archive/rpcs3
1
0
Fork 0
mirror of https://github.com/RPCS3/rpcs3.git synced 2025-07-06 15:01:28 +12:00
Code Issues Projects Releases Packages Wiki Activity Actions

Added support for a few more vector instructions

Browse source
This commit is contained in:
S Gopal Rajagopal 2014-09-15 19:56:15 +05:30
parent d8d0c0d2fe
commit eaa644dc20
4 changed files with 181 additions and 61 deletions
Download patch file Download diff file Expand all files Collapse all files

153
rpcs3/Emu/Cell/PPULLVMRecompiler.cpp
View file

@ -394,32 +394,37 @@ void PPULLVMRecompiler::VCMPBFP_(u32 vd, u32 va, u32 vb) {
}
void PPULLVMRecompiler::VCMPEQFP(u32 vd, u32 va, u32 vb) {
for (uint w = 0; w < 4; w++) {
m_ppu.VPR[vd]._u32[w] = m_ppu.VPR[va]._f[w] == m_ppu.VPR[vb]._f[w] ? 0xffffffff : 0;
}
auto va_v4f32 = GetVrAsFloatVec(va);
auto vb_v4f32 = GetVrAsFloatVec(vb);
auto cmp_v4i1 = m_ir_builder.CreateFCmpOEQ(va_v4f32, vb_v4f32);
auto cmp_v4i32 = m_ir_builder.CreateSExt(cmp_v4i1, VectorType::get(Type::getInt32Ty(m_llvm_context), 4));
SetVr(vd, cmp_v4i32);
}
void PPULLVMRecompiler::VCMPEQFP_(u32 vd, u32 va, u32 vb) {
int all_equal = 0x8;
int none_equal = 0x2;
VCMPEQFP(vd, va, vb);
for (uint w = 0; w < 4; w++) {
if (m_ppu.VPR[va]._f[w] == m_ppu.VPR[vb]._f[w]) {
m_ppu.VPR[vd]._u32[w] = 0xffffffff;
none_equal = 0;
} else {
m_ppu.VPR[vd]._u32[w] = 0;
all_equal = 0;
}
}
m_ppu.CR.cr6 = all_equal | none_equal;
auto vd_v16i8 = GetVrAsIntVec(vd, 8);
auto cr_i32 = (Value *)m_ir_builder.CreateLoad(m_cr);
cr_i32 = m_ir_builder.CreateAnd(cr_i32, 0xFFFFFF0F);
auto vd_mask_i32 = m_ir_builder.CreateCall(Intrinsic::getDeclaration(m_module, Intrinsic::x86_sse2_pmovmskb_128), vd_v16i8);
auto cmp_i1 = m_ir_builder.CreateICmpEQ(vd_mask_i32, m_ir_builder.getInt32(0));
auto cmp_i32 = m_ir_builder.CreateZExt(cmp_i1, Type::getInt32Ty(m_llvm_context));
cmp_i32 = m_ir_builder.CreateShl(cmp_i32, 5);
cr_i32 = m_ir_builder.CreateOr(cr_i32, cmp_i32);
cmp_i1 = m_ir_builder.CreateICmpEQ(vd_mask_i32, m_ir_builder.getInt32(0xFFFF));
cmp_i32 = m_ir_builder.CreateZExt(cmp_i1, Type::getInt32Ty(m_llvm_context));
cmp_i32 = m_ir_builder.CreateShl(cmp_i32, 7);
cr_i32 = m_ir_builder.CreateOr(cr_i32, cmp_i32);
m_ir_builder.CreateStore(cr_i32, m_cr);
}
void PPULLVMRecompiler::VCMPEQUB(u32 vd, u32 va, u32 vb) {
for (uint b = 0; b < 16; b++) {
m_ppu.VPR[vd]._u8[b] = m_ppu.VPR[va]._u8[b] == m_ppu.VPR[vb]._u8[b] ? 0xff : 0;
}
auto va_v16i8 = GetVrAsIntVec(va, 8);
auto vb_v16i8 = GetVrAsIntVec(va, 8);
auto cmp_v16i1 = m_ir_builder.CreateICmpEQ(va_v16i8, vb_v16i8);
auto cmp_v16i32 = m_ir_builder.CreateSExt(cmp_v16i1, VectorType::get(Type::getInt8Ty(m_llvm_context), 16));
SetVr(vd, cmp_v16i32);
}
void PPULLVMRecompiler::VCMPEQUB_(u32 vd, u32 va, u32 vb) {
@ -439,14 +444,15 @@ void PPULLVMRecompiler::VCMPEQUB_(u32 vd, u32 va, u32 vb) {
m_ppu.CR.cr6 = all_equal | none_equal;
}
void PPULLVMRecompiler::VCMPEQUH(u32 vd, u32 va, u32 vb) //nf
{
for (uint h = 0; h < 8; h++) {
m_ppu.VPR[vd]._u16[h] = m_ppu.VPR[va]._u16[h] == m_ppu.VPR[vb]._u16[h] ? 0xffff : 0;
}
void PPULLVMRecompiler::VCMPEQUH(u32 vd, u32 va, u32 vb) {
auto va_v8i16 = GetVrAsIntVec(va, 16);
auto vb_v8i16 = GetVrAsIntVec(va, 16);
auto cmp_v8i1 = m_ir_builder.CreateICmpEQ(va_v8i16, vb_v8i16);
auto cmp_v8i16 = m_ir_builder.CreateSExt(cmp_v8i1, VectorType::get(Type::getInt16Ty(m_llvm_context), 8));
SetVr(vd, cmp_v8i16);
}
void PPULLVMRecompiler::VCMPEQUH_(u32 vd, u32 va, u32 vb) //nf
{
void PPULLVMRecompiler::VCMPEQUH_(u32 vd, u32 va, u32 vb) {
int all_equal = 0x8;
int none_equal = 0x2;
@ -462,11 +468,15 @@ void PPULLVMRecompiler::VCMPEQUH_(u32 vd, u32 va, u32 vb) //nf
m_ppu.CR.cr6 = all_equal | none_equal;
}
void PPULLVMRecompiler::VCMPEQUW(u32 vd, u32 va, u32 vb) {
for (uint w = 0; w < 4; w++) {
m_ppu.VPR[vd]._u32[w] = m_ppu.VPR[va]._u32[w] == m_ppu.VPR[vb]._u32[w] ? 0xffffffff : 0;
}
auto va_v4i32 = GetVrAsIntVec(va, 32);
auto vb_v4i32 = GetVrAsIntVec(va, 32);
auto cmp_v4i1 = m_ir_builder.CreateICmpEQ(va_v4i32, vb_v4i32);
auto cmp_v4i32 = m_ir_builder.CreateSExt(cmp_v4i1, VectorType::get(Type::getInt32Ty(m_llvm_context), 4));
SetVr(vd, cmp_v4i32);
}
void PPULLVMRecompiler::VCMPEQUW_(u32 vd, u32 va, u32 vb) {
int all_equal = 0x8;
int none_equal = 0x2;
@ -483,11 +493,15 @@ void PPULLVMRecompiler::VCMPEQUW_(u32 vd, u32 va, u32 vb) {
m_ppu.CR.cr6 = all_equal | none_equal;
}
void PPULLVMRecompiler::VCMPGEFP(u32 vd, u32 va, u32 vb) {
for (uint w = 0; w < 4; w++) {
m_ppu.VPR[vd]._u32[w] = m_ppu.VPR[va]._f[w] >= m_ppu.VPR[vb]._f[w] ? 0xffffffff : 0;
}
auto va_v4f32 = GetVrAsFloatVec(va);
auto vb_v4f32 = GetVrAsFloatVec(vb);
auto cmp_v4i1 = m_ir_builder.CreateFCmpOGE(va_v4f32, vb_v4f32);
auto cmp_v4i32 = m_ir_builder.CreateSExt(cmp_v4i1, VectorType::get(Type::getInt32Ty(m_llvm_context), 4));
SetVr(vd, cmp_v4i32);
}
void PPULLVMRecompiler::VCMPGEFP_(u32 vd, u32 va, u32 vb) {
int all_ge = 0x8;
int none_ge = 0x2;
@ -504,11 +518,15 @@ void PPULLVMRecompiler::VCMPGEFP_(u32 vd, u32 va, u32 vb) {
m_ppu.CR.cr6 = all_ge | none_ge;
}
void PPULLVMRecompiler::VCMPGTFP(u32 vd, u32 va, u32 vb) {
for (uint w = 0; w < 4; w++) {
m_ppu.VPR[vd]._u32[w] = m_ppu.VPR[va]._f[w] > m_ppu.VPR[vb]._f[w] ? 0xffffffff : 0;
}
auto va_v4f32 = GetVrAsFloatVec(va);
auto vb_v4f32 = GetVrAsFloatVec(vb);
auto cmp_v4i1 = m_ir_builder.CreateFCmpOGT(va_v4f32, vb_v4f32);
auto cmp_v4i32 = m_ir_builder.CreateSExt(cmp_v4i1, VectorType::get(Type::getInt32Ty(m_llvm_context), 4));
SetVr(vd, cmp_v4i32);
}
void PPULLVMRecompiler::VCMPGTFP_(u32 vd, u32 va, u32 vb) {
int all_ge = 0x8;
int none_ge = 0x2;
@ -525,12 +543,15 @@ void PPULLVMRecompiler::VCMPGTFP_(u32 vd, u32 va, u32 vb) {
m_ppu.CR.cr6 = all_ge | none_ge;
}
void PPULLVMRecompiler::VCMPGTSB(u32 vd, u32 va, u32 vb) //nf
{
for (uint b = 0; b < 16; b++) {
m_ppu.VPR[vd]._u8[b] = m_ppu.VPR[va]._s8[b] > m_ppu.VPR[vb]._s8[b] ? 0xff : 0;
}
void PPULLVMRecompiler::VCMPGTSB(u32 vd, u32 va, u32 vb) {
auto va_v16i8 = GetVrAsIntVec(va, 8);
auto vb_v16i8 = GetVrAsIntVec(va, 8);
auto cmp_v16i1 = m_ir_builder.CreateICmpSGT(va_v16i8, vb_v16i8);
auto cmp_v16i32 = m_ir_builder.CreateSExt(cmp_v16i1, VectorType::get(Type::getInt8Ty(m_llvm_context), 16));
SetVr(vd, cmp_v16i32);
}
void PPULLVMRecompiler::VCMPGTSB_(u32 vd, u32 va, u32 vb) {
int all_gt = 0x8;
int none_gt = 0x2;
@ -547,11 +568,15 @@ void PPULLVMRecompiler::VCMPGTSB_(u32 vd, u32 va, u32 vb) {
m_ppu.CR.cr6 = all_gt | none_gt;
}
void PPULLVMRecompiler::VCMPGTSH(u32 vd, u32 va, u32 vb) {
for (uint h = 0; h < 8; h++) {
m_ppu.VPR[vd]._u16[h] = m_ppu.VPR[va]._s16[h] > m_ppu.VPR[vb]._s16[h] ? 0xffff : 0;
}
auto va_v8i16 = GetVrAsIntVec(va, 16);
auto vb_v8i16 = GetVrAsIntVec(va, 16);
auto cmp_v8i1 = m_ir_builder.CreateICmpSGT(va_v8i16, vb_v8i16);
auto cmp_v8i16 = m_ir_builder.CreateSExt(cmp_v8i1, VectorType::get(Type::getInt16Ty(m_llvm_context), 8));
SetVr(vd, cmp_v8i16);
}
void PPULLVMRecompiler::VCMPGTSH_(u32 vd, u32 va, u32 vb) {
int all_gt = 0x8;
int none_gt = 0x2;
@ -568,11 +593,15 @@ void PPULLVMRecompiler::VCMPGTSH_(u32 vd, u32 va, u32 vb) {
m_ppu.CR.cr6 = all_gt | none_gt;
}
void PPULLVMRecompiler::VCMPGTSW(u32 vd, u32 va, u32 vb) {
for (uint w = 0; w < 4; w++) {
m_ppu.VPR[vd]._u32[w] = m_ppu.VPR[va]._s32[w] > m_ppu.VPR[vb]._s32[w] ? 0xffffffff : 0;
}
auto va_v4i32 = GetVrAsIntVec(va, 32);
auto vb_v4i32 = GetVrAsIntVec(va, 32);
auto cmp_v4i1 = m_ir_builder.CreateICmpSGT(va_v4i32, vb_v4i32);
auto cmp_v4i32 = m_ir_builder.CreateSExt(cmp_v4i1, VectorType::get(Type::getInt32Ty(m_llvm_context), 4));
SetVr(vd, cmp_v4i32);
}
void PPULLVMRecompiler::VCMPGTSW_(u32 vd, u32 va, u32 vb) {
int all_gt = 0x8;
int none_gt = 0x2;
@ -589,11 +618,15 @@ void PPULLVMRecompiler::VCMPGTSW_(u32 vd, u32 va, u32 vb) {
m_ppu.CR.cr6 = all_gt | none_gt;
}
void PPULLVMRecompiler::VCMPGTUB(u32 vd, u32 va, u32 vb) {
for (uint b = 0; b < 16; b++) {
m_ppu.VPR[vd]._u8[b] = m_ppu.VPR[va]._u8[b] > m_ppu.VPR[vb]._u8[b] ? 0xff : 0;
}
auto va_v16i8 = GetVrAsIntVec(va, 8);
auto vb_v16i8 = GetVrAsIntVec(va, 8);
auto cmp_v16i1 = m_ir_builder.CreateICmpUGT(va_v16i8, vb_v16i8);
auto cmp_v16i32 = m_ir_builder.CreateSExt(cmp_v16i1, VectorType::get(Type::getInt8Ty(m_llvm_context), 16));
SetVr(vd, cmp_v16i32);
}
void PPULLVMRecompiler::VCMPGTUB_(u32 vd, u32 va, u32 vb) {
int all_gt = 0x8;
int none_gt = 0x2;
@ -610,11 +643,15 @@ void PPULLVMRecompiler::VCMPGTUB_(u32 vd, u32 va, u32 vb) {
m_ppu.CR.cr6 = all_gt | none_gt;
}
void PPULLVMRecompiler::VCMPGTUH(u32 vd, u32 va, u32 vb) {
for (uint h = 0; h < 8; h++) {
m_ppu.VPR[vd]._u16[h] = m_ppu.VPR[va]._u16[h] > m_ppu.VPR[vb]._u16[h] ? 0xffff : 0;
}
auto va_v8i16 = GetVrAsIntVec(va, 16);
auto vb_v8i16 = GetVrAsIntVec(va, 16);
auto cmp_v8i1 = m_ir_builder.CreateICmpUGT(va_v8i16, vb_v8i16);
auto cmp_v8i16 = m_ir_builder.CreateSExt(cmp_v8i1, VectorType::get(Type::getInt16Ty(m_llvm_context), 8));
SetVr(vd, cmp_v8i16);
}
void PPULLVMRecompiler::VCMPGTUH_(u32 vd, u32 va, u32 vb) {
int all_gt = 0x8;
int none_gt = 0x2;
@ -631,11 +668,15 @@ void PPULLVMRecompiler::VCMPGTUH_(u32 vd, u32 va, u32 vb) {
m_ppu.CR.cr6 = all_gt | none_gt;
}
void PPULLVMRecompiler::VCMPGTUW(u32 vd, u32 va, u32 vb) {
for (uint w = 0; w < 4; w++) {
m_ppu.VPR[vd]._u32[w] = m_ppu.VPR[va]._u32[w] > m_ppu.VPR[vb]._u32[w] ? 0xffffffff : 0;
}
auto va_v4i32 = GetVrAsIntVec(va, 32);
auto vb_v4i32 = GetVrAsIntVec(va, 32);
auto cmp_v4i1 = m_ir_builder.CreateICmpUGT(va_v4i32, vb_v4i32);
auto cmp_v4i32 = m_ir_builder.CreateSExt(cmp_v4i1, VectorType::get(Type::getInt32Ty(m_llvm_context), 4));
SetVr(vd, cmp_v4i32);
}
void PPULLVMRecompiler::VCMPGTUW_(u32 vd, u32 va, u32 vb) {
int all_gt = 0x8;
int none_gt = 0x2;
@ -652,6 +693,7 @@ void PPULLVMRecompiler::VCMPGTUW_(u32 vd, u32 va, u32 vb) {
m_ppu.CR.cr6 = all_gt | none_gt;
}
void PPULLVMRecompiler::VCTSXS(u32 vd, u32 uimm5, u32 vb) {
int nScale = 1 << uimm5;
@ -666,6 +708,7 @@ void PPULLVMRecompiler::VCTSXS(u32 vd, u32 uimm5, u32 vb) {
m_ppu.VPR[vd]._s32[w] = (int)result;
}
}
void PPULLVMRecompiler::VCTUXS(u32 vd, u32 uimm5, u32 vb) {
int nScale = 1 << uimm5;