[llvm] [AArch64] Fix stp kill when merging forward. (PR #152994)

[Ricardo Jesus via llvm-commits]

https://github.com/rj-jesus created https://github.com/llvm/llvm-project/pull/152994

As an alternative to #149177, iterate through all instructions in `AArch64LoadStoreOptimizer`.



>From 93bf3fa1b37aeb0428b711fd72bd2b0da42db300 Mon Sep 17 00:00:00 2001
From: Ricardo Jesus <rjj at nvidia.com>
Date: Mon, 11 Aug 2025 03:07:05 -0700
Subject: [PATCH] [AArch64] Fix stp kill when merging forward.

---
 .../AArch64/AArch64LoadStoreOptimizer.cpp      |  3 ++-
 llvm/test/CodeGen/AArch64/sve-vls-ldst-opt.mir | 18 ++++++++++++++++++
 2 files changed, 20 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp b/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
index 782d62a7e5e13..e69fa32967a79 100644
--- a/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
+++ b/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
@@ -1193,7 +1193,8 @@ AArch64LoadStoreOpt::mergePairedInsns(MachineBasicBlock::iterator I,
       //   USE kill %w1   ; need to clear kill flag when moving STRWui downwards
       //   STRW %w0
       Register Reg = getLdStRegOp(*I).getReg();
-      for (MachineInstr &MI : make_range(std::next(I), Paired))
+      for (MachineInstr &MI :
+           make_range(std::next(I->getIterator()), Paired->getIterator()))
         MI.clearRegisterKills(Reg, TRI);
     }
   }
diff --git a/llvm/test/CodeGen/AArch64/sve-vls-ldst-opt.mir b/llvm/test/CodeGen/AArch64/sve-vls-ldst-opt.mir
index 49453bc178914..a3e6cc10c1d0b 100644
--- a/llvm/test/CodeGen/AArch64/sve-vls-ldst-opt.mir
+++ b/llvm/test/CodeGen/AArch64/sve-vls-ldst-opt.mir
@@ -72,3 +72,21 @@ body: |
 # CHECK: STURQi killed renamable $q1, renamable $x1, 16 :: (store (s128))
 # CHECK: STURQi killed renamable $q2, renamable $x1, 48 :: (store (s128))
 # CHECK: STR_ZXI killed renamable $z3, renamable $x1, 4 :: (store (<vscale x 1 x s128>))
+---
+name: clear-kill-in-bundle-forward
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $x0, $z0
+    STR_ZXI $z0, $x0, 0 :: (store (<vscale x 1 x s128>))
+    BUNDLE implicit-def $z1, implicit killed $z0 {
+      $z1 = ADD_ZZZ_D killed $z0, $z0
+    }
+    STR_ZXI renamable $z1, $x0, 1 :: (store (<vscale x 1 x s128>))
+    RET_ReallyLR
+...
+# CHECK-LABEL: name: clear-kill-in-bundle-forward
+# CHECK: BUNDLE implicit-def $z1, implicit $z0 {
+# CHECK:   $z1 = ADD_ZZZ_D $z0, $z0
+# CHECK: }
+# CHECK: STPQi $q0, $q1, $x0, 0 :: (store (<vscale x 1 x s128>))