[PATCH] D125750: [InstCombine] fold fake floating point vector extract to shift+trunc.

[Jianjian Guan via Phabricator via llvm-commits]

jacquesguan created this revision.
jacquesguan added reviewers: spatel, lebedev.ri, RKSimon, dmgreen.
Herald added a subscriber: hiraditya.
Herald added a project: All.
jacquesguan requested review of this revision.
Herald added a project: LLVM.
Herald added a subscriber: llvm-commits.

This patch supports the FP part of D111082 <https://reviews.llvm.org/D111082>.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D125750

Files:
  llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
  llvm/test/Transforms/InstCombine/extractelement.ll


Index: llvm/test/Transforms/InstCombine/extractelement.ll
===================================================================
--- llvm/test/Transforms/InstCombine/extractelement.ll
+++ llvm/test/Transforms/InstCombine/extractelement.ll
@@ -407,19 +407,40 @@
   ret i8 %r
 }
 
-; negative test - can't have FP dest type without a cast
-
 define half @bitcast_fpvec_index0(i32 %x) {
-; ANY-LABEL: @bitcast_fpvec_index0(
-; ANY-NEXT:    [[V:%.*]] = bitcast i32 [[X:%.*]] to <2 x half>
-; ANY-NEXT:    [[R:%.*]] = extractelement <2 x half> [[V]], i64 0
-; ANY-NEXT:    ret half [[R]]
+; LE-LABEL: @bitcast_fpvec_index0(
+; LE-NEXT:    [[TMP1:%.*]] = trunc i32 [[X:%.*]] to i16
+; LE-NEXT:    [[R:%.*]] = bitcast i16 [[TMP1]] to half
+; LE-NEXT:    ret half [[R]]
+;
+; BE-LABEL: @bitcast_fpvec_index0(
+; BE-NEXT:    [[EXTELT_OFFSET:%.*]] = lshr i32 [[X:%.*]], 16
+; BE-NEXT:    [[TMP1:%.*]] = trunc i32 [[EXTELT_OFFSET]] to i16
+; BE-NEXT:    [[R:%.*]] = bitcast i16 [[TMP1]] to half
+; BE-NEXT:    ret half [[R]]
 ;
   %v = bitcast i32 %x to <2 x half>
   %r = extractelement <2 x half> %v, i8 0
   ret half %r
 }
 
+define half @bitcast_fpvec_index1(i32 %x) {
+; LE-LABEL: @bitcast_fpvec_index1(
+; LE-NEXT:    [[EXTELT_OFFSET:%.*]] = lshr i32 [[X:%.*]], 16
+; LE-NEXT:    [[TMP1:%.*]] = trunc i32 [[EXTELT_OFFSET]] to i16
+; LE-NEXT:    [[R:%.*]] = bitcast i16 [[TMP1]] to half
+; LE-NEXT:    ret half [[R]]
+;
+; BE-LABEL: @bitcast_fpvec_index1(
+; BE-NEXT:    [[TMP1:%.*]] = trunc i32 [[X:%.*]] to i16
+; BE-NEXT:    [[R:%.*]] = bitcast i16 [[TMP1]] to half
+; BE-NEXT:    ret half [[R]]
+;
+  %v = bitcast i32 %x to <2 x half>
+  %r = extractelement <2 x half> %v, i8 1
+  ret half %r
+}
+
 ; negative test - need constant index
 
 define i8 @bitcast_scalar_index_variable(i32 %x, i64 %y) {
Index: llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
===================================================================
--- llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
+++ llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
@@ -191,8 +191,7 @@
 
   // If we are casting an integer to vector and extracting a portion, that is
   // a shift-right and truncate.
-  // TODO: Allow FP dest type by casting the trunc to FP?
-  if (X->getType()->isIntegerTy() && DestTy->isIntegerTy() &&
+  if (X->getType()->isIntegerTy() &&
       isDesirableIntType(X->getType()->getPrimitiveSizeInBits())) {
     assert(isa<FixedVectorType>(Ext.getVectorOperand()->getType()) &&
            "Expected fixed vector type for bitcast from scalar integer");
@@ -205,6 +204,12 @@
     unsigned ShiftAmountC = ExtIndexC * DestTy->getPrimitiveSizeInBits();
     if (!ShiftAmountC || Ext.getVectorOperand()->hasOneUse()) {
       Value *Lshr = Builder.CreateLShr(X, ShiftAmountC, "extelt.offset");
+      if (DestTy->isFloatingPointTy()) {
+        Type *DstIntTy = IntegerType::getIntNTy(
+            Lshr->getContext(), DestTy->getPrimitiveSizeInBits());
+        Value *Trunc = Builder.CreateTrunc(Lshr, DstIntTy);
+        return new BitCastInst(Trunc, DestTy);
+      }
       return new TruncInst(Lshr, DestTy);
     }
   }


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D125750.429959.patch
Type: text/x-patch
Size: 3127 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20220517/71783ad3/attachment.bin>