Mirror/ppsspp
1
0 Fork 0
mirror of https://github.com/hrydgard/ppsspp.git synced 2026-10-01 14:58:14 +00:00
Code Issues Packages Projects Releases Wiki Activity
Files
b9ea98d2c5b8d07330af5efbbefd7b143d40d32b
ppsspp/Core/MIPS/x86/X64IRCompVec.cpp
T
Henrik RydgårdandClaude Opus 5.5 eaf55c467b IR: Fold ClampToZero into the 31-bit packs 
Vec4ClampToZero and Vec2ClampToZero only ever fed Vec4Pack31To8 and
Vec2Pack31To16, for vi2uc and vi2us. The packs now clamp negative lanes
to zero themselves, which saves an op and a vector temp, and lets x64
clamp with PACKUSWB's saturation after an arithmetic shift.

While at it, RISC-V compiles Vec2Unpack16To31, Vec2Pack31To16 and
Vec4Pack32To8, and LoongArch Vec2Unpack16To31, Vec2Pack31To16 and the
non-LSX Vec4Pack32To8, all of which went to the IR interpreter.
LoongArch's Vec2Pack32To16 and Vec2Unpack16To32 now take their scalar
path with LSX too, instead of falling back.

Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
2026-09-24 14:31:22 -06:00

389 lines
11 KiB
C++
Raw Blame History

// Copyright (c) 2023- PPSSPP Project.
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation, version 2.0 or later versions.
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU General Public License 2.0 for more details.
// A copy of the GPL 2.0 should have been included with the program.
// If not, see http://www.gnu.org/licenses/
// Official git repository and contact information can be found at
// https://github.com/hrydgard/ppsspp and http://www.ppsspp.org/.
#include "ppsspp_config.h"
#if PPSSPP_ARCH(X86) || PPSSPP_ARCH(AMD64)
#include <algorithm>
#include "Common/CPUDetect.h"
#include "Core/MemMap.h"
#include "Core/MIPS/x86/X64IRJit.h"
#include "Core/MIPS/x86/X64IRRegCache.h"
// This file contains compilation for vector instructions.
//
// All functions should have CONDITIONAL_DISABLE, so we can narrow things down to a file quickly.
// Currently known non working ones should have DISABLE. No flags because that's in IR already.
// #define CONDITIONAL_DISABLE { CompIR_Generic(inst); return; }
#define CONDITIONAL_DISABLE {}
#define DISABLE { CompIR_Generic(inst); return; }
#define INVALIDOP { _assert_msg_(false, "Invalid IR inst %d", (int)inst.op); CompIR_Generic(inst); return; }
namespace MIPSComp {
using namespace Gen;
using namespace X64IRJitConstants;
static bool Overlap(IRReg r1, int l1, IRReg r2, int l2) {
return r1 < r2 + l2 && r1 + l1 > r2;
}
void X64JitBackend::EmitVecConstants() {
static const float vec4InitData[8][4] = {
{ 0.0f, 0.0f, 0.0f, 0.0f },
{ 1.0f, 1.0f, 1.0f, 1.0f },
{ -1.0f, -1.0f, -1.0f, -1.0f },
{ 1.0f, 0.0f, 0.0f, 0.0f },
{ 0.0f, 1.0f, 0.0f, 0.0f },
{ 0.0f, 0.0f, 1.0f, 0.0f },
{ 0.0f, 0.0f, 0.0f, 1.0f },
};
constants.vec4InitValues = (const Float4Constant *)GetCodePointer();
for (size_t type = 0; type < ARRAY_SIZE(vec4InitData); ++type) {
for (int i = 0; i < 4; ++i) {
uint32_t val;
memcpy(&val, &vec4InitData[type][i], sizeof(val));
Write32(val);
}
}
}
void X64JitBackend::CompIR_VecArith(IRInst inst) {
CONDITIONAL_DISABLE;
switch (inst.op) {
case IROp::Vec4Add:
regs_.Map(inst);
if (inst.dest == inst.src1) {
ADDPS(regs_.FX(inst.dest), regs_.F(inst.src2));
} else if (inst.dest == inst.src2) {
ADDPS(regs_.FX(inst.dest), regs_.F(inst.src1));
} else if (cpu_info.bAVX) {
VADDPS(128, regs_.FX(inst.dest), regs_.FX(inst.src1), regs_.F(inst.src2));
} else {
MOVAPS(regs_.FX(inst.dest), regs_.F(inst.src1));
ADDPS(regs_.FX(inst.dest), regs_.F(inst.src2));
}
break;
case IROp::Vec4Sub:
if (inst.dest == inst.src1) {
regs_.Map(inst);
SUBPS(regs_.FX(inst.dest), regs_.F(inst.src2));
} else if (cpu_info.bAVX) {
regs_.Map(inst);
VSUBPS(128, regs_.FX(inst.dest), regs_.FX(inst.src1), regs_.F(inst.src2));
} else if (inst.dest == inst.src2) {
X64Reg tempReg = regs_.MapWithFPRTemp(inst);
MOVAPS(tempReg, regs_.F(inst.src2));
MOVAPS(regs_.FX(inst.dest), regs_.F(inst.src1));
SUBPS(regs_.FX(inst.dest), R(tempReg));
} else {
regs_.Map(inst);
MOVAPS(regs_.FX(inst.dest), regs_.F(inst.src1));
SUBPS(regs_.FX(inst.dest), regs_.F(inst.src2));
}
break;
case IROp::Vec4Mul:
regs_.Map(inst);
if (inst.dest == inst.src1) {
MULPS(regs_.FX(inst.dest), regs_.F(inst.src2));
} else if (inst.dest == inst.src2) {
MULPS(regs_.FX(inst.dest), regs_.F(inst.src1));
} else if (cpu_info.bAVX) {
VMULPS(128, regs_.FX(inst.dest), regs_.FX(inst.src1), regs_.F(inst.src2));
} else {
MOVAPS(regs_.FX(inst.dest), regs_.F(inst.src1));
MULPS(regs_.FX(inst.dest), regs_.F(inst.src2));
}
break;
case IROp::Vec4Div:
if (inst.dest == inst.src1) {
regs_.Map(inst);
DIVPS(regs_.FX(inst.dest), regs_.F(inst.src2));
} else if (cpu_info.bAVX) {
regs_.Map(inst);
VDIVPS(128, regs_.FX(inst.dest), regs_.FX(inst.src1), regs_.F(inst.src2));
} else if (inst.dest == inst.src2) {
X64Reg tempReg = regs_.MapWithFPRTemp(inst);
MOVAPS(tempReg, regs_.F(inst.src2));
MOVAPS(regs_.FX(inst.dest), regs_.F(inst.src1));
DIVPS(regs_.FX(inst.dest), R(tempReg));
} else {
regs_.Map(inst);
MOVAPS(regs_.FX(inst.dest), regs_.F(inst.src1));
DIVPS(regs_.FX(inst.dest), regs_.F(inst.src2));
}
break;
case IROp::Vec4Scale:
// TODO: Handle "aliasing" of sizes.
if (Overlap(inst.dest, 4, inst.src2, 1) || Overlap(inst.src1, 4, inst.src2, 1))
DISABLE;
regs_.Map(inst);
SHUFPS(regs_.FX(inst.src2), regs_.F(inst.src2), 0);
if (inst.dest == inst.src1) {
MULPS(regs_.FX(inst.dest), regs_.F(inst.src2));
} else if (inst.dest == inst.src2) {
MULPS(regs_.FX(inst.dest), regs_.F(inst.src1));
} else if (cpu_info.bAVX) {
VMULPS(128, regs_.FX(inst.dest), regs_.FX(inst.src1), regs_.F(inst.src2));
} else {
MOVAPS(regs_.FX(inst.dest), regs_.F(inst.src1));
MULPS(regs_.FX(inst.dest), regs_.F(inst.src2));
}
break;
case IROp::Vec4Neg:
regs_.Map(inst);
if (cpu_info.bAVX) {
VXORPS(128, regs_.FX(inst.dest), regs_.FX(inst.src1), M(constants.signBitAll)); // rip accessible
} else {
if (inst.dest != inst.src1)
MOVAPS(regs_.FX(inst.dest), regs_.F(inst.src1));
XORPS(regs_.FX(inst.dest), M(constants.signBitAll)); // rip accessible
}
break;
case IROp::Vec4Abs:
regs_.Map(inst);
if (cpu_info.bAVX) {
VANDPS(128, regs_.FX(inst.dest), regs_.FX(inst.src1), M(constants.noSignMask)); // rip accessible
} else {
if (inst.dest != inst.src1)
MOVAPS(regs_.FX(inst.dest), regs_.F(inst.src1));
ANDPS(regs_.FX(inst.dest), M(constants.noSignMask)); // rip accessible
}
break;
default:
INVALIDOP;
break;
}
}
void X64JitBackend::CompIR_VecAssign(IRInst inst) {
CONDITIONAL_DISABLE;
switch (inst.op) {
case IROp::Vec4Init:
regs_.Map(inst);
if (inst.src1 == (int)Vec4Init::AllZERO) {
XORPS(regs_.FX(inst.dest), regs_.F(inst.dest));
} else {
MOVAPS(regs_.FX(inst.dest), M(&constants.vec4InitValues[inst.src1])); // rip accessible
}
break;
case IROp::Vec4Shuffle:
if (regs_.GetFPRLaneCount(inst.src1) == 1 && (inst.src1 & 3) == 0 && inst.src2 == 0) {
// This is a broadcast. If dest == src1, this won't clear it.
regs_.SpillLockFPR(inst.src1);
regs_.MapVec4(inst.dest, MIPSMap::NOINIT);
} else {
regs_.Map(inst);
}
if (cpu_info.bAVX) {
VPERMILPS(128, regs_.FX(inst.dest), regs_.F(inst.src1), inst.src2);
} else {
if (inst.dest != inst.src1)
MOVAPS(regs_.FX(inst.dest), regs_.F(inst.src1));
SHUFPS(regs_.FX(inst.dest), regs_.F(inst.dest), inst.src2);
}
break;
case IROp::Vec4Blend:
if (cpu_info.bAVX) {
regs_.Map(inst);
VBLENDPS(128, regs_.FX(inst.dest), regs_.FX(inst.src1), regs_.F(inst.src2), (uint8_t)inst.constant);
} else if (cpu_info.bSSE4_1) {
regs_.Map(inst);
if (inst.dest == inst.src1) {
BLENDPS(regs_.FX(inst.dest), regs_.F(inst.src2), (uint8_t)inst.constant);
} else if (inst.dest == inst.src2) {
BLENDPS(regs_.FX(inst.dest), regs_.F(inst.src1), (uint8_t)~inst.constant);
} else {
MOVAPS(regs_.FX(inst.dest), regs_.F(inst.src1));
BLENDPS(regs_.FX(inst.dest), regs_.F(inst.src2), (uint8_t)inst.constant);
}
} else {
// Could use some shuffles...
DISABLE;
}
break;
case IROp::Vec4Mov:
regs_.Map(inst);
MOVAPS(regs_.FX(inst.dest), regs_.F(inst.src1));
break;
default:
INVALIDOP;
break;
}
}
void X64JitBackend::CompIR_VecHoriz(IRInst inst) {
CONDITIONAL_DISABLE;
switch (inst.op) {
case IROp::Vec4Dot:
{
// TODO: Handle "aliasing" of sizes. In theory it should be fine if not dirty...
if (Overlap(inst.dest, 1, inst.src1, 4) || Overlap(inst.dest, 1, inst.src2, 4))
DISABLE;
X64Reg tempReg = regs_.MapWithFPRTemp(inst);
if (inst.dest == inst.src1) {
MULPS(regs_.FX(inst.dest), regs_.F(inst.src2));
} else if (inst.dest == inst.src2) {
MULPS(regs_.FX(inst.dest), regs_.F(inst.src1));
} else if (cpu_info.bAVX) {
VMULPS(128, regs_.FX(inst.dest), regs_.FX(inst.src1), regs_.F(inst.src2));
} else if (cpu_info.bSSE4_1) {
MOVAPS(regs_.FX(inst.dest), regs_.F(inst.src1));
MULPS(regs_.FX(inst.dest), regs_.F(inst.src2));
}
// This shuffle can be done in one op for SSE3/AVX, but it's not always faster.
MOVAPS(tempReg, regs_.F(inst.dest));
SHUFPS(tempReg, regs_.F(inst.dest), VFPU_SWIZZLE(1, 0, 3, 2));
ADDPS(regs_.FX(inst.dest), R(tempReg));
MOVHLPS(tempReg, regs_.FX(inst.dest));
ADDSS(regs_.FX(inst.dest), R(tempReg));
break;
}
default:
INVALIDOP;
break;
}
}
void X64JitBackend::CompIR_VecPack(IRInst inst) {
CONDITIONAL_DISABLE;
// Only SSE2 here. A destination overlapping the source goes the slow way, it's rare.
switch (inst.op) {
case IROp::Vec4Pack32To8:
case IROp::Vec4Pack31To8:
{
if (Overlap(inst.dest, 1, inst.src1, 4)) {
DISABLE;
}
// The top byte of each lane, packed into one lane. For 31, the byte below the sign, with
// negative lanes clamped to zero: the arithmetic shift makes them negative, and PACKUSWB
// saturates them to 0.
X64Reg tempReg = regs_.MapWithFPRTemp(inst);
MOVDQA(tempReg, regs_.F(inst.src1));
if (inst.op == IROp::Vec4Pack31To8)
PSRAD(tempReg, 23);
else
PSRLD(tempReg, 24);
PACKSSDW(tempReg, R(tempReg));
PACKUSWB(tempReg, R(tempReg));
MOVDQA(regs_.FX(inst.dest), R(tempReg));
break;
}
case IROp::Vec2Pack32To16:
case IROp::Vec2Pack31To16:
{
if (Overlap(inst.dest, 1, inst.src1, 2)) {
DISABLE;
}
// The top 16 bits of each lane (for 31, the 16 below the sign, with negative lanes clamped
// to zero), packed into one lane. The arithmetic shift keeps them in PACKSSDW's range, so
// the bits come through unchanged.
X64Reg tempReg = regs_.MapWithFPRTemp(inst);
MOVDQA(tempReg, regs_.F(inst.src1));
if (inst.op == IROp::Vec2Pack31To16) {
PSRAD(tempReg, 31);
PANDN(tempReg, regs_.F(inst.src1));
PSLLD(tempReg, 1);
}
PSRAD(tempReg, 16);
PACKSSDW(tempReg, R(tempReg));
MOVDQA(regs_.FX(inst.dest), R(tempReg));
break;
}
case IROp::Vec4Unpack8To32:
{
if (Overlap(inst.dest, 4, inst.src1, 1)) {
DISABLE;
}
// Each byte into the top of a lane: duplicate bytes, then words, then shift the rest out.
X64Reg tempReg = regs_.MapWithFPRTemp(inst);
MOVDQA(tempReg, regs_.F(inst.src1));
PUNPCKLBW(tempReg, R(tempReg));
PUNPCKLWD(tempReg, R(tempReg));
PSLLD(tempReg, 24);
MOVDQA(regs_.FX(inst.dest), R(tempReg));
break;
}
case IROp::Vec2Unpack16To32:
case IROp::Vec2Unpack16To31:
{
if (Overlap(inst.dest, 2, inst.src1, 1)) {
DISABLE;
}
// Each half into the top of a lane, shifted down one more for 31.
X64Reg tempReg = regs_.MapWithFPRTemp(inst);
MOVDQA(tempReg, regs_.F(inst.src1));
PUNPCKLWD(tempReg, R(tempReg));
PSLLD(tempReg, 16);
if (inst.op == IROp::Vec2Unpack16To31)
PSRLD(tempReg, 1);
MOVDQA(regs_.FX(inst.dest), R(tempReg));
break;
}
case IROp::Vec4DuplicateUpperBitsAndShift1:
{
if (inst.dest != inst.src1 && Overlap(inst.dest, 4, inst.src1, 4)) {
DISABLE;
}
// v |= v >> 8; v |= v >> 16; v >>= 1, which turns 0xFF000000 into 0x7FFFFFFF.
X64Reg tempReg = regs_.MapWithFPRTemp(inst);
MOVDQA(tempReg, regs_.F(inst.src1));
PSRLD(tempReg, 8);
POR(tempReg, regs_.F(inst.src1));
MOVDQA(regs_.FX(inst.dest), R(tempReg));
PSRLD(tempReg, 16);
POR(regs_.FX(inst.dest), R(tempReg));
PSRLD(regs_.FX(inst.dest), 1);
break;
}
default:
INVALIDOP;
break;
}
}
} // namespace MIPSComp
#endif
Reference in new issue View Git Blame Copy Permalink