ShuriZma
/
pcsx2
mirror of https://github.com/PCSX2/pcsx2.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

GregMiscellaneous: zzogl-pg: 

* Fix wrong integer size.
* Enable sse2 version on windows too.


git-svn-id: http://pcsx2.googlecode.com/svn/branches/GregMiscellaneous@3822 96395faa-99c1-11dd-bbfe-3dabce05a288
This commit is contained in:
gregory.hainaut@gmail.com 2010-09-23 11:24:02 +00:00
parent 6e1b11c2b5
commit 79c4b1825d
1 changed files with 7 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
plugins/zzogl-pg/opengl/targets.cpp
View File

@ -26,7 +26,6 @@
#include "zerogs.h" #include "zerogs.h"
#include "targets.h" #include "targets.h"
#include "ZZoglShaders.h" #include "ZZoglShaders.h"
#include <emmintrin.h>
#include <xmmintrin.h> #include <xmmintrin.h>
#define RHA #define RHA
@ -3046,28 +3045,22 @@ __forceinline void update_4pixels_sse2(u32* src, Tdst* basepage, u32 i_msk, u32
u32* base_ptr; u32* base_ptr;
__m128i pixels; __m128i pixels;
// highly slow. memory -> register -> memory -> xmm ... // Directly load pixels into xmm registers
// Intel SSE4.1 support an instruction to load memory to a part of xmm // NOTE: Intel SSE4.1 support an instruction to load 32bits memory to a part of an xmm registers
#if 0
src_tmp[0] = src[RW((j<<6)+INDEX)];
src_tmp[1] = src[RW((j<<6)+INDEX+1)];
src_tmp[2] = src[RW((j<<6)+INDEX+2)];
src_tmp[3] = src[RW((j<<6)+INDEX+3)];
#endif
if (AA.x == 2) { if (AA.x == 2) {
// Note: pixels (32bits) are stored like that: // Note: pixels (32bits) are stored like that:
// p0 p0 p0 p0 p1 p1 p1 p1 // p0 p0 p0 p0 p1 p1 p1 p1
// p2 p2 p2 p2 p3 p3 p3 p3 // p2 p2 p2 p2 p3 p3 p3 p3
base_ptr = &src[(((j<<6)+INDEX)<<2)]; base_ptr = &src[(((j<<6)+INDEX)<<2)];
__m128i pixel_low = _mm_cvtsi32_si128(*(u64*)(base_ptr+3)); __m128i pixel_low = _mm_movpi64_epi64(*(__m64*)(base_ptr+3));
__m128i pixel_high = _mm_cvtsi32_si128(*(u64*)(base_ptr+11)); __m128i pixel_high = _mm_movpi64_epi64(*(__m64*)(base_ptr+11));
pixels = _mm_unpacklo_epi64(pixel_low, pixel_high); pixels = _mm_unpacklo_epi64(pixel_low, pixel_high);
} else if(AA.x ==1) { } else if(AA.x ==1) {
// Note: pixels (32bits) are stored like that: // Note: pixels (32bits) are stored like that:
// p0 p0 p1 p1 p2 p2 p3 p3 // p0 p0 p1 p1 p2 p2 p3 p3
base_ptr = &src[(((j<<6)+INDEX)<<1)]; base_ptr = &src[(((j<<6)+INDEX)<<1)];
__m128i pixel_low = _mm_cvtsi32_si128(*(u64*)(base_ptr+1)); __m128i pixel_low = _mm_movpi64_epi64(*(__m64*)(base_ptr+1));
__m128i pixel_high = _mm_cvtsi32_si128(*(u64*)(base_ptr+5)); __m128i pixel_high = _mm_movpi64_epi64(*(__m64*)(base_ptr+5));
pixels = _mm_unpacklo_epi64(pixel_low, pixel_high); pixels = _mm_unpacklo_epi64(pixel_low, pixel_high);
} else { } else {
base_ptr = &src[((j<<6)+INDEX)]; base_ptr = &src[((j<<6)+INDEX)];
@ -3179,7 +3172,7 @@ void Resolve_32b(const void* psrc, int fbp, int fbw, int fbh, u32 fbm)
// for(int j = fbw_div-1; j >= 0; --j) { // for(int j = fbw_div-1; j >= 0; --j) {
for(u32 j = 0 ; j < fbw_div; ++j) { for(u32 j = 0 ; j < fbw_div; ++j) {
Tdst* basepage = pPageOffset + (i_div + j) * 2048; Tdst* basepage = pPageOffset + (i_div + j) * 2048;
#if defined(ZEROGS_SSE2) && defined(__LINUX__) // need to port asm to windows world #ifdef ZEROGS_SSE2
update_4pixels_sse2<size, pageTable, Tdst, do_conversion, 0>(src, basepage, i_msk, j, mask, imask); update_4pixels_sse2<size, pageTable, Tdst, do_conversion, 0>(src, basepage, i_msk, j, mask, imask);
update_4pixels_sse2<size, pageTable, Tdst, do_conversion, 4>(src, basepage, i_msk, j, mask, imask); update_4pixels_sse2<size, pageTable, Tdst, do_conversion, 4>(src, basepage, i_msk, j, mask, imask);
update_4pixels_sse2<size, pageTable, Tdst, do_conversion, 8>(src, basepage, i_msk, j, mask, imask); update_4pixels_sse2<size, pageTable, Tdst, do_conversion, 8>(src, basepage, i_msk, j, mask, imask);