[PATCH] D89872: [SVE][AArch64] Fix TypeSize warning in GEP cost analysis

Joe Ellis via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Thu Oct 22 03:10:42 PDT 2020


joechrisellis updated this revision to Diff 299907.
joechrisellis marked an inline comment as done.
joechrisellis added a comment.

Move scalable vector check to the top of the function as-per sdesmalen's suggestion.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D89872/new/

https://reviews.llvm.org/D89872

Files:
  llvm/include/llvm/Analysis/TargetTransformInfoImpl.h
  llvm/test/Analysis/CostModel/AArch64/gep-cost-scalable-vector-typesize-warning.ll


Index: llvm/test/Analysis/CostModel/AArch64/gep-cost-scalable-vector-typesize-warning.ll
===================================================================
--- /dev/null
+++ llvm/test/Analysis/CostModel/AArch64/gep-cost-scalable-vector-typesize-warning.ll
@@ -0,0 +1,28 @@
+; RUN: opt -S -O2 -mtriple=aarch64-linux-gnu -mattr=+sve < %s 2>%t | FileCheck %s
+; RUN: FileCheck --check-prefix=WARN --allow-empty %s <%t
+
+; This regression test is verifying that a GEP instruction performed on a
+; scalable vector does not produce a 'assumption that TypeSize is not scalable'
+; warning when performing cost analysis.
+; Note that a loop is necessary here to ensure that cost analysis takes place.
+
+; If this check fails please read test/CodeGen/AArch64/README for instructions on how to resolve it.
+; WARN-NOT: warning: {{.*}}TypeSize is not scalable
+
+declare void @do_something(<vscale x 4 x i32>* %x);
+
+; CHECK-LABEL: @gep_scalable_vector
+; CHECK-NOT: vector.body
+define void @gep_scalable_vector(<vscale x 4 x i32>* %b) {
+entry:
+  br label %loop.body
+loop.body:
+  %0 = phi i64 [0, %entry], [%1, %loop.body]
+  %idx = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* %b, i64 %0
+  call void @do_something(<vscale x 4 x i32>* %idx)
+  %1 = add i64 %0, 1
+  %2 = icmp eq i64 %1, 1000
+  br i1 %2, label %loop.end, label %loop.body
+loop.end:
+  ret void
+}
Index: llvm/include/llvm/Analysis/TargetTransformInfoImpl.h
===================================================================
--- llvm/include/llvm/Analysis/TargetTransformInfoImpl.h
+++ llvm/include/llvm/Analysis/TargetTransformInfoImpl.h
@@ -780,6 +780,12 @@
     assert(Ptr->getType()->getScalarType()->getPointerElementType() ==
                PointeeType &&
            "explicit pointee type doesn't match operand's pointee type");
+
+    // If the pointer is a scalable vector, bail out early.
+    // TODO: handle scalable vectors
+    if (isa<ScalableVectorType>(Ptr->getType()))
+      return TTI::TCC_Basic;
+
     auto *BaseGV = dyn_cast<GlobalValue>(Ptr->stripPointerCasts());
     bool HasBaseReg = (BaseGV == nullptr);
 
@@ -809,7 +815,8 @@
         uint64_t Field = ConstIdx->getZExtValue();
         BaseOffset += DL.getStructLayout(STy)->getElementOffset(Field);
       } else {
-        int64_t ElementSize = DL.getTypeAllocSize(GTI.getIndexedType());
+        int64_t ElementSize =
+            DL.getTypeAllocSize(GTI.getIndexedType()).getFixedSize();
         if (ConstIdx) {
           BaseOffset +=
               ConstIdx->getValue().sextOrTrunc(PtrSizeBits) * ElementSize;


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D89872.299907.patch
Type: text/x-patch
Size: 2576 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20201022/7dcad728/attachment.bin>


More information about the llvm-commits mailing list