para renovar la suscripción. 
 
 Si eres un miembro Premium [[PREMIUM_URL] pulsa aquí] para recibir tu beneficio de transformación de voz.</notification>
-	<notification name="VoiceEffectsWillExpire">Una o más de tus transformaciones de voz caducarán en menos de [INTERVAL] días. 
-[[URL] Pulsa aquí] para renovar la suscripción 
-
-Si eres un miembro Premium [[PREMIUM_URL] pulsa aquí] para recibir tu beneficio de transformación de voz.</notification>
 	<notification name="VoiceEffectsNew">Están disponibles nuevas transformaciones de voz.</notification>
 	<notification name="Cannot enter parcel: not a group member">Sólo los miembros de un grupo determinado pueden visitar esta zona.</notification>
 	<notification name="Cannot enter parcel: banned">No puedes entrar en esta parcela, se te ha prohibido el acceso.</notification>
diff --git a/indra/newview/skins/default/xui/fr/notifications.xml b/indra/newview/skins/default/xui/fr/notifications.xml
index 587c88faad..17cf18633f 100644
--- a/indra/newview/skins/default/xui/fr/notifications.xml
+++ b/indra/newview/skins/default/xui/fr/notifications.xml
@@ -2451,10 +2451,6 @@ Si vous êtes un membre Premium,  [[PREMIUM_URL] cliquez ici] pour recevoir votr
 [[URL] Cliquez ici] pour renouveler votre abonnement. 
 
 Si vous êtes un membre Premium, [[PREMIUM_URL] cliquez ici] pour recevoir votre effet de voix.</notification>
-	<notification name="VoiceEffectsWillExpire">Au moins l'un de vos effets de voix expirera dans moins de [INTERVAL] jours.  
-[[URL] Cliquez ici] pour renouveler votre abonnement. 
-
-Si vous êtes un membre Premium,  [[PREMIUM_URL] cliquez ici] pour recevoir votre effet de voix.</notification>
 	<notification name="VoiceEffectsNew">De nouveaux effets de voix sont disponibles !</notification>
 	<notification name="Cannot enter parcel: not a group member">Seuls les membres d'un certain groupe peuvent visiter cette zone.</notification>
 	<notification name="Cannot enter parcel: banned">Vous ne pouvez pas pénétrer sur ce terrain car l'accès vous y est interdit.</notification>
diff --git a/indra/newview/skins/default/xui/it/notifications.xml b/indra/newview/skins/default/xui/it/notifications.xml
index f79cc1515b..1c40e7304a 100644
--- a/indra/newview/skins/default/xui/it/notifications.xml
+++ b/indra/newview/skins/default/xui/it/notifications.xml
@@ -2454,10 +2454,6 @@ Se sei un membro Premium, [[PREMIUM_URL] fai clic qui] per ricevere in regalo la
 [[URL] Fai clic qui] per rinnovare l'abbonamento. 
 
 Se sei un membro Premium, [[PREMIUM_URL] fai clic qui] per ricevere in regalo la manipolazione vocale.</notification>
-	<notification name="VoiceEffectsWillExpire">Almeno una delle tue manipolazioni vocali scadrà tra meno di [INTERVAL] giorni. 
-[[URL] Fai clic qui] per rinnovare l'abbonamento. 
-
-Se sei un membro Premium, [[PREMIUM_URL] fai clic qui] per ricevere in regalo la manipolazione vocale.</notification>
 	<notification name="VoiceEffectsNew">Sono disponibili nuove manipolazioni vocali.</notification>
 	<notification name="Cannot enter parcel: not a group member">Soltanto i membri di un determinato gruppo possono visitare questa zona.</notification>
 	<notification name="Cannot enter parcel: banned">Non puoi entrare nel terreno, sei stato bloccato.</notification>
diff --git a/indra/newview/skins/default/xui/ja/notifications.xml b/indra/newview/skins/default/xui/ja/notifications.xml
index 123e95df04..fbd56e118c 100644
--- a/indra/newview/skins/default/xui/ja/notifications.xml
+++ b/indra/newview/skins/default/xui/ja/notifications.xml
@@ -4663,17 +4663,6 @@ Webページにリンクすると、他人がこの場所に簡単にアクセ�
 			voice
 		</tag>
 	</notification>
-	<notification name="VoiceEffectsWillExpire">ボイスモーフィング効果の１つ、または複数の有効期限が[INTERVAL]日以内に終了します。
-期限を延長・更新するには[[URL] ここ]をクリックしてください。
-
-プレミアム会員の方は、[[PREMIUM_URL] ここ]をクリックしてボイスモーフィング特典をお受け取りください。
-    <tag>
-			fail
-		</tag>
-		<tag>
-			voice
-		</tag>
-	</notification>
 	<notification name="VoiceEffectsNew">新しいボイスモーフィング効果が登場！
     <tag>
 			voice
diff --git a/indra/newview/skins/default/xui/pl/notifications.xml b/indra/newview/skins/default/xui/pl/notifications.xml
index e668c6cc20..17c11bc75f 100644
--- a/indra/newview/skins/default/xui/pl/notifications.xml
+++ b/indra/newview/skins/default/xui/pl/notifications.xml
@@ -3118,11 +3118,6 @@ Jeśli jesteś użytkownikiem premium, to [[PREMIUM_URL] kliknij tutaj] aby otrz
 [[URL] Kliknij tutaj] aby odnowić subskrypcję.
 Jeśli jesteś użytkownikiem premium, to [[PREMIUM_URL] kliknij tutaj] aby otrzymać swój perk Przekształceń Głosu.
 	</notification>
-	<notification name="VoiceEffectsWillExpire">
-		Jedno lub więcej z Twoich Przekształceń Głosu wygaśnie za mniej niż [INTERVAL] dni.
-[[URL] Kliknij tutaj] aby odnowić subskrypcję.
-Jeśli jesteś użytkownikiem premium, to [[PREMIUM_URL] kliknij tutaj] aby otrzymać swój perk Przekształceń Głosu.
-	</notification>
 	<notification name="VoiceEffectsNew">
 		Nowe Przekształcenia Głosu są dostępne!
 	</notification>
diff --git a/indra/newview/skins/default/xui/pt/notifications.xml b/indra/newview/skins/default/xui/pt/notifications.xml
index a3220bca54..0390239669 100644
--- a/indra/newview/skins/default/xui/pt/notifications.xml
+++ b/indra/newview/skins/default/xui/pt/notifications.xml
@@ -2441,10 +2441,6 @@ Se você é um Membro Premium, [[PREMIUM_URL] clique aqui] para receber o seu ap
 [[URL] Clique aqui] para renovar o serviço. 
 
 Se você é um Membro Premium, [[PREMIUM_URL] clique aqui] para receber o seu app de distorção de voz.</notification>
-	<notification name="VoiceEffectsWillExpire">Uma ou mais das suas distorções de voz tem vencimento em menos de [INTERVAL] dias. 
-[[URL] Clique aqui] para renovar o serviço. 
-
-Se você é um Membro Premium, [[PREMIUM_URL] clique aqui] para receber o seu app de distorção de voz.</notification>
 	<notification name="VoiceEffectsNew">Novas Distorções de voz!</notification>
 	<notification name="Cannot enter parcel: not a group member">Só membros de um grupo podem acessar esta área.</notification>
 	<notification name="Cannot enter parcel: banned">Você não pode entrar nessa terra, você foi banido.</notification>
diff --git a/indra/newview/skins/default/xui/ru/notifications.xml b/indra/newview/skins/default/xui/ru/notifications.xml
index e75fd1fd82..bde18edc23 100644
--- a/indra/newview/skins/default/xui/ru/notifications.xml
+++ b/indra/newview/skins/default/xui/ru/notifications.xml
@@ -3232,12 +3232,6 @@
 
 Если вы - владелец премиум-аккаунта, [[PREMIUM_URL] щелкните здесь], чтобы получить право на анимационное изменение голоса.
 	</notification>
-	<notification name="VoiceEffectsWillExpire">
-		Срок действия одного или нескольких ваших типов анимационного изменения голоса истекает через [INTERVAL] дней или раньше. 
-[[URL] Щелкните здесь], чтобы обновить подписку. 
-
-Если вы - владелец премиум-аккаунта, [[PREMIUM_URL] щелкните здесь], чтобы получить право на анимационное изменение голоса.
-	</notification>
 	<notification name="VoiceEffectsNew">
 		Появились новые типы изменения голоса!
 	</notification>
diff --git a/indra/newview/skins/default/xui/tr/notifications.xml b/indra/newview/skins/default/xui/tr/notifications.xml
index 17d2969d19..30aa0c0342 100644
--- a/indra/newview/skins/default/xui/tr/notifications.xml
+++ b/indra/newview/skins/default/xui/tr/notifications.xml
@@ -3232,12 +3232,6 @@ Aboneliğinizi yenilemek için [[URL] buraya tıklayın].
 
 Özel Üye iseniz, ses dönüştürme özelliğini almak için [[PREMIUM_URL] buraya tıklayın].
 	</notification>
-	<notification name="VoiceEffectsWillExpire">
-		Ses Dönüşümlerinizden birinin ya da daha fazlasının süresi [INTERVAL] günden daha az bir zamanda dolacak. 
-Aboneliğinizi yenilemek için [[URL] buraya tıklayın]. 
-
-Özel Üye iseniz, ses dönüştürme özelliğini almak için [[PREMIUM_URL] buraya tıklayın].
-	</notification>
 	<notification name="VoiceEffectsNew">
 		Yeni Ses Şekilleri kullanılabilir!
 	</notification>
diff --git a/indra/newview/skins/default/xui/zh/notifications.xml b/indra/newview/skins/default/xui/zh/notifications.xml
index 4d0f1cb85b..3ebea7dc27 100644
--- a/indra/newview/skins/default/xui/zh/notifications.xml
+++ b/indra/newview/skins/default/xui/zh/notifications.xml
@@ -3216,12 +3216,6 @@ SHA1 指紋：[MD5_DIGEST]
 
 付費用戶請[[PREMIUM_URL] 點按這裡]領取免費變聲工具。
 	</notification>
-	<notification name="VoiceEffectsWillExpire">
-		至少一個你訂用的變聲效果將在 [INTERVAL] 天後到期。
-[[URL] 點按這裡]繼續訂用。
-
-付費用戶請[[PREMIUM_URL] 點按這裡]領取免費變聲工具。
-	</notification>
 	<notification name="VoiceEffectsNew">
 		新的變聲效果上市了！
 	</notification>