[RISCV] Don't outline lo operand #80920
Open
+295
−3
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…h lo. Copied from llvm/test/CodeGen/RISCV/machineoutliner-pcrel-lo.mir.
|
@llvm/pr-subscribers-backend-risc-v Author: Jim Lin (tclin914) ChangesIf lo and hi operand put in separate section, it has the same issue like pcrel-hi and pcrel-lo. Refer from: https://reviews.llvm.org/D132528. Full diff: https://github.com/llvm/llvm-project/pull/80920.diff 2 Files Affected:
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
index 89eb71d917428..1864fbb63a147 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -2488,9 +2488,10 @@ RISCVInstrInfo::getOutliningTypeImpl(MachineBasicBlock::iterator &MBBI,
// Make sure the operands don't reference something unsafe.
for (const auto &MO : MI.operands()) {
- // pcrel-hi and pcrel-lo can't put in separate sections, filter that out
- // if any possible.
- if (MO.getTargetFlags() == RISCVII::MO_PCREL_LO &&
+ // pcrel-hi/hi and pcrel-lo/lo can't put in separate sections, filter that
+ // out if any possible.
+ if ((MO.getTargetFlags() == RISCVII::MO_PCREL_LO ||
+ MO.getTargetFlags() == RISCVII::MO_LO) &&
(MI.getMF()->getTarget().getFunctionSections() || F.hasComdat() ||
F.hasSection()))
return outliner::InstrType::Illegal;
diff --git a/llvm/test/CodeGen/RISCV/machineoutliner-lo.mir b/llvm/test/CodeGen/RISCV/machineoutliner-lo.mir
new file mode 100644
index 0000000000000..5b35a318e0c13
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/machineoutliner-lo.mir
@@ -0,0 +1,291 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -march=riscv32 -x mir -run-pass=machine-outliner -simplify-mir -verify-machineinstrs < %s \
+# RUN: | FileCheck %s
+# RUN: llc -march=riscv64 -x mir -run-pass=machine-outliner -simplify-mir -verify-machineinstrs < %s \
+# RUN: | FileCheck %s
+# RUN: llc -march=riscv32 -x mir -run-pass=machine-outliner -simplify-mir --function-sections -verify-machineinstrs < %s \
+# RUN: | FileCheck -check-prefix=CHECK-FS %s
+# RUN: llc -march=riscv64 -x mir -run-pass=machine-outliner -simplify-mir --function-sections -verify-machineinstrs < %s \
+# RUN: | FileCheck -check-prefix=CHECK-FS %s
+
+--- |
+ ; Cannot outline instructions with lo operands if function section
+ ; enabled.
+ @bar = dso_local local_unnamed_addr global i32 0, align 4
+ define i32 @foo(i32 %a, i32 %b) { ret i32 0 }
+
+ $foo2 = comdat any
+ define i32 @foo2(i32 %a, i32 %b) comdat { ret i32 0 }
+
+ define i32 @foo3(i32 %a, i32 %b) section ".abc" { ret i32 0 }
+...
+---
+name: foo
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: name: foo
+ ; CHECK: bb.0:
+ ; CHECK-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: $x5 = PseudoCALLReg target-flags(riscv-call) @OUTLINED_FUNCTION_0, implicit-def $x5, implicit-def $x10, implicit-def $x11, implicit-def $x12, implicit $x10, implicit $x11, implicit $x13
+ ; CHECK-NEXT: PseudoBR %bb.3
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.1:
+ ; CHECK-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: $x5 = PseudoCALLReg target-flags(riscv-call) @OUTLINED_FUNCTION_0, implicit-def $x5, implicit-def $x10, implicit-def $x11, implicit-def $x12, implicit $x10, implicit $x11, implicit $x13
+ ; CHECK-NEXT: PseudoBR %bb.3
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.2:
+ ; CHECK-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: $x5 = PseudoCALLReg target-flags(riscv-call) @OUTLINED_FUNCTION_0, implicit-def $x5, implicit-def $x10, implicit-def $x11, implicit-def $x12, implicit $x10, implicit $x11, implicit $x13
+ ; CHECK-NEXT: PseudoBR %bb.3
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.3:
+ ; CHECK-NEXT: PseudoRET
+ ; CHECK-FS-LABEL: name: foo
+ ; CHECK-FS: bb.0:
+ ; CHECK-FS-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: $x5 = PseudoCALLReg target-flags(riscv-call) @OUTLINED_FUNCTION_0, implicit-def $x5, implicit-def $x10, implicit-def $x11, implicit-def $x12, implicit $x10, implicit $x11
+ ; CHECK-FS-NEXT: $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ ; CHECK-FS-NEXT: PseudoBR %bb.3
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: bb.1:
+ ; CHECK-FS-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: $x5 = PseudoCALLReg target-flags(riscv-call) @OUTLINED_FUNCTION_0, implicit-def $x5, implicit-def $x10, implicit-def $x11, implicit-def $x12, implicit $x10, implicit $x11
+ ; CHECK-FS-NEXT: $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ ; CHECK-FS-NEXT: PseudoBR %bb.3
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: bb.2:
+ ; CHECK-FS-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: $x5 = PseudoCALLReg target-flags(riscv-call) @OUTLINED_FUNCTION_0, implicit-def $x5, implicit-def $x10, implicit-def $x11, implicit-def $x12, implicit $x10, implicit $x11
+ ; CHECK-FS-NEXT: $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ ; CHECK-FS-NEXT: PseudoBR %bb.3
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: bb.3:
+ ; CHECK-FS-NEXT: PseudoRET
+ bb.0:
+ liveins: $x10, $x11, $x13
+
+ $x11 = ORI $x11, 1023
+ $x12 = ADDI $x10, 17
+ $x11 = AND $x12, $x11
+ $x10 = SUB $x10, $x11
+ $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ PseudoBR %bb.3
+
+ bb.1:
+ liveins: $x10, $x11, $x13
+
+ $x11 = ORI $x11, 1023
+ $x12 = ADDI $x10, 17
+ $x11 = AND $x12, $x11
+ $x10 = SUB $x10, $x11
+ $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ PseudoBR %bb.3
+
+ bb.2:
+ liveins: $x10, $x11, $x13
+
+ $x11 = ORI $x11, 1023
+ $x12 = ADDI $x10, 17
+ $x11 = AND $x12, $x11
+ $x10 = SUB $x10, $x11
+ $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ PseudoBR %bb.3
+
+ bb.3:
+ PseudoRET
+...
+---
+name: foo2
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: name: foo2
+ ; CHECK: bb.0:
+ ; CHECK-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: $x5 = PseudoCALLReg target-flags(riscv-call) @OUTLINED_FUNCTION_1, implicit-def $x5, implicit-def $x10, implicit-def $x11, implicit-def $x12, implicit $x10, implicit $x11
+ ; CHECK-NEXT: $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ ; CHECK-NEXT: PseudoBR %bb.3
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.1:
+ ; CHECK-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: $x5 = PseudoCALLReg target-flags(riscv-call) @OUTLINED_FUNCTION_1, implicit-def $x5, implicit-def $x10, implicit-def $x11, implicit-def $x12, implicit $x10, implicit $x11
+ ; CHECK-NEXT: $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ ; CHECK-NEXT: PseudoBR %bb.3
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.2:
+ ; CHECK-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: $x5 = PseudoCALLReg target-flags(riscv-call) @OUTLINED_FUNCTION_1, implicit-def $x5, implicit-def $x10, implicit-def $x11, implicit-def $x12, implicit $x10, implicit $x11
+ ; CHECK-NEXT: $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ ; CHECK-NEXT: PseudoBR %bb.3
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.3:
+ ; CHECK-NEXT: PseudoRET
+ ; CHECK-FS-LABEL: name: foo2
+ ; CHECK-FS: bb.0:
+ ; CHECK-FS-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: $x5 = PseudoCALLReg target-flags(riscv-call) @OUTLINED_FUNCTION_0, implicit-def $x5, implicit-def $x10, implicit-def $x11, implicit-def $x12, implicit $x10, implicit $x11
+ ; CHECK-FS-NEXT: $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ ; CHECK-FS-NEXT: PseudoBR %bb.3
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: bb.1:
+ ; CHECK-FS-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: $x5 = PseudoCALLReg target-flags(riscv-call) @OUTLINED_FUNCTION_0, implicit-def $x5, implicit-def $x10, implicit-def $x11, implicit-def $x12, implicit $x10, implicit $x11
+ ; CHECK-FS-NEXT: $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ ; CHECK-FS-NEXT: PseudoBR %bb.3
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: bb.2:
+ ; CHECK-FS-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: $x5 = PseudoCALLReg target-flags(riscv-call) @OUTLINED_FUNCTION_0, implicit-def $x5, implicit-def $x10, implicit-def $x11, implicit-def $x12, implicit $x10, implicit $x11
+ ; CHECK-FS-NEXT: $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ ; CHECK-FS-NEXT: PseudoBR %bb.3
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: bb.3:
+ ; CHECK-FS-NEXT: PseudoRET
+ bb.0:
+ liveins: $x10, $x11, $x13
+
+ $x11 = ORI $x11, 1023
+ $x12 = ADDI $x10, 17
+ $x11 = AND $x12, $x11
+ $x10 = SUB $x10, $x11
+ $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ PseudoBR %bb.3
+
+ bb.1:
+ liveins: $x10, $x11, $x13
+
+ $x11 = ORI $x11, 1023
+ $x12 = ADDI $x10, 17
+ $x11 = AND $x12, $x11
+ $x10 = SUB $x10, $x11
+ $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ PseudoBR %bb.3
+
+ bb.2:
+ liveins: $x10, $x11, $x13
+
+ $x11 = ORI $x11, 1023
+ $x12 = ADDI $x10, 17
+ $x11 = AND $x12, $x11
+ $x10 = SUB $x10, $x11
+ $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ PseudoBR %bb.3
+
+ bb.3:
+ PseudoRET
+...
+---
+name: foo3
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: name: foo3
+ ; CHECK: bb.0:
+ ; CHECK-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: $x11 = ORI $x11, 1023
+ ; CHECK-NEXT: $x12 = ADDI $x10, 17
+ ; CHECK-NEXT: $x11 = AND $x12, $x11
+ ; CHECK-NEXT: $x10 = SUB $x10, $x11
+ ; CHECK-NEXT: $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ ; CHECK-NEXT: PseudoBR %bb.3
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.1:
+ ; CHECK-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: $x11 = ORI $x11, 1023
+ ; CHECK-NEXT: $x12 = ADDI $x10, 17
+ ; CHECK-NEXT: $x11 = AND $x12, $x11
+ ; CHECK-NEXT: $x10 = SUB $x10, $x11
+ ; CHECK-NEXT: $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ ; CHECK-NEXT: PseudoBR %bb.3
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.2:
+ ; CHECK-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: $x11 = ORI $x11, 1023
+ ; CHECK-NEXT: $x12 = ADDI $x10, 17
+ ; CHECK-NEXT: $x11 = AND $x12, $x11
+ ; CHECK-NEXT: $x10 = SUB $x10, $x11
+ ; CHECK-NEXT: $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ ; CHECK-NEXT: PseudoBR %bb.3
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.3:
+ ; CHECK-NEXT: PseudoRET
+ ; CHECK-FS-LABEL: name: foo3
+ ; CHECK-FS: bb.0:
+ ; CHECK-FS-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: $x11 = ORI $x11, 1023
+ ; CHECK-FS-NEXT: $x12 = ADDI $x10, 17
+ ; CHECK-FS-NEXT: $x11 = AND $x12, $x11
+ ; CHECK-FS-NEXT: $x10 = SUB $x10, $x11
+ ; CHECK-FS-NEXT: $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ ; CHECK-FS-NEXT: PseudoBR %bb.3
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: bb.1:
+ ; CHECK-FS-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: $x11 = ORI $x11, 1023
+ ; CHECK-FS-NEXT: $x12 = ADDI $x10, 17
+ ; CHECK-FS-NEXT: $x11 = AND $x12, $x11
+ ; CHECK-FS-NEXT: $x10 = SUB $x10, $x11
+ ; CHECK-FS-NEXT: $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ ; CHECK-FS-NEXT: PseudoBR %bb.3
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: bb.2:
+ ; CHECK-FS-NEXT: liveins: $x10, $x11, $x13
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: $x11 = ORI $x11, 1023
+ ; CHECK-FS-NEXT: $x12 = ADDI $x10, 17
+ ; CHECK-FS-NEXT: $x11 = AND $x12, $x11
+ ; CHECK-FS-NEXT: $x10 = SUB $x10, $x11
+ ; CHECK-FS-NEXT: $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ ; CHECK-FS-NEXT: PseudoBR %bb.3
+ ; CHECK-FS-NEXT: {{ $}}
+ ; CHECK-FS-NEXT: bb.3:
+ ; CHECK-FS-NEXT: PseudoRET
+ bb.0:
+ liveins: $x10, $x11, $x13
+
+ $x11 = ORI $x11, 1023
+ $x12 = ADDI $x10, 17
+ $x11 = AND $x12, $x11
+ $x10 = SUB $x10, $x11
+ $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ PseudoBR %bb.3
+
+ bb.1:
+ liveins: $x10, $x11, $x13
+
+ $x11 = ORI $x11, 1023
+ $x12 = ADDI $x10, 17
+ $x11 = AND $x12, $x11
+ $x10 = SUB $x10, $x11
+ $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ PseudoBR %bb.3
+
+ bb.2:
+ liveins: $x10, $x11, $x13
+
+ $x11 = ORI $x11, 1023
+ $x12 = ADDI $x10, 17
+ $x11 = AND $x12, $x11
+ $x10 = SUB $x10, $x11
+ $x11 = LW killed renamable $x13, target-flags(riscv-lo) <mcsymbol .Lhi1> :: (dereferenceable load (s32) from @bar)
+ PseudoBR %bb.3
+
+ bb.3:
+ PseudoRET
+...
|
I don't get it. IIUC, lo/hi will be resolved to absolute address of symbol, the position of lo/hi doesn't matter. |
I agree this isn't an issue. The lo/hi relocations both point to the same symbol. pcrel was an issue because the pcrel-lo relocation contains a symbol that points to the address of an auipc instruction and the pcrel-hi relocation points to the symbol we want the address of. So there are 2 symbols involved. |
Yes, this issue isn't all the same with pcrel, I met incorrect codegen if lo/hi are not in the same section and compiled with lto option. I will look more deeply. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
If lo and hi operand put in separate section, it has the same issue like pcrel-hi and pcrel-lo.
Refer from: https://reviews.llvm.org/D132528.