[PATCH] D104239: [SVE] Fix cast<FixedVectorType> in truncateToMinimalBitwidths

Dylan Fleming via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Thu Jul 1 04:22:47 PDT 2021


DylanFleming-arm updated this revision to Diff 355831.
DylanFleming-arm added a comment.

Fixed issue where loop wasn't correctly vectorized with scalable vectors


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D104239/new/

https://reviews.llvm.org/D104239

Files:
  llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
  llvm/test/Transforms/LoopVectorize/scalable-trunc-min-bitwidth.ll


Index: llvm/test/Transforms/LoopVectorize/scalable-trunc-min-bitwidth.ll
===================================================================
--- /dev/null
+++ llvm/test/Transforms/LoopVectorize/scalable-trunc-min-bitwidth.ll
@@ -0,0 +1,76 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -loop-vectorize -instsimplify -scalable-vectorization=on -force-target-supports-scalable-vectors -S | FileCheck %s
+
+define void @trunc_minimal_bitwidth(i8* %bptr, i16* %hptr, i32 %sptr, i64 %dptr) {
+; CHECK-LABEL: @trunc_minimal_bitwidth(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[HPTR1:%.*]] = bitcast i16* [[HPTR:%.*]] to i8*
+; CHECK-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 4
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[DPTR:%.*]], [[TMP1]]
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; CHECK:       vector.memcheck:
+; CHECK-NEXT:    [[SCEVGEP:%.*]] = getelementptr i16, i16* [[HPTR]], i64 [[DPTR]]
+; CHECK-NEXT:    [[SCEVGEP2:%.*]] = bitcast i16* [[SCEVGEP]] to i8*
+; CHECK-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i8, i8* [[BPTR:%.*]], i64 1
+; CHECK-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[HPTR1]], [[SCEVGEP3]]
+; CHECK-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[BPTR]], [[SCEVGEP2]]
+; CHECK-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; CHECK-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 4
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[DPTR]], [[TMP3]]
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[DPTR]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT4:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[SPTR:%.*]], i32 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT5:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT4]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP4:%.*]] = trunc <vscale x 4 x i32> [[BROADCAST_SPLAT5]] to <vscale x 4 x i16>
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, i16* [[HPTR]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP6:%.*]] = bitcast i16* [[TMP5]] to <vscale x 4 x i16>*
+; CHECK-NEXT:    store <vscale x 4 x i16> [[TMP4]], <vscale x 4 x i16>* [[TMP6]], align 2, !alias.scope !0, !noalias !3
+; CHECK-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
+; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[DPTR]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[CONV21:%.*]] = trunc i32 [[SPTR]] to i16
+; CHECK-NEXT:    [[ARRAYIDX23:%.*]] = getelementptr inbounds i16, i16* [[HPTR]], i64 [[INDVARS_IV]]
+; CHECK-NEXT:    store i16 [[CONV21]], i16* [[ARRAYIDX23]], align 2
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[DPTR]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK:       for.exit:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %for.body
+
+for.body:
+  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
+  %0 = load i8, i8* %bptr, align 1
+  %conv = zext i8 %0 to i32
+  %conv21 = trunc i32 %sptr to i16
+  %arrayidx23 = getelementptr inbounds i16, i16* %hptr, i64 %indvars.iv
+  store i16 %conv21, i16* %arrayidx23, align 2
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond.not = icmp eq i64 %indvars.iv.next, %dptr
+  br i1 %exitcond.not, label %for.exit, label %for.body, !llvm.loop !0
+
+for.exit:
+  ret void
+}
+
+!0 = !{!0, !1, !2}
+!1 = !{!"llvm.loop.vectorize.width", i32 4}
+!2 = !{!"llvm.loop.vectorize.scalable.enable", i1 true}
\ No newline at end of file
Index: llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
===================================================================
--- llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -3951,9 +3951,8 @@
       Type *OriginalTy = I->getType();
       Type *ScalarTruncatedTy =
           IntegerType::get(OriginalTy->getContext(), KV.second);
-      auto *TruncatedTy = FixedVectorType::get(
-          ScalarTruncatedTy,
-          cast<FixedVectorType>(OriginalTy)->getNumElements());
+      auto *TruncatedTy = VectorType::get(
+          ScalarTruncatedTy, cast<VectorType>(OriginalTy)->getElementCount());
       if (TruncatedTy == OriginalTy)
         continue;
 


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D104239.355831.patch
Type: text/x-patch
Size: 5518 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20210701/316417cf/attachment.bin>


More information about the llvm-commits mailing list