archive/rpcs3
1
0
Fork 0
mirror of https://github.com/RPCS3/rpcs3.git synced 2025-07-11 01:08:39 +12:00
Code Issues Projects Releases Packages Wiki Activity Actions

rsx: Improve use of CPU vector extensions

Browse source 
- Allow use of intrinsics when SSSE3 and SSSE4.1 are not available in the build target environment
- Properly separate SSE4.1 code from SSSE3 code for some older proceessors without SSE4.1
This commit is contained in:
kd-11 2019-09-11 17:41:57 +03:00 • committed by kd-11
parent 27af75fe71
commit 60845daf45
1 changed files with 150 additions and 60 deletions
Download patch file Download diff file Expand all files Collapse all files

210
rpcs3/Emu/RSX/Common/BufferUtils.cpp
View file

@ -8,15 +8,56 @@
#define DEBUG_VERTEX_STREAMING 0 #define DEBUG_VERTEX_STREAMING 0
const bool s_use_ssse3 = #if defined(_MSC_VER)
#ifdef _MSC_VER #define __SSSE3__ 1
utils::has_ssse3(); #define __SSE4_1__ 1
#elif __SSSE3__
true;
#else #else
false; #define __sse_intrin static FORCE_INLINE
#define _mm_shuffle_epi8(opa, opb) opb #endif // _MSC_VER
#endif
// NOTE: Clang does not allow to redefine missing intrinsics
#ifndef __SSSE3__
__sse_intrin __m128i __mm_shuffle_epi8(__m128i opd, __m128i opa)
{
__asm__("pshufb %1, %0" : "+x" (opd) : "xm" (opa));
return opd;
}
#else
#define __mm_shuffle_epi8 _mm_shuffle_epi8
#endif // __SSSE3__
#ifndef __SSE4_1__
__sse_intrin __m128i __mm_max_epu32(__m128i opd, __m128i opa)
{
__asm__("pmaxud %1, %0" : "+x" (opd) : "xm" (opa));
return opd;
}
__sse_intrin __m128i __mm_min_epu32(__m128i opd, __m128i opa)
{
__asm__("pminud %1, %0" : "+x" (opd) : "xm" (opa));
return opd;
}
__sse_intrin __m128i __mm_max_epu16(__m128i opd, __m128i opa)
{
__asm__("pmaxuw %1, %0" : "+x" (opd) : "xm" (opa));
return opd;
}
__sse_intrin __m128i __mm_min_epu16(__m128i opd, __m128i opa)
{
__asm__("pminuw %1, %0" : "+x" (opd) : "xm" (opa));
return opd;
}
#else
#define __mm_max_epu32 _mm_max_epu32
#define __mm_min_epu32 _mm_min_epu32
#define __mm_max_epu16 _mm_max_epu16
#define __mm_min_epu16 _mm_min_epu16
#endif // __SSE4_1__
#undef __sse_intrin
const bool s_use_ssse3 = utils::has_ssse3();
const bool s_use_sse4_1 = utils::has_sse41();
namespace namespace
{ {
@ -74,7 +115,7 @@ namespace
for (u32 i = 0; i < iterations; ++i) for (u32 i = 0; i < iterations; ++i)
{ {
const __m128i vector = _mm_loadu_si128(src_ptr); const __m128i vector = _mm_loadu_si128(src_ptr);
const __m128i shuffled_vector = _mm_shuffle_epi8(vector, mask); const __m128i shuffled_vector = __mm_shuffle_epi8(vector, mask);
_mm_stream_si128(dst_ptr, shuffled_vector); _mm_stream_si128(dst_ptr, shuffled_vector);
src_ptr++; src_ptr++;
@ -125,7 +166,7 @@ namespace
for (u32 i = 0; i < iterations; ++i) for (u32 i = 0; i < iterations; ++i)
{ {
const __m128i vector = _mm_loadu_si128(src_ptr); const __m128i vector = _mm_loadu_si128(src_ptr);
const __m128i shuffled_vector = _mm_shuffle_epi8(vector, mask); const __m128i shuffled_vector = __mm_shuffle_epi8(vector, mask);
_mm_stream_si128(dst_ptr, shuffled_vector); _mm_stream_si128(dst_ptr, shuffled_vector);
src_ptr++; src_ptr++;
@ -185,7 +226,7 @@ namespace
for (u32 i = 0; i < iterations; ++i) for (u32 i = 0; i < iterations; ++i)
{ {
const __m128i vector = _mm_loadu_si128((__m128i*)src_ptr); const __m128i vector = _mm_loadu_si128((__m128i*)src_ptr);
const __m128i shuffled_vector = _mm_shuffle_epi8(vector, mask); const __m128i shuffled_vector = __mm_shuffle_epi8(vector, mask);
_mm_storeu_si128((__m128i*)dst_ptr, shuffled_vector); _mm_storeu_si128((__m128i*)dst_ptr, shuffled_vector);
src_ptr += src_stride; src_ptr += src_stride;
@ -249,7 +290,7 @@ namespace
for (u32 i = 0; i < iterations; ++i) for (u32 i = 0; i < iterations; ++i)
{ {
const __m128i vector = _mm_loadu_si128((__m128i*)src_ptr); const __m128i vector = _mm_loadu_si128((__m128i*)src_ptr);
const __m128i shuffled_vector = _mm_shuffle_epi8(vector, mask); const __m128i shuffled_vector = __mm_shuffle_epi8(vector, mask);
_mm_storeu_si128((__m128i*)dst_ptr, shuffled_vector); _mm_storeu_si128((__m128i*)dst_ptr, shuffled_vector);
src_ptr += src_stride; src_ptr += src_stride;
@ -580,42 +621,65 @@ namespace
for (unsigned n = 0; n < iterations; ++n) for (unsigned n = 0; n < iterations; ++n)
{ {
const __m128i raw = _mm_loadu_si128(src_stream++); const __m128i raw = _mm_loadu_si128(src_stream++);
const __m128i value = _mm_shuffle_epi8(raw, mask); const __m128i value = __mm_shuffle_epi8(raw, mask);
max = _mm_max_epu16(max, value); max = __mm_max_epu16(max, value);
min = _mm_min_epu16(min, value); min = __mm_min_epu16(min, value);
_mm_storeu_si128(dst_stream++, value); _mm_storeu_si128(dst_stream++, value);
} }
const __m128i mask_step1 = _mm_set_epi8( if (s_use_sse4_1)
0, 0, 0, 0, 0, 0, 0, 0, {
0xF, 0xE, 0xD, 0xC, 0xB, 0xA, 0x9, 0x8); const __m128i mask_step1 = _mm_set_epi8(
0, 0, 0, 0, 0, 0, 0, 0,
0xF, 0xE, 0xD, 0xC, 0xB, 0xA, 0x9, 0x8);
const __m128i mask_step2 = _mm_set_epi8( const __m128i mask_step2 = _mm_set_epi8(
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0x7, 0x6, 0x5, 0x4); 0, 0, 0, 0, 0x7, 0x6, 0x5, 0x4);
const __m128i mask_step3 = _mm_set_epi8( const __m128i mask_step3 = _mm_set_epi8(
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0x3, 0x2); 0, 0, 0, 0, 0, 0, 0x3, 0x2);
__m128i tmp = _mm_shuffle_epi8(min, mask_step1); __m128i tmp = __mm_shuffle_epi8(min, mask_step1);
min = _mm_min_epu16(min, tmp); min = __mm_min_epu16(min, tmp);
tmp = _mm_shuffle_epi8(min, mask_step2); tmp = __mm_shuffle_epi8(min, mask_step2);
min = _mm_min_epu16(min, tmp); min = __mm_min_epu16(min, tmp);
tmp = _mm_shuffle_epi8(min, mask_step3); tmp = __mm_shuffle_epi8(min, mask_step3);
min = _mm_min_epu16(min, tmp); min = __mm_min_epu16(min, tmp);
tmp = _mm_shuffle_epi8(max, mask_step1); tmp = __mm_shuffle_epi8(max, mask_step1);
max = _mm_max_epu16(max, tmp); max = __mm_max_epu16(max, tmp);
tmp = _mm_shuffle_epi8(max, mask_step2); tmp = __mm_shuffle_epi8(max, mask_step2);
max = _mm_max_epu16(max, tmp); max = __mm_max_epu16(max, tmp);
tmp = _mm_shuffle_epi8(max, mask_step3); tmp = __mm_shuffle_epi8(max, mask_step3);
max = _mm_max_epu16(max, tmp); max = __mm_max_epu16(max, tmp);
const u16 min_index = u16(_mm_cvtsi128_si32(min) & 0xFFFF); const u16 min_index = u16(_mm_cvtsi128_si32(min) & 0xFFFF);
const u16 max_index = u16(_mm_cvtsi128_si32(max) & 0xFFFF); const u16 max_index = u16(_mm_cvtsi128_si32(max) & 0xFFFF);
return std::make_tuple(min_index, max_index, count); return std::make_tuple(min_index, max_index, count);
}
else
{
// Manual min-max
alignas(16) u16 _min[8];
alignas(16) u16 _max[8];
_mm_store_si128((__m128i*)_min, min);
_mm_store_si128((__m128i*)_max, max);
u16 min_index = _min[0];
u16 max_index = _max[0];
for (int i = 1; i < 8; ++i)
{
min_index = std::min(min_index, _min[i]);
max_index = std::max(max_index, _max[i]);
}
return std::make_tuple(min_index, max_index, count);
}
} }
static static
@ -637,35 +701,61 @@ namespace
for (unsigned n = 0; n < iterations; ++n) for (unsigned n = 0; n < iterations; ++n)
{ {
const __m128i raw = _mm_loadu_si128(src_stream++); const __m128i raw = _mm_loadu_si128(src_stream++);
const __m128i value = _mm_shuffle_epi8(raw, mask); const __m128i value = __mm_shuffle_epi8(raw, mask);
max = _mm_max_epu32(max, value); max = __mm_max_epu32(max, value);
min = _mm_min_epu32(min, value); min = __mm_min_epu32(min, value);
_mm_storeu_si128(dst_stream++, value); _mm_storeu_si128(dst_stream++, value);
} }
// Aggregate min-max if (s_use_sse4_1)
const __m128i mask_step1 = _mm_set_epi8( {
0, 0, 0, 0, 0, 0, 0, 0, // Aggregate min-max
0xF, 0xE, 0xD, 0xC, 0xB, 0xA, 0x9, 0x8); const __m128i mask_step1 = _mm_set_epi8(
0, 0, 0, 0, 0, 0, 0, 0,
0xF, 0xE, 0xD, 0xC, 0xB, 0xA, 0x9, 0x8);
const __m128i mask_step2 = _mm_set_epi8( const __m128i mask_step2 = _mm_set_epi8(
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0x7, 0x6, 0x5, 0x4); 0, 0, 0, 0, 0x7, 0x6, 0x5, 0x4);
__m128i tmp = _mm_shuffle_epi8(min, mask_step1); // a1, a2, a3, a4
min = _mm_min_epu16(min, tmp); // a1, a2, a1, a2
tmp = _mm_shuffle_epi8(min, mask_step2); // mAX
min = _mm_min_epu16(min, tmp); __m128i tmp = __mm_shuffle_epi8(min, mask_step1);
min = __mm_min_epu32(min, tmp);
tmp = __mm_shuffle_epi8(min, mask_step2);
min = __mm_min_epu32(min, tmp);
tmp = _mm_shuffle_epi8(max, mask_step1); tmp = __mm_shuffle_epi8(max, mask_step1);
max = _mm_max_epu16(max, tmp); max = __mm_max_epu32(max, tmp);
tmp = _mm_shuffle_epi8(max, mask_step2); tmp = __mm_shuffle_epi8(max, mask_step2);
max = _mm_max_epu16(max, tmp); max = __mm_max_epu32(max, tmp);
const u32 min_index = u32(_mm_cvtsi128_si32(min)); const u32 min_index = u32(_mm_cvtsi128_si32(min));
const u32 max_index = u32(_mm_cvtsi128_si32(max)); const u32 max_index = u32(_mm_cvtsi128_si32(max));
return std::make_tuple(min_index, max_index, count); return std::make_tuple(min_index, max_index, count);
}
else
{
// Manual min-max
alignas(16) u32 _min[4];
alignas(16) u32 _max[4];
_mm_store_si128((__m128i*)_min, min);
_mm_store_si128((__m128i*)_max, max);
u32 min_index = _min[0];
u32 max_index = _max[0];
for (int i = 1; i < 4; ++i)
{
min_index = std::min(min_index, _min[i]);
max_index = std::max(max_index, _max[i]);
}
return std::make_tuple(min_index, max_index, count);
}
} }
template<typename T> template<typename T>