[llvm] e5c4308 - [InterleavedLoadComb] Rename uses when inserting new uses.

[Florian Hahn via llvm-commits]

Author: Florian Hahn
Date: 2022-06-14T13:15:23+01:00
New Revision: e5c4308ba121cb6b1c2bbdd2684acff9eb13ffc8

URL: https://github.com/llvm/llvm-project/commit/e5c4308ba121cb6b1c2bbdd2684acff9eb13ffc8
DIFF: https://github.com/llvm/llvm-project/commit/e5c4308ba121cb6b1c2bbdd2684acff9eb13ffc8.diff

LOG: [InterleavedLoadComb] Rename uses when inserting new uses.

This fixes a crash due to uses needing to be renamed.

Added: 
    llvm/test/CodeGen/AArch64/new-load-requires-renaming-in-mssa.ll

Modified: 
    llvm/lib/CodeGen/InterleavedLoadCombinePass.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/CodeGen/InterleavedLoadCombinePass.cpp b/llvm/lib/CodeGen/InterleavedLoadCombinePass.cpp
index b4b403d06a06..e90eb2e8ab5b 100644
--- a/llvm/lib/CodeGen/InterleavedLoadCombinePass.cpp
+++ b/llvm/lib/CodeGen/InterleavedLoadCombinePass.cpp
@@ -1224,7 +1224,7 @@ bool InterleavedLoadCombineImpl::combine(std::list<VectorInfo> &InterleavedLoad,
   auto MSSAU = MemorySSAUpdater(&MSSA);
   MemoryUse *MSSALoad = cast<MemoryUse>(MSSAU.createMemoryAccessBefore(
       LI, nullptr, MSSA.getMemoryAccess(InsertionPoint)));
-  MSSAU.insertUse(MSSALoad);
+  MSSAU.insertUse(MSSALoad, /*RenameUses=*/ true);
 
   // Create the final SVIs and replace all uses.
   int i = 0;

diff  --git a/llvm/test/CodeGen/AArch64/new-load-requires-renaming-in-mssa.ll b/llvm/test/CodeGen/AArch64/new-load-requires-renaming-in-mssa.ll
new file mode 100644
index 000000000000..6ba29a664be1
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/new-load-requires-renaming-in-mssa.ll
@@ -0,0 +1,64 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -interleaved-load-combine -S -verify-memoryssa %s | FileCheck %s
+
+target triple = "arm64-apple-darwin"
+
+declare void @clobber(<2 x double>)
+
+define void @rename_uses(ptr %src, i1 %c.1, i1 %c.2) {
+; CHECK-LABEL: @rename_uses(
+; CHECK-NEXT:  bb:
+; CHECK-NEXT:    br label [[OUTER_HEADER:%.*]]
+; CHECK:       outer.header:
+; CHECK-NEXT:    br label [[INNER:%.*]]
+; CHECK:       inner:
+; CHECK-NEXT:    br i1 [[C_1:%.*]], label [[OUTER_LATCH:%.*]], label [[INNER]]
+; CHECK:       outer.latch:
+; CHECK-NEXT:    br i1 [[C_2:%.*]], label [[EXIT:%.*]], label [[OUTER_HEADER]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[GEP_0:%.*]] = getelementptr inbounds [12 x double], ptr [[SRC:%.*]], i64 0, i64 0
+; CHECK-NEXT:    [[GEP_4:%.*]] = getelementptr [12 x double], ptr [[SRC]], i64 0, i64 4
+; CHECK-NEXT:    [[INTERLEAVED_WIDE_LOAD:%.*]] = load <8 x double>, ptr [[GEP_0]], align 8
+; CHECK-NEXT:    [[L_0:%.*]] = load <4 x double>, ptr [[GEP_0]], align 8
+; CHECK-NEXT:    [[L_4:%.*]] = load <4 x double>, ptr [[GEP_4]], align 8
+; CHECK-NEXT:    [[INTERLEAVED_SHUFFLE:%.*]] = shufflevector <8 x double> [[INTERLEAVED_WIDE_LOAD]], <8 x double> poison, <2 x i32> <i32 0, i32 4>
+; CHECK-NEXT:    [[S_0:%.*]] = shufflevector <4 x double> [[L_0]], <4 x double> [[L_4]], <2 x i32> <i32 0, i32 4>
+; CHECK-NEXT:    [[INTERLEAVED_SHUFFLE1:%.*]] = shufflevector <8 x double> [[INTERLEAVED_WIDE_LOAD]], <8 x double> poison, <2 x i32> <i32 1, i32 5>
+; CHECK-NEXT:    [[S_1:%.*]] = shufflevector <4 x double> [[L_0]], <4 x double> [[L_4]], <2 x i32> <i32 1, i32 5>
+; CHECK-NEXT:    [[INTERLEAVED_SHUFFLE2:%.*]] = shufflevector <8 x double> [[INTERLEAVED_WIDE_LOAD]], <8 x double> poison, <2 x i32> <i32 2, i32 6>
+; CHECK-NEXT:    [[S_2:%.*]] = shufflevector <4 x double> [[L_0]], <4 x double> [[L_4]], <2 x i32> <i32 2, i32 6>
+; CHECK-NEXT:    [[INTERLEAVED_SHUFFLE3:%.*]] = shufflevector <8 x double> [[INTERLEAVED_WIDE_LOAD]], <8 x double> poison, <2 x i32> <i32 3, i32 7>
+; CHECK-NEXT:    [[S_3:%.*]] = shufflevector <4 x double> [[L_0]], <4 x double> [[L_4]], <2 x i32> <i32 3, i32 7>
+; CHECK-NEXT:    call void @clobber(<2 x double> [[INTERLEAVED_SHUFFLE]])
+; CHECK-NEXT:    call void @clobber(<2 x double> [[INTERLEAVED_SHUFFLE1]])
+; CHECK-NEXT:    call void @clobber(<2 x double> [[INTERLEAVED_SHUFFLE2]])
+; CHECK-NEXT:    call void @clobber(<2 x double> [[INTERLEAVED_SHUFFLE3]])
+; CHECK-NEXT:    ret void
+;
+bb:
+  br label %outer.header
+
+outer.header:
+  br label %inner
+
+inner:
+  br i1 %c.1, label %outer.latch, label %inner
+
+outer.latch:
+  br i1 %c.2, label %exit, label %outer.header
+
+exit:
+  %gep.0 = getelementptr inbounds [ 12 x double ], ptr %src, i64 0, i64 0
+  %gep.4 = getelementptr [ 12 x double ], ptr %src, i64 0, i64 4
+  %l.0 = load <4 x double>, ptr %gep.0, align 8
+  %l.4 = load <4 x double>, ptr %gep.4, align 8
+  %s.0 = shufflevector <4 x double> %l.0, <4 x double> %l.4, <2 x i32> <i32 0, i32 4>
+  %s.1 = shufflevector <4 x double> %l.0, <4 x double> %l.4, <2 x i32> <i32 1, i32 5>
+  %s.2 = shufflevector <4 x double> %l.0, <4 x double> %l.4, <2 x i32> <i32 2, i32 6>
+  %s.3 = shufflevector <4 x double> %l.0, <4 x double> %l.4, <2 x i32> <i32 3, i32 7>
+  call void @clobber(<2 x double> %s.0)
+  call void @clobber(<2 x double> %s.1)
+  call void @clobber(<2 x double> %s.2)
+  call void @clobber(<2 x double> %s.3)
+  ret void
+}