template<class _mm> void TimgFilterNoiseMplayer::Tprocess::lineNoiseAvg_simd(uint8_t *dst,const uint8_t *src, int len, int8_t **shift_)
{
if (_mm::align && (intptr_t(src)&15 || intptr_t(dst)&15)) {
lineNoiseAvg_simd<typename _mm::T64>(dst,src,len,shift_);
return;
}
const int mmx_len=len&(~(_mm::size-1));
int8_t *shift2[3]= {shift_[0]+mmx_len, shift_[1]+mmx_len, shift_[2]+mmx_len};
for (int x=-mmx_len; x<0; x+=_mm::size) {
//".balign 16 \n\t"
typename _mm::__m mm0,mm1,mm2,mm3;
movdqu (mm1,shift2[0]+mmx_len+x);
movq (mm0,src+mmx_len+x);
typename _mm::__m shift1_8;
movVqu(shift1_8, shift2[1]+mmx_len+x);
paddb (mm1,shift1_8);
typename _mm::__m shift2_8;
movVqu(shift2_8, shift2[2]+mmx_len+x);
paddb (mm1,shift2_8);
movq (mm2,mm0);
movq (mm3,mm1);
punpcklbw (mm0,mm0);
punpckhbw (mm2,mm2);
punpcklbw (mm1,mm1);
punpckhbw (mm3,mm3);
pmulhw (mm1,mm0);
pmulhw (mm3,mm2);
paddw (mm1,mm1);
paddw (mm3,mm3);
paddw (mm1,mm0);
paddw (mm3,mm2);
psrlw (mm1,8);
psrlw (mm3,8);
packuswb (mm1,mm3);
movq (dst+mmx_len+x,mm1);
}
if (mmx_len!=len) {
lineNoiseAvg_C(dst+mmx_len, src+mmx_len, len-mmx_len, shift2);
}
}
void GSDrawScanlineCodeGenerator::lerp16(const Xmm& a, const Xmm& b, const Xmm& f, int shift)
{
#if _M_SSE >= 0x500
vpsubw(a, b);
modulate16(a, f, shift);
vpaddw(a, b);
#else
psubw(a, b);
modulate16(a, f, shift);
paddw(a, b);
#endif
}
void GSDrawScanlineCodeGenerator::lerp16_4(const Xmm& a, const Xmm& b, const Xmm& f)
{
#if _M_SSE >= 0x500
vpsubw(a, b);
vpmullw(a, f);
vpsraw(a, 4);
vpaddw(a, b);
#else
psubw(a, b);
pmullw(a, f);
psraw(a, 4);
paddw(a, b);
#endif
}
int main()
{
int rval;
mmx_t ma;
mmx_t mb;
movq_r2r(mm0, mm1);
rval = mmx_ok();
/* Announce return value of mmx_ok() */
// printf("Value returned from init was %x.", rval);
// printf(" (Indicates MMX %s available)\n\n",(rval)? "is" : "not");
// fflush(stdout); fflush(stderr);
// if(rval)
{
/* PADD *****************************************************/
ma.q = 0x1111111180000000LL;
mb.q = 0x7fffffff00000001LL;
paddd(ma, mb);
fprintf(stdout, "paddd: mb.q is %016llx\n", mb.q);
fprintf(stderr, "paddd: mb.q is 9111111080000001\n");
fflush(stdout); fflush(stderr);
ma.q = 0x0001000100010001LL;
mb.q = 0x80007fffffff0001LL;
paddw(ma, mb);
fprintf(stdout, "paddw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "paddw: mb.q is 8001800000000002\n");
fflush(stdout); fflush(stderr);
ma.q = 0x80007fffffff0001LL;
mb.q = 0x0001000100010000LL;
paddw(ma, mb);
fprintf(stdout, "paddw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "paddw: mb.q is 8001800000000001\n");
fflush(stdout); fflush(stderr);
ma.q = 0x01010101807fff01LL;
mb.q = 0x807fff0101010101LL;
paddb(ma, mb);
fprintf(stdout, "paddb: mb.q is %016llx\n", mb.q);
fprintf(stderr, "paddb: mb.q is 8180000281800002\n");
fflush(stdout); fflush(stderr);
/* PADDS ****************************************************/
ma.q = 0x0001000100010001LL;
mb.q = 0x80007fffffff0001LL;
paddsw(ma, mb);
fprintf(stdout, "paddsw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "paddsw: mb.q is 80017fff00000002\n");
ma.q = 0x80007fffffff0001LL;
mb.q = 0x0001000100010000LL;
paddsw(ma, mb);
fprintf(stdout, "paddsw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "paddsw: mb.q is 80017fff00000001\n");
ma.q = 0x01010101807fff01LL;
mb.q = 0x807fff0101010101LL;
paddsb(ma, mb);
fprintf(stdout, "paddsb: mb.q is %016llx\n", mb.q);
fprintf(stderr, "paddsb: mb.q is 817f0002817f0002\n");
fflush(stdout); fflush(stderr);
/* PADDUS ***************************************************/
ma.q = 0x0001000100010001LL;
mb.q = 0x80007fffffff0001LL;
paddusw(ma, mb);
fprintf(stdout, "paddusw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "paddusw: mb.q is 80018000ffff0002\n");
fflush(stdout); fflush(stderr);
ma.q = 0x80007fffffff0001LL;
mb.q = 0x0001000100010000LL;
paddusw(ma, mb);
fprintf(stdout, "paddusw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "paddusw: mb.q is 80018000ffff0001\n");
fflush(stdout); fflush(stderr);
ma.q = 0x01010101807fff01LL;
mb.q = 0x807fff0101010101LL;
paddusb(ma, mb);
fprintf(stdout, "paddusb: mb.q is %016llx\n", mb.q);
fprintf(stderr, "paddusb: mb.q is 8180ff028180ff02\n");
fflush(stdout); fflush(stderr);
/* PSUB *****************************************************/
ma.q = 0x7fffffff00000001LL;
mb.q = 0x1111111180000000LL;
psubd(ma, mb);
fprintf(stdout, "psubd: mb.q is %016llx\n", mb.q);
fprintf(stderr, "psubd: mb.q is 911111127fffffff\n");
fflush(stdout); fflush(stderr);
ma.q = 0x80007fffffff0001LL;
mb.q = 0x0001000100010001LL;
psubw(ma, mb);
fprintf(stdout, "psubw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "psubw: mb.q is 8001800200020000\n");
fflush(stdout); fflush(stderr);
ma.q = 0x0001000100010000LL;
mb.q = 0x80007fffffff0001LL;
psubw(ma, mb);
fprintf(stdout, "psubw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "psubw: mb.q is 7fff7ffefffe0001\n");
fflush(stdout); fflush(stderr);
ma.q = 0x807fff0101010101LL;
mb.q = 0x01010101807fff01LL;
psubb(ma, mb);
fprintf(stdout, "psubb: mb.q is %016llx\n", mb.q);
fprintf(stderr, "psubb: mb.q is 818202007f7efe00\n");
fflush(stdout); fflush(stderr);
/* PSUBS ****************************************************/
ma.q = 0x80007fffffff0001LL;
mb.q = 0x0001000100010001LL;
psubsw(ma, mb);
fprintf(stdout, "psubsw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "psubsw: mb.q is 7fff800200020000\n");
fflush(stdout); fflush(stderr);
ma.q = 0x0001000100010000LL;
mb.q = 0x80007fffffff0001LL;
psubsw(ma, mb);
fprintf(stdout, "psubsw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "psubsw: mb.q is 80007ffefffe0001\n");
fflush(stdout); fflush(stderr);
ma.q = 0x807fff0101010101LL;
mb.q = 0x01010101807fff01LL;
psubsb(ma, mb);
fprintf(stdout, "psubsb: mb.q is %016llx\n", mb.q);
fprintf(stderr, "psubsb: mb.q is 7f820200807efe00\n");
fflush(stdout); fflush(stderr);
/* PSUBUS ***************************************************/
ma.q = 0x80007fffffff0001LL;
mb.q = 0x0001000100010001LL;
psubusw(ma, mb);
fprintf(stdout, "psubusw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "psubusw: mb.q is 0000000000000000\n");
fflush(stdout); fflush(stderr);
ma.q = 0x0001000100010000LL;
mb.q = 0x80007fffffff0001LL;
psubusw(ma, mb);
fprintf(stdout, "psubusw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "psubusw: mb.q is 7fff7ffefffe0001\n");
fflush(stdout); fflush(stderr);
ma.q = 0x807fff0101010101LL;
mb.q = 0x01010101807fff01LL;
psubusb(ma, mb);
fprintf(stdout, "psubusb: mb.q is %016llx\n", mb.q);
fprintf(stderr, "psubusb: mb.q is 000000007f7efe00\n");
fflush(stdout); fflush(stderr);
/* PMUL *****************************************************/
ma.q = 0x8000ffff00ff0000LL;
mb.q = 0x0200ffff00ffffffLL;
pmulhw(ma, mb);
fprintf(stdout, "pmulhw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "pmulhw: mb.q is ff00000000000000\n");
fflush(stdout); fflush(stderr);
mb.q = 0x0200ffff00ffffffLL;
pmullw(ma, mb);
fprintf(stdout, "pmullw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "pmullw: mb.q is 00000001fe010000\n");
fflush(stdout); fflush(stderr);
/* PMADD ****************************************************/
ma.q = 0x8000345680007f34LL;
mb.q = 0x93234a27ffff1707LL;
pmaddwd(ma, mb);
fprintf(stdout, "pmaddwd: mb.q is %016llx\n", mb.q);
fprintf(stderr, "pmaddwd: mb.q is 4597551a0b71a66c\n");
fflush(stdout); fflush(stderr);
/* PCMPEQ ***************************************************/
ma.q = 0x800034568f237f34LL;
mb.q = 0x93009a568f237f34LL;
pcmpeqd(ma, mb);
fprintf(stdout, "pcmpeqd: mb.q is %016llx\n", mb.q);
fprintf(stderr, "pcmpeqd: mb.q is 00000000ffffffff\n");
fflush(stdout); fflush(stderr);
mb.q = 0x93009a568f237f34LL;
pcmpeqw(ma, mb);
fprintf(stdout, "pcmpeqw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "pcmpeqw: mb.q is 00000000ffffffff\n");
fflush(stdout); fflush(stderr);
mb.q = 0x93009a568f237f34LL;
pcmpeqb(ma, mb);
fprintf(stdout, "pcmpeqb: mb.q is %016llx\n", mb.q);
fprintf(stderr, "pcmpeqb: mb.q is 00ff00ffffffffff\n");
fflush(stdout); fflush(stderr);
/* PCMPGT ***************************************************/
ma.q = 0x666688884477aaffLL;
mb.q = 0x1234567890abcdefLL;
pcmpgtd(ma, mb);
fprintf(stdout, "pcmpgtd: mb.q is %016llx\n", mb.q);
fprintf(stderr, "pcmpgtd: mb.q is 0000000000000000\n");
fflush(stdout); fflush(stderr);
mb.q = 0x1234567890abcdefLL;
pcmpgtw(ma, mb);
fprintf(stdout, "pcmpgtw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "pcmpgtw: mb.q is 0000ffff0000ffff\n");
fflush(stdout); fflush(stderr);
mb.q = 0x1234567890abcdefLL;
pcmpgtb(ma, mb);
fprintf(stdout, "pcmpgtb: mb.q is %016llx\n", mb.q);
fprintf(stderr, "pcmpgtb: mb.q is 0000ffff0000ff00\n");
fflush(stdout); fflush(stderr);
/* PACKSS ***************************************************/
ma.q = 0x00012222000abbbbLL;
mb.q = 0x0000888800003333LL;
packssdw(ma, mb);
fprintf(stdout, "packssdw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "packssdw: mb.q is 7fff7fff7fff3333\n");
fflush(stdout); fflush(stderr);
ma.q = 0x00aa00dd01009999LL;
mb.q = 0x0011002200330044LL;
packsswb(ma, mb);
fprintf(stdout, "packsswb: mb.q is %016llx\n", mb.q);
fprintf(stderr, "packsswb: mb.q is 7f7f7f8011223344\n");
fflush(stdout); fflush(stderr);
/* PACKUS ***************************************************/
ma.q = 0x00aa00dd01009999LL;
mb.q = 0x0011002200330044LL;
packuswb(ma, mb);
fprintf(stdout, "packuswb: mb.q is %016llx\n", mb.q);
fprintf(stderr, "packuswb: mb.q is aaddff0011223344\n");
fflush(stdout); fflush(stderr);
/* PUNPCKH **************************************************/
ma.q = 0x090a0b0c0d0e0f00LL;
mb.q = 0x0102030405060708LL;
punpckhdq(ma, mb);
fprintf(stdout, "punpckhdq: mb.q is %016llx\n", mb.q);
fprintf(stderr, "punpckhdq: mb.q is 090a0b0c01020304\n");
fflush(stdout); fflush(stderr);
mb.q = 0x0102030405060708LL;
punpckhwd(ma, mb);
fprintf(stdout, "punpckhwd: mb.q is %016llx\n", mb.q);
fprintf(stderr, "punpckhwd: mb.q is 090a01020b0c0304\n");
fflush(stdout); fflush(stderr);
mb.q = 0x0102030405060708LL;
punpckhbw(ma, mb);
fprintf(stdout, "punpckhbw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "punpckhbw: mb.q is 09010a020b030c04\n");
fflush(stdout); fflush(stderr);
/* PUNPCKL **************************************************/
ma.q = 0x090a0b0c0d0e0f00LL;
mb.q = 0x0102030405060708LL;
punpckldq(ma, mb);
fprintf(stdout, "punpckldq: mb.q is %016llx\n", mb.q);
fprintf(stderr, "punpckldq: mb.q is 0d0e0f0005060708\n");
fflush(stdout); fflush(stderr);
mb.q = 0x0102030405060708LL;
punpcklwd(ma, mb);
fprintf(stdout, "punpcklwd: mb.q is %016llx\n", mb.q);
fprintf(stderr, "punpcklwd: mb.q is 0d0e05060f000708\n");
fflush(stdout); fflush(stderr);
mb.q = 0x0102030405060708LL;
punpcklbw(ma, mb);
fprintf(stdout, "punpcklbw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "punpcklbw: mb.q is 0d050e060f070008\n");
fflush(stdout); fflush(stderr);
/* PAND, PANDN, POR, PXOR ***********************************/
ma.q = 0x5555555555555555LL;
mb.q = 0x3333333333333333LL;
pand(ma, mb);
fprintf(stdout, "pand: mb.q is %016llx\n", mb.q);
fprintf(stderr, "pand: mb.q is 1111111111111111\n");
fflush(stdout); fflush(stderr);
mb.q = 0x3333333333333333LL;
pandn(ma, mb);
fprintf(stdout, "pandn: mb.q is %016llx\n", mb.q);
fprintf(stderr, "pandn: mb.q is 4444444444444444\n");
fflush(stdout); fflush(stderr);
mb.q = 0x3333333333333333LL;
por(ma, mb);
fprintf(stdout, "por: mb.q is %016llx\n", mb.q);
fprintf(stderr, "por: mb.q is 7777777777777777\n");
fflush(stdout); fflush(stderr);
mb.q = 0x3333333333333333LL;
pxor(ma, mb);
fprintf(stdout, "pxor: mb.q is %016llx\n", mb.q);
fprintf(stderr, "pxor: mb.q is 6666666666666666\n");
fflush(stdout); fflush(stderr);
/* PSLL *****************************************************/
ma.q = 0x0000000000000018LL;
mb.q = 0x0123456789abcdefLL;
psllq(ma, mb);
fprintf(stdout, "psllq: mb.q is %016llx\n", mb.q);
fprintf(stderr, "psllq: mb.q is 6789abcdef000000\n");
fflush(stdout); fflush(stderr);
mb.q = 0x0123456789abcdefLL;
pslld(ma, mb);
fprintf(stdout, "pslld: mb.q is %016llx\n", mb.q);
fprintf(stderr, "pslld: mb.q is 67000000ef000000\n");
fflush(stdout); fflush(stderr);
mb.q = 0x0123456789abcdefLL;
psllw(ma, mb);
fprintf(stdout, "psllw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "psllw: mb.q is 0000000000000000\n");
fflush(stdout); fflush(stderr);
/* PSRL *****************************************************/
ma.q = 0x0000000000000018LL;
mb.q = 0x0123456789abcdefLL;
psrlq(ma, mb);
fprintf(stdout, "psrlq: mb.q is %016llx\n", mb.q);
fprintf(stderr, "psrlq: mb.q is 0000000123456789\n");
fflush(stdout); fflush(stderr);
mb.q = 0x0123456789abcdefLL;
psrld(ma, mb);
fprintf(stdout, "psrld: mb.q is %016llx\n", mb.q);
fprintf(stderr, "psrld: mb.q is 0000000100000089\n");
fflush(stdout); fflush(stderr);
mb.q = 0x0123456789abcdefLL;
psrlw(ma, mb);
fprintf(stdout, "psrlw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "psrlw: mb.q is 0000000000000000\n");
fflush(stdout); fflush(stderr);
/* PSRA *****************************************************/
ma.q = 0x0000000000000018LL;
mb.q = 0x0123456789abcdefLL;
psrad(ma, mb);
fprintf(stdout, "psrad: mb.q is %016llx\n", mb.q);
fprintf(stderr, "psrad: mb.q is 00000001ffffff89\n");
fflush(stdout); fflush(stderr);
mb.q = 0x0123456789abcdefLL;
psraw(ma, mb);
fprintf(stdout, "psraw: mb.q is %016llx\n", mb.q);
fprintf(stderr, "psraw: mb.q is 00000000ffffffff\n");
fflush(stdout); fflush(stderr);
/* Exit MXX *************************************************/
emms();
}
/* Clean-up and exit nicely */
exit(0);
}
void GPUDrawScanlineCodeGenerator::lerp16(const Xmm& a, const Xmm& b, const Operand& f)
{
psubw(a, b);
modulate16<shift>(a, f);
paddw(a, b);
}
void GPUDrawScanlineCodeGenerator::SampleTexture()
{
if(!m_sel.tme)
{
return;
}
if(m_sel.tlu)
{
mov(edx, ptr[&m_local.gd->clut]);
}
// xmm2 = s
// xmm3 = t
// xmm7 = test
// xmm0, xmm4, xmm5, xmm6 = free
// xmm1 = used
if(m_sel.ltf)
{
// GSVector4i u = s.sub16(GSVector4i(0x00200020)); // - 0.125f
// GSVector4i v = t.sub16(GSVector4i(0x00200020)); // - 0.125f
mov(eax, 0x00200020);
movd(xmm0, eax);
pshufd(xmm0, xmm0, _MM_SHUFFLE(0, 0, 0, 0));
psubw(xmm2, xmm0);
psubw(xmm3, xmm0);
// GSVector4i uf = (u & GSVector4i::x00ff()) << 7;
// GSVector4i vf = (v & GSVector4i::x00ff()) << 7;
movdqa(xmm0, xmm2);
psllw(xmm0, 8);
psrlw(xmm0, 1);
movdqa(ptr[&m_local.temp.uf], xmm0);
if(!m_sel.sprite)
{
movdqa(xmm0, xmm3);
psllw(xmm0, 8);
psrlw(xmm0, 1);
movdqa(ptr[&m_local.temp.vf], xmm0);
}
}
// GSVector4i u0 = s.srl16(8);
// GSVector4i v0 = t.srl16(8);
psrlw(xmm2, 8);
psrlw(xmm3, 8);
// xmm2 = u
// xmm3 = v
// xmm7 = test
// xmm0, xmm4, xmm5, xmm6 = free
// xmm1 = used
if(m_sel.ltf)
{
// GSVector4i u1 = u0.add16(GSVector4i::x0001());
// GSVector4i v1 = v0.add16(GSVector4i::x0001());
movdqa(xmm4, xmm2);
movdqa(xmm5, xmm3);
pcmpeqd(xmm0, xmm0);
psrlw(xmm0, 15);
paddw(xmm4, xmm0);
paddw(xmm5, xmm0);
if(m_sel.twin)
{
// u0 = (u0 & m_local.twin[0].u).add16(m_local.twin[1].u);
// v0 = (v0 & m_local.twin[0].v).add16(m_local.twin[1].v);
// u1 = (u1 & m_local.twin[0].u).add16(m_local.twin[1].u);
// v1 = (v1 & m_local.twin[0].v).add16(m_local.twin[1].v);
movdqa(xmm0, ptr[&m_local.twin[0].u]);
movdqa(xmm6, ptr[&m_local.twin[1].u]);
pand(xmm2, xmm0);
paddw(xmm2, xmm6);
pand(xmm4, xmm0);
paddw(xmm4, xmm6);
movdqa(xmm0, ptr[&m_local.twin[0].v]);
movdqa(xmm6, ptr[&m_local.twin[1].v]);
pand(xmm3, xmm0);
paddw(xmm3, xmm6);
pand(xmm5, xmm0);
paddw(xmm5, xmm6);
}
else
{
// u0 = u0.min_i16(m_local.twin[2].u);
// v0 = v0.min_i16(m_local.twin[2].v);
// u1 = u1.min_i16(m_local.twin[2].u);
// v1 = v1.min_i16(m_local.twin[2].v);
// TODO: if(!sprite) clamp16 else:
movdqa(xmm0, ptr[&m_local.twin[2].u]);
movdqa(xmm6, ptr[&m_local.twin[2].v]);
pminsw(xmm2, xmm0);
pminsw(xmm3, xmm6);
pminsw(xmm4, xmm0);
pminsw(xmm5, xmm6);
}
// xmm2 = u0
// xmm3 = v0
// xmm4 = u1
// xmm5 = v1
// xmm7 = test
// xmm0, xmm6 = free
// xmm1 = used
// GSVector4i addr00 = v0.sll16(8) | u0;
// GSVector4i addr01 = v0.sll16(8) | u1;
// GSVector4i addr10 = v1.sll16(8) | u0;
// GSVector4i addr11 = v1.sll16(8) | u1;
psllw(xmm3, 8);
movdqa(xmm0, xmm3);
por(xmm3, xmm2);
por(xmm0, xmm4);
psllw(xmm5, 8);
movdqa(xmm6, xmm5);
por(xmm5, xmm2);
por(xmm6, xmm4);
// xmm3 = addr00
// xmm0 = addr01
// xmm5 = addr10
// xmm6 = addr11
// xmm7 = test
// xmm2, xmm4 = free
// xmm1 = used
ReadTexel(xmm2, xmm3);
ReadTexel(xmm4, xmm0);
ReadTexel(xmm3, xmm5);
ReadTexel(xmm5, xmm6);
// xmm2 = c00
// xmm4 = c01
// xmm3 = c10
// xmm5 = c11
// xmm7 = test
// xmm0, xmm6 = free
// xmm1 = used
// spill (TODO)
movdqa(ptr[&m_local.temp.fd], xmm1);
movdqa(ptr[&m_local.temp.test], xmm7);
// xmm2 = c00
// xmm4 = c01
// xmm3 = c10
// xmm5 = c11
// xmm0, xmm1, xmm6, xmm7 = free
movdqa(xmm1, xmm2);
psllw(xmm1, 11);
psrlw(xmm1, 8);
movdqa(xmm0, xmm4);
psllw(xmm0, 11);
psrlw(xmm0, 8);
lerp16<0>(xmm0, xmm1, ptr[&m_local.temp.uf]);
movdqa(xmm6, xmm2);
psllw(xmm6, 6);
psrlw(xmm6, 11);
psllw(xmm6, 3);
movdqa(xmm1, xmm4);
psllw(xmm1, 6);
psrlw(xmm1, 11);
psllw(xmm1, 3);
lerp16<0>(xmm1, xmm6, ptr[&m_local.temp.uf]);
movdqa(xmm7, xmm2);
psllw(xmm7, 1);
psrlw(xmm7, 11);
psllw(xmm7, 3);
movdqa(xmm6, xmm4);
psllw(xmm6, 1);
psrlw(xmm6, 11);
psllw(xmm6, 3);
lerp16<0>(xmm6, xmm7, ptr[&m_local.temp.uf]);
psraw(xmm2, 15);
psrlw(xmm2, 8);
psraw(xmm4, 15);
psrlw(xmm4, 8);
lerp16<0>(xmm4, xmm2, ptr[&m_local.temp.uf]);
// xmm0 = r00
// xmm1 = g00
// xmm6 = b00
// xmm4 = a00
// xmm3 = c10
// xmm5 = c11
// xmm2, xmm7 = free
movdqa(xmm7, xmm3);
psllw(xmm7, 11);
psrlw(xmm7, 8);
movdqa(xmm2, xmm5);
psllw(xmm2, 11);
psrlw(xmm2, 8);
lerp16<0>(xmm2, xmm7, ptr[&m_local.temp.uf]);
lerp16<0>(xmm2, xmm0, ptr[&m_local.temp.vf]);
// xmm2 = r
// xmm1 = g00
// xmm6 = b00
// xmm4 = a00
// xmm3 = c10
// xmm5 = c11
// xmm0, xmm7 = free
movdqa(xmm7, xmm3);
psllw(xmm7, 6);
psrlw(xmm7, 11);
psllw(xmm7, 3);
movdqa(xmm0, xmm5);
psllw(xmm0, 6);
psrlw(xmm0, 11);
psllw(xmm0, 3);
lerp16<0>(xmm0, xmm7, ptr[&m_local.temp.uf]);
lerp16<0>(xmm0, xmm1, ptr[&m_local.temp.vf]);
// xmm2 = r
// xmm0 = g
// xmm6 = b00
// xmm4 = a00
// xmm3 = c10
// xmm5 = c11
// xmm1, xmm7 = free
movdqa(xmm7, xmm3);
psllw(xmm7, 1);
psrlw(xmm7, 11);
psllw(xmm7, 3);
movdqa(xmm1, xmm5);
psllw(xmm1, 1);
psrlw(xmm1, 11);
psllw(xmm1, 3);
lerp16<0>(xmm1, xmm7, ptr[&m_local.temp.uf]);
lerp16<0>(xmm1, xmm6, ptr[&m_local.temp.vf]);
// xmm2 = r
// xmm0 = g
// xmm1 = b
// xmm4 = a00
// xmm3 = c10
// xmm5 = c11
// xmm6, xmm7 = free
psraw(xmm3, 15);
psrlw(xmm3, 8);
psraw(xmm5, 15);
psrlw(xmm5, 8);
lerp16<0>(xmm5, xmm3, ptr[&m_local.temp.uf]);
lerp16<0>(xmm5, xmm4, ptr[&m_local.temp.vf]);
// xmm2 = r
// xmm0 = g
// xmm1 = b
// xmm5 = a
// xmm3, xmm4, xmm6, xmm7 = free
// TODO
movdqa(xmm3, xmm5); // a
movdqa(xmm4, xmm2); // r
movdqa(xmm6, xmm1); // b
movdqa(xmm5, xmm0); // g
// reload test
movdqa(xmm7, ptr[&m_local.temp.test]);
// xmm4 = r
// xmm5 = g
// xmm6 = b
// xmm3 = a
// xmm7 = test
// xmm0, xmm1, xmm2 = free
// test |= (c[0] | c[1] | c[2] | c[3]).eq16(GSVector4i::zero()); // mask out blank pixels (not perfect)
movdqa(xmm1, xmm3);
por(xmm1, xmm4);
movdqa(xmm2, xmm5);
por(xmm2, xmm6);
por(xmm1, xmm2);
pxor(xmm0, xmm0);
pcmpeqw(xmm1, xmm0);
por(xmm7, xmm1);
// a = a.gt16(GSVector4i::zero());
pcmpgtw(xmm3, xmm0);
// reload fd
movdqa(xmm1, ptr[&m_local.temp.fd]);
}
else
{
if(m_sel.twin)
{
// u = (u & m_local.twin[0].u).add16(m_local.twin[1].u);
// v = (v & m_local.twin[0].v).add16(m_local.twin[1].v);
pand(xmm2, ptr[&m_local.twin[0].u]);
paddw(xmm2, ptr[&m_local.twin[1].u]);
pand(xmm3, ptr[&m_local.twin[0].v]);
paddw(xmm3, ptr[&m_local.twin[1].v]);
}
else
{
// u = u.min_i16(m_local.twin[2].u);
// v = v.min_i16(m_local.twin[2].v);
// TODO: if(!sprite) clamp16 else:
pminsw(xmm2, ptr[&m_local.twin[2].u]);
pminsw(xmm3, ptr[&m_local.twin[2].v]);
}
// xmm2 = u
// xmm3 = v
// xmm7 = test
// xmm0, xmm4, xmm5, xmm6 = free
// xmm1 = used
// GSVector4i addr = v.sll16(8) | u;
psllw(xmm3, 8);
por(xmm3, xmm2);
// xmm3 = addr
// xmm7 = test
// xmm0, xmm2, xmm4, xmm5, xmm6 = free
// xmm1 = used
ReadTexel(xmm6, xmm3);
// xmm3 = c00
// xmm7 = test
// xmm0, xmm2, xmm4, xmm5, xmm6 = free
// xmm1 = used
// test |= c00.eq16(GSVector4i::zero()); // mask out blank pixels
pxor(xmm0, xmm0);
pcmpeqw(xmm0, xmm6);
por(xmm7, xmm0);
// c[0] = (c00 << 3) & 0x00f800f8;
// c[1] = (c00 >> 2) & 0x00f800f8;
// c[2] = (c00 >> 7) & 0x00f800f8;
// c[3] = c00.sra16(15);
movdqa(xmm3, xmm6);
psraw(xmm3, 15); // a
pcmpeqd(xmm0, xmm0);
psrlw(xmm0, 11);
psllw(xmm0, 3); // 0x00f8
movdqa(xmm4, xmm6);
psllw(xmm4, 3);
pand(xmm4, xmm0); // r
movdqa(xmm5, xmm6);
psrlw(xmm5, 2);
pand(xmm5, xmm0); // g
psrlw(xmm6, 7);
pand(xmm6, xmm0); // b
}
}
void GPUDrawScanlineCodeGenerator::Step()
{
// steps -= 8;
sub(ecx, 8);
// fb += 8;
add(edi, 8 * sizeof(uint16));
if(m_sel.tme)
{
// GSVector4i st = m_local.d8.st;
movdqa(xmm4, ptr[&m_local.d8.st]);
// s = s.add16(st.xxxx());
// t = t.add16(st.yyyy());
pshufd(xmm2, xmm4, _MM_SHUFFLE(0, 0, 0, 0));
paddw(xmm2, ptr[&m_local.temp.s]);
movdqa(ptr[&m_local.temp.s], xmm2);
// TODO: if(!sprite) ... else reload t
pshufd(xmm3, xmm4, _MM_SHUFFLE(1, 1, 1, 1));
paddw(xmm3, ptr[&m_local.temp.t]);
movdqa(ptr[&m_local.temp.t], xmm3);
}
if(m_sel.tfx != 3) // != decal
{
if(m_sel.iip)
{
// GSVector4i c = m_local.d8.c;
// r = r.add16(c.xxxx());
// g = g.add16(c.yyyy());
// b = b.add16(c.zzzz());
movdqa(xmm6, ptr[&m_local.d8.c]);
pshufd(xmm4, xmm6, _MM_SHUFFLE(0, 0, 0, 0));
pshufd(xmm5, xmm6, _MM_SHUFFLE(1, 1, 1, 1));
pshufd(xmm6, xmm6, _MM_SHUFFLE(2, 2, 2, 2));
paddw(xmm4, ptr[&m_local.temp.r]);
paddw(xmm5, ptr[&m_local.temp.g]);
paddw(xmm6, ptr[&m_local.temp.b]);
movdqa(ptr[&m_local.temp.r], xmm4);
movdqa(ptr[&m_local.temp.g], xmm5);
movdqa(ptr[&m_local.temp.b], xmm6);
}
else
{
movdqa(xmm4, ptr[&m_local.temp.r]);
movdqa(xmm5, ptr[&m_local.temp.g]);
movdqa(xmm6, ptr[&m_local.temp.b]);
}
}
}
void GPUDrawScanlineCodeGenerator::Init()
{
mov(eax, dword[esp + _top]);
// uint16* fb = (uint16*)m_global.vm + (top << (10 + sel.scalex)) + left;
mov(edi, eax);
shl(edi, 10 + m_sel.scalex);
add(edi, edx);
lea(edi, ptr[edi * 2 + (size_t)m_local.gd->vm]);
// int steps = pixels - 8;
sub(ecx, 8);
if(m_sel.dtd)
{
// dither = GSVector4i::load<false>(&m_dither[top & 3][left & 3]);
and(eax, 3);
shl(eax, 5);
and(edx, 3);
shl(edx, 1);
movdqu(xmm0, ptr[eax + edx + (size_t)m_dither]);
movdqa(ptr[&m_local.temp.dither], xmm0);
}
mov(edx, dword[esp + _v]);
if(m_sel.tme)
{
mov(esi, dword[&m_local.gd->tex]);
// GSVector4i vt = GSVector4i(v.t).xxzzl();
cvttps2dq(xmm4, ptr[edx + offsetof(GSVertexSW, t)]);
pshuflw(xmm4, xmm4, _MM_SHUFFLE(2, 2, 0, 0));
// s = vt.xxxx().add16(m_local.d.s);
// t = vt.yyyy().add16(m_local.d.t);
pshufd(xmm2, xmm4, _MM_SHUFFLE(0, 0, 0, 0));
pshufd(xmm3, xmm4, _MM_SHUFFLE(1, 1, 1, 1));
paddw(xmm2, ptr[&m_local.d.s]);
if(!m_sel.sprite)
{
paddw(xmm3, ptr[&m_local.d.t]);
}
else
{
if(m_sel.ltf)
{
movdqa(xmm0, xmm3);
psllw(xmm0, 8);
psrlw(xmm0, 1);
movdqa(ptr[&m_local.temp.vf], xmm0);
}
}
movdqa(ptr[&m_local.temp.s], xmm2);
movdqa(ptr[&m_local.temp.t], xmm3);
}
if(m_sel.tfx != 3) // != decal
{
// GSVector4i vc = GSVector4i(v.c).xxzzlh();
cvttps2dq(xmm6, ptr[edx + offsetof(GSVertexSW, c)]);
pshuflw(xmm6, xmm6, _MM_SHUFFLE(2, 2, 0, 0));
pshufhw(xmm6, xmm6, _MM_SHUFFLE(2, 2, 0, 0));
// r = vc.xxxx();
// g = vc.yyyy();
// b = vc.zzzz();
pshufd(xmm4, xmm6, _MM_SHUFFLE(0, 0, 0, 0));
pshufd(xmm5, xmm6, _MM_SHUFFLE(1, 1, 1, 1));
pshufd(xmm6, xmm6, _MM_SHUFFLE(2, 2, 2, 2));
if(m_sel.iip)
{
// r = r.add16(m_local.d.r);
// g = g.add16(m_local.d.g);
// b = b.add16(m_local.d.b);
paddw(xmm4, ptr[&m_local.d.r]);
paddw(xmm5, ptr[&m_local.d.g]);
paddw(xmm6, ptr[&m_local.d.b]);
}
movdqa(ptr[&m_local.temp.r], xmm4);
movdqa(ptr[&m_local.temp.g], xmm5);
movdqa(ptr[&m_local.temp.b], xmm6);
}
}
void GSDrawScanlineCodeGenerator::SampleTexture()
{
if(!m_sel.fb || m_sel.tfx == TFX_NONE)
{
return;
}
mov(ebx, dword[&m_env.tex]);
// ebx = tex
if(!m_sel.fst)
{
// TODO: move these into Init/Step too?
cvttps2dq(xmm2, xmm2);
cvttps2dq(xmm3, xmm3);
if(m_sel.ltf)
{
// u -= 0x8000;
// v -= 0x8000;
mov(eax, 0x8000);
movd(xmm4, eax);
pshufd(xmm4, xmm4, _MM_SHUFFLE(0, 0, 0, 0));
psubd(xmm2, xmm4);
psubd(xmm3, xmm4);
}
}
// xmm2 = u
// xmm3 = v
if(m_sel.ltf)
{
// GSVector4i uf = u.xxzzlh().srl16(1);
movdqa(xmm0, xmm2);
pshuflw(xmm0, xmm0, _MM_SHUFFLE(2, 2, 0, 0));
pshufhw(xmm0, xmm0, _MM_SHUFFLE(2, 2, 0, 0));
psrlw(xmm0, 1);
movdqa(xmmword[&m_env.temp.uf], xmm0);
if(!m_sel.sprite)
{
// GSVector4i vf = v.xxzzlh().srl16(1);
movdqa(xmm1, xmm3);
pshuflw(xmm1, xmm1, _MM_SHUFFLE(2, 2, 0, 0));
pshufhw(xmm1, xmm1, _MM_SHUFFLE(2, 2, 0, 0));
psrlw(xmm1, 1);
movdqa(xmmword[&m_env.temp.vf], xmm1);
}
}
// GSVector4i uv0 = u.sra32(16).ps32(v.sra32(16));
psrad(xmm2, 16);
psrad(xmm3, 16);
packssdw(xmm2, xmm3);
if(m_sel.ltf)
{
// GSVector4i uv1 = uv0.add16(GSVector4i::x0001());
movdqa(xmm3, xmm2);
pcmpeqd(xmm1, xmm1);
psrlw(xmm1, 15);
paddw(xmm3, xmm1);
// uv0 = Wrap(uv0);
// uv1 = Wrap(uv1);
Wrap(xmm2, xmm3);
}
else
{
// uv0 = Wrap(uv0);
Wrap(xmm2);
}
// xmm2 = uv0
// xmm3 = uv1 (ltf)
// xmm0, xmm1, xmm4, xmm5, xmm6 = free
// xmm7 = used
// GSVector4i y0 = uv0.uph16() << tw;
// GSVector4i x0 = uv0.upl16();
pxor(xmm0, xmm0);
movd(xmm1, ptr[&m_env.tw]);
movdqa(xmm4, xmm2);
punpckhwd(xmm2, xmm0);
punpcklwd(xmm4, xmm0);
pslld(xmm2, xmm1);
// xmm0 = 0
// xmm1 = tw
// xmm2 = y0
// xmm3 = uv1 (ltf)
// xmm4 = x0
// xmm5, xmm6 = free
// xmm7 = used
if(m_sel.ltf)
{
// GSVector4i y1 = uv1.uph16() << tw;
// GSVector4i x1 = uv1.upl16();
movdqa(xmm6, xmm3);
punpckhwd(xmm3, xmm0);
punpcklwd(xmm6, xmm0);
pslld(xmm3, xmm1);
// xmm2 = y0
// xmm3 = y1
// xmm4 = x0
// xmm6 = x1
// xmm0, xmm5, xmm6 = free
// xmm7 = used
// GSVector4i addr00 = y0 + x0;
// GSVector4i addr01 = y0 + x1;
// GSVector4i addr10 = y1 + x0;
// GSVector4i addr11 = y1 + x1;
movdqa(xmm5, xmm2);
paddd(xmm5, xmm4);
paddd(xmm2, xmm6);
movdqa(xmm0, xmm3);
paddd(xmm0, xmm4);
paddd(xmm3, xmm6);
// xmm5 = addr00
// xmm2 = addr01
// xmm0 = addr10
// xmm3 = addr11
// xmm1, xmm4, xmm6 = free
// xmm7 = used
// c00 = addr00.gather32_32((const uint32/uint8*)tex[, clut]);
// c01 = addr01.gather32_32((const uint32/uint8*)tex[, clut]);
// c10 = addr10.gather32_32((const uint32/uint8*)tex[, clut]);
// c11 = addr11.gather32_32((const uint32/uint8*)tex[, clut]);
ReadTexel(xmm6, xmm5, xmm1, xmm4);
// xmm2, xmm5, xmm1 = free
ReadTexel(xmm4, xmm2, xmm5, xmm1);
// xmm0, xmm2, xmm5 = free
ReadTexel(xmm1, xmm0, xmm2, xmm5);
// xmm3, xmm0, xmm2 = free
ReadTexel(xmm5, xmm3, xmm0, xmm2);
// xmm6 = c00
// xmm4 = c01
// xmm1 = c10
// xmm5 = c11
// xmm0, xmm2, xmm3 = free
// xmm7 = used
movdqa(xmm0, xmmword[&m_env.temp.uf]);
// GSVector4i rb00 = c00 & mask;
// GSVector4i ga00 = (c00 >> 8) & mask;
movdqa(xmm2, xmm6);
psllw(xmm2, 8);
psrlw(xmm2, 8);
psrlw(xmm6, 8);
// GSVector4i rb01 = c01 & mask;
// GSVector4i ga01 = (c01 >> 8) & mask;
movdqa(xmm3, xmm4);
psllw(xmm3, 8);
psrlw(xmm3, 8);
psrlw(xmm4, 8);
// xmm0 = uf
// xmm2 = rb00
// xmm3 = rb01
// xmm6 = ga00
// xmm4 = ga01
// xmm1 = c10
// xmm5 = c11
// xmm7 = used
// rb00 = rb00.lerp16<0>(rb01, uf);
// ga00 = ga00.lerp16<0>(ga01, uf);
lerp16<0>(xmm3, xmm2, xmm0);
lerp16<0>(xmm4, xmm6, xmm0);
// xmm0 = uf
// xmm3 = rb00
// xmm4 = ga00
// xmm1 = c10
// xmm5 = c11
// xmm2, xmm6 = free
// xmm7 = used
// GSVector4i rb10 = c10 & mask;
// GSVector4i ga10 = (c10 >> 8) & mask;
movdqa(xmm2, xmm1);
psllw(xmm1, 8);
psrlw(xmm1, 8);
psrlw(xmm2, 8);
// GSVector4i rb11 = c11 & mask;
// GSVector4i ga11 = (c11 >> 8) & mask;
movdqa(xmm6, xmm5);
psllw(xmm5, 8);
psrlw(xmm5, 8);
psrlw(xmm6, 8);
// xmm0 = uf
// xmm3 = rb00
// xmm4 = ga00
// xmm1 = rb10
// xmm5 = rb11
// xmm2 = ga10
// xmm6 = ga11
// xmm7 = used
// rb10 = rb10.lerp16<0>(rb11, uf);
// ga10 = ga10.lerp16<0>(ga11, uf);
lerp16<0>(xmm5, xmm1, xmm0);
lerp16<0>(xmm6, xmm2, xmm0);
// xmm3 = rb00
// xmm4 = ga00
// xmm5 = rb10
// xmm6 = ga10
// xmm0, xmm1, xmm2 = free
// xmm7 = used
// rb00 = rb00.lerp16<0>(rb10, vf);
// ga00 = ga00.lerp16<0>(ga10, vf);
movdqa(xmm0, xmmword[&m_env.temp.vf]);
lerp16<0>(xmm5, xmm3, xmm0);
lerp16<0>(xmm6, xmm4, xmm0);
}
else
{
// GSVector4i addr00 = y0 + x0;
paddd(xmm2, xmm4);
// c00 = addr00.gather32_32((const uint32/uint8*)tex[, clut]);
ReadTexel(xmm5, xmm2, xmm0, xmm1);
// GSVector4i mask = GSVector4i::x00ff();
// c[0] = c00 & mask;
// c[1] = (c00 >> 8) & mask;
movdqa(xmm6, xmm5);
psllw(xmm5, 8);
psrlw(xmm5, 8);
psrlw(xmm6, 8);
}
}
void GSDrawScanlineCodeGenerator::Step()
{
// steps -= 4;
sub(ecx, 4);
// fza_offset++;
add(edi, 8);
if(!m_sel.sprite)
{
// z += m_env.d4.z;
if(m_sel.zb)
{
movaps(xmm0, xmmword[&m_env.temp.z]);
addps(xmm0, xmmword[&m_env.d4.z]);
movaps(xmmword[&m_env.temp.z], xmm0);
}
// f = f.add16(m_env.d4.f);
if(m_sel.fwrite && m_sel.fge)
{
movdqa(xmm1, xmmword[&m_env.temp.f]);
paddw(xmm1, xmmword[&m_env.d4.f]);
movdqa(xmmword[&m_env.temp.f], xmm1);
}
}
else
{
if(m_sel.ztest)
{
movdqa(xmm0, xmmword[&m_env.p.z]);
}
}
if(m_sel.fb)
{
if(m_sel.tfx != TFX_NONE)
{
if(m_sel.fst)
{
// GSVector4i st = m_env.d4.st;
// si += st.xxxx();
// if(!sprite) ti += st.yyyy();
movdqa(xmm4, xmmword[&m_env.d4.st]);
pshufd(xmm2, xmm4, _MM_SHUFFLE(0, 0, 0, 0));
paddd(xmm2, xmmword[&m_env.temp.s]);
movdqa(xmmword[&m_env.temp.s], xmm2);
if(!m_sel.sprite)
{
pshufd(xmm3, xmm4, _MM_SHUFFLE(1, 1, 1, 1));
paddd(xmm3, xmmword[&m_env.temp.t]);
movdqa(xmmword[&m_env.temp.t], xmm3);
}
else
{
movdqa(xmm3, xmmword[&m_env.temp.t]);
}
}
else
{
// GSVector4 stq = m_env.d4.stq;
// s += stq.xxxx();
// t += stq.yyyy();
// q += stq.zzzz();
movaps(xmm2, xmmword[&m_env.d4.stq]);
movaps(xmm3, xmm2);
movaps(xmm4, xmm2);
shufps(xmm2, xmm2, _MM_SHUFFLE(0, 0, 0, 0));
shufps(xmm3, xmm3, _MM_SHUFFLE(1, 1, 1, 1));
shufps(xmm4, xmm4, _MM_SHUFFLE(2, 2, 2, 2));
addps(xmm2, xmmword[&m_env.temp.s]);
addps(xmm3, xmmword[&m_env.temp.t]);
addps(xmm4, xmmword[&m_env.temp.q]);
movaps(xmmword[&m_env.temp.s], xmm2);
movaps(xmmword[&m_env.temp.t], xmm3);
movaps(xmmword[&m_env.temp.q], xmm4);
rcpps(xmm4, xmm4);
mulps(xmm2, xmm4);
mulps(xmm3, xmm4);
}
}
if(!(m_sel.tfx == TFX_DECAL && m_sel.tcc))
{
if(m_sel.iip)
{
// GSVector4i c = m_env.d4.c;
// rb = rb.add16(c.xxxx());
// ga = ga.add16(c.yyyy());
movdqa(xmm7, xmmword[&m_env.d4.c]);
pshufd(xmm5, xmm7, _MM_SHUFFLE(0, 0, 0, 0));
pshufd(xmm6, xmm7, _MM_SHUFFLE(1, 1, 1, 1));
paddw(xmm5, xmmword[&m_env.temp.rb]);
paddw(xmm6, xmmword[&m_env.temp.ga]);
movdqa(xmmword[&m_env.temp.rb], xmm5);
movdqa(xmmword[&m_env.temp.ga], xmm6);
}
else
{
if(m_sel.tfx == TFX_NONE)
{
movdqa(xmm5, xmmword[&m_env.c.rb]);
movdqa(xmm6, xmmword[&m_env.c.ga]);
}
}
}
}
// test = m_test[7 + (steps & (steps >> 31))];
mov(edx, ecx);
sar(edx, 31);
and(edx, ecx);
shl(edx, 4);
movdqa(xmm7, xmmword[edx + (size_t)&m_test[7]]);
}
void GSDrawScanlineCodeGenerator::Init(int params)
{
const int _top = params + 4;
const int _v = params + 8;
// int skip = left & 3;
mov(ebx, edx);
and(edx, 3);
// left -= skip;
sub(ebx, edx);
// int steps = right - left - 4;
sub(ecx, ebx);
sub(ecx, 4);
// GSVector4i test = m_test[skip] | m_test[7 + (steps & (steps >> 31))];
shl(edx, 4);
movdqa(xmm7, xmmword[edx + (size_t)&m_test[0]]);
mov(eax, ecx);
sar(eax, 31);
and(eax, ecx);
shl(eax, 4);
por(xmm7, xmmword[eax + (size_t)&m_test[7]]);
// GSVector2i* fza_base = &m_env.fzbr[top];
mov(esi, dword[esp + _top]);
lea(esi, ptr[esi * 8]);
add(esi, dword[&m_env.fzbr]);
// GSVector2i* fza_offset = &m_env.fzbc[left >> 2];
lea(edi, ptr[ebx * 2]);
add(edi, dword[&m_env.fzbc]);
if(!m_sel.sprite && (m_sel.fwrite && m_sel.fge || m_sel.zb) || m_sel.fb && (m_sel.edge || m_sel.tfx != TFX_NONE || m_sel.iip))
{
// edx = &m_env.d[skip]
shl(edx, 4);
lea(edx, ptr[edx + (size_t)m_env.d]);
// ebx = &v
mov(ebx, dword[esp + _v]);
}
if(!m_sel.sprite)
{
if(m_sel.fwrite && m_sel.fge || m_sel.zb)
{
movaps(xmm0, xmmword[ebx + 16]); // v.p
if(m_sel.fwrite && m_sel.fge)
{
// f = GSVector4i(vp).zzzzh().zzzz().add16(m_env.d[skip].f);
cvttps2dq(xmm1, xmm0);
pshufhw(xmm1, xmm1, _MM_SHUFFLE(2, 2, 2, 2));
pshufd(xmm1, xmm1, _MM_SHUFFLE(2, 2, 2, 2));
paddw(xmm1, xmmword[edx + 16 * 6]);
movdqa(xmmword[&m_env.temp.f], xmm1);
}
if(m_sel.zb)
{
// z = vp.zzzz() + m_env.d[skip].z;
shufps(xmm0, xmm0, _MM_SHUFFLE(2, 2, 2, 2));
addps(xmm0, xmmword[edx]);
movaps(xmmword[&m_env.temp.z], xmm0);
}
}
}
else
{
if(m_sel.ztest)
{
movdqa(xmm0, xmmword[&m_env.p.z]);
}
}
if(m_sel.fb)
{
if(m_sel.edge || m_sel.tfx != TFX_NONE)
{
movaps(xmm4, xmmword[ebx + 32]); // v.t
}
if(m_sel.edge)
{
pshufhw(xmm3, xmm4, _MM_SHUFFLE(2, 2, 2, 2));
pshufd(xmm3, xmm3, _MM_SHUFFLE(3, 3, 3, 3));
psrlw(xmm3, 9);
movdqa(xmmword[&m_env.temp.cov], xmm3);
}
if(m_sel.tfx != TFX_NONE)
{
if(m_sel.fst)
{
// GSVector4i vti(vt);
cvttps2dq(xmm4, xmm4);
// si = vti.xxxx() + m_env.d[skip].si;
// ti = vti.yyyy(); if(!sprite) ti += m_env.d[skip].ti;
pshufd(xmm2, xmm4, _MM_SHUFFLE(0, 0, 0, 0));
pshufd(xmm3, xmm4, _MM_SHUFFLE(1, 1, 1, 1));
paddd(xmm2, xmmword[edx + 16 * 7]);
if(!m_sel.sprite)
{
paddd(xmm3, xmmword[edx + 16 * 8]);
}
else
{
if(m_sel.ltf)
{
movdqa(xmm4, xmm3);
pshuflw(xmm4, xmm4, _MM_SHUFFLE(2, 2, 0, 0));
pshufhw(xmm4, xmm4, _MM_SHUFFLE(2, 2, 0, 0));
psrlw(xmm4, 1);
movdqa(xmmword[&m_env.temp.vf], xmm4);
}
}
movdqa(xmmword[&m_env.temp.s], xmm2);
movdqa(xmmword[&m_env.temp.t], xmm3);
}
else
{
// s = vt.xxxx() + m_env.d[skip].s;
// t = vt.yyyy() + m_env.d[skip].t;
// q = vt.zzzz() + m_env.d[skip].q;
movaps(xmm2, xmm4);
movaps(xmm3, xmm4);
shufps(xmm2, xmm2, _MM_SHUFFLE(0, 0, 0, 0));
shufps(xmm3, xmm3, _MM_SHUFFLE(1, 1, 1, 1));
shufps(xmm4, xmm4, _MM_SHUFFLE(2, 2, 2, 2));
addps(xmm2, xmmword[edx + 16 * 1]);
addps(xmm3, xmmword[edx + 16 * 2]);
addps(xmm4, xmmword[edx + 16 * 3]);
movaps(xmmword[&m_env.temp.s], xmm2);
movaps(xmmword[&m_env.temp.t], xmm3);
movaps(xmmword[&m_env.temp.q], xmm4);
rcpps(xmm4, xmm4);
mulps(xmm2, xmm4);
mulps(xmm3, xmm4);
}
}
if(!(m_sel.tfx == TFX_DECAL && m_sel.tcc))
{
if(m_sel.iip)
{
// GSVector4i vc = GSVector4i(v.c);
cvttps2dq(xmm6, xmmword[ebx]); // v.c
// vc = vc.upl16(vc.zwxy());
pshufd(xmm5, xmm6, _MM_SHUFFLE(1, 0, 3, 2));
punpcklwd(xmm6, xmm5);
// rb = vc.xxxx().add16(m_env.d[skip].rb);
// ga = vc.zzzz().add16(m_env.d[skip].ga);
pshufd(xmm5, xmm6, _MM_SHUFFLE(0, 0, 0, 0));
pshufd(xmm6, xmm6, _MM_SHUFFLE(2, 2, 2, 2));
paddw(xmm5, xmmword[edx + 16 * 4]);
paddw(xmm6, xmmword[edx + 16 * 5]);
movdqa(xmmword[&m_env.temp.rb], xmm5);
movdqa(xmmword[&m_env.temp.ga], xmm6);
}
else
{
if(m_sel.tfx == TFX_NONE)
{
movdqa(xmm5, xmmword[&m_env.c.rb]);
movdqa(xmm6, xmmword[&m_env.c.ga]);
}
}
}
}
}
void GSDrawScanlineCodeGenerator::WriteFrame(int params)
{
const int _top = params + 4;
if(!m_sel.fwrite)
{
return;
}
if(m_sel.colclamp == 0)
{
// c[0] &= 0x000000ff;
// c[1] &= 0x000000ff;
pcmpeqd(xmm7, xmm7);
psrlw(xmm7, 8);
pand(xmm5, xmm7);
pand(xmm6, xmm7);
}
if(m_sel.fpsm == 2 && m_sel.dthe)
{
mov(eax, dword[esp + _top]);
and(eax, 3);
shl(eax, 5);
paddw(xmm5, xmmword[eax + (size_t)&m_env.dimx[0]]);
paddw(xmm6, xmmword[eax + (size_t)&m_env.dimx[1]]);
}
// GSVector4i fs = c[0].upl16(c[1]).pu16(c[0].uph16(c[1]));
movdqa(xmm7, xmm5);
punpcklwd(xmm5, xmm6);
punpckhwd(xmm7, xmm6);
packuswb(xmm5, xmm7);
if(m_sel.fba && m_sel.fpsm != 1)
{
// fs |= 0x80000000;
pcmpeqd(xmm7, xmm7);
pslld(xmm7, 31);
por(xmm5, xmm7);
}
if(m_sel.fpsm == 2)
{
// GSVector4i rb = fs & 0x00f800f8;
// GSVector4i ga = fs & 0x8000f800;
mov(eax, 0x00f800f8);
movd(xmm6, eax);
pshufd(xmm6, xmm6, _MM_SHUFFLE(0, 0, 0, 0));
mov(eax, 0x8000f800);
movd(xmm7, eax);
pshufd(xmm7, xmm7, _MM_SHUFFLE(0, 0, 0, 0));
movdqa(xmm4, xmm5);
pand(xmm4, xmm6);
pand(xmm5, xmm7);
// fs = (ga >> 16) | (rb >> 9) | (ga >> 6) | (rb >> 3);
movdqa(xmm6, xmm4);
movdqa(xmm7, xmm5);
psrld(xmm4, 3);
psrld(xmm6, 9);
psrld(xmm5, 6);
psrld(xmm7, 16);
por(xmm5, xmm4);
por(xmm7, xmm6);
por(xmm5, xmm7);
}
if(m_sel.rfb)
{
// fs = fs.blend(fd, fm);
blend(xmm5, xmm2, xmm3); // TODO: could be skipped in certain cases, depending on fpsm and fm
}
bool fast = m_sel.rfb && m_sel.fpsm < 2;
WritePixel(xmm5, xmm0, ebx, dl, fast, m_sel.fpsm);
}
void GSDrawScanlineCodeGenerator::AlphaBlend()
{
if(!m_sel.fwrite)
{
return;
}
if(m_sel.abe == 0 && m_sel.aa1 == 0)
{
return;
}
if((m_sel.aba != m_sel.abb) && (m_sel.aba == 1 || m_sel.abb == 1 || m_sel.abc == 1) || m_sel.abd == 1)
{
switch(m_sel.fpsm)
{
case 0:
case 1:
// c[2] = fd & mask;
// c[3] = (fd >> 8) & mask;
movdqa(xmm0, xmm2);
movdqa(xmm1, xmm2);
psllw(xmm0, 8);
psrlw(xmm0, 8);
psrlw(xmm1, 8);
break;
case 2:
// c[2] = ((fd & 0x7c00) << 9) | ((fd & 0x001f) << 3);
// c[3] = ((fd & 0x8000) << 8) | ((fd & 0x03e0) >> 2);
movdqa(xmm0, xmm2);
movdqa(xmm1, xmm2);
movdqa(xmm4, xmm2);
pcmpeqd(xmm7, xmm7);
psrld(xmm7, 27); // 0x0000001f
pand(xmm0, xmm7);
pslld(xmm0, 3);
pslld(xmm7, 10); // 0x00007c00
pand(xmm4, xmm7);
pslld(xmm4, 9);
por(xmm0, xmm4);
movdqa(xmm4, xmm1);
psrld(xmm7, 5); // 0x000003e0
pand(xmm1, xmm7);
psrld(xmm1, 2);
psllw(xmm7, 10); // 0x00008000
pand(xmm4, xmm7);
pslld(xmm4, 8);
por(xmm1, xmm4);
break;
}
}
// xmm5, xmm6 = src rb, ga
// xmm0, xmm1 = dst rb, ga
// xmm2, xmm3 = used
// xmm4, xmm7 = free
if(m_sel.pabe || (m_sel.aba != m_sel.abb) && (m_sel.abb == 0 || m_sel.abd == 0))
{
movdqa(xmm4, xmm5);
}
if(m_sel.aba != m_sel.abb)
{
// rb = c[aba * 2 + 0];
switch(m_sel.aba)
{
case 0:
break;
case 1:
movdqa(xmm5, xmm0);
break;
case 2:
pxor(xmm5, xmm5);
break;
}
// rb = rb.sub16(c[abb * 2 + 0]);
switch(m_sel.abb)
{
case 0:
psubw(xmm5, xmm4);
break;
case 1:
psubw(xmm5, xmm0);
break;
case 2:
break;
}
if(!(m_sel.fpsm == 1 && m_sel.abc == 1))
{
// GSVector4i a = abc < 2 ? c[abc * 2 + 1].yywwlh().sll16(7) : m_env.afix;
switch(m_sel.abc)
{
case 0:
case 1:
movdqa(xmm7, m_sel.abc ? xmm1 : xmm6);
pshuflw(xmm7, xmm7, _MM_SHUFFLE(3, 3, 1, 1));
pshufhw(xmm7, xmm7, _MM_SHUFFLE(3, 3, 1, 1));
psllw(xmm7, 7);
break;
case 2:
movdqa(xmm7, xmmword[&m_env.afix]);
break;
}
// rb = rb.modulate16<1>(a);
modulate16<1>(xmm5, xmm7);
}
// rb = rb.add16(c[abd * 2 + 0]);
switch(m_sel.abd)
{
case 0:
paddw(xmm5, xmm4);
break;
case 1:
paddw(xmm5, xmm0);
break;
case 2:
break;
}
}
else
{
// rb = c[abd * 2 + 0];
switch(m_sel.abd)
{
case 0:
break;
case 1:
movdqa(xmm5, xmm0);
break;
case 2:
pxor(xmm5, xmm5);
break;
}
}
if(m_sel.pabe)
{
// mask = (c[1] << 8).sra32(31);
movdqa(xmm0, xmm6);
pslld(xmm0, 8);
psrad(xmm0, 31);
// rb = c[0].blend8(rb, mask);
blend8r(xmm5, xmm4);
}
// xmm6 = src ga
// xmm1 = dst ga
// xmm5 = rb
// xmm7 = a
// xmm2, xmm3 = used
// xmm0, xmm4 = free
movdqa(xmm4, xmm6);
if(m_sel.aba != m_sel.abb)
{
// ga = c[aba * 2 + 1];
switch(m_sel.aba)
{
case 0:
break;
case 1:
movdqa(xmm6, xmm1);
break;
case 2:
pxor(xmm6, xmm6);
break;
}
// ga = ga.sub16(c[abeb * 2 + 1]);
switch(m_sel.abb)
{
case 0:
psubw(xmm6, xmm4);
break;
case 1:
psubw(xmm6, xmm1);
break;
case 2:
break;
}
if(!(m_sel.fpsm == 1 && m_sel.abc == 1))
{
// ga = ga.modulate16<1>(a);
modulate16<1>(xmm6, xmm7);
}
// ga = ga.add16(c[abd * 2 + 1]);
switch(m_sel.abd)
{
case 0:
paddw(xmm6, xmm4);
break;
case 1:
paddw(xmm6, xmm1);
break;
case 2:
break;
}
}
else
{
// ga = c[abd * 2 + 1];
switch(m_sel.abd)
{
case 0:
break;
case 1:
movdqa(xmm6, xmm1);
break;
case 2:
pxor(xmm6, xmm6);
break;
}
}
// xmm4 = src ga
// xmm5 = rb
// xmm6 = ga
// xmm2, xmm3 = used
// xmm0, xmm1, xmm7 = free
if(m_sel.pabe)
{
if(!m_cpu.has(util::Cpu::tSSE41))
{
// doh, previous blend8r overwrote xmm0 (sse41 uses pblendvb)
movdqa(xmm0, xmm4);
pslld(xmm0, 8);
psrad(xmm0, 31);
}
psrld(xmm0, 16); // zero out high words to select the source alpha in blend (so it also does mix16)
// ga = c[1].blend8(ga, mask).mix16(c[1]);
blend8r(xmm6, xmm4);
}
else
{
if(m_sel.fpsm != 1) // TODO: fm == 0xffxxxxxx
{
mix16(xmm6, xmm4, xmm7);
}
}
}
void GSDrawScanlineCodeGenerator::ColorTFX()
{
if(!m_sel.fwrite)
{
return;
}
switch(m_sel.tfx)
{
case TFX_MODULATE:
// GSVector4i rb = iip ? rbf : m_env.c.rb;
// rbt = rbt.modulate16<1>(rb).clamp8();
modulate16<1>(xmm5, xmmword[m_sel.iip ? &m_env.temp.rb : &m_env.c.rb]);
clamp16(xmm5, xmm1);
break;
case TFX_DECAL:
break;
case TFX_HIGHLIGHT:
case TFX_HIGHLIGHT2:
if(m_sel.tfx == TFX_HIGHLIGHT2 && m_sel.tcc)
{
// GSVector4i ga = iip ? gaf : m_env.c.ga;
movdqa(xmm2, xmmword[m_sel.iip ? &m_env.temp.ga : &m_env.c.ga]);
}
// gat = gat.modulate16<1>(ga).add16(af).clamp8().mix16(gat);
movdqa(xmm1, xmm6);
modulate16<1>(xmm6, xmm2);
pshuflw(xmm2, xmm2, _MM_SHUFFLE(3, 3, 1, 1));
pshufhw(xmm2, xmm2, _MM_SHUFFLE(3, 3, 1, 1));
psrlw(xmm2, 7);
paddw(xmm6, xmm2);
clamp16(xmm6, xmm0);
mix16(xmm6, xmm1, xmm0);
// GSVector4i rb = iip ? rbf : m_env.c.rb;
// rbt = rbt.modulate16<1>(rb).add16(af).clamp8();
modulate16<1>(xmm5, xmmword[m_sel.iip ? &m_env.temp.rb : &m_env.c.rb]);
paddw(xmm5, xmm2);
clamp16(xmm5, xmm0);
break;
case TFX_NONE:
// rbt = iip ? rb.srl16(7) : rb;
if(m_sel.iip)
{
psrlw(xmm5, 7);
}
break;
}
}
int main()
{
int rval, i, co, tmp; //, j;
mmx_t ma, mb; //, mm0, mm1, *pm0, *pm1;
tmp = calc_cpu_speed();
printf(" Calculating CPU-Speed....running at %d MHz\n", tmp);
printf(" Calculating CPU-OFFSET...found : ");
co = calc_cpu_ofs();
printf(" %d\n",co);
rval = 1; //mmx_ok(); /* Announce return value of mmx_ok() */
printf(" *** Bogus message since we are emulating so MMX does allways exist ***\n");
printf(" Value returned from init was %x.", rval);
printf(" (Indicates MMX %s available)\n\n",(rval)? "is" : "not");
if(rval) {
/* PADD *****************************************************/
ma.q = 0x1111111180000000LL;
mb.q = 0x7fffffff00000001LL;
mmx_regdump(ma);
mmx_regdump(mb);
paddd( &ma, &mb);
printf("paddd: mb.q is %016llx\n", mb.q);
mmx_regdump(mb);
paddw( &ma, &mb);
mmx_regdump(ma);
mmx_regdump(mb);
ma.q = 0x0000000000000008LL;
mb.q = 0x0001000200030004LL;
mmx_regdump(ma);
mmx_regdump(mb);
psllw( &ma, &mb);
mmx_regdump(ma);
mmx_regdump(mb);
ma.q = 0x0000000000000000LL;
mb.q = 0x888044a87f06fe80LL;
mmx_regdump(ma);
mmx_regdump(mb);
// packuswb( &ma,&mb);
ma.q = 0x00aa00dd01009999LL;
mb.q = 0x0011002200330044LL;
mmx_regdump(ma);
mmx_regdump(mb);
packuswb( &ma, &mb);
// punpckhdq(&ma,&mb);
mmx_regdump(ma);
mmx_regdump(mb);
/* punpckhdq */
ma.q = 0x090a0b0c0d0e0f00LL;
mb.q = 0x0102030405060708LL;
mmx_regdump(ma);
mmx_regdump(mb);
punpckhdq( &ma, &mb);
mmx_regdump(ma);
mmx_regdump(mb);
/* punpckhwd */
ma.q = 0x090a0b0c0d0e0f00LL;
mb.q = 0x0102030405060708LL;
mmx_regdump(ma);
mmx_regdump(mb);
punpckhwd( &ma, &mb);
mmx_regdump(ma);
mmx_regdump(mb);
/* punpckhbw */
ma.q = 0x090a0b0c0d0e0f00LL;
mb.q = 0x0102030405060708LL;
mmx_regdump(ma);
mmx_regdump(mb);
punpckhbw( &ma, &mb);
mmx_regdump(ma);
mmx_regdump(mb);
/* punpckldq */
ma.q = 0x090a0b0c0d0e0f00LL;
mb.q = 0x0102030405060708LL;
mmx_regdump(ma);
mmx_regdump(mb);
punpckldq( &ma, &mb);
mmx_regdump(ma);
mmx_regdump(mb);
/* punpcklwd */
ma.q = 0x090a0b0c0d0e0f00LL;
mb.q = 0x0102030405060708LL;
mmx_regdump(ma);
mmx_regdump(mb);
punpcklwd( &ma, &mb);
mmx_regdump(ma);
mmx_regdump(mb);
/* punpcklbw */
ma.q = 0x090a0b0c0d0e0f00LL;
mb.q = 0x0102030405060708LL;
mmx_regdump(ma);
mmx_regdump(mb);
punpcklbw( &ma, &mb);
mmx_regdump(ma);
mmx_regdump(mb);
/* pmaddwd */
ma.q = 0x8000800080008000LL;
mb.q = 0x8000800080008000LL;
mmx_regdump(ma);
mmx_regdump(mb);
pmaddwd( &ma, &mb);
mmx_regdump(ma);
mmx_regdump(mb);
/* packsswb */
ma.q = 0x00aa00dd01009999LL;
mb.q = 0x0011002200330044LL;
mmx_regdump(ma);
mmx_regdump(mb);
packsswb( &ma, &mb);
mmx_regdump(ma);
mmx_regdump(mb);
/* packsswb */
ma.q = 0x007e7f00ef9dff88LL;
mb.q = 0xff020085007e81cfLL;
mmx_regdump(ma);
mmx_regdump(mb);
packsswb( &ma, &mb);
mmx_regdump(ma);
mmx_regdump(mb);
for (i=0; i< 0x0F; i++){
printf ("%d. Test running\n", i);
GET_TSC(tsc1);
mb.w[0] = i;
ma.w[0] = i;
GET_TSC(tsc2);
tick_dump(co);
GET_TSC(tsc1); /* Let's test some real asm */
asm("
// pushl %eax\n
// movl $0, %eax\n
// cpuid\n
packsswb %mm0, %mm1\n
pmaddwd %mm0, %mm1\n
punpcklbw %mm0, %mm1\n
punpcklbw %mm0, %mm1\n
pmaddwd %mm0, %mm1\n
punpcklbw %mm0, %mm1\n
// popl %eax\n
");
GET_TSC(tsc2);
tick_dump(co);
ma.q = 0x8000800080008000LL;
mb.q = 0x8000800080008000LL;
GET_TSC(tsc1); /* Here some emulated routines */
pmaddwd( &ma, &mb);
// packsswb( &ma, &mb);
GET_TSC(tsc2);
tick_dump(co);
// mmx_regdump(mb);
}
}
exit(0); /* Clean-up and exit nicely */
}