[llvm] 098d3f9 - [InstSimplify] Simplify to vector constants when possible

[Arthur Eubanks via llvm-commits]

Author: Arthur Eubanks
Date: 2020-08-26T11:40:36-07:00
New Revision: 098d3f98276de90b6e1468031bd3858615240bb7

URL: https://github.com/llvm/llvm-project/commit/098d3f98276de90b6e1468031bd3858615240bb7
DIFF: https://github.com/llvm/llvm-project/commit/098d3f98276de90b6e1468031bd3858615240bb7.diff

LOG: [InstSimplify] Simplify to vector constants when possible

InstSimplify should do all transformations that ConstProp does, but
one thing that ConstProp does that InstSimplify wouldn't is inline
vector instructions that are constants, e.g. into a ret.

Previously vector instructions wouldn't be inlined in InstSimplify
because llvm::Simplify*Instruction() would return nullptr for specific
instructions, such as vector instructions that were actually constants,
if it couldn't simplify them.

This changes SimplifyInsertElementInst, SimplifyExtractElementInst, and
SimplifyShuffleVectorInst to return a vector constant when possible.

Reviewed By: efriedma

Differential Revision: https://reviews.llvm.org/D85946

Added: 
    

Modified: 
    llvm/lib/Analysis/InstructionSimplify.cpp
    llvm/test/Analysis/ConstantFolding/vscale-shufflevector.ll
    llvm/test/Transforms/InstSimplify/vscale.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp
index 9b5bb37a0ff75..887079445eccf 100644
--- a/llvm/lib/Analysis/InstructionSimplify.cpp
+++ b/llvm/lib/Analysis/InstructionSimplify.cpp
@@ -4302,7 +4302,7 @@ Value *llvm::SimplifyInsertElementInst(Value *Vec, Value *Val, Value *Idx,
   auto *ValC = dyn_cast<Constant>(Val);
   auto *IdxC = dyn_cast<Constant>(Idx);
   if (VecC && ValC && IdxC)
-    return ConstantFoldInsertElementInstruction(VecC, ValC, IdxC);
+    return ConstantExpr::getInsertElement(VecC, ValC, IdxC);
 
   // For fixed-length vector, fold into undef if index is out of bounds.
   if (auto *CI = dyn_cast<ConstantInt>(Idx)) {
@@ -4367,7 +4367,7 @@ static Value *SimplifyExtractElementInst(Value *Vec, Value *Idx,
   auto *VecVTy = cast<VectorType>(Vec->getType());
   if (auto *CVec = dyn_cast<Constant>(Vec)) {
     if (auto *CIdx = dyn_cast<Constant>(Idx))
-      return ConstantFoldExtractElementInstruction(CVec, CIdx);
+      return ConstantExpr::getExtractElement(CVec, CIdx);
 
     // The index is not relevant if our vector is a splat.
     if (auto *Splat = CVec->getSplatValue())
@@ -4565,8 +4565,8 @@ static Value *SimplifyShuffleVectorInst(Value *Op0, Value *Op1,
   // If all operands are constant, constant fold the shuffle. This
   // transformation depends on the value of the mask which is not known at
   // compile time for scalable vectors
-  if (!Scalable && Op0Const && Op1Const)
-    return ConstantFoldShuffleVectorInstruction(Op0Const, Op1Const, Mask);
+  if (Op0Const && Op1Const)
+    return ConstantExpr::getShuffleVector(Op0Const, Op1Const, Mask);
 
   // Canonicalization: if only one input vector is constant, it shall be the
   // second one. This transformation depends on the value of the mask which

diff  --git a/llvm/test/Analysis/ConstantFolding/vscale-shufflevector.ll b/llvm/test/Analysis/ConstantFolding/vscale-shufflevector.ll
index dc3b66e18f871..9c1f6730122e2 100644
--- a/llvm/test/Analysis/ConstantFolding/vscale-shufflevector.ll
+++ b/llvm/test/Analysis/ConstantFolding/vscale-shufflevector.ll
@@ -15,9 +15,7 @@ target triple = "aarch64"
 ; the compiler. It happens to be the case that this will be the result.
 
 ; CHECK-LABEL: define <vscale x 8 x i1> @vscale_version()
-; CHECK-NEXT: %splatter = insertelement <vscale x 8 x i1> undef, i1 true, i32 0
-; CHECK-NEXT: %foo = shufflevector <vscale x 8 x i1> %splatter, <vscale x 8 x i1> undef, <vscale x 8 x i32> zeroinitializer
-; CHECK-NEXT: ret <vscale x 8 x i1> %foo
+; CHECK-NEXT: ret <vscale x 8 x i1> shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> undef, i1 true, i32 0), <vscale x 8 x i1> undef, <vscale x 8 x i32> zeroinitializer)
 
 define <vscale x 8 x i1> @vscale_version() {
   %splatter = insertelement <vscale x 8 x i1> undef, i1 true, i32 0

diff  --git a/llvm/test/Transforms/InstSimplify/vscale.ll b/llvm/test/Transforms/InstSimplify/vscale.ll
index dd4ca47a52eca..802cb99001672 100644
--- a/llvm/test/Transforms/InstSimplify/vscale.ll
+++ b/llvm/test/Transforms/InstSimplify/vscale.ll
@@ -51,6 +51,23 @@ define <vscale x 4 x i32> @insert_extract_element_same_vec_idx_1(<vscale x 4 x i
   ret <vscale x 4 x i32> %r
 }
 
+define <vscale x 4 x i32> @insertelement_inline_to_ret() {
+; CHECK-LABEL: @insertelement_inline_to_ret(
+; CHECK-NEXT:    ret <vscale x 4 x i32> insertelement (<vscale x 4 x i32> undef, i32 1, i32 0)
+;
+  %i = insertelement <vscale x 4 x i32> undef, i32 1, i32 0
+  ret <vscale x 4 x i32> %i
+}
+
+define <vscale x 4 x i32> @insertelement_shufflevector_inline_to_ret() {
+; CHECK-LABEL: @insertelement_shufflevector_inline_to_ret(
+; CHECK-NEXT:    ret <vscale x 4 x i32> shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> undef, i32 1, i32 0), <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer)
+;
+  %i = insertelement <vscale x 4 x i32> undef, i32 1, i32 0
+  %i2 = shufflevector <vscale x 4 x i32> %i, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  ret <vscale x 4 x i32> %i2
+}
+
 ; extractelement
 
 define i32 @extractelement_idx_undef(<vscale x 4 x i32> %a) {
@@ -120,6 +137,16 @@ define <vscale x 2 x i1> @cmp_le_smax_always_true(<vscale x 2 x i64> %x) {
    ret <vscale x 2 x i1> %cmp
 }
 
+define <vscale x 4 x float> @bitcast() {
+; CHECK-LABEL: @bitcast(
+; CHECK-NEXT:    ret <vscale x 4 x float> bitcast (<vscale x 4 x i32> shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> undef, i32 1, i32 0), <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer) to <vscale x 4 x float>)
+;
+  %i1 = insertelement <vscale x 4 x i32> undef, i32 1, i32 0
+  %i2 = shufflevector <vscale x 4 x i32> %i1, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %i3 = bitcast <vscale x 4 x i32> %i2 to <vscale x 4 x float>
+  ret <vscale x 4 x float> %i3
+}
+
 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
 ;; Memory Access and Addressing Operations
 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;