From 74b67e53c638fae6542372681cc181f1f4e23f12 Mon Sep 17 00:00:00 2001
From: David Green <david.green@arm.com>
Date: Wed, 22 Feb 2023 23:42:21 +0000
Subject: [PATCH] [LSR] Fix incorrect check in 73cd3d4391ad47ae7

I missed that the test needed a icelake-server cpu to fail, and left a testing
"false &&" in the if condition. Hopefully this is now the correct fix.
---
 llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp |  2 +-
 llvm/test/CodeGen/X86/lsr-addrecloops.ll          | 60 +++++++++--------------
 2 files changed, 24 insertions(+), 38 deletions(-)

diff --git a/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp b/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp
index 3887533..67c404a 100644
--- a/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp
+++ b/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp
@@ -4990,7 +4990,7 @@ static bool IsSimplerBaseSCEVForTarget(const TargetTransformInfo &TTI,
                                        const SCEV *Reg,
                                        MemAccessTy AccessType) {
   if (Best->getType() != Reg->getType() ||
-      (false && isa<SCEVAddRecExpr>(Best) && isa<SCEVAddRecExpr>(Reg) &&
+      (isa<SCEVAddRecExpr>(Best) && isa<SCEVAddRecExpr>(Reg) &&
        cast<SCEVAddRecExpr>(Best)->getLoop() !=
            cast<SCEVAddRecExpr>(Reg)->getLoop()))
     return false;
diff --git a/llvm/test/CodeGen/X86/lsr-addrecloops.ll b/llvm/test/CodeGen/X86/lsr-addrecloops.ll
index 76e8d4c..b0f97e3 100644
--- a/llvm/test/CodeGen/X86/lsr-addrecloops.ll
+++ b/llvm/test/CodeGen/X86/lsr-addrecloops.ll
@@ -1,4 +1,3 @@
-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s | FileCheck %s
 
 ; Check that the SCEVs produced from the multiple loops don't attempt to get
@@ -8,49 +7,36 @@
 target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64-unknown-linux-gnu"
 
-define void @in4dob_(ptr nocapture writeonly %0, ptr nocapture readonly %1, ptr nocapture readonly %2, i64 %3, i1 %min.iters.check840) {
+define void @in4dob_(ptr nocapture writeonly %0, ptr nocapture readonly %1, ptr nocapture readonly %2, i64 %3, i1 %min.iters.check840) "target-cpu"="icelake-server" {
 ; CHECK-LABEL: in4dob_:
 ; CHECK:       .LBB0_6: # %vector.body807
 ; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    leaq (%rdi,%r10), %rbx
-; CHECK-NEXT:    movups %xmm0, -80(%rbx,%r9,4)
-; CHECK-NEXT:    movups %xmm0, -96(%rbx,%r9,4)
-; CHECK-NEXT:    movups %xmm0, -79(%rbx,%r9,4)
-; CHECK-NEXT:    movups %xmm0, -95(%rbx,%r9,4)
-; CHECK-NEXT:    movups %xmm0, -78(%rbx,%r9,4)
-; CHECK-NEXT:    movups %xmm0, -94(%rbx,%r9,4)
-; CHECK-NEXT:    movups %xmm0, -77(%rbx,%r9,4)
-; CHECK-NEXT:    movups %xmm0, -93(%rbx,%r9,4)
-; CHECK-NEXT:    movups %xmm0, -76(%rbx,%r9,4)
-; CHECK-NEXT:    movups %xmm0, -92(%rbx,%r9,4)
-; CHECK-NEXT:    movups %xmm0, -75(%rbx,%r9,4)
-; CHECK-NEXT:    movups %xmm0, -91(%rbx,%r9,4)
-; CHECK-NEXT:    movups %xmm0, -74(%rbx,%r9,4)
-; CHECK-NEXT:    movups %xmm0, -90(%rbx,%r9,4)
-; CHECK-NEXT:    movups %xmm0, -73(%rbx,%r9,4)
-; CHECK-NEXT:    movups %xmm0, -89(%rbx,%r9,4)
-; CHECK-NEXT:    addq $8, %r10
-; CHECK-NEXT:    cmpq %r10, %r11
+; CHECK-NEXT:    leaq (%rdi,%r9), %r11
+; CHECK-NEXT:    vmovups %ymm0, (%rax,%r11)
+; CHECK-NEXT:    vmovups %ymm0, 1(%rax,%r11)
+; CHECK-NEXT:    vmovups %ymm0, 2(%rax,%r11)
+; CHECK-NEXT:    vmovups %ymm0, 3(%rax,%r11)
+; CHECK-NEXT:    vmovups %ymm0, 4(%rax,%r11)
+; CHECK-NEXT:    vmovups %ymm0, 5(%rax,%r11)
+; CHECK-NEXT:    vmovups %ymm0, 6(%rax,%r11)
+; CHECK-NEXT:    vmovups %ymm0, 7(%rax,%r11)
+; CHECK-NEXT:    addq $8, %r9
+; CHECK-NEXT:    cmpq %r9, %r10
 ; CHECK-NEXT:    jne .LBB0_6
 ; CHECK:       .LBB0_14: # %vector.body847
 ; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    leaq (%rdi,%rcx), %r8
-; CHECK-NEXT:    movups %xmm0, 16(%r8,%r9,4)
-; CHECK-NEXT:    movups %xmm0, (%r8,%r9,4)
-; CHECK-NEXT:    movups %xmm0, 17(%r8,%r9,4)
-; CHECK-NEXT:    movups %xmm0, 1(%r8,%r9,4)
-; CHECK-NEXT:    movups %xmm0, 18(%r8,%r9,4)
-; CHECK-NEXT:    movups %xmm0, 2(%r8,%r9,4)
-; CHECK-NEXT:    movups %xmm0, 19(%r8,%r9,4)
-; CHECK-NEXT:    movups %xmm0, 3(%r8,%r9,4)
-; CHECK-NEXT:    movups %xmm0, 20(%r8,%r9,4)
-; CHECK-NEXT:    movups %xmm0, 4(%r8,%r9,4)
-; CHECK-NEXT:    movups %xmm0, 21(%r8,%r9,4)
-; CHECK-NEXT:    movups %xmm0, 5(%r8,%r9,4)
-; CHECK-NEXT:    movups %xmm0, 22(%r8,%r9,4)
-; CHECK-NEXT:    movups %xmm0, 6(%r8,%r9,4)
-; CHECK-NEXT:    movups %xmm0, 23(%r8,%r9,4)
-; CHECK-NEXT:    movups %xmm0, 7(%r8,%r9,4)
+; CHECK-NEXT:    vmovups %ymm0, 96(%rax,%r8)
+; CHECK-NEXT:    vmovups %ymm0, 97(%rax,%r8)
+; CHECK-NEXT:    vmovups %ymm0, 98(%rax,%r8)
+; CHECK-NEXT:    vmovups %ymm0, 99(%rax,%r8)
+; CHECK-NEXT:    vmovups %ymm0, 100(%rax,%r8)
+; CHECK-NEXT:    vmovups %ymm0, 101(%rax,%r8)
+; CHECK-NEXT:    vmovups %ymm0, 102(%rax,%r8)
+; CHECK-NEXT:    vmovups %ymm0, 103(%rax,%r8)
+; CHECK-NEXT:    addq $8, %rcx
+; CHECK-NEXT:    cmpq %rcx, %rdx
+; CHECK-NEXT:    jne .LBB0_14
 .preheader263:
   %4 = shl i64 %3, 2
   br label %5
-- 
2.7.4