vertexjit: Cut a few instrs from morph on x86.

This commit is contained in:
Unknown W. Brackets committed 2014-03-22 15:56:30 -07:00
1 parent 162f229294
commit 0da5caf11a
1 file changed
+39 -27
+39 -27
View File
@@ -1050,28 +1050,32 @@ void VertexDecoderJitCache::Jit_PosFloatSkin() {
}
void VertexDecoderJitCache::Jit_AnyS8Morph(int srcoff, int dstoff) {
// TODO: Optimize the first one to skip an ADDPS.
XORPS(fpScratchReg, R(fpScratchReg));
MOV(PTRBITS, R(tempReg1), ImmPtr(&gstate_c.morphWeights[0]));
// Sum into fpScratchReg.
bool first = true;
for (int n = 0; n < dec_->morphcount; ++n) {
const X64Reg reg = first ? fpScratchReg : fpScratchReg2;
// Okay, first convert to floats.
XORPS(fpScratchReg3, R(fpScratchReg3));
MOVD_xmm(fpScratchReg2, MDisp(srcReg, dec_->onesize_ * n + srcoff));
PUNPCKLBW(fpScratchReg2, R(fpScratchReg3));
PUNPCKLWD(fpScratchReg2, R(fpScratchReg3));
PSLLD(fpScratchReg2, 24);
PSRAD(fpScratchReg2, 24); // Ugly sign extension, can be done faster in SSE4
CVTDQ2PS(fpScratchReg2, R(fpScratchReg2));
MOVD_xmm(reg, MDisp(srcReg, dec_->onesize_ * n + srcoff));
PUNPCKLBW(reg, R(fpScratchReg3));
PUNPCKLWD(reg, R(fpScratchReg3));
PSLLD(reg, 24);
PSRAD(reg, 24); // Ugly sign extension, can be done faster in SSE4
CVTDQ2PS(reg, R(reg));
// Now, It's time to multiply by the weight and 1.0f/127.0f.
MOVSS(fpScratchReg3, MDisp(tempReg1, sizeof(float) * n));
MULSS(fpScratchReg3, M(by127));
SHUFPS(fpScratchReg3, R(fpScratchReg3), _MM_SHUFFLE(0, 0, 0, 0));
MULPS(fpScratchReg2, R(fpScratchReg3));
ADDPS(fpScratchReg, R(fpScratchReg2));
MULPS(reg, R(fpScratchReg3));
if (!first) {
ADDPS(fpScratchReg, R(fpScratchReg2));
} else {
first = false;
}
}
// TODO: Is it okay that we're over-writing by 4 bytes? Probably...
@@ -1079,27 +1083,31 @@ void VertexDecoderJitCache::Jit_AnyS8Morph(int srcoff, int dstoff) {
}
void VertexDecoderJitCache::Jit_AnyS16Morph(int srcoff, int dstoff) {
// TODO: Optimize the first one to skip an ADDPS.
XORPS(fpScratchReg, R(fpScratchReg));
MOV(PTRBITS, R(tempReg1), ImmPtr(&gstate_c.morphWeights[0]));
// Sum into fpScratchReg.
bool first = true;
for (int n = 0; n < dec_->morphcount; ++n) {
const X64Reg reg = first ? fpScratchReg : fpScratchReg2;
// Okay, first convert to floats.
XORPS(fpScratchReg3, R(fpScratchReg3));
MOVQ_xmm(fpScratchReg2, MDisp(srcReg, dec_->onesize_ * n + srcoff));
PUNPCKLWD(fpScratchReg2, R(fpScratchReg3));
PSLLD(fpScratchReg2, 16);
PSRAD(fpScratchReg2, 16); // Ugly sign extension, can be done faster in SSE4
CVTDQ2PS(fpScratchReg2, R(fpScratchReg2));
MOVQ_xmm(reg, MDisp(srcReg, dec_->onesize_ * n + srcoff));
PUNPCKLWD(reg, R(fpScratchReg3));
PSLLD(reg, 16);
PSRAD(reg, 16); // Ugly sign extension, can be done faster in SSE4
CVTDQ2PS(reg, R(reg));
// Now, It's time to multiply by the weight and 1.0f/32767.0f.
MOVSS(fpScratchReg3, MDisp(tempReg1, sizeof(float) * n));
MULSS(fpScratchReg3, M(by32767));
SHUFPS(fpScratchReg3, R(fpScratchReg3), _MM_SHUFFLE(0, 0, 0, 0));
MULPS(fpScratchReg2, R(fpScratchReg3));
ADDPS(fpScratchReg, R(fpScratchReg2));
MULPS(reg, R(fpScratchReg3));
if (!first) {
ADDPS(fpScratchReg, R(fpScratchReg2));
} else {
first = false;
}
}
// TODO: Is it okay that we're over-writing by 4 bytes? Probably...
@@ -1107,17 +1115,21 @@ void VertexDecoderJitCache::Jit_AnyS16Morph(int srcoff, int dstoff) {
}
void VertexDecoderJitCache::Jit_AnyFloatMorph(int srcoff, int dstoff) {
// TODO: Optimize the first one to skip an ADDPS.
XORPS(fpScratchReg, R(fpScratchReg));
MOV(PTRBITS, R(tempReg1), ImmPtr(&gstate_c.morphWeights[0]));
// Sum into fpScratchReg.
bool first = true;
for (int n = 0; n < dec_->morphcount; ++n) {
MOVUPS(fpScratchReg2, MDisp(srcReg, dec_->onesize_ * n + srcoff));
const X64Reg reg = first ? fpScratchReg : fpScratchReg2;
MOVUPS(reg, MDisp(srcReg, dec_->onesize_ * n + srcoff));
MOVSS(fpScratchReg3, MDisp(tempReg1, sizeof(float) * n));
SHUFPS(fpScratchReg3, R(fpScratchReg3), _MM_SHUFFLE(0, 0, 0, 0));
MULPS(fpScratchReg2, R(fpScratchReg3));
ADDPS(fpScratchReg, R(fpScratchReg2));
MULPS(reg, R(fpScratchReg3));
if (!first) {
ADDPS(fpScratchReg, R(fpScratchReg2));
} else {
first = false;
}
}
// TODO: Is it okay that we're over-writing by 4 bytes? Probably...