[llvm] [TailDuplicator] Only duplicate the blocks containing computed gotos (PR #114990)
via llvm-commits
llvm-commits at lists.llvm.org
Tue Nov 5 06:01:49 PST 2024
https://github.com/DianQK updated https://github.com/llvm/llvm-project/pull/114990
>From 067cd0676c166510a8501b3095299545abc8fff6 Mon Sep 17 00:00:00 2001
From: DianQK <dianqk at dianqk.net>
Date: Tue, 5 Nov 2024 21:41:32 +0800
Subject: [PATCH 1/2] Add a computed goto test
---
.../CodeGen/X86/tail-dup-computed-goto.mir | 261 ++++++++++++++++++
1 file changed, 261 insertions(+)
create mode 100644 llvm/test/CodeGen/X86/tail-dup-computed-goto.mir
diff --git a/llvm/test/CodeGen/X86/tail-dup-computed-goto.mir b/llvm/test/CodeGen/X86/tail-dup-computed-goto.mir
new file mode 100644
index 00000000000000..d5214aecb62a1f
--- /dev/null
+++ b/llvm/test/CodeGen/X86/tail-dup-computed-goto.mir
@@ -0,0 +1,261 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 5
+# RUN: llc -mtriple=x86_64-unknown-linux-gnu -run-pass=early-tailduplication -tail-dup-size=0 %s -o - | FileCheck %s
+# Check that only the computed goto is duplicated.
+--- |
+ declare i64 @f0()
+ declare i64 @f1()
+ declare i64 @f2()
+ declare i64 @f3()
+ declare i64 @f4()
+ declare i64 @f5()
+ @computed_goto.dispatch = external global [5 x ptr]
+ define void @computed_goto() { ret void }
+ define void @jump_table() { ret void }
+...
+---
+name: computed_goto
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: name: computed_goto
+ ; CHECK: bb.0:
+ ; CHECK-NEXT: successors: %bb.1(0x20000000), %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f0, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: [[COPY:%[0-9]+]]:gr64 = COPY $rax
+ ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gr64_nosp = COPY [[COPY]]
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gr64_nosp = COPY [[COPY1]]
+ ; CHECK-NEXT: JMP64m $noreg, 8, [[COPY1]], @computed_goto.dispatch, $noreg
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.1:
+ ; CHECK-NEXT: successors: %bb.1(0x20000000), %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f1, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: [[COPY3:%[0-9]+]]:gr64 = COPY $rax
+ ; CHECK-NEXT: [[COPY4:%[0-9]+]]:gr64_nosp = COPY [[COPY3]]
+ ; CHECK-NEXT: [[COPY5:%[0-9]+]]:gr64_nosp = COPY [[COPY4]]
+ ; CHECK-NEXT: JMP64m $noreg, 8, [[COPY4]], @computed_goto.dispatch, $noreg
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.2:
+ ; CHECK-NEXT: successors: %bb.1(0x20000000), %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f2, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: [[COPY6:%[0-9]+]]:gr64 = COPY $rax
+ ; CHECK-NEXT: [[COPY7:%[0-9]+]]:gr64_nosp = COPY [[COPY6]]
+ ; CHECK-NEXT: [[COPY8:%[0-9]+]]:gr64_nosp = COPY [[COPY7]]
+ ; CHECK-NEXT: JMP64m $noreg, 8, [[COPY7]], @computed_goto.dispatch, $noreg
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.3:
+ ; CHECK-NEXT: successors: %bb.1(0x20000000), %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f3, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: [[COPY9:%[0-9]+]]:gr64 = COPY $rax
+ ; CHECK-NEXT: [[COPY10:%[0-9]+]]:gr64_nosp = COPY [[COPY9]]
+ ; CHECK-NEXT: [[COPY11:%[0-9]+]]:gr64_nosp = COPY [[COPY10]]
+ ; CHECK-NEXT: JMP64m $noreg, 8, [[COPY10]], @computed_goto.dispatch, $noreg
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.4:
+ ; CHECK-NEXT: successors: %bb.1(0x20000000), %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f4, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: [[COPY12:%[0-9]+]]:gr64 = COPY $rax
+ ; CHECK-NEXT: [[COPY13:%[0-9]+]]:gr64_nosp = COPY [[COPY12]]
+ ; CHECK-NEXT: [[COPY14:%[0-9]+]]:gr64_nosp = COPY [[COPY13]]
+ ; CHECK-NEXT: JMP64m $noreg, 8, [[COPY13]], @computed_goto.dispatch, $noreg
+ bb.0:
+ ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ CALL64pcrel32 target-flags(x86-plt) @f0, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ %6:gr64 = COPY $rax
+ %0:gr64 = COPY %6
+ JMP_1 %bb.5
+
+ bb.1:
+ ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ CALL64pcrel32 target-flags(x86-plt) @f1, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ %10:gr64 = COPY $rax
+ %1:gr64 = COPY %10
+ JMP_1 %bb.5
+
+ bb.2:
+ ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ CALL64pcrel32 target-flags(x86-plt) @f2, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ %9:gr64 = COPY $rax
+ %2:gr64 = COPY %9
+ JMP_1 %bb.5
+
+ bb.3:
+ ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ CALL64pcrel32 target-flags(x86-plt) @f3, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ %8:gr64 = COPY $rax
+ %3:gr64 = COPY %8
+ JMP_1 %bb.5
+
+ bb.4:
+ ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ CALL64pcrel32 target-flags(x86-plt) @f4, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ %7:gr64 = COPY $rax
+ %4:gr64 = COPY %7
+
+ bb.5:
+ successors: %bb.1, %bb.2, %bb.3, %bb.4
+
+ %5:gr64_nosp = PHI %0, %bb.0, %4, %bb.4, %3, %bb.3, %2, %bb.2, %1, %bb.1
+ JMP64m $noreg, 8, %5, @computed_goto.dispatch, $noreg
+
+...
+---
+name: jump_table
+tracksRegLiveness: true
+jumpTable:
+ kind: block-address
+ entries:
+ - id: 0
+ blocks: [ '%bb.2', '%bb.3', '%bb.4', '%bb.5', '%bb.6' ]
+body: |
+ ; CHECK-LABEL: name: jump_table
+ ; CHECK: bb.0:
+ ; CHECK-NEXT: successors: %bb.3(0x1999999a), %bb.4(0x1999999a), %bb.5(0x1999999a), %bb.6(0x1999999a), %bb.7(0x1999999a)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f0, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: [[COPY:%[0-9]+]]:gr64 = COPY $rax
+ ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gr64 = COPY [[COPY]]
+ ; CHECK-NEXT: [[DEC64r:%[0-9]+]]:gr64_nosp = DEC64r [[COPY1]], implicit-def dead $eflags
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gr64 = COPY [[COPY1]]
+ ; CHECK-NEXT: JMP64m $noreg, 8, [[DEC64r]], %jump-table.0, $noreg :: (load (s64) from jump-table)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.3:
+ ; CHECK-NEXT: successors: %bb.3(0x1999999a), %bb.4(0x1999999a), %bb.5(0x1999999a), %bb.6(0x1999999a), %bb.7(0x1999999a)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f1, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: [[COPY3:%[0-9]+]]:gr64 = COPY $rax
+ ; CHECK-NEXT: [[COPY4:%[0-9]+]]:gr64 = COPY [[COPY3]]
+ ; CHECK-NEXT: [[DEC64r1:%[0-9]+]]:gr64_nosp = DEC64r [[COPY4]], implicit-def dead $eflags
+ ; CHECK-NEXT: [[COPY5:%[0-9]+]]:gr64 = COPY [[COPY4]]
+ ; CHECK-NEXT: JMP64m $noreg, 8, [[DEC64r1]], %jump-table.0, $noreg :: (load (s64) from jump-table)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.4:
+ ; CHECK-NEXT: successors: %bb.3(0x1999999a), %bb.4(0x1999999a), %bb.5(0x1999999a), %bb.6(0x1999999a), %bb.7(0x1999999a)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f2, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: [[COPY6:%[0-9]+]]:gr64 = COPY $rax
+ ; CHECK-NEXT: [[COPY7:%[0-9]+]]:gr64 = COPY [[COPY6]]
+ ; CHECK-NEXT: [[DEC64r2:%[0-9]+]]:gr64_nosp = DEC64r [[COPY7]], implicit-def dead $eflags
+ ; CHECK-NEXT: [[COPY8:%[0-9]+]]:gr64 = COPY [[COPY7]]
+ ; CHECK-NEXT: JMP64m $noreg, 8, [[DEC64r2]], %jump-table.0, $noreg :: (load (s64) from jump-table)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.5:
+ ; CHECK-NEXT: successors: %bb.3(0x1999999a), %bb.4(0x1999999a), %bb.5(0x1999999a), %bb.6(0x1999999a), %bb.7(0x1999999a)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f3, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: [[COPY9:%[0-9]+]]:gr64 = COPY $rax
+ ; CHECK-NEXT: [[COPY10:%[0-9]+]]:gr64 = COPY [[COPY9]]
+ ; CHECK-NEXT: [[DEC64r3:%[0-9]+]]:gr64_nosp = DEC64r [[COPY10]], implicit-def dead $eflags
+ ; CHECK-NEXT: [[COPY11:%[0-9]+]]:gr64 = COPY [[COPY10]]
+ ; CHECK-NEXT: JMP64m $noreg, 8, [[DEC64r3]], %jump-table.0, $noreg :: (load (s64) from jump-table)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.6:
+ ; CHECK-NEXT: successors: %bb.3(0x1999999a), %bb.4(0x1999999a), %bb.5(0x1999999a), %bb.6(0x1999999a), %bb.7(0x1999999a)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f4, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: [[COPY12:%[0-9]+]]:gr64 = COPY $rax
+ ; CHECK-NEXT: [[COPY13:%[0-9]+]]:gr64 = COPY [[COPY12]]
+ ; CHECK-NEXT: [[DEC64r4:%[0-9]+]]:gr64_nosp = DEC64r [[COPY13]], implicit-def dead $eflags
+ ; CHECK-NEXT: [[COPY14:%[0-9]+]]:gr64 = COPY [[COPY13]]
+ ; CHECK-NEXT: JMP64m $noreg, 8, [[DEC64r4]], %jump-table.0, $noreg :: (load (s64) from jump-table)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.7:
+ ; CHECK-NEXT: successors: %bb.3(0x1999999a), %bb.4(0x1999999a), %bb.5(0x1999999a), %bb.6(0x1999999a), %bb.7(0x1999999a)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f5, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ ; CHECK-NEXT: [[COPY15:%[0-9]+]]:gr64 = COPY $rax
+ ; CHECK-NEXT: [[COPY16:%[0-9]+]]:gr64 = COPY [[COPY15]]
+ ; CHECK-NEXT: [[DEC64r5:%[0-9]+]]:gr64_nosp = DEC64r [[COPY16]], implicit-def dead $eflags
+ ; CHECK-NEXT: [[COPY17:%[0-9]+]]:gr64 = COPY [[COPY16]]
+ ; CHECK-NEXT: JMP64m $noreg, 8, [[DEC64r5]], %jump-table.0, $noreg :: (load (s64) from jump-table)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.8:
+ bb.0:
+ ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ CALL64pcrel32 target-flags(x86-plt) @f0, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ %7:gr64 = COPY $rax
+ %0:gr64 = COPY %7
+
+ bb.1:
+ %1:gr64 = PHI %0, %bb.0, %6, %bb.6, %5, %bb.5, %4, %bb.4, %3, %bb.3, %2, %bb.2
+ %8:gr64_nosp = DEC64r %1, implicit-def dead $eflags
+
+ bb.8:
+ successors: %bb.2(0x1999999a), %bb.3(0x1999999a), %bb.4(0x1999999a), %bb.5(0x1999999a), %bb.6(0x1999999a)
+
+ JMP64m $noreg, 8, %8, %jump-table.0, $noreg :: (load (s64) from jump-table)
+
+ bb.2:
+ ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ CALL64pcrel32 target-flags(x86-plt) @f1, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ %13:gr64 = COPY $rax
+ %2:gr64 = COPY %13
+ JMP_1 %bb.1
+
+ bb.3:
+ ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ CALL64pcrel32 target-flags(x86-plt) @f2, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ %12:gr64 = COPY $rax
+ %3:gr64 = COPY %12
+ JMP_1 %bb.1
+
+ bb.4:
+ ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ CALL64pcrel32 target-flags(x86-plt) @f3, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ %11:gr64 = COPY $rax
+ %4:gr64 = COPY %11
+ JMP_1 %bb.1
+
+ bb.5:
+ ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ CALL64pcrel32 target-flags(x86-plt) @f4, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ %10:gr64 = COPY $rax
+ %5:gr64 = COPY %10
+ JMP_1 %bb.1
+
+ bb.6:
+ ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ CALL64pcrel32 target-flags(x86-plt) @f5, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+ ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+ %9:gr64 = COPY $rax
+ %6:gr64 = COPY %9
+ JMP_1 %bb.1
+
+ bb.7:
+
+...
>From 58b5757a497810fba408ba777f87489b08261920 Mon Sep 17 00:00:00 2001
From: DianQK <dianqk at dianqk.net>
Date: Tue, 5 Nov 2024 21:15:09 +0800
Subject: [PATCH 2/2] [TailDuplicator] Only duplicate the blocks containing
computed gotos
---
llvm/include/llvm/CodeGen/MachineInstr.h | 8 ++-
llvm/lib/CodeGen/TailDuplicator.cpp | 22 ++++---
.../CodeGen/X86/tail-dup-computed-goto.mir | 64 +++++++++----------
3 files changed, 49 insertions(+), 45 deletions(-)
diff --git a/llvm/include/llvm/CodeGen/MachineInstr.h b/llvm/include/llvm/CodeGen/MachineInstr.h
index ead6bbe1d5f641..6d268628ec14b4 100644
--- a/llvm/include/llvm/CodeGen/MachineInstr.h
+++ b/llvm/include/llvm/CodeGen/MachineInstr.h
@@ -986,8 +986,12 @@ class MachineInstr
/// Return true if this is an indirect branch, such as a
/// branch through a register.
- bool isIndirectBranch(QueryType Type = AnyInBundle) const {
- return hasProperty(MCID::IndirectBranch, Type);
+ bool isIndirectBranch(QueryType Type = AnyInBundle,
+ bool IncludeJumpTable = true) const {
+ return hasProperty(MCID::IndirectBranch, Type) &&
+ (IncludeJumpTable || !llvm::any_of(operands(), [](const auto &Op) {
+ return Op.isJTI();
+ }));
}
/// Return true if this is a branch which may fall
diff --git a/llvm/lib/CodeGen/TailDuplicator.cpp b/llvm/lib/CodeGen/TailDuplicator.cpp
index 3f2e1511d403a0..988c58beac20f6 100644
--- a/llvm/lib/CodeGen/TailDuplicator.cpp
+++ b/llvm/lib/CodeGen/TailDuplicator.cpp
@@ -603,17 +603,19 @@ bool TailDuplicator::shouldTailDuplicate(bool IsSimple,
TailBB.canFallThrough())
return false;
- // If the target has hardware branch prediction that can handle indirect
- // branches, duplicating them can often make them predictable when there
- // are common paths through the code. The limit needs to be high enough
- // to allow undoing the effects of tail merging and other optimizations
- // that rearrange the predecessors of the indirect branch.
-
- bool HasIndirectbr = false;
+ // Only duplicate the blocks containing computed gotos. This basically
+ // unfactors computed gotos that were factored early on in the compilation
+ // process to speed up edge based data flow. If we do not unfactor them again,
+ // it can seriously pessimize code with many computed jumps in the source
+ // code, such as interpreters.
+ bool HasComputedGoto = false;
if (!TailBB.empty())
- HasIndirectbr = TailBB.back().isIndirectBranch();
+ HasComputedGoto = TailBB.back().isIndirectBranch(
+ /*Type=*/MachineInstr::AnyInBundle,
+ // Jump tables are not considered computed gotos.
+ /*IncludeJumpTable=*/false);
- if (HasIndirectbr && PreRegAlloc)
+ if (HasComputedGoto && PreRegAlloc)
MaxDuplicateCount = TailDupIndirectBranchSize;
// Check the instructions in the block to determine whether tail-duplication
@@ -685,7 +687,7 @@ bool TailDuplicator::shouldTailDuplicate(bool IsSimple,
}
}
- if (HasIndirectbr && PreRegAlloc)
+ if (HasComputedGoto && PreRegAlloc)
return true;
if (IsSimple)
diff --git a/llvm/test/CodeGen/X86/tail-dup-computed-goto.mir b/llvm/test/CodeGen/X86/tail-dup-computed-goto.mir
index d5214aecb62a1f..b25e17ca82d4dc 100644
--- a/llvm/test/CodeGen/X86/tail-dup-computed-goto.mir
+++ b/llvm/test/CodeGen/X86/tail-dup-computed-goto.mir
@@ -128,76 +128,74 @@ jumpTable:
body: |
; CHECK-LABEL: name: jump_table
; CHECK: bb.0:
- ; CHECK-NEXT: successors: %bb.3(0x1999999a), %bb.4(0x1999999a), %bb.5(0x1999999a), %bb.6(0x1999999a), %bb.7(0x1999999a)
+ ; CHECK-NEXT: successors: %bb.1(0x80000000)
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f0, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
; CHECK-NEXT: [[COPY:%[0-9]+]]:gr64 = COPY $rax
; CHECK-NEXT: [[COPY1:%[0-9]+]]:gr64 = COPY [[COPY]]
- ; CHECK-NEXT: [[DEC64r:%[0-9]+]]:gr64_nosp = DEC64r [[COPY1]], implicit-def dead $eflags
- ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gr64 = COPY [[COPY1]]
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.1:
+ ; CHECK-NEXT: successors: %bb.2(0x80000000)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[PHI:%[0-9]+]]:gr64 = PHI [[COPY1]], %bb.0, %3, %bb.7, %4, %bb.6, %5, %bb.5, %6, %bb.4, %7, %bb.3
+ ; CHECK-NEXT: [[DEC64r:%[0-9]+]]:gr64_nosp = DEC64r [[PHI]], implicit-def dead $eflags
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.2:
+ ; CHECK-NEXT: successors: %bb.3(0x1999999a), %bb.4(0x1999999a), %bb.5(0x1999999a), %bb.6(0x1999999a), %bb.7(0x1999999a)
+ ; CHECK-NEXT: {{ $}}
; CHECK-NEXT: JMP64m $noreg, 8, [[DEC64r]], %jump-table.0, $noreg :: (load (s64) from jump-table)
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: bb.3:
- ; CHECK-NEXT: successors: %bb.3(0x1999999a), %bb.4(0x1999999a), %bb.5(0x1999999a), %bb.6(0x1999999a), %bb.7(0x1999999a)
+ ; CHECK-NEXT: successors: %bb.1(0x80000000)
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f1, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
- ; CHECK-NEXT: [[COPY3:%[0-9]+]]:gr64 = COPY $rax
- ; CHECK-NEXT: [[COPY4:%[0-9]+]]:gr64 = COPY [[COPY3]]
- ; CHECK-NEXT: [[DEC64r1:%[0-9]+]]:gr64_nosp = DEC64r [[COPY4]], implicit-def dead $eflags
- ; CHECK-NEXT: [[COPY5:%[0-9]+]]:gr64 = COPY [[COPY4]]
- ; CHECK-NEXT: JMP64m $noreg, 8, [[DEC64r1]], %jump-table.0, $noreg :: (load (s64) from jump-table)
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gr64 = COPY $rax
+ ; CHECK-NEXT: [[COPY3:%[0-9]+]]:gr64 = COPY [[COPY2]]
+ ; CHECK-NEXT: JMP_1 %bb.1
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: bb.4:
- ; CHECK-NEXT: successors: %bb.3(0x1999999a), %bb.4(0x1999999a), %bb.5(0x1999999a), %bb.6(0x1999999a), %bb.7(0x1999999a)
+ ; CHECK-NEXT: successors: %bb.1(0x80000000)
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f2, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
- ; CHECK-NEXT: [[COPY6:%[0-9]+]]:gr64 = COPY $rax
- ; CHECK-NEXT: [[COPY7:%[0-9]+]]:gr64 = COPY [[COPY6]]
- ; CHECK-NEXT: [[DEC64r2:%[0-9]+]]:gr64_nosp = DEC64r [[COPY7]], implicit-def dead $eflags
- ; CHECK-NEXT: [[COPY8:%[0-9]+]]:gr64 = COPY [[COPY7]]
- ; CHECK-NEXT: JMP64m $noreg, 8, [[DEC64r2]], %jump-table.0, $noreg :: (load (s64) from jump-table)
+ ; CHECK-NEXT: [[COPY4:%[0-9]+]]:gr64 = COPY $rax
+ ; CHECK-NEXT: [[COPY5:%[0-9]+]]:gr64 = COPY [[COPY4]]
+ ; CHECK-NEXT: JMP_1 %bb.1
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: bb.5:
- ; CHECK-NEXT: successors: %bb.3(0x1999999a), %bb.4(0x1999999a), %bb.5(0x1999999a), %bb.6(0x1999999a), %bb.7(0x1999999a)
+ ; CHECK-NEXT: successors: %bb.1(0x80000000)
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f3, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
- ; CHECK-NEXT: [[COPY9:%[0-9]+]]:gr64 = COPY $rax
- ; CHECK-NEXT: [[COPY10:%[0-9]+]]:gr64 = COPY [[COPY9]]
- ; CHECK-NEXT: [[DEC64r3:%[0-9]+]]:gr64_nosp = DEC64r [[COPY10]], implicit-def dead $eflags
- ; CHECK-NEXT: [[COPY11:%[0-9]+]]:gr64 = COPY [[COPY10]]
- ; CHECK-NEXT: JMP64m $noreg, 8, [[DEC64r3]], %jump-table.0, $noreg :: (load (s64) from jump-table)
+ ; CHECK-NEXT: [[COPY6:%[0-9]+]]:gr64 = COPY $rax
+ ; CHECK-NEXT: [[COPY7:%[0-9]+]]:gr64 = COPY [[COPY6]]
+ ; CHECK-NEXT: JMP_1 %bb.1
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: bb.6:
- ; CHECK-NEXT: successors: %bb.3(0x1999999a), %bb.4(0x1999999a), %bb.5(0x1999999a), %bb.6(0x1999999a), %bb.7(0x1999999a)
+ ; CHECK-NEXT: successors: %bb.1(0x80000000)
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f4, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
- ; CHECK-NEXT: [[COPY12:%[0-9]+]]:gr64 = COPY $rax
- ; CHECK-NEXT: [[COPY13:%[0-9]+]]:gr64 = COPY [[COPY12]]
- ; CHECK-NEXT: [[DEC64r4:%[0-9]+]]:gr64_nosp = DEC64r [[COPY13]], implicit-def dead $eflags
- ; CHECK-NEXT: [[COPY14:%[0-9]+]]:gr64 = COPY [[COPY13]]
- ; CHECK-NEXT: JMP64m $noreg, 8, [[DEC64r4]], %jump-table.0, $noreg :: (load (s64) from jump-table)
+ ; CHECK-NEXT: [[COPY8:%[0-9]+]]:gr64 = COPY $rax
+ ; CHECK-NEXT: [[COPY9:%[0-9]+]]:gr64 = COPY [[COPY8]]
+ ; CHECK-NEXT: JMP_1 %bb.1
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: bb.7:
- ; CHECK-NEXT: successors: %bb.3(0x1999999a), %bb.4(0x1999999a), %bb.5(0x1999999a), %bb.6(0x1999999a), %bb.7(0x1999999a)
+ ; CHECK-NEXT: successors: %bb.1(0x80000000)
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f5, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
- ; CHECK-NEXT: [[COPY15:%[0-9]+]]:gr64 = COPY $rax
- ; CHECK-NEXT: [[COPY16:%[0-9]+]]:gr64 = COPY [[COPY15]]
- ; CHECK-NEXT: [[DEC64r5:%[0-9]+]]:gr64_nosp = DEC64r [[COPY16]], implicit-def dead $eflags
- ; CHECK-NEXT: [[COPY17:%[0-9]+]]:gr64 = COPY [[COPY16]]
- ; CHECK-NEXT: JMP64m $noreg, 8, [[DEC64r5]], %jump-table.0, $noreg :: (load (s64) from jump-table)
+ ; CHECK-NEXT: [[COPY10:%[0-9]+]]:gr64 = COPY $rax
+ ; CHECK-NEXT: [[COPY11:%[0-9]+]]:gr64 = COPY [[COPY10]]
+ ; CHECK-NEXT: JMP_1 %bb.1
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: bb.8:
bb.0:
More information about the llvm-commits
mailing list