Revert r129401 for now. Clang is using the old way of doing things.
llvm-svn: 129403
diff --git a/llvm/lib/Target/X86/X86InstrInfo.cpp b/llvm/lib/Target/X86/X86InstrInfo.cpp
index 85ab916..21df57c 100644
--- a/llvm/lib/Target/X86/X86InstrInfo.cpp
+++ b/llvm/lib/Target/X86/X86InstrInfo.cpp
@@ -916,6 +916,7 @@
case X86::MOVSDrm:
case X86::MOVAPSrm:
case X86::MOVUPSrm:
+ case X86::MOVUPSrm_Int:
case X86::MOVAPDrm:
case X86::MOVDQArm:
case X86::MMX_MOVD64rm:
@@ -2844,9 +2845,11 @@
case X86::FsMOVAPDrm:
case X86::MOVAPSrm:
case X86::MOVUPSrm:
+ case X86::MOVUPSrm_Int:
case X86::MOVAPDrm:
case X86::MOVDQArm:
case X86::MOVDQUrm:
+ case X86::MOVDQUrm_Int:
break;
}
switch (Opc2) {
@@ -2866,9 +2869,11 @@
case X86::FsMOVAPDrm:
case X86::MOVAPSrm:
case X86::MOVUPSrm:
+ case X86::MOVUPSrm_Int:
case X86::MOVAPDrm:
case X86::MOVDQArm:
case X86::MOVDQUrm:
+ case X86::MOVDQUrm_Int:
break;
}
diff --git a/llvm/lib/Target/X86/X86InstrSSE.td b/llvm/lib/Target/X86/X86InstrSSE.td
index c36e975..8f08e68 100644
--- a/llvm/lib/Target/X86/X86InstrSSE.td
+++ b/llvm/lib/Target/X86/X86InstrSSE.td
@@ -329,6 +329,15 @@
// Intrinsic forms of MOVUPS/D load and store
let isAsmParserOnly = 0 in {
+ let canFoldAsLoad = 1, isReMaterializable = 1 in
+ def VMOVUPSrm_Int : VPSI<0x10, MRMSrcMem, (outs VR128:$dst),
+ (ins f128mem:$src),
+ "movups\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse_loadu_ps addr:$src))]>, VEX;
+ def VMOVUPDrm_Int : VPDI<0x10, MRMSrcMem, (outs VR128:$dst),
+ (ins f128mem:$src),
+ "movupd\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_loadu_pd addr:$src))]>, VEX;
def VMOVUPSmr_Int : VPSI<0x11, MRMDestMem, (outs),
(ins f128mem:$dst, VR128:$src),
"movups\t{$src, $dst|$dst, $src}",
@@ -338,6 +347,13 @@
"movupd\t{$src, $dst|$dst, $src}",
[(int_x86_sse2_storeu_pd addr:$dst, VR128:$src)]>, VEX;
}
+let canFoldAsLoad = 1, isReMaterializable = 1 in
+def MOVUPSrm_Int : PSI<0x10, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
+ "movups\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse_loadu_ps addr:$src))]>;
+def MOVUPDrm_Int : PDI<0x10, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
+ "movupd\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_loadu_pd addr:$src))]>;
def MOVUPSmr_Int : PSI<0x11, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src),
"movups\t{$src, $dst|$dst, $src}",
@@ -2213,12 +2229,22 @@
// Intrinsic forms of MOVDQU load and store
let isAsmParserOnly = 0 in {
+let canFoldAsLoad = 1 in
+def VMOVDQUrm_Int : I<0x6F, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
+ "vmovdqu\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_loadu_dq addr:$src))]>,
+ XS, VEX, Requires<[HasAVX]>;
def VMOVDQUmr_Int : I<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src),
"vmovdqu\t{$src, $dst|$dst, $src}",
[(int_x86_sse2_storeu_dq addr:$dst, VR128:$src)]>,
XS, VEX, Requires<[HasAVX]>;
}
+let canFoldAsLoad = 1 in
+def MOVDQUrm_Int : I<0x6F, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
+ "movdqu\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_loadu_dq addr:$src))]>,
+ XS, Requires<[HasSSE2]>;
def MOVDQUmr_Int : I<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src),
"movdqu\t{$src, $dst|$dst, $src}",
[(int_x86_sse2_storeu_dq addr:$dst, VR128:$src)]>,
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index 726105f..875e9ca 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -537,7 +537,11 @@
break;
case Intrinsic::ppc_altivec_lvx:
case Intrinsic::ppc_altivec_lvxl:
- // Turn PPC lvx -> load if the pointer is known aligned.
+ case Intrinsic::x86_sse_loadu_ps:
+ case Intrinsic::x86_sse2_loadu_pd:
+ case Intrinsic::x86_sse2_loadu_dq:
+ // Turn PPC lvx -> load if the pointer is known aligned.
+ // Turn X86 loadups -> load if the pointer is known aligned.
if (getOrEnforceKnownAlignment(II->getArgOperand(0), 16, TD) >= 16) {
Value *Ptr = Builder->CreateBitCast(II->getArgOperand(0),
PointerType::getUnqual(II->getType()));
diff --git a/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp b/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp
index f6e2c88..87e78fa 100644
--- a/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp
+++ b/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp
@@ -572,6 +572,9 @@
switch (II->getIntrinsicID()) {
default: break;
case Intrinsic::prefetch:
+ case Intrinsic::x86_sse2_loadu_dq:
+ case Intrinsic::x86_sse2_loadu_pd:
+ case Intrinsic::x86_sse_loadu_ps:
case Intrinsic::x86_sse_storeu_ps:
case Intrinsic::x86_sse2_storeu_pd:
case Intrinsic::x86_sse2_storeu_dq:
diff --git a/llvm/lib/VMCore/AutoUpgrade.cpp b/llvm/lib/VMCore/AutoUpgrade.cpp
index 4541f38..9e551bb 100644
--- a/llvm/lib/VMCore/AutoUpgrade.cpp
+++ b/llvm/lib/VMCore/AutoUpgrade.cpp
@@ -527,12 +527,6 @@
// or 0.
NewFn = 0;
return true;
- } else if (Name.compare(5, 16, "x86.sse.loadu.ps", 16) == 0 ||
- Name.compare(5, 17, "x86.sse2.loadu.dq", 17) == 0 ||
- Name.compare(5, 17, "x86.sse2.loadu.pd", 17) == 0) {
- // Calls to these instructions are transformed into unaligned loads.
- NewFn = 0;
- return true;
} else if (Name.compare(5, 17, "x86.ssse3.pshuf.w", 17) == 0) {
// This is an SSE/MMX instruction.
const Type *X86_MMXTy = VectorType::getX86_MMXTy(FTy->getContext());
@@ -952,29 +946,7 @@
// Remove upgraded instruction.
CI->eraseFromParent();
-
- } else if (F->getName() == "llvm.x86.sse.loadu.ps" ||
- F->getName() == "llvm.x86.sse2.loadu.dq" ||
- F->getName() == "llvm.x86.sse2.loadu.pd") {
- // Convert to a native, unaligned load.
- const Type *VecTy = CI->getType();
- const Type *IntTy = IntegerType::get(C, 128);
- IRBuilder<> Builder(C);
- Builder.SetInsertPoint(CI->getParent(), CI);
-
- Value *BC = Builder.CreateBitCast(CI->getArgOperand(0),
- PointerType::getUnqual(IntTy),
- "cast");
- LoadInst *LI = Builder.CreateLoad(BC, CI->getName());
- LI->setAlignment(1); // Unaligned load.
- BC = Builder.CreateBitCast(LI, VecTy, "new.cast");
-
- // Fix up all the uses with our new load.
- if (!CI->use_empty())
- CI->replaceAllUsesWith(BC);
-
- // Remove intrinsic.
- CI->eraseFromParent();
+
} else {
llvm_unreachable("Unknown function for CallInst upgrade.");
}