Vendor import of llvm release_40 branch r294123:vendor/llvm/llvm-release_40-r294123

https://llvm.org/svn/llvm-project/llvm/branches/release_40@294123

15 files changed, 415 insertions, 113 deletions

diff --git a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 680f62fa91bc..fd156fa7dd07 100644
--- a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -8123,9 +8123,12 @@ SDValue DAGCombiner::visitFADDForFMACombine(SDNode *N) {
   }
 
   // More folding opportunities when target permits.
-  if ((AllowFusion || HasFMAD)  && Aggressive) {
+  if (Aggressive) {
     // fold (fadd (fma x, y, (fmul u, v)), z) -> (fma x, y (fma u, v, z))
-    if (N0.getOpcode() == PreferredFusedOpcode &&
+    // FIXME: The UnsafeAlgebra flag should be propagated to FMA/FMAD, but FMF
+    // are currently only supported on binary nodes.
+    if (Options.UnsafeFPMath &&
+        N0.getOpcode() == PreferredFusedOpcode &&
         N0.getOperand(2).getOpcode() == ISD::FMUL &&
         N0->hasOneUse() && N0.getOperand(2)->hasOneUse()) {
       return DAG.getNode(PreferredFusedOpcode, SL, VT,
@@ -8137,7 +8140,10 @@ SDValue DAGCombiner::visitFADDForFMACombine(SDNode *N) {
     }
 
     // fold (fadd x, (fma y, z, (fmul u, v)) -> (fma y, z (fma u, v, x))
-    if (N1->getOpcode() == PreferredFusedOpcode &&
+    // FIXME: The UnsafeAlgebra flag should be propagated to FMA/FMAD, but FMF
+    // are currently only supported on binary nodes.
+    if (Options.UnsafeFPMath &&
+        N1->getOpcode() == PreferredFusedOpcode &&
         N1.getOperand(2).getOpcode() == ISD::FMUL &&
         N1->hasOneUse() && N1.getOperand(2)->hasOneUse()) {
       return DAG.getNode(PreferredFusedOpcode, SL, VT,
@@ -8367,10 +8373,13 @@ SDValue DAGCombiner::visitFSUBForFMACombine(SDNode *N) {
   }
 
   // More folding opportunities when target permits.
-  if ((AllowFusion || HasFMAD) && Aggressive) {
+  if (Aggressive) {
     // fold (fsub (fma x, y, (fmul u, v)), z)
     //   -> (fma x, y (fma u, v, (fneg z)))
-    if (N0.getOpcode() == PreferredFusedOpcode &&
+    // FIXME: The UnsafeAlgebra flag should be propagated to FMA/FMAD, but FMF
+    // are currently only supported on binary nodes.
+    if (Options.UnsafeFPMath &&
+        N0.getOpcode() == PreferredFusedOpcode &&
         N0.getOperand(2).getOpcode() == ISD::FMUL &&
         N0->hasOneUse() && N0.getOperand(2)->hasOneUse()) {
       return DAG.getNode(PreferredFusedOpcode, SL, VT,
@@ -8384,7 +8393,10 @@ SDValue DAGCombiner::visitFSUBForFMACombine(SDNode *N) {
 
     // fold (fsub x, (fma y, z, (fmul u, v)))
     //   -> (fma (fneg y), z, (fma (fneg u), v, x))
-    if (N1.getOpcode() == PreferredFusedOpcode &&
+    // FIXME: The UnsafeAlgebra flag should be propagated to FMA/FMAD, but FMF
+    // are currently only supported on binary nodes.
+    if (Options.UnsafeFPMath &&
+        N1.getOpcode() == PreferredFusedOpcode &&
         N1.getOperand(2).getOpcode() == ISD::FMUL) {
       SDValue N20 = N1.getOperand(2).getOperand(0);
       SDValue N21 = N1.getOperand(2).getOperand(1);
diff --git a/lib/Transforms/InstCombine/InstCombineCompares.cpp b/lib/Transforms/InstCombine/InstCombineCompares.cpp
index bca1304534b6..428f94bb5e93 100644
--- a/lib/Transforms/InstCombine/InstCombineCompares.cpp
+++ b/lib/Transforms/InstCombine/InstCombineCompares.cpp
@@ -4039,11 +4039,6 @@ Instruction *InstCombiner::foldICmpUsingKnownBits(ICmpInst &I) {
         Constant *CMinus1 = ConstantInt::get(Op0->getType(), *CmpC - 1);
         return new ICmpInst(ICmpInst::ICMP_EQ, Op0, CMinus1);
       }
-      // (x <u 2147483648) -> (x >s -1)  -> true if sign bit clear
-      if (CmpC->isMinSignedValue()) {
-        Constant *AllOnes = Constant::getAllOnesValue(Op0->getType());
-        return new ICmpInst(ICmpInst::ICMP_SGT, Op0, AllOnes);
-      }
     }
     break;
   }
@@ -4063,11 +4058,6 @@ Instruction *InstCombiner::foldICmpUsingKnownBits(ICmpInst &I) {
       if (*CmpC == Op0Max - 1)
         return new ICmpInst(ICmpInst::ICMP_EQ, Op0,
                             ConstantInt::get(Op1->getType(), *CmpC + 1));
-
-      // (x >u 2147483647) -> (x <s 0)  -> true if sign bit set
-      if (CmpC->isMaxSignedValue())
-        return new ICmpInst(ICmpInst::ICMP_SLT, Op0,
-                            Constant::getNullValue(Op0->getType()));
     }
     break;
   }
@@ -4299,6 +4289,27 @@ Instruction *InstCombiner::visitICmpInst(ICmpInst &I) {
           (SI->getOperand(2) == Op0 && SI->getOperand(1) == Op1))
         return nullptr;
 
+  // FIXME: We only do this after checking for min/max to prevent infinite
+  // looping caused by a reverse canonicalization of these patterns for min/max.
+  // FIXME: The organization of folds is a mess. These would naturally go into
+  // canonicalizeCmpWithConstant(), but we can't move all of the above folds
+  // down here after the min/max restriction.
+  ICmpInst::Predicate Pred = I.getPredicate();
+  const APInt *C;
+  if (match(Op1, m_APInt(C))) {
+    // For i32: x >u 2147483647 -> x <s 0  -> true if sign bit set
+    if (Pred == ICmpInst::ICMP_UGT && C->isMaxSignedValue()) {
+      Constant *Zero = Constant::getNullValue(Op0->getType());
+      return new ICmpInst(ICmpInst::ICMP_SLT, Op0, Zero);
+    }
+
+    // For i32: x <u 2147483648 -> x >s -1  -> true if sign bit clear
+    if (Pred == ICmpInst::ICMP_ULT && C->isMinSignedValue()) {
+      Constant *AllOnes = Constant::getAllOnesValue(Op0->getType());
+      return new ICmpInst(ICmpInst::ICMP_SGT, Op0, AllOnes);
+    }
+  }
+
   if (Instruction *Res = foldICmpInstWithConstant(I))
     return Res;
 
diff --git a/lib/Transforms/Scalar/LoopStrengthReduce.cpp b/lib/Transforms/Scalar/LoopStrengthReduce.cpp
index 01728ae680de..194587a85e7c 100644
--- a/lib/Transforms/Scalar/LoopStrengthReduce.cpp
+++ b/lib/Transforms/Scalar/LoopStrengthReduce.cpp
@@ -158,8 +158,9 @@ struct MemAccessTy {
 
   bool operator!=(MemAccessTy Other) const { return !(*this == Other); }
 
-  static MemAccessTy getUnknown(LLVMContext &Ctx) {
-    return MemAccessTy(Type::getVoidTy(Ctx), UnknownAddressSpace);
+  static MemAccessTy getUnknown(LLVMContext &Ctx,
+                                unsigned AS = UnknownAddressSpace) {
+    return MemAccessTy(Type::getVoidTy(Ctx), AS);
   }
 };
 
@@ -2279,8 +2280,10 @@ bool LSRInstance::reconcileNewOffset(LSRUse &LU, int64_t NewOffset,
   // TODO: Be less conservative when the type is similar and can use the same
   // addressing modes.
   if (Kind == LSRUse::Address) {
-    if (AccessTy != LU.AccessTy)
-      NewAccessTy = MemAccessTy::getUnknown(AccessTy.MemTy->getContext());
+    if (AccessTy.MemTy != LU.AccessTy.MemTy) {
+      NewAccessTy = MemAccessTy::getUnknown(AccessTy.MemTy->getContext(),
+                                            AccessTy.AddrSpace);
+    }
   }
 
   // Conservatively assume HasBaseReg is true for now.
diff --git a/test/CodeGen/AMDGPU/fma-combine.ll b/test/CodeGen/AMDGPU/fma-combine.ll
index ab1b6d96a8d8..50c5a5abf7fa 100644
--- a/test/CodeGen/AMDGPU/fma-combine.ll
+++ b/test/CodeGen/AMDGPU/fma-combine.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -fp-contract=fast < %s | FileCheck -check-prefix=SI-NOFMA -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs -fp-contract=fast < %s | FileCheck -check-prefix=SI-NOFMA -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -fp-contract=fast -enable-no-infs-fp-math -mattr=+fp32-denormals < %s | FileCheck -check-prefix=SI-FMA -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -fp-contract=fast < %s | FileCheck -check-prefix=SI-NOFMA -check-prefix=SI-SAFE -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs -fp-contract=fast < %s | FileCheck -check-prefix=SI-NOFMA -check-prefix=SI-SAFE -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -fp-contract=fast -enable-no-infs-fp-math -enable-unsafe-fp-math -mattr=+fp32-denormals < %s | FileCheck -check-prefix=SI-FMA -check-prefix=SI-UNSAFE -check-prefix=SI -check-prefix=FUNC %s
 
 ; Note: The SI-FMA conversions of type x * (y + 1) --> x * y + x would be
 ; beneficial even without fp32 denormals, but they do require no-infs-fp-math
@@ -308,8 +308,14 @@ define void @combine_to_fma_fsub_2_f64_2uses_mul(double addrspace(1)* noalias %o
 ; SI-DAG: buffer_load_dwordx2 [[Z:v\[[0-9]+:[0-9]+\]]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:16{{$}}
 ; SI-DAG: buffer_load_dwordx2 [[U:v\[[0-9]+:[0-9]+\]]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:24{{$}}
 ; SI-DAG: buffer_load_dwordx2 [[V:v\[[0-9]+:[0-9]+\]]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:32{{$}}
-; SI: v_fma_f64 [[FMA0:v\[[0-9]+:[0-9]+\]]], [[U]], [[V]], -[[Z]]
-; SI: v_fma_f64 [[RESULT:v\[[0-9]+:[0-9]+\]]], [[X]], [[Y]], [[FMA0]]
+
+; SI-SAFE: v_mul_f64 [[TMP0:v\[[0-9]+:[0-9]+\]]], [[U]], [[V]]
+; SI-SAFE: v_fma_f64 [[TMP1:v\[[0-9]+:[0-9]+\]]], [[X]], [[Y]], [[TMP0]]
+; SI-SAFE: v_add_f64 [[RESULT:v\[[0-9]+:[0-9]+\]]], [[TMP1]], -[[Z]]
+
+; SI-UNSAFE: v_fma_f64 [[FMA0:v\[[0-9]+:[0-9]+\]]], [[U]], [[V]], -[[Z]]
+; SI-UNSAFE: v_fma_f64 [[RESULT:v\[[0-9]+:[0-9]+\]]], [[X]], [[Y]], [[FMA0]]
+
 ; SI: buffer_store_dwordx2 [[RESULT]]
 define void @aggressive_combine_to_fma_fsub_0_f64(double addrspace(1)* noalias %out, double addrspace(1)* noalias %in) #1 {
   %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
@@ -343,8 +349,14 @@ define void @aggressive_combine_to_fma_fsub_0_f64(double addrspace(1)* noalias %
 ; SI-DAG: buffer_load_dwordx2 [[Z:v\[[0-9]+:[0-9]+\]]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:16{{$}}
 ; SI-DAG: buffer_load_dwordx2 [[U:v\[[0-9]+:[0-9]+\]]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:24{{$}}
 ; SI-DAG: buffer_load_dwordx2 [[V:v\[[0-9]+:[0-9]+\]]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:32{{$}}
-; SI: v_fma_f64 [[FMA0:v\[[0-9]+:[0-9]+\]]], -[[U]], [[V]], [[X]]
-; SI: v_fma_f64 [[RESULT:v\[[0-9]+:[0-9]+\]]], -[[Y]], [[Z]], [[FMA0]]
+
+; SI-SAFE: v_mul_f64 [[TMP0:v\[[0-9]+:[0-9]+\]]], [[U]], [[V]]
+; SI-SAFE: v_fma_f64 [[TMP1:v\[[0-9]+:[0-9]+\]]], [[Y]], [[Z]], [[TMP0]]
+; SI-SAFE: v_add_f64 [[RESULT:v\[[0-9]+:[0-9]+\]]], [[X]], -[[TMP1]]
+
+; SI-UNSAFE: v_fma_f64 [[FMA0:v\[[0-9]+:[0-9]+\]]], -[[U]], [[V]], [[X]]
+; SI-UNSAFE: v_fma_f64 [[RESULT:v\[[0-9]+:[0-9]+\]]], -[[Y]], [[Z]], [[FMA0]]
+
 ; SI: buffer_store_dwordx2 [[RESULT]]
 define void @aggressive_combine_to_fma_fsub_1_f64(double addrspace(1)* noalias %out, double addrspace(1)* noalias %in) #1 {
   %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
diff --git a/test/CodeGen/AMDGPU/mad-combine.ll b/test/CodeGen/AMDGPU/mad-combine.ll
index d141281f36b8..9caba32cbac0 100644
--- a/test/CodeGen/AMDGPU/mad-combine.ll
+++ b/test/CodeGen/AMDGPU/mad-combine.ll
@@ -1,12 +1,12 @@
 ; Make sure we still form mad even when unsafe math or fp-contract is allowed instead of fma.
 
-; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=SI-STD -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -fp-contract=fast < %s | FileCheck -check-prefix=SI -check-prefix=SI-STD -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=SI -check-prefix=SI-STD -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=SI-STD  -check-prefix=SI-STD-SAFE -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -fp-contract=fast < %s | FileCheck -check-prefix=SI -check-prefix=SI-STD -check-prefix=SI-STD-SAFE -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=SI -check-prefix=SI-STD -check-prefix=SI-STD-UNSAFE -check-prefix=FUNC %s
 
 ; Make sure we don't form mad with denormals
-; RUN: llc -march=amdgcn -mcpu=tahiti -mattr=+fp32-denormals -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=SI-DENORM -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=verde -mattr=+fp32-denormals -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=SI-DENORM-SLOWFMAF -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=tahiti -mattr=+fp32-denormals -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=SI-DENORM -check-prefix=SI-DENORM-FASTFMAF -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=verde -mattr=+fp32-denormals -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=SI-DENORM -check-prefix=SI-DENORM-SLOWFMAF -check-prefix=FUNC %s
 
 declare i32 @llvm.amdgcn.workitem.id.x() #0
 declare float @llvm.fabs.f32(float) #0
@@ -21,7 +21,7 @@ declare float @llvm.fmuladd.f32(float, float, float) #0
 
 ; SI-STD: v_mac_f32_e32 [[C]], [[B]], [[A]]
 
-; SI-DENORM: v_fma_f32 [[RESULT:v[0-9]+]], [[A]], [[B]], [[C]]
+; SI-DENORM-FASTFMAF: v_fma_f32 [[RESULT:v[0-9]+]], [[A]], [[B]], [[C]]
 
 ; SI-DENORM-SLOWFMAF-NOT: v_fma
 ; SI-DENORM-SLOWFMAF-NOT: v_mad
@@ -58,8 +58,8 @@ define void @combine_to_mad_f32_0(float addrspace(1)* noalias %out, float addrsp
 ; SI-STD-DAG: v_mac_f32_e32 [[C]], [[B]], [[A]]
 ; SI-STD-DAG: v_mac_f32_e32 [[D]], [[B]], [[A]]
 
-; SI-DENORM-DAG: v_fma_f32 [[RESULT0:v[0-9]+]], [[A]], [[B]], [[C]]
-; SI-DENORM-DAG: v_fma_f32 [[RESULT1:v[0-9]+]], [[A]], [[B]], [[D]]
+; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT0:v[0-9]+]], [[A]], [[B]], [[C]]
+; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT1:v[0-9]+]], [[A]], [[B]], [[D]]
 
 ; SI-DENORM-SLOWFMAF: v_mul_f32_e32 [[TMP:v[0-9]+]], [[B]], [[A]]
 ; SI-DENORM-SLOWFMAF-DAG: v_add_f32_e32 [[RESULT0:v[0-9]+]], [[C]], [[TMP]]
@@ -100,7 +100,7 @@ define void @combine_to_mad_f32_0_2use(float addrspace(1)* noalias %out, float a
 ; SI-DAG: buffer_load_dword [[C:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8{{$}}
 
 ; SI-STD: v_mac_f32_e32 [[C]], [[B]], [[A]]
-; SI-DENORM: v_fma_f32 [[RESULT:v[0-9]+]], [[A]], [[B]], [[C]]
+; SI-DENORM-FASTFMAF: v_fma_f32 [[RESULT:v[0-9]+]], [[A]], [[B]], [[C]]
 
 ; SI-DENORM-SLOWFMAF: v_mul_f32_e32 [[TMP:v[0-9]+]], [[B]], [[A]]
 ; SI-DENORM-SLOWFMAF: v_add_f32_e32 [[RESULT:v[0-9]+]], [[TMP]], [[C]]
@@ -131,7 +131,7 @@ define void @combine_to_mad_f32_1(float addrspace(1)* noalias %out, float addrsp
 ; SI-DAG: buffer_load_dword [[C:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8{{$}}
 
 ; SI-STD: v_mad_f32 [[RESULT:v[0-9]+]], [[A]], [[B]], -[[C]]
-; SI-DENORM: v_fma_f32 [[RESULT:v[0-9]+]], [[A]], [[B]], -[[C]]
+; SI-DENORM-FASTFMAF: v_fma_f32 [[RESULT:v[0-9]+]], [[A]], [[B]], -[[C]]
 
 ; SI-DENORM-SLOWFMAF: v_mul_f32_e32 [[TMP:v[0-9]+]], [[B]], [[A]]
 ; SI-DENORM-SLOWFMAF: v_subrev_f32_e32 [[RESULT:v[0-9]+]], [[C]], [[TMP]]
@@ -164,8 +164,8 @@ define void @combine_to_mad_fsub_0_f32(float addrspace(1)* noalias %out, float a
 ; SI-STD-DAG: v_mad_f32 [[RESULT0:v[0-9]+]], [[A]], [[B]], -[[C]]
 ; SI-STD-DAG: v_mad_f32 [[RESULT1:v[0-9]+]], [[A]], [[B]], -[[D]]
 
-; SI-DENORM-DAG: v_fma_f32 [[RESULT0:v[0-9]+]], [[A]], [[B]], -[[C]]
-; SI-DENORM-DAG: v_fma_f32 [[RESULT1:v[0-9]+]], [[A]], [[B]], -[[D]]
+; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT0:v[0-9]+]], [[A]], [[B]], -[[C]]
+; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT1:v[0-9]+]], [[A]], [[B]], -[[D]]
 
 ; SI-DENORM-SLOWFMAF: v_mul_f32_e32 [[TMP:v[0-9]+]], [[B]], [[A]]
 ; SI-DENORM-SLOWFMAF-DAG: v_subrev_f32_e32 [[RESULT0:v[0-9]+]], [[C]], [[TMP]]
@@ -203,7 +203,7 @@ define void @combine_to_mad_fsub_0_f32_2use(float addrspace(1)* noalias %out, fl
 ; SI-DAG: buffer_load_dword [[C:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8{{$}}
 
 ; SI-STD: v_mad_f32 [[RESULT:v[0-9]+]], -[[A]], [[B]], [[C]]
-; SI-DENORM: v_fma_f32 [[RESULT:v[0-9]+]], -[[A]], [[B]], [[C]]
+; SI-DENORM-FASTFMAF: v_fma_f32 [[RESULT:v[0-9]+]], -[[A]], [[B]], [[C]]
 
 ; SI-DENORM-SLOWFMAF: v_mul_f32_e32 [[TMP:v[0-9]+]], [[B]], [[A]]
 ; SI-DENORM-SLOWFMAF: v_subrev_f32_e32 [[RESULT:v[0-9]+]], [[TMP]], [[C]]
@@ -235,8 +235,8 @@ define void @combine_to_mad_fsub_1_f32(float addrspace(1)* noalias %out, float a
 ; SI-STD-DAG: v_mad_f32 [[RESULT0:v[0-9]+]], -[[A]], [[B]], [[C]]
 ; SI-STD-DAG: v_mad_f32 [[RESULT1:v[0-9]+]], -[[A]], [[B]], [[D]]
 
-; SI-DENORM-DAG: v_fma_f32 [[RESULT0:v[0-9]+]], -[[A]], [[B]], [[C]]
-; SI-DENORM-DAG: v_fma_f32 [[RESULT1:v[0-9]+]], -[[A]], [[B]], [[D]]
+; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT0:v[0-9]+]], -[[A]], [[B]], [[C]]
+; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT1:v[0-9]+]], -[[A]], [[B]], [[D]]
 
 ; SI-DENORM-SLOWFMAF: v_mul_f32_e32 [[TMP:v[0-9]+]], [[B]], [[A]]
 ; SI-DENORM-SLOWFMAF-DAG: v_subrev_f32_e32 [[RESULT0:v[0-9]+]], [[TMP]], [[C]]
@@ -275,7 +275,7 @@ define void @combine_to_mad_fsub_1_f32_2use(float addrspace(1)* noalias %out, fl
 
 ; SI-STD: v_mad_f32 [[RESULT:v[0-9]+]], [[A]], -[[B]], -[[C]]
 
-; SI-DENORM: v_fma_f32 [[RESULT:v[0-9]+]], -[[A]], [[B]], -[[C]]
+; SI-DENORM-FASTFMAF: v_fma_f32 [[RESULT:v[0-9]+]], -[[A]], [[B]], -[[C]]
 
 ; SI-DENORM-SLOWFMAF: v_mul_f32_e64 [[TMP:v[0-9]+]], [[A]], -[[B]]
 ; SI-DENORM-SLOWFMAF: v_subrev_f32_e32 [[RESULT:v[0-9]+]], [[C]], [[TMP]]
@@ -309,8 +309,8 @@ define void @combine_to_mad_fsub_2_f32(float addrspace(1)* noalias %out, float a
 ; SI-STD-DAG: v_mad_f32 [[RESULT0:v[0-9]+]], [[A]], -[[B]], -[[C]]
 ; SI-STD-DAG: v_mad_f32 [[RESULT1:v[0-9]+]], [[A]], -[[B]], -[[D]]
 
-; SI-DENORM-DAG: v_fma_f32 [[RESULT0:v[0-9]+]], -[[A]], [[B]], -[[C]]
-; SI-DENORM-DAG: v_fma_f32 [[RESULT1:v[0-9]+]], -[[A]], [[B]], -[[D]]
+; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT0:v[0-9]+]], -[[A]], [[B]], -[[C]]
+; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT1:v[0-9]+]], -[[A]], [[B]], -[[D]]
 
 ; SI-DENORM-SLOWFMAF: v_mul_f32_e64 [[TMP:v[0-9]+]], [[A]], -[[B]]
 ; SI-DENORM-SLOWFMAF-DAG: v_subrev_f32_e32 [[RESULT0:v[0-9]+]], [[C]], [[TMP]]
@@ -352,8 +352,8 @@ define void @combine_to_mad_fsub_2_f32_2uses_neg(float addrspace(1)* noalias %ou
 ; SI-STD-DAG: v_mad_f32 [[RESULT0:v[0-9]+]], -[[A]], [[B]], -[[C]]
 ; SI-STD-DAG: v_mad_f32 [[RESULT1:v[0-9]+]], [[A]], [[B]], -[[D]]
 
-; SI-DENORM-DAG: v_fma_f32 [[RESULT0:v[0-9]+]], -[[A]], [[B]], -[[C]]
-; SI-DENORM-DAG: v_fma_f32 [[RESULT1:v[0-9]+]], [[A]], [[B]], -[[D]]
+; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT0:v[0-9]+]], -[[A]], [[B]], -[[C]]
+; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT1:v[0-9]+]], [[A]], [[B]], -[[D]]
 
 ; SI-DENORM-SLOWFMAF: v_mul_f32_e32 [[TMP:v[0-9]+]], [[B]], [[A]]
 ; SI-DENORM-SLOWFMAF-DAG: v_sub_f32_e64 [[RESULT0:v[0-9]+]], -[[TMP]], [[C]]
@@ -399,12 +399,9 @@ define void @combine_to_mad_fsub_2_f32_2uses_mul(float addrspace(1)* noalias %ou
 ; SI-STD: v_fma_f32 [[TMP1:v[0-9]+]], [[A]], [[B]], [[TMP0]]
 ; SI-STD: v_subrev_f32_e32 [[RESULT:v[0-9]+]], [[C]], [[TMP1]]
 
-; SI-DENORM: v_fma_f32 [[TMP0:v[0-9]+]], [[D]], [[E]], -[[C]]
-; SI-DENORM: v_fma_f32 [[RESULT:v[0-9]+]], [[A]], [[B]], [[TMP0]]
-
-; SI-DENORM-SLOWFMAF: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[E]], [[D]]
-; SI-DENORM-SLOWFMAF: v_fma_f32 [[TMP1:v[0-9]+]], [[A]], [[B]], [[TMP0]]
-; SI-DENORM-SLOWFMAF: v_subrev_f32_e32 [[RESULT1:v[0-9]+]], [[C]], [[TMP1]]
+; SI-DENORM: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[E]], [[D]]
+; SI-DENORM: v_fma_f32 [[TMP1:v[0-9]+]], [[A]], [[B]], [[TMP0]]
+; SI-DENORM: v_subrev_f32_e32 [[RESULT1:v[0-9]+]], [[C]], [[TMP1]]
 
 ; SI: buffer_store_dword [[RESULT]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
 define void @aggressive_combine_to_mad_fsub_0_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) #1 {
@@ -444,12 +441,9 @@ define void @aggressive_combine_to_mad_fsub_0_f32(float addrspace(1)* noalias %o
 ; SI-STD: v_fma_f32 [[TMP1:v[0-9]+]], [[B]], [[C]], [[TMP0]]
 ; SI-STD: v_subrev_f32_e32 [[RESULT:v[0-9]+]], [[TMP1]], [[A]]
 
-; SI-DENORM: v_fma_f32 [[TMP0:v[0-9]+]], -[[D]], [[E]], [[A]]
-; SI-DENORM: v_fma_f32 [[RESULT:v[0-9]+]], -[[B]], [[C]], [[TMP0]]
-
-; SI-DENORM-SLOWFMAF: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[E]], [[D]]
-; SI-DENORM-SLOWFMAF: v_fma_f32 [[TMP1:v[0-9]+]], [[B]], [[C]], [[TMP0]]
-; SI-DENORM-SLOWFMAF: v_subrev_f32_e32 [[RESULT:v[0-9]+]], [[TMP1]], [[A]]
+; SI-DENORM: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[E]], [[D]]
+; SI-DENORM: v_fma_f32 [[TMP1:v[0-9]+]], [[B]], [[C]], [[TMP0]]
+; SI-DENORM: v_subrev_f32_e32 [[RESULT:v[0-9]+]], [[TMP1]], [[A]]
 
 ; SI: buffer_store_dword [[RESULT]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
 ; SI: s_endpgm
@@ -485,19 +479,23 @@ define void @aggressive_combine_to_mad_fsub_1_f32(float addrspace(1)* noalias %o
 ; SI-DAG: buffer_load_dword [[D:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:12{{$}}
 ; SI-DAG: buffer_load_dword [[E:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:16{{$}}
 
-; SI-STD: v_mad_f32 [[TMP:v[0-9]+]], [[D]], [[E]], -[[C]]
-; SI-STD: v_mac_f32_e32 [[TMP]], [[B]], [[A]]
+; SI-STD-SAFE: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[E]], [[D]]
+; SI-STD-SAFE: v_mac_f32_e32 [[TMP0]], [[B]], [[A]]
+; SI-STD-SAFE: v_subrev_f32_e32 [[RESULT:v[0-9]+]], [[C]], [[TMP0]]
+
+; SI-STD-UNSAFE: v_mad_f32 [[RESULT:v[0-9]+]], [[D]], [[E]], -[[C]]
+; SI-STD-UNSAFE: v_mac_f32_e32 [[RESULT]], [[B]], [[A]]
 
-; SI-DENORM: v_fma_f32 [[TMP:v[0-9]+]], [[D]], [[E]], -[[C]]
-; SI-DENORM: v_fma_f32 [[RESULT:v[0-9]+]], [[A]], [[B]], [[TMP]]
+; SI-DENORM-FASTFMAF: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[E]], [[D]]
+; SI-DENORM-FASTFMAF: v_fma_f32 [[TMP1:v[0-9]+]], [[A]], [[B]], [[TMP0]]
+; SI-DENORM-FASTFMAF: v_subrev_f32_e32 [[RESULT:v[0-9]+]], [[C]], [[TMP1]]
 
 ; SI-DENORM-SLOWFMAF-DAG: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[E]], [[D]]
 ; SI-DENORM-SLOWFMAF-DAG: v_mul_f32_e32 [[TMP1:v[0-9]+]], [[B]], [[A]]
 ; SI-DENORM-SLOWFMAF: v_add_f32_e32 [[TMP2:v[0-9]+]], [[TMP0]], [[TMP1]]
 ; SI-DENORM-SLOWFMAF: v_subrev_f32_e32 [[RESULT:v[0-9]+]], [[C]], [[TMP2]]
 
-; SI-DENORM: buffer_store_dword [[RESULT]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
-; SI-STD: buffer_store_dword [[TMP]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
+; SI: buffer_store_dword [[RESULT]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
 ; SI: s_endpgm
 define void @aggressive_combine_to_mad_fsub_2_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) #1 {
   %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
@@ -532,11 +530,16 @@ define void @aggressive_combine_to_mad_fsub_2_f32(float addrspace(1)* noalias %o
 ; SI-DAG: buffer_load_dword [[D:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:12{{$}}
 ; SI-DAG: buffer_load_dword [[E:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:16{{$}}
 
-; SI-STD: v_mad_f32 [[TMP:v[0-9]+]], -[[D]], [[E]], [[A]]
-; SI-STD: v_mad_f32 [[RESULT:v[0-9]+]], -[[B]], [[C]], [[TMP]]
+; SI-STD-SAFE: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[E]], [[D]]
+; SI-STD-SAFE: v_mac_f32_e32 [[TMP0]], [[C]], [[B]]
+; SI-STD-SAFE: v_subrev_f32_e32 [[RESULT:v[0-9]+]], [[TMP0]], [[A]]
+
+; SI-STD-UNSAFE: v_mad_f32 [[TMP:v[0-9]+]], -[[D]], [[E]], [[A]]
+; SI-STD-UNSAFE: v_mad_f32 [[RESULT:v[0-9]+]], -[[B]], [[C]], [[TMP]]
 
-; SI-DENORM: v_fma_f32 [[TMP:v[0-9]+]], -[[D]], [[E]], [[A]]
-; SI-DENORM: v_fma_f32 [[RESULT:v[0-9]+]], -[[B]], [[C]], [[TMP]]
+; SI-DENORM-FASTFMAF: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[E]], [[D]]
+; SI-DENORM-FASTFMAF: v_fma_f32 [[TMP1:v[0-9]+]], [[B]], [[C]], [[TMP0]]
+; SI-DENORM-FASTFMAF: v_subrev_f32_e32 [[RESULT:v[0-9]+]], [[TMP1]], [[A]]
 
 ; SI-DENORM-SLOWFMAF-DAG: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[E]], [[D]]
 ; SI-DENORM-SLOWFMAF-DAG: v_mul_f32_e32 [[TMP1:v[0-9]+]], [[C]], [[B]]
diff --git a/test/CodeGen/NVPTX/fma-assoc.ll b/test/CodeGen/NVPTX/fma-assoc.ll
index 80a08a86316c..df86d476efdc 100644
--- a/test/CodeGen/NVPTX/fma-assoc.ll
+++ b/test/CodeGen/NVPTX/fma-assoc.ll
@@ -1,9 +1,10 @@
-; RUN: llc < %s -march=nvptx -mcpu=sm_20 -fp-contract=fast | FileCheck %s
+; RUN: llc < %s -march=nvptx -mcpu=sm_20 -fp-contract=fast | FileCheck %s -check-prefix=CHECK
+; RUN: llc < %s -march=nvptx -mcpu=sm_20 -fp-contract=fast -enable-unsafe-fp-math | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-UNSAFE
 
 define ptx_device float @t1_f32(float %x, float %y, float %z,
                                 float %u, float %v) {
-; CHECK: fma.rn.f32 %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}};
-; CHECK: fma.rn.f32 %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}};
+; CHECK-UNSAFE: fma.rn.f32 %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}};
+; CHECK-UNSAFE: fma.rn.f32 %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}};
 ; CHECK: ret;
   %a = fmul float %x, %y
   %b = fmul float %u, %v
@@ -14,8 +15,8 @@ define ptx_device float @t1_f32(float %x, float %y, float %z,
 
 define ptx_device double @t1_f64(double %x, double %y, double %z,
                                  double %u, double %v) {
-; CHECK: fma.rn.f64 %fd{{[0-9]+}}, %fd{{[0-9]+}}, %fd{{[0-9]+}}, %fd{{[0-9]+}};
-; CHECK: fma.rn.f64 %fd{{[0-9]+}}, %fd{{[0-9]+}}, %fd{{[0-9]+}}, %fd{{[0-9]+}};
+; CHECK-UNSAFE: fma.rn.f64 %fd{{[0-9]+}}, %fd{{[0-9]+}}, %fd{{[0-9]+}}, %fd{{[0-9]+}};
+; CHECK-UNSAFE: fma.rn.f64 %fd{{[0-9]+}}, %fd{{[0-9]+}}, %fd{{[0-9]+}}, %fd{{[0-9]+}};
 ; CHECK: ret;
   %a = fmul double %x, %y
   %b = fmul double %u, %v
diff --git a/test/CodeGen/PowerPC/fma-assoc.ll b/test/CodeGen/PowerPC/fma-assoc.ll
index e44bfc65242e..5080e5b250e9 100644
--- a/test/CodeGen/PowerPC/fma-assoc.ll
+++ b/test/CodeGen/PowerPC/fma-assoc.ll
@@ -1,5 +1,7 @@
-; RUN: llc -verify-machineinstrs < %s -march=ppc32 -fp-contract=fast -mattr=-vsx -disable-ppc-vsx-fma-mutation=false | FileCheck %s
-; RUN: llc -verify-machineinstrs < %s -mtriple=powerpc64-unknown-linux-gnu -fp-contract=fast -mattr=+vsx -mcpu=pwr7 -disable-ppc-vsx-fma-mutation=false | FileCheck -check-prefix=CHECK-VSX %s
+; RUN: llc -verify-machineinstrs < %s -march=ppc32 -fp-contract=fast -mattr=-vsx -disable-ppc-vsx-fma-mutation=false | FileCheck -check-prefix=CHECK -check-prefix=CHECK-SAFE %s
+; RUN: llc -verify-machineinstrs < %s -mtriple=powerpc64-unknown-linux-gnu -fp-contract=fast -mattr=+vsx -mcpu=pwr7 -disable-ppc-vsx-fma-mutation=false | FileCheck -check-prefix=CHECK-VSX -check-prefix=CHECK-VSX-SAFE %s
+; RUN: llc -verify-machineinstrs < %s -march=ppc32 -fp-contract=fast -enable-unsafe-fp-math -mattr=-vsx -disable-ppc-vsx-fma-mutation=false | FileCheck -check-prefix=CHECK -check-prefix=CHECK-UNSAFE %s
+; RUN: llc -verify-machineinstrs < %s -mtriple=powerpc64-unknown-linux-gnu -fp-contract=fast -enable-unsafe-fp-math -mattr=+vsx -mcpu=pwr7 -disable-ppc-vsx-fma-mutation=false | FileCheck -check-prefix=CHECK-VSX -check-prefix=CHECK-UNSAFE-VSX %s
 
 define double @test_FMADD_ASSOC1(double %A, double %B, double %C,
                                  double %D, double %E) {
@@ -8,16 +10,28 @@ define double @test_FMADD_ASSOC1(double %A, double %B, double %C,
   %H = fadd double %F, %G         ; <double> [#uses=1]
   %I = fadd double %H, %E         ; <double> [#uses=1]
   ret double %I
-; CHECK-LABEL: test_FMADD_ASSOC1:
-; CHECK: fmadd
-; CHECK-NEXT: fmadd
-; CHECK-NEXT: blr
+; CHECK-SAFE-LABEL: test_FMADD_ASSOC1:
+; CHECK-SAFE: fmul
+; CHECK-SAFE-NEXT: fmadd
+; CHECK-SAFE-NEXT: fadd
+; CHECK-SAFE-NEXT: blr
 
-; CHECK-VSX-LABEL: test_FMADD_ASSOC1:
-; CHECK-VSX: xsmaddmdp
-; CHECK-VSX-NEXT: xsmaddadp
-; CHECK-VSX-NEXT: fmr
-; CHECK-VSX-NEXT: blr
+; CHECK-UNSAFE-LABEL: test_FMADD_ASSOC1:
+; CHECK-UNSAFE: fmadd
+; CHECK-UNSAFE-NEXT: fmadd
+; CHECK-UNSAFE-NEXT: blr
+
+; CHECK-VSX-SAFE-LABEL: test_FMADD_ASSOC1:
+; CHECK-VSX-SAFE: xsmuldp
+; CHECK-VSX-SAFE-NEXT: xsmaddadp
+; CHECK-VSX-SAFE-NEXT: xsadddp
+; CHECK-VSX-SAFE-NEXT: blr
+
+; CHECK-VSX-UNSAFE-LABEL: test_FMADD_ASSOC1:
+; CHECK-VSX-UNSAFE: xsmaddmdp
+; CHECK-VSX-UNSAFE-NEXT: xsmaddadp
+; CHECK-VSX-UNSAFE-NEXT: fmr
+; CHECK-VSX-UNSAFE-NEXT: blr
 }
 
 define double @test_FMADD_ASSOC2(double %A, double %B, double %C,
@@ -27,16 +41,28 @@ define double @test_FMADD_ASSOC2(double %A, double %B, double %C,
   %H = fadd double %F, %G         ; <double> [#uses=1]
   %I = fadd double %E, %H         ; <double> [#uses=1]
   ret double %I
-; CHECK-LABEL: test_FMADD_ASSOC2:
-; CHECK: fmadd
-; CHECK-NEXT: fmadd
-; CHECK-NEXT: blr
+; CHECK-SAFE-LABEL: test_FMADD_ASSOC2:
+; CHECK-SAFE: fmul
+; CHECK-SAFE-NEXT: fmadd
+; CHECK-SAFE-NEXT: fadd
+; CHECK-SAFE-NEXT: blr
 
-; CHECK-VSX-LABEL: test_FMADD_ASSOC2:
-; CHECK-VSX: xsmaddmdp
-; CHECK-VSX-NEXT: xsmaddadp
-; CHECK-VSX-NEXT: fmr
-; CHECK-VSX-NEXT: blr
+; CHECK-UNSAFE-LABEL: test_FMADD_ASSOC2:
+; CHECK-UNSAFE: fmadd
+; CHECK-UNSAFE-NEXT: fmadd
+; CHECK-UNSAFE-NEXT: blr
+
+; CHECK-VSX-SAFE-LABEL: test_FMADD_ASSOC2:
+; CHECK-VSX-SAFE: xsmuldp
+; CHECK-VSX-SAFE-NEXT: xsmaddadp
+; CHECK-VSX-SAFE-NEXT: xsadddp
+; CHECK-VSX-SAFE-NEXT: blr
+
+; CHECK-VSX-UNSAFE-LABEL: test_FMADD_ASSOC2:
+; CHECK-VSX-UNSAFE: xsmaddmdp
+; CHECK-VSX-UNSAFE-NEXT: xsmaddadp
+; CHECK-VSX-UNSAFE-NEXT: fmr
+; CHECK-VSX-UNSAFE-NEXT: blr
 }
 
 define double @test_FMSUB_ASSOC1(double %A, double %B, double %C,
@@ -46,16 +72,28 @@ define double @test_FMSUB_ASSOC1(double %A, double %B, double %C,
   %H = fadd double %F, %G         ; <double> [#uses=1]
   %I = fsub double %H, %E         ; <double> [#uses=1]
   ret double %I
-; CHECK-LABEL: test_FMSUB_ASSOC1:
-; CHECK: fmsub
-; CHECK-NEXT: fmadd
-; CHECK-NEXT: blr
+; CHECK-SAFE-LABEL: test_FMSUB_ASSOC1:
+; CHECK-SAFE: fmul
+; CHECK-SAFE-NEXT: fmadd
+; CHECK-SAFE-NEXT: fsub
+; CHECK-SAFE-NEXT: blr
 
-; CHECK-VSX-LABEL: test_FMSUB_ASSOC1:
-; CHECK-VSX: xsmsubmdp
-; CHECK-VSX-NEXT: xsmaddadp
-; CHECK-VSX-NEXT: fmr
-; CHECK-VSX-NEXT: blr
+; CHECK-UNSAFE-LABEL: test_FMSUB_ASSOC1:
+; CHECK-UNSAFE: fmsub
+; CHECK-UNSAFE-NEXT: fmadd
+; CHECK-UNSAFE-NEXT: blr
+
+; CHECK-SAFE-VSX-LABEL: test_FMSUB_ASSOC1:
+; CHECK-SAFE-VSX: xsmuldp
+; CHECK-SAFE-VSX-NEXT: xsmaddadp
+; CHECK-SAFE-VSX-NEXT: xssubdp
+; CHECK-SAFE-VSX-NEXT: blr
+
+; CHECK-UNSAFE-VSX-LABEL: test_FMSUB_ASSOC1:
+; CHECK-UNSAFE-VSX: xsmsubmdp
+; CHECK-UNSAFE-VSX-NEXT: xsmaddadp
+; CHECK-UNSAFE-VSX-NEXT: fmr
+; CHECK-UNSAFE-VSX-NEXT: blr
 }
 
 define double @test_FMSUB_ASSOC2(double %A, double %B, double %C,
@@ -65,16 +103,28 @@ define double @test_FMSUB_ASSOC2(double %A, double %B, double %C,
   %H = fadd double %F, %G         ; <double> [#uses=1]
   %I = fsub double %E, %H         ; <double> [#uses=1]
   ret double %I
-; CHECK-LABEL: test_FMSUB_ASSOC2:
-; CHECK: fnmsub
-; CHECK-NEXT: fnmsub
-; CHECK-NEXT: blr
+; CHECK-SAFE-LABEL: test_FMSUB_ASSOC2:
+; CHECK-SAFE: fmul
+; CHECK-SAFE-NEXT: fmadd
+; CHECK-SAFE-NEXT: fsub
+; CHECK-SAFE-NEXT: blr
 
-; CHECK-VSX-LABEL: test_FMSUB_ASSOC2:
-; CHECK-VSX: xsnmsubmdp
-; CHECK-VSX-NEXT: xsnmsubadp
-; CHECK-VSX-NEXT: fmr
-; CHECK-VSX-NEXT: blr
+; CHECK-UNSAFE-LABEL: test_FMSUB_ASSOC2:
+; CHECK-UNSAFE: fnmsub
+; CHECK-UNSAFE-NEXT: fnmsub
+; CHECK-UNSAFE-NEXT: blr
+
+; CHECK-SAFE-VSX-LABEL: test_FMSUB_ASSOC2:
+; CHECK-SAFE-VSX: xsmuldp
+; CHECK-SAFE-VSX-NEXT: xsmaddadp
+; CHECK-SAFE-VSX-NEXT: xssubdp
+; CHECK-SAFE-VSX-NEXT: blr
+
+; CHECK-UNSAFE-VSX-LABEL: test_FMSUB_ASSOC2:
+; CHECK-UNSAFE-VSX: xsnmsubmdp
+; CHECK-UNSAFE-VSX-NEXT: xsnmsubadp
+; CHECK-UNSAFE-VSX-NEXT: fmr
+; CHECK-UNSAFE-VSX-NEXT: blr
 }
 
 define double @test_FMADD_ASSOC_EXT1(float %A, float %B, double %C,
diff --git a/test/Object/Inputs/phdr-note.elf-x86-64 b/test/Object/Inputs/phdr-note.elf-x86-64
new file mode 100755
index 000000000000..2825c6840ca8
--- /dev/null
+++ b/test/Object/Inputs/phdr-note.elf-x86-64
diff --git a/test/Object/Inputs/phdrs.elf-x86-64 b/test/Object/Inputs/phdrs.elf-x86-64
new file mode 100755
index 000000000000..078574ab5801
--- /dev/null
+++ b/test/Object/Inputs/phdrs.elf-x86-64
diff --git a/test/Transforms/InstCombine/minmax-fold.ll b/test/Transforms/InstCombine/minmax-fold.ll
index e39f7f491cbc..bf46cefd8ab3 100644
--- a/test/Transforms/InstCombine/minmax-fold.ll
+++ b/test/Transforms/InstCombine/minmax-fold.ll
@@ -339,3 +339,84 @@ define i32 @test75(i32 %x) {
   ret i32 %retval
 }
 
+; The next 3 min tests should canonicalize to the same form...and not infinite loop.
+
+define double @PR31751_umin1(i32 %x) {
+; CHECK-LABEL: @PR31751_umin1(
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 %x, 2147483647
+; CHECK-NEXT:    [[CONV1:%.*]] = select i1 [[TMP1]], i32 %x, i32 2147483647
+; CHECK-NEXT:    [[TMP2:%.*]] = sitofp i32 [[CONV1]] to double
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %cmp = icmp slt i32 %x, 0
+  %sel = select i1 %cmp, i32 2147483647, i32 %x
+  %conv = sitofp i32 %sel to double
+  ret double %conv
+}
+
+define double @PR31751_umin2(i32 %x) {
+; CHECK-LABEL: @PR31751_umin2(
+; CHECK-NEXT:    [[CMP:%.*]] = icmp ult i32 %x, 2147483647
+; CHECK-NEXT:    [[SEL:%.*]] = select i1 [[CMP]], i32 %x, i32 2147483647
+; CHECK-NEXT:    [[CONV:%.*]] = sitofp i32 [[SEL]] to double
+; CHECK-NEXT:    ret double [[CONV]]
+;
+  %cmp = icmp ult i32 %x, 2147483647
+  %sel = select i1 %cmp, i32 %x, i32 2147483647
+  %conv = sitofp i32 %sel to double
+  ret double %conv
+}
+
+define double @PR31751_umin3(i32 %x) {
+; CHECK-LABEL: @PR31751_umin3(
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult i32 %x, 2147483647
+; CHECK-NEXT:    [[SEL:%.*]] = select i1 [[TMP1]], i32 %x, i32 2147483647
+; CHECK-NEXT:    [[CONV:%.*]] = sitofp i32 [[SEL]] to double
+; CHECK-NEXT:    ret double [[CONV]]
+;
+  %cmp = icmp ugt i32 %x, 2147483647
+  %sel = select i1 %cmp, i32 2147483647, i32 %x
+  %conv = sitofp i32 %sel to double
+  ret double %conv
+}
+
+; The next 3 max tests should canonicalize to the same form...and not infinite loop.
+
+define double @PR31751_umax1(i32 %x) {
+; CHECK-LABEL: @PR31751_umax1(
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ugt i32 %x, -2147483648
+; CHECK-NEXT:    [[CONV1:%.*]] = select i1 [[TMP1]], i32 %x, i32 -2147483648
+; CHECK-NEXT:    [[TMP2:%.*]] = sitofp i32 [[CONV1]] to double
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %cmp = icmp sgt i32 %x, -1
+  %sel = select i1 %cmp, i32 2147483648, i32 %x
+  %conv = sitofp i32 %sel to double
+  ret double %conv
+}
+
+define double @PR31751_umax2(i32 %x) {
+; CHECK-LABEL: @PR31751_umax2(
+; CHECK-NEXT:    [[CMP:%.*]] = icmp ugt i32 %x, -2147483648
+; CHECK-NEXT:    [[SEL:%.*]] = select i1 [[CMP]], i32 %x, i32 -2147483648
+; CHECK-NEXT:    [[CONV:%.*]] = sitofp i32 [[SEL]] to double
+; CHECK-NEXT:    ret double [[CONV]]
+;
+  %cmp = icmp ugt i32 %x, 2147483648
+  %sel = select i1 %cmp, i32 %x, i32 2147483648
+  %conv = sitofp i32 %sel to double
+  ret double %conv
+}
+
+define double @PR31751_umax3(i32 %x) {
+; CHECK-LABEL: @PR31751_umax3(
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ugt i32 %x, -2147483648
+; CHECK-NEXT:    [[SEL:%.*]] = select i1 [[TMP1]], i32 %x, i32 -2147483648
+; CHECK-NEXT:    [[CONV:%.*]] = sitofp i32 [[SEL]] to double
+; CHECK-NEXT:    ret double [[CONV]]
+;
+  %cmp = icmp ult i32 %x, 2147483648
+  %sel = select i1 %cmp, i32 2147483648, i32 %x
+  %conv = sitofp i32 %sel to double
+  ret double %conv
+}
diff --git a/test/Transforms/LoopStrengthReduce/AMDGPU/preserve-addrspace-assert.ll b/test/Transforms/LoopStrengthReduce/AMDGPU/preserve-addrspace-assert.ll
new file mode 100644
index 000000000000..b3b696d42c59
--- /dev/null
+++ b/test/Transforms/LoopStrengthReduce/AMDGPU/preserve-addrspace-assert.ll
@@ -0,0 +1,54 @@
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -loop-reduce %s | FileCheck %s
+
+; Test for assert resulting from inconsistent isLegalAddressingMode
+; answers when the address space was dropped from the query.
+
+target datalayout = "e-p:32:32-p1:64:64-p2:64:64-p3:32:32-p4:64:64-p5:32:32-p24:64:64-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64"
+
+%0 = type { i32, double, i32, float }
+
+; CHECK-LABEL: @lsr_crash_preserve_addrspace_unknown_type(
+; CHECK: %tmp4 = bitcast %0 addrspace(3)* %tmp to double addrspace(3)*
+; CHECK: %scevgep5 = getelementptr double, double addrspace(3)* %tmp4, i32 1
+; CHEC: load double, double addrspace(3)* %scevgep5
+
+; CHECK: %scevgep = getelementptr i32, i32 addrspace(3)* %tmp1, i32 4
+; CHECK:%tmp14 = load i32, i32 addrspace(3)* %scevgep
+define void @lsr_crash_preserve_addrspace_unknown_type() #0 {
+bb:
+  br label %bb1
+
+bb1:                                              ; preds = %bb17, %bb
+  %tmp = phi %0 addrspace(3)* [ undef, %bb ], [ %tmp18, %bb17 ]
+  %tmp2 = getelementptr inbounds %0, %0 addrspace(3)* %tmp, i64 0, i32 1
+  %tmp3 = load double, double addrspace(3)* %tmp2, align 8
+  br label %bb4
+
+bb4:                                              ; preds = %bb1
+  br i1 undef, label %bb8, label %bb5
+
+bb5:                                              ; preds = %bb4
+  unreachable
+
+bb8:                                              ; preds = %bb4
+  %tmp9 = getelementptr inbounds %0, %0 addrspace(3)* %tmp, i64 0, i32 0
+  %tmp10 = load i32, i32 addrspace(3)* %tmp9, align 4
+  %tmp11 = icmp eq i32 0, %tmp10
+  br i1 %tmp11, label %bb12, label %bb17
+
+bb12:                                             ; preds = %bb8
+  %tmp13 = getelementptr inbounds %0, %0 addrspace(3)* %tmp, i64 0, i32 2
+  %tmp14 = load i32, i32 addrspace(3)* %tmp13, align 4
+  %tmp15 = icmp eq i32 0, %tmp14
+  br i1 %tmp15, label %bb16, label %bb17
+
+bb16:                                             ; preds = %bb12
+  unreachable
+
+bb17:                                             ; preds = %bb12, %bb8
+  %tmp18 = getelementptr inbounds %0, %0 addrspace(3)* %tmp, i64 2
+  br label %bb1
+}
+
+attributes #0 = { nounwind }
+attributes #1 = { nounwind readnone }
diff --git a/test/tools/llvm-objdump/X86/openbsd-headers.test b/test/tools/llvm-objdump/X86/openbsd-headers.test
new file mode 100644
index 000000000000..48f496ca616e
--- /dev/null
+++ b/test/tools/llvm-objdump/X86/openbsd-headers.test
@@ -0,0 +1,20 @@
+## openbsd-phdrs.elf-x86-64 was generated using GNU ld (GNU Binutils for Ubuntu) 2.26.1.
+## llvm-mc -filetype=obj -triple=x86_64-pc-linux test.s -o main.o
+## ld -script linker.script main.o -o openbsd-phdrs.elf-x86-64
+##
+## test.s is an empty file.
+## linker.script:
+## PHDRS { text PT_LOAD FILEHDR PHDRS; foo 0x65a3dbe6; bar 0x65a3dbe7; zed 0x65a41be6; }
+##   Where 0x65a3dbe6 is the value of PT_OPENBSD_RANDOMIZE,
+##         0x65a3dbe7 is the value of PT_OPENBSD_WXNEEDED,
+##         0x65a41be6 is the value of PT_OPENBSD_BOOTDATA
+## SECTIONS { . = SIZEOF_HEADERS; .all : { *(.*) } : text }
+RUN: llvm-objdump -p %p/../../../Object/Inputs/openbsd-phdrs.elf-x86-64 \
+RUN:     | FileCheck %s
+
+CHECK: OPENBSD_RANDOMIZE off    0x0000000000000000 vaddr 0x0000000000000000 paddr 0x0000000000000000 align 2**3
+CHECK-NEXT:      filesz 0x0000000000000000 memsz 0x0000000000000000 flags ---
+CHECK-NEXT: OPENBSD_WXNEEDED off    0x0000000000000000 vaddr 0x0000000000000000 paddr 0x0000000000000000 align 2**3
+CHECK-NEXT:      filesz 0x0000000000000000 memsz 0x0000000000000000 flags ---
+CHECK-NEXT: OPENBSD_BOOTDATA off    0x0000000000000000 vaddr 0x0000000000000000 paddr 0x0000000000000000 align 2**3
+CHECK-NEXT:      filesz 0x0000000000000000 memsz 0x0000000000000000 flags ---
diff --git a/test/tools/llvm-objdump/X86/phdrs.test b/test/tools/llvm-objdump/X86/phdrs.test
new file mode 100644
index 000000000000..646f6af8733a
--- /dev/null
+++ b/test/tools/llvm-objdump/X86/phdrs.test
@@ -0,0 +1,32 @@
+## phdrs.elf-x86-64 was generated using lld (3.9).
+## llvm-mc -filetype=obj -triple=x86_64-unknown-linux test.s -o test.o
+## lld test.o -o phdrs.elf-x86-64
+##
+## test.s:
+## .global _start
+## _start:
+##
+## .global d
+## .section .foo,"awT",@progbits
+## d:
+## .long 2
+##
+RUN: llvm-objdump -p %p/../../../Object/Inputs/phdrs.elf-x86-64 \
+RUN:     | FileCheck %s
+
+CHECK: RELRO off    0x0000000000001000 vaddr 0x0000000000201000 paddr 0x0000000000201000 align 2**0
+CHECK-NEXT:      filesz 0x0000000000000004 memsz 0x0000000000001000 flags r--
+
+## phdr-note.elf-x86-64 was generated using lld (3.9).
+## llvm-mc -filetype=obj -triple=x86_64-pc-linux test.s -o test.o
+## lld test.o -o phdr-note.elf-x86-64 -shared
+##
+## test.s:
+##       .section        .note.test,"a",@note
+##       .quad 42
+
+RUN: llvm-objdump -p %p/../../../Object/Inputs/phdr-note.elf-x86-64 \
+RUN:     | FileCheck %s --check-prefix=NOTE
+
+NOTE: NOTE off    0x0000000000000200 vaddr 0x0000000000000200 paddr 0x0000000000000200 align 2**0
+NOTE-NEXT:       filesz 0x0000000000000008 memsz 0x0000000000000008 flags r--
diff --git a/tools/llvm-objdump/ELFDump.cpp b/tools/llvm-objdump/ELFDump.cpp
index aa6aa94ccd9e..7f5fe5a9d3b8 100644
--- a/tools/llvm-objdump/ELFDump.cpp
+++ b/tools/llvm-objdump/ELFDump.cpp
@@ -36,6 +36,9 @@ template <class ELFT> void printProgramHeaders(const ELFFile<ELFT> *o) {
     case ELF::PT_GNU_EH_FRAME:
       outs() << "EH_FRAME ";
       break;
+    case ELF::PT_GNU_RELRO:
+      outs() << "   RELRO ";
+      break;
     case ELF::PT_GNU_STACK:
       outs() << "   STACK ";
       break;
@@ -45,6 +48,18 @@ template <class ELFT> void printProgramHeaders(const ELFFile<ELFT> *o) {
     case ELF::PT_LOAD:
       outs() << "    LOAD ";
       break;
+    case ELF::PT_NOTE:
+      outs() << "    NOTE ";
+      break;
+    case ELF::PT_OPENBSD_BOOTDATA:
+      outs() << "    OPENBSD_BOOTDATA ";
+      break;
+    case ELF::PT_OPENBSD_RANDOMIZE:
+      outs() << "    OPENBSD_RANDOMIZE ";
+      break;
+    case ELF::PT_OPENBSD_WXNEEDED:
+      outs() << "    OPENBSD_WXNEEDED ";
+      break;
     case ELF::PT_PHDR:
       outs() << "    PHDR ";
       break;
diff --git a/utils/release/test-release.sh b/utils/release/test-release.sh
index 642bb670e7e2..b0c771579802 100755
--- a/utils/release/test-release.sh
+++ b/utils/release/test-release.sh
@@ -36,6 +36,7 @@ do_libs="yes"
 do_libunwind="yes"
 do_test_suite="yes"
 do_openmp="yes"
+do_lld="yes"
 do_lldb="no"
 do_polly="no"
 BuildDir="`pwd`"
@@ -64,6 +65,7 @@ function usage() {
     echo " -no-libunwind        Disable check-out & build libunwind"
     echo " -no-test-suite       Disable check-out & build test-suite"
     echo " -no-openmp           Disable check-out & build libomp"
+    echo " -no-lld              Disable check-out & build lld"
     echo " -lldb                Enable check-out & build lldb"
     echo " -no-lldb             Disable check-out & build lldb (default)"
     echo " -polly               Enable check-out & build Polly"
@@ -143,6 +145,9 @@ while [ $# -gt 0 ]; do
         -no-openmp )
             do_openmp="no"
             ;;
+        -no-lld )
+            do_lld="no"
+            ;;
         -lldb )
             do_lldb="yes"
             ;;
@@ -225,6 +230,9 @@ esac
 if [ $do_openmp = "yes" ]; then
   projects="$projects openmp"
 fi
+if [ $do_lld = "yes" ]; then
+  projects="$projects lld"
+fi
 if [ $do_lldb = "yes" ]; then
   projects="$projects lldb"
 fi
@@ -297,7 +305,7 @@ function export_sources() {
         cfe)
             projsrc=llvm.src/tools/clang
             ;;
-        lldb|polly)
+        lld|lldb|polly)
             projsrc=llvm.src/tools/$proj
             ;;
         clang-tools-extra)