From 6ac6736994240d9789a81bf585468bef50805fd8 Mon Sep 17 00:00:00 2001 From: Graham Madarasz Date: Mon, 11 Mar 2013 16:00:25 -0700 Subject: Move 16b aligned memcpy and alignment utilities to llmem in llcommon for easier use elsewhere --- indra/llmath/llvector4a.cpp | 50 +-------------------------------------------- 1 file changed, 1 insertion(+), 49 deletions(-) (limited to 'indra/llmath/llvector4a.cpp') diff --git a/indra/llmath/llvector4a.cpp b/indra/llmath/llvector4a.cpp index 6edeb0fefe..570fa41a43 100644 --- a/indra/llmath/llvector4a.cpp +++ b/indra/llmath/llvector4a.cpp @@ -41,55 +41,7 @@ extern const LLVector4a LL_V4A_EPSILON = reinterpret_cast ( F /*static */void LLVector4a::memcpyNonAliased16(F32* __restrict dst, const F32* __restrict src, size_t bytes) { - assert(src != NULL); - assert(dst != NULL); - assert(bytes > 0); - assert((bytes % sizeof(F32))== 0); - ll_assert_aligned(src,16); - ll_assert_aligned(dst,16); - assert(bytes%16==0); - - F32* end = dst + (bytes / sizeof(F32) ); - - if (bytes > 64) - { - F32* begin_64 = LL_NEXT_ALIGNED_ADDRESS_64(dst); - - //at least 64 (16*4) bytes before the end of the destination, switch to 16 byte copies - F32* end_64 = end-16; - - _mm_prefetch((char*)begin_64, _MM_HINT_NTA); - _mm_prefetch((char*)begin_64 + 64, _MM_HINT_NTA); - _mm_prefetch((char*)begin_64 + 128, _MM_HINT_NTA); - _mm_prefetch((char*)begin_64 + 192, _MM_HINT_NTA); - - while (dst < begin_64) - { - copy4a(dst, src); - dst += 4; - src += 4; - } - - while (dst < end_64) - { - _mm_prefetch((char*)src + 512, _MM_HINT_NTA); - _mm_prefetch((char*)dst + 512, _MM_HINT_NTA); - copy4a(dst, src); - copy4a(dst+4, src+4); - copy4a(dst+8, src+8); - copy4a(dst+12, src+12); - - dst += 16; - src += 16; - } - } - - while (dst < end) - { - copy4a(dst, src); - dst += 4; - src += 4; - } + ll_memcpy_nonaliased_aligned_16((char*)dst, (char*)src, bytes); } void LLVector4a::setRotated( const LLRotation& rot, const LLVector4a& vec ) -- cgit v1.2.3