void GSDrawScanlineCodeGenerator::clamp16(const Xmm& a, const Xmm& temp)
{
#if _M_SSE >= 0x500
vpackuswb(a, a);
vpmovzxbw(a, a);
#elif _M_SSE >= 0x401
packuswb(a, a);
pmovzxbw(a, a);
#else
packuswb(a, a);
pxor(temp, temp);
punpcklbw(a, temp);
#endif
}
void GSDrawScanlineCodeGenerator::WriteFrame()
{
if(!m_sel.fwrite)
{
return;
}
if(m_sel.colclamp == 0)
{
// c[0] &= 0x00ff00ff;
// c[1] &= 0x00ff00ff;
vpcmpeqd(xmm15, xmm15);
vpsrlw(xmm15, 8);
vpand(xmm2, xmm15);
vpand(xmm3, xmm15);
}
if(m_sel.fpsm == 2 && m_sel.dthe)
{
mov(rax, r8);
and(rax, 3);
shl(rax, 5);
vpaddw(xmm2, ptr[r12 + rax + offsetof(GSScanlineGlobalData, dimx[0])]);
vpaddw(xmm3, ptr[r12 + rax + offsetof(GSScanlineGlobalData, dimx[1])]);
}
// GSVector4i fs = c[0].upl16(c[1]).pu16(c[0].uph16(c[1]));
vpunpckhwd(xmm15, xmm2, xmm3);
vpunpcklwd(xmm2, xmm3);
vpackuswb(xmm2, xmm15);
if(m_sel.fba && m_sel.fpsm != 1)
{
// fs |= 0x80000000;
vpcmpeqd(xmm15, xmm15);
vpslld(xmm15, 31);
vpor(xmm2, xmm15);
}
// xmm2 = fs
// xmm4 = fm
// xmm6 = fd
if(m_sel.fpsm == 2)
{
// GSVector4i rb = fs & 0x00f800f8;
// GSVector4i ga = fs & 0x8000f800;
mov(eax, 0x00f800f8);
vmovd(xmm0, eax);
vpshufd(xmm0, xmm0, _MM_SHUFFLE(0, 0, 0, 0));
mov(eax, 0x8000f800);
vmovd(xmm1, eax);
vpshufd(xmm1, xmm1, _MM_SHUFFLE(0, 0, 0, 0));
vpand(xmm0, xmm2);
vpand(xmm1, xmm2);
// fs = (ga >> 16) | (rb >> 9) | (ga >> 6) | (rb >> 3);
vpsrld(xmm2, xmm0, 9);
vpsrld(xmm0, 3);
vpsrld(xmm3, xmm1, 16);
vpsrld(xmm1, 6);
vpor(xmm0, xmm1);
vpor(xmm2, xmm3);
vpor(xmm2, xmm0);
}
if(m_sel.rfb)
{
// fs = fs.blend(fd, fm);
blend(xmm2, xmm6, xmm4); // TODO: could be skipped in certain cases, depending on fpsm and fm
}
bool fast = m_sel.rfb && m_sel.fpsm < 2;
WritePixel(xmm2, rbx, dl, fast, m_sel.fpsm, 0);
}
void GSDrawScanlineCodeGenerator::clamp16(const Ymm& a, const Ymm& temp)
{
vpackuswb(a, a);
vpermq(a, a, _MM_SHUFFLE(3, 1, 2, 0)); // this sucks
vpmovzxbw(a, a);
}
