void GSDrawScanlineCodeGenerator::ReadTexel(const Xmm& dst, const Xmm& addr, uint8 i)
{
if(!m_cpu.has(util::Cpu::tSSE41) && i > 0)
{
ASSERT(0);
}
if(i == 0) movd(eax, addr);
else pextrd(eax, addr, i);
if(m_sel.tlu) movzx(eax, byte[ebx + eax]);
const Address& src = m_sel.tlu ? ptr[eax * 4 + (size_t)m_env.clut] : ptr[ebx + eax * 4];
if(i == 0) movd(dst, src);
else pinsrd(dst, src, i);
}
WriteMMX()
{
#ifdef XBYAK32
mov(ecx, ptr [esp + 4]);
#endif
movd(mm0, ecx);
ret();
}
void rom(int n)
{
switch(n)
{
case 0:
fetch();
break;
case 1:
iopr('+'); // Immediate operation iopr
break;
case 2:
iopr('-');
break;
case 3:
iopr('^');
break;
case 4:
iopr('&');
break;
case 5:
iopr('|');
break;
case 6:
iopr('~');
break;
case 7: // STOP function
CP++; // To be finished later
return;
case 8:
// ret(); return flag function
break;
case 9:
nopr('+');
break;
case 10:
nopr('-');
break;
case 11:
nopr('^');
break;
case 12:
nopr('&');
break;
case 13:
nopr('|');
break;
case 14:
nopr('~');
break;
case 15:
movs();
break;
case 16:
movd();
break;
}
}
void GPUDrawScanlineCodeGenerator::ReadTexel(const Xmm& dst, const Xmm& addr)
{
for(int i = 0; i < 8; i++)
{
pextrw(eax, addr, (uint8)i);
if(m_sel.tlu) movzx(eax, byte[esi + eax]);
const Address& src = m_sel.tlu ? ptr[edx + eax * 2] : ptr[esi + eax * 2];
if(i == 0) movd(dst, src);
else pinsrw(dst, src, (uint8)i);
}
}
void GSDrawScanlineCodeGenerator::WritePixel(const Xmm& src, const Xmm& temp, const Reg32& addr, uint8 i, int psm)
{
static const int offsets[4] = {0, 2, 8, 10};
Address dst = ptr[addr * 2 + (size_t)m_env.vm + offsets[i] * 2];
if(m_cpu.has(util::Cpu::tSSE41))
{
switch(psm)
{
case 0:
if(i == 0) movd(dst, src);
else pextrd(dst, src, i);
break;
case 1:
if(i == 0) movd(eax, src);
else pextrd(eax, src, i);
xor(eax, dst);
and(eax, 0xffffff);
xor(dst, eax);
break;
case 2:
pextrw(eax, src, i * 2);
mov(dst, ax);
break;
}
}
else
{
switch(psm)
{
case 0:
if(i == 0) movd(dst, src);
else {
pshufd(temp, src, _MM_SHUFFLE(i, i, i, i));
movd(dst, temp);
}
break;
case 1:
if(i == 0) movd(eax, src);
else {
pshufd(temp, src, _MM_SHUFFLE(i, i, i, i));
movd(eax, temp);
}
xor(eax, dst);
and(eax, 0xffffff);
xor(dst, eax);
break;
case 2:
pextrw(eax, src, i * 2);
mov(dst, ax);
break;
}
}
}
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 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::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);
}
ReadMMX()
{
movd(eax, mm0);
ret();
}
