void GSDrawScanlineCodeGenerator::WriteZBuf()
{
if(!m_sel.zwrite)
{
return;
}
movdqa(xmm1, xmmword[!m_sel.sprite ? &m_env.temp.zs : &m_env.p.z]);
bool fast = false;
if(m_sel.ztest && m_sel.zpsm < 2)
{
// zs = zs.blend8(zd, zm);
movdqa(xmm0, xmm4);
movdqa(xmm7, xmmword[&m_env.temp.zd]);
blend8(xmm1, xmm7);
fast = true;
}
WritePixel(xmm1, xmm0, ebp, dh, fast, m_sel.zpsm);
}
void GPUDrawScanlineCodeGenerator::WriteFrame()
{
// GSVector4i fs = r | g | b | (m_sel.md ? GSVector4i(0x80008000) : m_sel.tme ? a : 0);
pcmpeqd(xmm0, xmm0);
if(m_sel.md || m_sel.tme)
{
movdqa(xmm2, xmm0);
psllw(xmm2, 15);
}
psrlw(xmm0, 11);
psllw(xmm0, 3);
// xmm0 = 0x00f8
// xmm2 = 0x8000 (md)
// GSVector4i r = (c[0] & 0x00f800f8) >> 3;
pand(xmm4, xmm0);
psrlw(xmm4, 3);
// GSVector4i g = (c[1] & 0x00f800f8) << 2;
pand(xmm5, xmm0);
psllw(xmm5, 2);
por(xmm4, xmm5);
// GSVector4i b = (c[2] & 0x00f800f8) << 7;
pand(xmm6, xmm0);
psllw(xmm6, 7);
por(xmm4, xmm6);
if(m_sel.md)
{
// GSVector4i a = GSVector4i(0x80008000);
por(xmm4, xmm2);
}
else if(m_sel.tme)
{
// GSVector4i a = (c[3] << 8) & 0x80008000;
psllw(xmm3, 8);
pand(xmm3, xmm2);
por(xmm4, xmm3);
}
// fs = fs.blend8(fd, test);
movdqa(xmm0, xmm7);
blend8(xmm4, xmm1);
// GSVector4i::store<false>(fb, fs);
// movdqu(ptr[edi], xmm4);
movq(qword[edi], xmm4);
movhps(qword[edi + 8], xmm4);
}
void GPUDrawScanlineCodeGenerator::AlphaBlend()
{
if(!m_sel.abe)
{
return;
}
// xmm1 = fd
// xmm3 = a
// xmm4 = r
// xmm5 = g
// xmm6 = b
// xmm7 = test
// xmm0, xmm2 = free
// GSVector4i r = (fd & 0x001f001f) << 3;
pcmpeqd(xmm0, xmm0);
psrlw(xmm0, 11); // 0x001f
movdqa(xmm2, xmm1);
pand(xmm2, xmm0);
psllw(xmm2, 3);
switch(m_sel.abr)
{
case 0:
// r = r.avg8(c[0]);
pavgb(xmm2, xmm4);
break;
case 1:
// r = r.addus8(c[0]);
paddusb(xmm2, xmm4);
break;
case 2:
// r = r.subus8(c[0]);
psubusb(xmm2, xmm4);
break;
case 3:
// r = r.addus8(c[0].srl16(2));
movdqa(xmm0, xmm4);
psrlw(xmm0, 2);
paddusb(xmm2, xmm0);
break;
}
if(m_sel.tme)
{
movdqa(xmm0, xmm3);
blend8(xmm4, xmm2);
}
else
{
movdqa(xmm4, xmm2);
}
// GSVector4i g = (d & 0x03e003e0) >> 2;
pcmpeqd(xmm0, xmm0);
psrlw(xmm0, 11);
psllw(xmm0, 5); // 0x03e0
movdqa(xmm2, xmm1);
pand(xmm2, xmm0);
psrlw(xmm2, 2);
switch(m_sel.abr)
{
case 0:
// g = g.avg8(c[2]);
pavgb(xmm2, xmm5);
break;
case 1:
// g = g.addus8(c[2]);
paddusb(xmm2, xmm5);
break;
case 2:
// g = g.subus8(c[2]);
psubusb(xmm2, xmm5);
break;
case 3:
// g = g.addus8(c[2].srl16(2));
movdqa(xmm0, xmm5);
psrlw(xmm0, 2);
paddusb(xmm2, xmm0);
break;
}
if(m_sel.tme)
{
movdqa(xmm0, xmm3);
blend8(xmm5, xmm2);
}
else
{
movdqa(xmm5, xmm2);
}
// GSVector4i b = (d & 0x7c007c00) >> 7;
pcmpeqd(xmm0, xmm0);
psrlw(xmm0, 11);
psllw(xmm0, 10); // 0x7c00
movdqa(xmm2, xmm1);
pand(xmm2, xmm0);
psrlw(xmm2, 7);
switch(m_sel.abr)
{
case 0:
// b = b.avg8(c[2]);
pavgb(xmm2, xmm6);
break;
case 1:
// b = b.addus8(c[2]);
paddusb(xmm2, xmm6);
break;
case 2:
// b = b.subus8(c[2]);
psubusb(xmm2, xmm6);
break;
case 3:
// b = b.addus8(c[2].srl16(2));
movdqa(xmm0, xmm6);
psrlw(xmm0, 2);
paddusb(xmm2, xmm0);
break;
}
if(m_sel.tme)
{
movdqa(xmm0, xmm3);
blend8(xmm6, xmm2);
}
else
{
movdqa(xmm6, xmm2);
}
}
void GSDrawScanlineCodeGenerator::AlphaTFX()
{
if(!m_sel.fb)
{
return;
}
switch(m_sel.tfx)
{
case TFX_MODULATE:
// GSVector4i ga = iip ? gaf : m_env.c.ga;
movdqa(xmm4, xmmword[m_sel.iip ? &m_env.temp.ga : &m_env.c.ga]);
// gat = gat.modulate16<1>(ga).clamp8();
modulate16<1>(xmm6, xmm4);
clamp16(xmm6, xmm3);
// if(!tcc) gat = gat.mix16(ga.srl16(7));
if(!m_sel.tcc)
{
psrlw(xmm4, 7);
mix16(xmm6, xmm4, xmm3);
}
break;
case TFX_DECAL:
// if(!tcc) gat = gat.mix16(ga.srl16(7));
if(!m_sel.tcc)
{
// GSVector4i ga = iip ? gaf : m_env.c.ga;
movdqa(xmm4, xmmword[m_sel.iip ? &m_env.temp.ga : &m_env.c.ga]);
psrlw(xmm4, 7);
mix16(xmm6, xmm4, xmm3);
}
break;
case TFX_HIGHLIGHT:
// GSVector4i ga = iip ? gaf : m_env.c.ga;
movdqa(xmm4, xmmword[m_sel.iip ? &m_env.temp.ga : &m_env.c.ga]);
movdqa(xmm2, xmm4);
// gat = gat.mix16(!tcc ? ga.srl16(7) : gat.addus8(ga.srl16(7)));
psrlw(xmm4, 7);
if(m_sel.tcc)
{
paddusb(xmm4, xmm6);
}
mix16(xmm6, xmm4, xmm3);
break;
case TFX_HIGHLIGHT2:
// if(!tcc) gat = gat.mix16(ga.srl16(7));
if(!m_sel.tcc)
{
// GSVector4i ga = iip ? gaf : m_env.c.ga;
movdqa(xmm4, xmmword[m_sel.iip ? &m_env.temp.ga : &m_env.c.ga]);
movdqa(xmm2, xmm4);
psrlw(xmm4, 7);
mix16(xmm6, xmm4, xmm3);
}
break;
case TFX_NONE:
// gat = iip ? ga.srl16(7) : ga;
if(m_sel.iip)
{
psrlw(xmm6, 7);
}
break;
}
if(m_sel.aa1)
{
// gs_user figure 3-2: anti-aliasing after tfx, before tests, modifies alpha
// FIXME: bios config screen cubes
if(!m_sel.abe)
{
// a = cov
if(m_sel.edge)
{
movdqa(xmm0, xmmword[&m_env.temp.cov]);
}
else
{
pcmpeqd(xmm0, xmm0);
psllw(xmm0, 15);
psrlw(xmm0, 8);
}
mix16(xmm6, xmm0, xmm1);
}
else
{
// a = a == 0x80 ? cov : a
pcmpeqd(xmm0, xmm0);
psllw(xmm0, 15);
psrlw(xmm0, 8);
if(m_sel.edge)
{
movdqa(xmm1, xmmword[&m_env.temp.cov]);
}
else
{
movdqa(xmm1, xmm0);
}
pcmpeqw(xmm0, xmm6);
psrld(xmm0, 16);
pslld(xmm0, 16);
blend8(xmm6, xmm1);
}
}
}
void GSDrawScanlineCodeGenerator::Wrap(const Xmm& uv)
{
// xmm0, xmm1, xmm4, xmm5, xmm6 = free
int wms_clamp = ((m_sel.wms + 1) >> 1) & 1;
int wmt_clamp = ((m_sel.wmt + 1) >> 1) & 1;
int region = ((m_sel.wms | m_sel.wmt) >> 1) & 1;
if(wms_clamp == wmt_clamp)
{
if(wms_clamp)
{
if(region)
{
pmaxsw(uv, xmmword[&m_env.t.min]);
}
else
{
pxor(xmm0, xmm0);
pmaxsw(uv, xmm0);
}
pminsw(uv, xmmword[&m_env.t.max]);
}
else
{
pand(uv, xmmword[&m_env.t.min]);
if(region)
{
por(uv, xmmword[&m_env.t.max]);
}
}
}
else
{
movdqa(xmm1, uv);
movdqa(xmm4, xmmword[&m_env.t.min]);
movdqa(xmm5, xmmword[&m_env.t.max]);
// GSVector4i clamp = t.sat_i16(m_env.t.min, m_env.t.max);
pmaxsw(uv, xmm4);
pminsw(uv, xmm5);
// GSVector4i repeat = (t & m_env.t.min) | m_env.t.max;
pand(xmm1, xmm4);
if(region)
{
por(xmm1, xmm5);
}
// clamp.blend8(repeat, m_env.t.mask);
movdqa(xmm0, xmmword[&m_env.t.mask]);
blend8(uv, xmm1);
}
}
