Merge pull request #3261 from MerryMage/DPH
shader_jit_x64_compiler: Use haddps for horizontal summation
This commit is contained in:
commit
aecd2b85fe
|
@ -387,13 +387,8 @@ void JitShader::Compile_DP4(Instruction instr) {
|
||||||
|
|
||||||
Compile_SanitizedMul(SRC1, SRC2, SCRATCH);
|
Compile_SanitizedMul(SRC1, SRC2, SCRATCH);
|
||||||
|
|
||||||
movaps(SRC2, SRC1);
|
haddps(SRC1, SRC1);
|
||||||
shufps(SRC1, SRC1, _MM_SHUFFLE(2, 3, 0, 1)); // XYZW -> ZWXY
|
haddps(SRC1, SRC1);
|
||||||
addps(SRC1, SRC2);
|
|
||||||
|
|
||||||
movaps(SRC2, SRC1);
|
|
||||||
shufps(SRC1, SRC1, _MM_SHUFFLE(0, 1, 2, 3)); // XYZW -> WZYX
|
|
||||||
addps(SRC1, SRC2);
|
|
||||||
|
|
||||||
Compile_DestEnable(instr, SRC1);
|
Compile_DestEnable(instr, SRC1);
|
||||||
}
|
}
|
||||||
|
@ -419,13 +414,8 @@ void JitShader::Compile_DPH(Instruction instr) {
|
||||||
|
|
||||||
Compile_SanitizedMul(SRC1, SRC2, SCRATCH);
|
Compile_SanitizedMul(SRC1, SRC2, SCRATCH);
|
||||||
|
|
||||||
movaps(SRC2, SRC1);
|
haddps(SRC1, SRC1);
|
||||||
shufps(SRC1, SRC1, _MM_SHUFFLE(2, 3, 0, 1)); // XYZW -> ZWXY
|
haddps(SRC1, SRC1);
|
||||||
addps(SRC1, SRC2);
|
|
||||||
|
|
||||||
movaps(SRC2, SRC1);
|
|
||||||
shufps(SRC1, SRC1, _MM_SHUFFLE(0, 1, 2, 3)); // XYZW -> WZYX
|
|
||||||
addps(SRC1, SRC2);
|
|
||||||
|
|
||||||
Compile_DestEnable(instr, SRC1);
|
Compile_DestEnable(instr, SRC1);
|
||||||
}
|
}
|
||||||
|
|
Reference in New Issue