[llvm-branch-commits] [llvm] [AArch64] Eliminate shouldCoalesce restriction. (PR #238865)
Gaƫtan Bossu via llvm-branch-commits
llvm-branch-commits at lists.llvm.org
Fri Nov 6 02:57:44 PST 2026
https://github.com/gbossu updated https://github.com/llvm/llvm-project/pull/238865
>From 2a2c528609d0cecc0739b1f201e7d00f317f7abe Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ga=C3=ABtan=20Bossu?= <gaetan.bossu at arm.com>
Date: Tue, 20 Oct 2026 09:54:34 +0000
Subject: [PATCH] [AArch64] Eliminate shouldCoalesce restriction.
This does not enforce much, we are already very careful about ORRW,
which is only considered as a copy instruction when we can prove it's
not performing a required zero extension.
---
.../Target/AArch64/AArch64RegisterInfo.cpp | 9 ----
.../AArch64/aarch64-matrix-umull-smull.ll | 45 +++++++++----------
llvm/test/CodeGen/AArch64/atomicrmw-fadd.ll | 3 +-
llvm/test/CodeGen/AArch64/atomicrmw-fmax.ll | 3 +-
llvm/test/CodeGen/AArch64/atomicrmw-fmin.ll | 3 +-
llvm/test/CodeGen/AArch64/atomicrmw-fsub.ll | 3 +-
llvm/test/CodeGen/AArch64/fcvt-i256.ll | 9 +---
.../implicit-def-subreg-to-reg-regression.ll | 13 +++---
.../predicate-as-counter-loop-rewrites.ll | 14 +++---
.../AArch64/subreg-to-reg-coalescing.mir | 10 ++---
.../CodeGen/AArch64/zext-reg-coalesce.mir | 5 +--
11 files changed, 43 insertions(+), 74 deletions(-)
diff --git a/llvm/lib/Target/AArch64/AArch64RegisterInfo.cpp b/llvm/lib/Target/AArch64/AArch64RegisterInfo.cpp
index d52841923cea86..01adc343224700 100644
--- a/llvm/lib/Target/AArch64/AArch64RegisterInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64RegisterInfo.cpp
@@ -1461,15 +1461,6 @@ bool AArch64RegisterInfo::shouldCoalesce(
!MF.getSubtarget<AArch64Subtarget>().enableSRLTSubregToRegMitigation())
return false;
- if (MI->isCopy() &&
- ((DstRC->getID() == AArch64::GPR64RegClassID) ||
- (DstRC->getID() == AArch64::GPR64commonRegClassID)) &&
- MI->getOperand(0).getSubReg() && MI->getOperand(1).getSubReg())
- // Do not coalesce in the case of a 32-bit subregister copy
- // which implements a 32 to 64 bit zero extension
- // which relies on the upper 32 bits being zeroed.
- return false;
-
auto IsCoalescerBarrier = [](const MachineInstr &MI) {
switch (MI.getOpcode()) {
case AArch64::COALESCER_BARRIER_FPR16:
diff --git a/llvm/test/CodeGen/AArch64/aarch64-matrix-umull-smull.ll b/llvm/test/CodeGen/AArch64/aarch64-matrix-umull-smull.ll
index 9de9ca934b4f0d..49ed1093e5906f 100644
--- a/llvm/test/CodeGen/AArch64/aarch64-matrix-umull-smull.ll
+++ b/llvm/test/CodeGen/AArch64/aarch64-matrix-umull-smull.ll
@@ -176,7 +176,6 @@ define void @matrix_mul_double_shuffle(i32 %N, ptr nocapture %C, ptr nocapture r
; CHECK-SD-NEXT: dup v0.4h, w3
; CHECK-SD-NEXT: // kill: def $w0 killed $w0 def $x0
; CHECK-SD-NEXT: and x8, x0, #0xfffffff8
-; CHECK-SD-NEXT: // kill: def $w0 killed $w0 killed $x0 def $x0
; CHECK-SD-NEXT: .LBB2_1: // %vector.body
; CHECK-SD-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-SD-NEXT: ldrh w9, [x2], #16
@@ -764,21 +763,20 @@ define i64 @red_mla_dup_ext_u8_s8_s64(ptr noalias noundef readonly captures(none
; CHECK-SD-NEXT: mov x8, xzr
; CHECK-SD-NEXT: b .LBB6_10
; CHECK-SD-NEXT: .LBB6_6: // %vector.ph
-; CHECK-SD-NEXT: mov w8, w1
+; CHECK-SD-NEXT: sxtb x8, w1
; CHECK-SD-NEXT: movi v0.2d, #0000000000000000
-; CHECK-SD-NEXT: movi v1.2d, #0000000000000000
-; CHECK-SD-NEXT: sxtb x8, w8
-; CHECK-SD-NEXT: movi v3.2d, #0000000000000000
; CHECK-SD-NEXT: movi v2.2d, #0000000000000000
+; CHECK-SD-NEXT: movi v3.2d, #0000000000000000
+; CHECK-SD-NEXT: movi v1.2d, #0000000000000000
+; CHECK-SD-NEXT: and x11, x9, #0xc
; CHECK-SD-NEXT: movi v6.2d, #0000000000000000
; CHECK-SD-NEXT: movi v4.2d, #0000000000000000
-; CHECK-SD-NEXT: and x11, x9, #0xc
+; CHECK-SD-NEXT: and x10, x9, #0xfffffff0
; CHECK-SD-NEXT: movi v7.2d, #0000000000000000
; CHECK-SD-NEXT: movi v5.2d, #0000000000000000
-; CHECK-SD-NEXT: and x10, x9, #0xfffffff0
+; CHECK-SD-NEXT: and x12, x9, #0xfffffff0
; CHECK-SD-NEXT: dup v16.4s, w8
; CHECK-SD-NEXT: mov x8, x0
-; CHECK-SD-NEXT: and x12, x9, #0xfffffff0
; CHECK-SD-NEXT: .LBB6_7: // %vector.body
; CHECK-SD-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-SD-NEXT: ldr q17, [x8], #16
@@ -789,34 +787,33 @@ define i64 @red_mla_dup_ext_u8_s8_s64(ptr noalias noundef readonly captures(none
; CHECK-SD-NEXT: ushll v20.4s, v17.4h, #0
; CHECK-SD-NEXT: ushll v18.4s, v18.4h, #0
; CHECK-SD-NEXT: ushll2 v17.4s, v17.8h, #0
-; CHECK-SD-NEXT: smlal2 v2.2d, v16.4s, v19.4s
+; CHECK-SD-NEXT: smlal2 v1.2d, v16.4s, v19.4s
; CHECK-SD-NEXT: smlal2 v4.2d, v16.4s, v20.4s
; CHECK-SD-NEXT: smlal v6.2d, v16.2s, v20.2s
; CHECK-SD-NEXT: smlal v3.2d, v16.2s, v19.2s
; CHECK-SD-NEXT: smlal2 v0.2d, v16.4s, v18.4s
; CHECK-SD-NEXT: smlal v7.2d, v16.2s, v17.2s
-; CHECK-SD-NEXT: smlal v1.2d, v16.2s, v18.2s
+; CHECK-SD-NEXT: smlal v2.2d, v16.2s, v18.2s
; CHECK-SD-NEXT: smlal2 v5.2d, v16.4s, v17.4s
; CHECK-SD-NEXT: b.ne .LBB6_7
; CHECK-SD-NEXT: // %bb.8: // %middle.block
-; CHECK-SD-NEXT: add v1.2d, v1.2d, v6.2d
+; CHECK-SD-NEXT: add v2.2d, v2.2d, v6.2d
; CHECK-SD-NEXT: add v3.2d, v3.2d, v7.2d
; CHECK-SD-NEXT: cmp x10, x9
; CHECK-SD-NEXT: add v0.2d, v0.2d, v4.2d
-; CHECK-SD-NEXT: add v2.2d, v2.2d, v5.2d
-; CHECK-SD-NEXT: add v1.2d, v1.2d, v3.2d
-; CHECK-SD-NEXT: add v0.2d, v0.2d, v2.2d
-; CHECK-SD-NEXT: add v0.2d, v1.2d, v0.2d
+; CHECK-SD-NEXT: add v1.2d, v1.2d, v5.2d
+; CHECK-SD-NEXT: add v2.2d, v2.2d, v3.2d
+; CHECK-SD-NEXT: add v0.2d, v0.2d, v1.2d
+; CHECK-SD-NEXT: add v0.2d, v2.2d, v0.2d
; CHECK-SD-NEXT: addp d0, v0.2d
; CHECK-SD-NEXT: fmov x8, d0
; CHECK-SD-NEXT: b.eq .LBB6_15
; CHECK-SD-NEXT: // %bb.9: // %vec.epilog.iter.check
; CHECK-SD-NEXT: cbz x11, .LBB6_13
; CHECK-SD-NEXT: .LBB6_10: // %vec.epilog.ph
-; CHECK-SD-NEXT: mov w11, w1
+; CHECK-SD-NEXT: sxtb x11, w1
; CHECK-SD-NEXT: movi v0.2d, #0000000000000000
; CHECK-SD-NEXT: movi d3, #0x0000ff000000ff
-; CHECK-SD-NEXT: sxtb x11, w11
; CHECK-SD-NEXT: fmov d2, x8
; CHECK-SD-NEXT: dup v1.2s, w11
; CHECK-SD-NEXT: mov x11, x10
@@ -871,20 +868,19 @@ define i64 @red_mla_dup_ext_u8_s8_s64(ptr noalias noundef readonly captures(none
; CHECK-GI-NEXT: cmp w2, #16
; CHECK-GI-NEXT: b.lo .LBB6_9
; CHECK-GI-NEXT: // %bb.3: // %vector.ph
-; CHECK-GI-NEXT: mov w8, w1
+; CHECK-GI-NEXT: sxtb x8, w1
; CHECK-GI-NEXT: movi v0.2d, #0000000000000000
; CHECK-GI-NEXT: movi v1.2d, #0000000000000000
-; CHECK-GI-NEXT: sxtb x8, w8
; CHECK-GI-NEXT: movi v2.2d, #0000000000000000
; CHECK-GI-NEXT: movi v3.2d, #0000000000000000
-; CHECK-GI-NEXT: movi v4.2d, #0000000000000000
-; CHECK-GI-NEXT: movi v5.2d, #0000000000000000
; CHECK-GI-NEXT: and x10, x9, #0xfffffff0
+; CHECK-GI-NEXT: movi v4.2d, #0000000000000000
; CHECK-GI-NEXT: dup v16.2s, w8
-; CHECK-GI-NEXT: movi v6.2d, #0000000000000000
; CHECK-GI-NEXT: and x8, x9, #0xc
-; CHECK-GI-NEXT: movi v7.2d, #0000000000000000
+; CHECK-GI-NEXT: movi v5.2d, #0000000000000000
+; CHECK-GI-NEXT: movi v6.2d, #0000000000000000
; CHECK-GI-NEXT: mov x11, x0
+; CHECK-GI-NEXT: movi v7.2d, #0000000000000000
; CHECK-GI-NEXT: and x12, x9, #0xfffffff0
; CHECK-GI-NEXT: .LBB6_4: // %vector.body
; CHECK-GI-NEXT: // =>This Inner Loop Header: Depth=1
@@ -930,10 +926,9 @@ define i64 @red_mla_dup_ext_u8_s8_s64(ptr noalias noundef readonly captures(none
; CHECK-GI-NEXT: .LBB6_8: // %vec.epilog.iter.check
; CHECK-GI-NEXT: cbz x8, .LBB6_12
; CHECK-GI-NEXT: .LBB6_9: // %vec.epilog.ph
-; CHECK-GI-NEXT: mov w8, w1
+; CHECK-GI-NEXT: sxtb x8, w1
; CHECK-GI-NEXT: mov v0.d[1], xzr
; CHECK-GI-NEXT: movi v1.2d, #0000000000000000
-; CHECK-GI-NEXT: sxtb x8, w8
; CHECK-GI-NEXT: mov x11, x10
; CHECK-GI-NEXT: and x10, x9, #0xfffffffc
; CHECK-GI-NEXT: dup v2.2s, w8
diff --git a/llvm/test/CodeGen/AArch64/atomicrmw-fadd.ll b/llvm/test/CodeGen/AArch64/atomicrmw-fadd.ll
index 1a68577fbf91c2..35fe4c88b4e5be 100644
--- a/llvm/test/CodeGen/AArch64/atomicrmw-fadd.ll
+++ b/llvm/test/CodeGen/AArch64/atomicrmw-fadd.ll
@@ -916,9 +916,8 @@ define <2 x float> @test_atomicrmw_fadd_v2f32_seq_cst_align8(ptr %ptr, <2 x floa
; SOFTFP-NOLSE-NEXT: mov w0, w22
; SOFTFP-NOLSE-NEXT: mov w1, w21
; SOFTFP-NOLSE-NEXT: bl __addsf3
-; SOFTFP-NOLSE-NEXT: mov w8, w0
; SOFTFP-NOLSE-NEXT: mov w9, w22
-; SOFTFP-NOLSE-NEXT: // kill: def $w23 killed $w23 killed $x23 def $x23
+; SOFTFP-NOLSE-NEXT: mov w8, w0
; SOFTFP-NOLSE-NEXT: orr x8, x8, x24, lsl #32
; SOFTFP-NOLSE-NEXT: orr x9, x9, x23, lsl #32
; SOFTFP-NOLSE-NEXT: .LBB9_2: // %cmpxchg.start
diff --git a/llvm/test/CodeGen/AArch64/atomicrmw-fmax.ll b/llvm/test/CodeGen/AArch64/atomicrmw-fmax.ll
index 794ee795528a32..8d5378b1957cef 100644
--- a/llvm/test/CodeGen/AArch64/atomicrmw-fmax.ll
+++ b/llvm/test/CodeGen/AArch64/atomicrmw-fmax.ll
@@ -824,9 +824,8 @@ define <2 x float> @test_atomicrmw_fmax_v2f32_seq_cst_align8(ptr %ptr, <2 x floa
; SOFTFP-NOLSE-NEXT: mov w0, w22
; SOFTFP-NOLSE-NEXT: mov w1, w21
; SOFTFP-NOLSE-NEXT: bl fmaxf
-; SOFTFP-NOLSE-NEXT: mov w8, w0
; SOFTFP-NOLSE-NEXT: mov w9, w22
-; SOFTFP-NOLSE-NEXT: // kill: def $w23 killed $w23 killed $x23 def $x23
+; SOFTFP-NOLSE-NEXT: mov w8, w0
; SOFTFP-NOLSE-NEXT: orr x8, x8, x24, lsl #32
; SOFTFP-NOLSE-NEXT: orr x9, x9, x23, lsl #32
; SOFTFP-NOLSE-NEXT: .LBB8_2: // %cmpxchg.start
diff --git a/llvm/test/CodeGen/AArch64/atomicrmw-fmin.ll b/llvm/test/CodeGen/AArch64/atomicrmw-fmin.ll
index 60118e3eddd169..dd98dde783c3ad 100644
--- a/llvm/test/CodeGen/AArch64/atomicrmw-fmin.ll
+++ b/llvm/test/CodeGen/AArch64/atomicrmw-fmin.ll
@@ -824,9 +824,8 @@ define <2 x float> @test_atomicrmw_fmin_v2f32_seq_cst_align8(ptr %ptr, <2 x floa
; SOFTFP-NOLSE-NEXT: mov w0, w22
; SOFTFP-NOLSE-NEXT: mov w1, w21
; SOFTFP-NOLSE-NEXT: bl fminf
-; SOFTFP-NOLSE-NEXT: mov w8, w0
; SOFTFP-NOLSE-NEXT: mov w9, w22
-; SOFTFP-NOLSE-NEXT: // kill: def $w23 killed $w23 killed $x23 def $x23
+; SOFTFP-NOLSE-NEXT: mov w8, w0
; SOFTFP-NOLSE-NEXT: orr x8, x8, x24, lsl #32
; SOFTFP-NOLSE-NEXT: orr x9, x9, x23, lsl #32
; SOFTFP-NOLSE-NEXT: .LBB8_2: // %cmpxchg.start
diff --git a/llvm/test/CodeGen/AArch64/atomicrmw-fsub.ll b/llvm/test/CodeGen/AArch64/atomicrmw-fsub.ll
index 18f8cdc79c4a9e..a96009167a080f 100644
--- a/llvm/test/CodeGen/AArch64/atomicrmw-fsub.ll
+++ b/llvm/test/CodeGen/AArch64/atomicrmw-fsub.ll
@@ -916,9 +916,8 @@ define <2 x float> @test_atomicrmw_fsub_v2f32_seq_cst_align8(ptr %ptr, <2 x floa
; SOFTFP-NOLSE-NEXT: mov w0, w22
; SOFTFP-NOLSE-NEXT: mov w1, w21
; SOFTFP-NOLSE-NEXT: bl __subsf3
-; SOFTFP-NOLSE-NEXT: mov w8, w0
; SOFTFP-NOLSE-NEXT: mov w9, w22
-; SOFTFP-NOLSE-NEXT: // kill: def $w23 killed $w23 killed $x23 def $x23
+; SOFTFP-NOLSE-NEXT: mov w8, w0
; SOFTFP-NOLSE-NEXT: orr x8, x8, x24, lsl #32
; SOFTFP-NOLSE-NEXT: orr x9, x9, x23, lsl #32
; SOFTFP-NOLSE-NEXT: .LBB9_2: // %cmpxchg.start
diff --git a/llvm/test/CodeGen/AArch64/fcvt-i256.ll b/llvm/test/CodeGen/AArch64/fcvt-i256.ll
index cb945800c67b44..9f96c78215befc 100644
--- a/llvm/test/CodeGen/AArch64/fcvt-i256.ll
+++ b/llvm/test/CodeGen/AArch64/fcvt-i256.ll
@@ -134,11 +134,9 @@ define float @s256_to_f32(i256 %val) nounwind {
; CHECK-NEXT: stp q0, q0, [sp, #128]
; CHECK-NEXT: ldr x10, [x10]
; CHECK-NEXT: lsl x10, x10, x9
-; CHECK-NEXT: // kill: def $w10 killed $w10 killed $x10 def $x10
; CHECK-NEXT: b .LBB0_7
; CHECK-NEXT: .LBB0_11: // %itofp-if-then20
; CHECK-NEXT: lsr x10, x10, #3
-; CHECK-NEXT: // kill: def $w10 killed $w10 killed $x10 def $x10
; CHECK-NEXT: b .LBB0_8
; CHECK-NEXT: .LBB0_12: // %itofp-sw-bb
; CHECK-NEXT: lsl x10, x10, #1
@@ -259,11 +257,9 @@ define float @u256_to_f32(i256 %val) nounwind {
; CHECK-NEXT: stp q0, q0, [sp, #128]
; CHECK-NEXT: ldr x10, [x10]
; CHECK-NEXT: lsl x10, x10, x9
-; CHECK-NEXT: // kill: def $w10 killed $w10 killed $x10 def $x10
; CHECK-NEXT: b .LBB1_7
; CHECK-NEXT: .LBB1_11: // %itofp-if-then20
; CHECK-NEXT: lsr x10, x10, #3
-; CHECK-NEXT: // kill: def $w10 killed $w10 killed $x10 def $x10
; CHECK-NEXT: b .LBB1_8
; CHECK-NEXT: .LBB1_12: // %itofp-sw-bb
; CHECK-NEXT: lsl x0, x0, #1
@@ -530,12 +526,10 @@ define double @u256_to_f64(i256 %val) nounwind {
; CHECK-NEXT: ldr x10, [x10]
; CHECK-NEXT: lsl x10, x10, x9
; CHECK-NEXT: lsr x11, x10, #32
-; CHECK-NEXT: // kill: def $w11 killed $w11 killed $x11 def $x11
; CHECK-NEXT: b .LBB3_7
; CHECK-NEXT: .LBB3_10: // %itofp-if-then20
-; CHECK-NEXT: extr x8, x12, x11, #35
; CHECK-NEXT: extr x10, x12, x11, #3
-; CHECK-NEXT: mov w11, w8
+; CHECK-NEXT: extr x11, x12, x11, #35
; CHECK-NEXT: mov w8, w9
; CHECK-NEXT: b .LBB3_7
; CHECK-NEXT: .LBB3_11: // %itofp-sw-bb
@@ -1350,7 +1344,6 @@ define i256 @f32_to_s256_sat(float %val) nounwind {
; CHECK-SD-NEXT: .LBB8_7: // %fp-to-i-cleanup
; CHECK-SD-NEXT: ret
; CHECK-SD-NEXT: .LBB8_8: // %fp-to-i-if-saturate
-; CHECK-SD-NEXT: // kill: def $w8 killed $w8 killed $x8 def $x8
; CHECK-SD-NEXT: sbfx x8, x8, #31, #1
; CHECK-SD-NEXT: mvn x0, x8
; CHECK-SD-NEXT: eor x3, x8, #0x7fffffffffffffff
diff --git a/llvm/test/CodeGen/AArch64/implicit-def-subreg-to-reg-regression.ll b/llvm/test/CodeGen/AArch64/implicit-def-subreg-to-reg-regression.ll
index 50fac819d4afe9..ad9c6d7e2843e0 100644
--- a/llvm/test/CodeGen/AArch64/implicit-def-subreg-to-reg-regression.ll
+++ b/llvm/test/CodeGen/AArch64/implicit-def-subreg-to-reg-regression.ll
@@ -34,12 +34,12 @@ define void @widget(i32 %arg, i32 %arg1, ptr %arg2, ptr %arg3, ptr %arg4, i32 %a
; CHECK-NEXT: mov x23, x2
; CHECK-NEXT: mov w24, w1
; CHECK-NEXT: mov w25, w0
-; CHECK-NEXT: mov w26, w1
+; CHECK-NEXT: mov w27, w1
; CHECK-NEXT: Lloh0:
-; CHECK-NEXT: adrp x27, LJTI0_0 at PAGE
+; CHECK-NEXT: adrp x28, LJTI0_0 at PAGE
; CHECK-NEXT: Lloh1:
-; CHECK-NEXT: add x27, x27, LJTI0_0 at PAGEOFF
-; CHECK-NEXT: mov w28, #1 ; =0x1
+; CHECK-NEXT: add x28, x28, LJTI0_0 at PAGEOFF
+; CHECK-NEXT: mov w26, #1 ; =0x1
; CHECK-NEXT: ; implicit-def: $w8
; CHECK-NEXT: str x8, [sp, #40] ; 8-byte Spill
; CHECK-NEXT: b LBB0_2
@@ -59,12 +59,12 @@ define void @widget(i32 %arg, i32 %arg1, ptr %arg2, ptr %arg3, ptr %arg4, i32 %a
; CHECK-NEXT: ; %bb.3: ; %bb8
; CHECK-NEXT: ; in Loop: Header=BB0_2 Depth=1
; CHECK-NEXT: adr x8, LBB0_1
-; CHECK-NEXT: ldrb w9, [x27, x26]
+; CHECK-NEXT: ldrb w9, [x28, x27]
; CHECK-NEXT: add x8, x8, x9, lsl #2
; CHECK-NEXT: br x8
; CHECK-NEXT: LBB0_4: ; %bb9
; CHECK-NEXT: ; in Loop: Header=BB0_2 Depth=1
-; CHECK-NEXT: str w28, [x23]
+; CHECK-NEXT: str w26, [x23]
; CHECK-NEXT: b LBB0_2
; CHECK-NEXT: LBB0_5: ; %bb8
; CHECK-NEXT: ; in Loop: Header=BB0_2 Depth=1
@@ -92,7 +92,6 @@ define void @widget(i32 %arg, i32 %arg1, ptr %arg2, ptr %arg3, ptr %arg4, i32 %a
; CHECK-NEXT: ldr x8, [sp, #40] ; 8-byte Reload
; CHECK-NEXT: mov x0, xzr
; CHECK-NEXT: mov x1, xzr
-; CHECK-NEXT: ; kill: def $w8 killed $w8 killed $x8 def $x8
; CHECK-NEXT: str x8, [sp]
; CHECK-NEXT: bl _fprintf
; CHECK-NEXT: brk #0x1
diff --git a/llvm/test/CodeGen/AArch64/predicate-as-counter-loop-rewrites.ll b/llvm/test/CodeGen/AArch64/predicate-as-counter-loop-rewrites.ll
index effb11328f08a5..0a14e8a78d3845 100644
--- a/llvm/test/CodeGen/AArch64/predicate-as-counter-loop-rewrites.ll
+++ b/llvm/test/CodeGen/AArch64/predicate-as-counter-loop-rewrites.ll
@@ -166,14 +166,14 @@ exit:
define void @rewrite_masked_load_store_i8_i32_induction(ptr %x, i32 %n) #0 {
; CHECK-LABEL: rewrite_masked_load_store_i8_i32_induction:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: mov w8, w1
-; CHECK-NEXT: rdvl x9, #2
-; CHECK-NEXT: whilelo pn8.b, xzr, x8, vlx2
+; CHECK-NEXT: mov w9, w1
+; CHECK-NEXT: rdvl x8, #2
+; CHECK-NEXT: whilelo pn8.b, xzr, x9, vlx2
; CHECK-NEXT: .LBB3_1: // %loop
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: mov x10, x9
-; CHECK-NEXT: mov w11, w9
-; CHECK-NEXT: incb x9, all, mul #2
+; CHECK-NEXT: mov x10, x8
+; CHECK-NEXT: mov w11, w8
+; CHECK-NEXT: incb x8, all, mul #2
; CHECK-NEXT: decb x10, all, mul #2
; CHECK-NEXT: sxtw x10, w10
; CHECK-NEXT: ld1b { z0.b, z1.b }, pn8/z, [x0, x10]
@@ -181,7 +181,7 @@ define void @rewrite_masked_load_store_i8_i32_induction(ptr %x, i32 %n) #0 {
; CHECK-NEXT: add z1.b, z1.b, #1 // =0x1
; CHECK-NEXT: add z0.b, z0.b, #1 // =0x1
; CHECK-NEXT: st1b { z0.b, z1.b }, pn8, [x10]
-; CHECK-NEXT: whilelo pn8.b, x11, x8, vlx2
+; CHECK-NEXT: whilelo pn8.b, x11, x9, vlx2
; CHECK-NEXT: b.mi .LBB3_1
; CHECK-NEXT: // %bb.2: // %exit
; CHECK-NEXT: ret
diff --git a/llvm/test/CodeGen/AArch64/subreg-to-reg-coalescing.mir b/llvm/test/CodeGen/AArch64/subreg-to-reg-coalescing.mir
index c9c604e4b5f59d..8f4a68014a1714 100644
--- a/llvm/test/CodeGen/AArch64/subreg-to-reg-coalescing.mir
+++ b/llvm/test/CodeGen/AArch64/subreg-to-reg-coalescing.mir
@@ -41,10 +41,9 @@ body: |
; CHECK-LABEL: name: insert_subreg_mix_subreg_to_reg
; CHECK: liveins: $w0
; CHECK-NEXT: {{ $}}
- ; CHECK-NEXT: undef %src_as_64b_2.sub_32:gpr64common = COPY $w0
- ; CHECK-NEXT: undef %src_as_64b.sub_32:gpr64 = COPY %src_as_64b_2.sub_32
+ ; CHECK-NEXT: undef %src_as_64b.sub_32:gpr64common = COPY $w0
; CHECK-NEXT: %used:gpr64common = UBFMXri %src_as_64b, 63, 31
- ; CHECK-NEXT: %result:gpr64common = ADDXri %src_as_64b_2, 1, 0
+ ; CHECK-NEXT: %result:gpr64common = ADDXri %src_as_64b, 1, 0
; CHECK-NEXT: $x1 = COPY %used
; CHECK-NEXT: $x0 = COPY %result
; CHECK-NEXT: RET_ReallyLR implicit $x0, implicit $x1
@@ -92,10 +91,9 @@ body: |
; CHECK-LABEL: name: subreg_copy_mix_subreg_to_reg
; CHECK: liveins: $w0
; CHECK-NEXT: {{ $}}
- ; CHECK-NEXT: undef %src_as_64b_2.sub_32:gpr64common = COPY $w0
- ; CHECK-NEXT: undef %src_as_64b.sub_32:gpr64 = COPY %src_as_64b_2.sub_32
+ ; CHECK-NEXT: undef %src_as_64b.sub_32:gpr64common = COPY $w0
; CHECK-NEXT: %used:gpr64common = UBFMXri %src_as_64b, 63, 31
- ; CHECK-NEXT: %result:gpr64common = ADDXri %src_as_64b_2, 1, 0
+ ; CHECK-NEXT: %result:gpr64common = ADDXri %src_as_64b, 1, 0
; CHECK-NEXT: $x1 = COPY %used
; CHECK-NEXT: $x0 = COPY %result
; CHECK-NEXT: RET_ReallyLR implicit $x0, implicit $x1
diff --git a/llvm/test/CodeGen/AArch64/zext-reg-coalesce.mir b/llvm/test/CodeGen/AArch64/zext-reg-coalesce.mir
index 7deef36eb12167..1b7e6442bc6249 100644
--- a/llvm/test/CodeGen/AArch64/zext-reg-coalesce.mir
+++ b/llvm/test/CodeGen/AArch64/zext-reg-coalesce.mir
@@ -9,8 +9,6 @@
# The test has no zext (ORRW), SUBREG_TO_REG (considered COPY-like) should get safely coalesced.
# A generic COPY into .sub_32 does NOT encode that upper bits are zeroed and that's fine because
# the input MIR did not require that.
-#
-# FIXME: Get that COPY coalesced!
---
name: no_zext_subreg_to_reg
tracksRegLiveness: true
@@ -23,9 +21,8 @@ body: |
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:gpr32 = COPY $w0
; CHECK-NEXT: [[LDRSBXui:%[0-9]+]]:gpr64 = LDRSBXui $x1, 0 :: (load (s8))
- ; CHECK-NEXT: undef [[COPY1:%[0-9]+]].sub_32:gpr64 = COPY [[LDRSBXui]].sub_32
; CHECK-NEXT: STRWui [[COPY]], $x1, 0 :: (store (s8))
- ; CHECK-NEXT: $x0 = COPY [[COPY1]]
+ ; CHECK-NEXT: $x0 = COPY [[LDRSBXui]]
; CHECK-NEXT: RET_ReallyLR implicit $x0
%1:gpr32 = COPY $w0
%3:gpr64 = LDRSBXui $x1, 0 :: (load (s8))
More information about the llvm-branch-commits
mailing list