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 MacroAssembler::fast_log(XMMRegister xmm0, XMMRegister xmm1, XMMRegister xmm2, XMMRegister xmm3, XMMRegister xmm4, XMMRegister xmm5, XMMRegister xmm6, XMMRegister xmm7, Register eax, Register ecx, Register edx, Register tmp1, Register tmp2) {
Label L_2TAG_PACKET_0_0_2, L_2TAG_PACKET_1_0_2, L_2TAG_PACKET_2_0_2, L_2TAG_PACKET_3_0_2;
Label L_2TAG_PACKET_4_0_2, L_2TAG_PACKET_5_0_2, L_2TAG_PACKET_6_0_2, L_2TAG_PACKET_7_0_2;
Label L_2TAG_PACKET_8_0_2;
Label L_2TAG_PACKET_12_0_2, L_2TAG_PACKET_13_0_2, B1_3, B1_5, start;
assert_different_registers(tmp1, tmp2, eax, ecx, edx);
jmp(start);
address L_tbl = (address)_L_tbl;
address log2 = (address)_log2;
address coeff = (address)_coeff;
bind(start);
subq(rsp, 24);
movsd(Address(rsp, 0), xmm0);
mov64(rax, 0x3ff0000000000000);
movdq(xmm2, rax);
mov64(rdx, 0x77f0000000000000);
movdq(xmm3, rdx);
movl(ecx, 32768);
movdl(xmm4, rcx);
mov64(tmp1, 0xffffe00000000000);
movdq(xmm5, tmp1);
movdqu(xmm1, xmm0);
pextrw(eax, xmm0, 3);
por(xmm0, xmm2);
movl(ecx, 16352);
psrlq(xmm0, 27);
lea(tmp2, ExternalAddress(L_tbl));
psrld(xmm0, 2);
rcpps(xmm0, xmm0);
psllq(xmm1, 12);
pshufd(xmm6, xmm5, 228);
psrlq(xmm1, 12);
subl(eax, 16);
cmpl(eax, 32736);
jcc(Assembler::aboveEqual, L_2TAG_PACKET_0_0_2);
bind(L_2TAG_PACKET_1_0_2);
paddd(xmm0, xmm4);
por(xmm1, xmm3);
movdl(edx, xmm0);
psllq(xmm0, 29);
pand(xmm5, xmm1);
pand(xmm0, xmm6);
subsd(xmm1, xmm5);
mulpd(xmm5, xmm0);
andl(eax, 32752);
subl(eax, ecx);
cvtsi2sdl(xmm7, eax);
mulsd(xmm1, xmm0);
movq(xmm6, ExternalAddress(log2)); // 0xfefa3800UL, 0x3fa62e42UL
movdqu(xmm3, ExternalAddress(coeff)); // 0x92492492UL, 0x3fc24924UL, 0x00000000UL, 0xbfd00000UL
subsd(xmm5, xmm2);
andl(edx, 16711680);
shrl(edx, 12);
movdqu(xmm0, Address(tmp2, edx));
movdqu(xmm4, ExternalAddress(16 + coeff)); // 0x3d6fb175UL, 0xbfc5555eUL, 0x55555555UL, 0x3fd55555UL
addsd(xmm1, xmm5);
movdqu(xmm2, ExternalAddress(32 + coeff)); // 0x9999999aUL, 0x3fc99999UL, 0x00000000UL, 0xbfe00000UL
mulsd(xmm6, xmm7);
movddup(xmm5, xmm1);
mulsd(xmm7, ExternalAddress(8 + log2)); // 0x93c76730UL, 0x3ceef357UL
mulsd(xmm3, xmm1);
addsd(xmm0, xmm6);
mulpd(xmm4, xmm5);
mulpd(xmm5, xmm5);
movddup(xmm6, xmm0);
addsd(xmm0, xmm1);
addpd(xmm4, xmm2);
mulpd(xmm3, xmm5);
subsd(xmm6, xmm0);
mulsd(xmm4, xmm1);
pshufd(xmm2, xmm0, 238);
addsd(xmm1, xmm6);
mulsd(xmm5, xmm5);
addsd(xmm7, xmm2);
addpd(xmm4, xmm3);
addsd(xmm1, xmm7);
mulpd(xmm4, xmm5);
addsd(xmm1, xmm4);
pshufd(xmm5, xmm4, 238);
addsd(xmm1, xmm5);
addsd(xmm0, xmm1);
jmp(B1_5);
bind(L_2TAG_PACKET_0_0_2);
movq(xmm0, Address(rsp, 0));
movq(xmm1, Address(rsp, 0));
addl(eax, 16);
cmpl(eax, 32768);
jcc(Assembler::aboveEqual, L_2TAG_PACKET_2_0_2);
cmpl(eax, 16);
jcc(Assembler::below, L_2TAG_PACKET_3_0_2);
bind(L_2TAG_PACKET_4_0_2);
addsd(xmm0, xmm0);
jmp(B1_5);
bind(L_2TAG_PACKET_5_0_2);
jcc(Assembler::above, L_2TAG_PACKET_4_0_2);
cmpl(edx, 0);
jcc(Assembler::above, L_2TAG_PACKET_4_0_2);
jmp(L_2TAG_PACKET_6_0_2);
bind(L_2TAG_PACKET_3_0_2);
xorpd(xmm1, xmm1);
addsd(xmm1, xmm0);
movdl(edx, xmm1);
psrlq(xmm1, 32);
movdl(ecx, xmm1);
orl(edx, ecx);
cmpl(edx, 0);
jcc(Assembler::equal, L_2TAG_PACKET_7_0_2);
xorpd(xmm1, xmm1);
movl(eax, 18416);
pinsrw(xmm1, eax, 3);
mulsd(xmm0, xmm1);
movdqu(xmm1, xmm0);
pextrw(eax, xmm0, 3);
por(xmm0, xmm2);
psrlq(xmm0, 27);
movl(ecx, 18416);
psrld(xmm0, 2);
rcpps(xmm0, xmm0);
psllq(xmm1, 12);
pshufd(xmm6, xmm5, 228);
psrlq(xmm1, 12);
jmp(L_2TAG_PACKET_1_0_2);
bind(L_2TAG_PACKET_2_0_2);
movdl(edx, xmm1);
psrlq(xmm1, 32);
movdl(ecx, xmm1);
addl(ecx, ecx);
cmpl(ecx, -2097152);
jcc(Assembler::aboveEqual, L_2TAG_PACKET_5_0_2);
orl(edx, ecx);
cmpl(edx, 0);
jcc(Assembler::equal, L_2TAG_PACKET_7_0_2);
bind(L_2TAG_PACKET_6_0_2);
xorpd(xmm1, xmm1);
xorpd(xmm0, xmm0);
movl(eax, 32752);
pinsrw(xmm1, eax, 3);
mulsd(xmm0, xmm1);
movl(Address(rsp, 16), 3);
jmp(L_2TAG_PACKET_8_0_2);
bind(L_2TAG_PACKET_7_0_2);
xorpd(xmm1, xmm1);
xorpd(xmm0, xmm0);
movl(eax, 49136);
pinsrw(xmm0, eax, 3);
divsd(xmm0, xmm1);
movl(Address(rsp, 16), 2);
bind(L_2TAG_PACKET_8_0_2);
movq(Address(rsp, 8), xmm0);
bind(B1_3);
movq(xmm0, Address(rsp, 8));
bind(B1_5);
addq(rsp, 24);
}
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]);
}
}
}
}
}
