[llvm] [ARM] Fix -mno-omit-leaf-frame-pointer flag doesn't works on 32-bit ARM (PR #109628)

via llvm-commits llvm-commits at lists.llvm.org
Mon Sep 23 01:02:05 PDT 2024 

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-backend-arm

Author: gxlayer (guoxin049)

<details>
<summary>Changes</summary>

The -mno-omit-leaf-frame-pointer flag works on 32-bit ARM architectures and addresses the bug reported in #<!-- -->108019

---
Full diff: https://github.com/llvm/llvm-project/pull/109628.diff


20 Files Affected:

- (modified) llvm/include/llvm/Target/TargetOptions.h (+4) 
- (modified) llvm/lib/CodeGen/TargetOptionsImpl.cpp (+13) 
- (modified) llvm/lib/Target/ARM/ARMFrameLowering.cpp (+2-1) 
- (modified) llvm/test/CodeGen/ARM/arm-shrink-wrapping.ll (+1-1) 
- (modified) llvm/test/CodeGen/ARM/call-tc.ll (+2-2) 
- (modified) llvm/test/CodeGen/ARM/debug-frame.ll (+1-1) 
- (modified) llvm/test/CodeGen/ARM/ehabi.ll (+1-1) 
- (modified) llvm/test/CodeGen/ARM/frame-chain.ll (+7-4) 
- (modified) llvm/test/CodeGen/ARM/ifcvt5.ll (+1-1) 
- (modified) llvm/test/CodeGen/ARM/ldrd.ll (+2-2) 
- (modified) llvm/test/CodeGen/ARM/stack-size-section.ll (+1-1) 
- (modified) llvm/test/CodeGen/ARM/v7k-abi-align.ll (+2-2) 
- (modified) llvm/test/CodeGen/Thumb/frame-chain.ll (+10-6) 
- (modified) llvm/test/CodeGen/Thumb2/frame-pointer.ll (+1-1) 
- (modified) llvm/test/CodeGen/Thumb2/frameless.ll (+2-2) 
- (modified) llvm/test/CodeGen/Thumb2/frameless2.ll (+1-1) 
- (modified) llvm/test/CodeGen/Thumb2/machine-licm.ll (+2-2) 
- (modified) llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/arm_generated_funcs.ll (+1-1) 
- (modified) llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/arm_generated_funcs.ll.generated.expected (+1-1) 
- (modified) llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/arm_generated_funcs.ll.nogenerated.expected (+1-1) 


``````````diff
diff --git a/llvm/include/llvm/Target/TargetOptions.h b/llvm/include/llvm/Target/TargetOptions.h
index 94e0fa2404d6fc..37cff4ef57541e 100644
--- a/llvm/include/llvm/Target/TargetOptions.h
+++ b/llvm/include/llvm/Target/TargetOptions.h
@@ -161,6 +161,10 @@ namespace llvm {
     /// DisableFramePointerElim - This returns true if frame pointer elimination
     /// optimization should be disabled for the given machine function.
     bool DisableFramePointerElim(const MachineFunction &MF) const;
+    
+    /// EnableLeafFramePointerElim - This returns true if leaf frame pointer elimination
+    /// optimization should be disabled for the given machine function.
+    bool EnableLeafFramePointerElim(const MachineFunction &MF) const;
 
     /// FramePointerIsReserved - This returns true if the frame pointer must
     /// always either point to a new frame record or be un-modified in the given
diff --git a/llvm/lib/CodeGen/TargetOptionsImpl.cpp b/llvm/lib/CodeGen/TargetOptionsImpl.cpp
index 5bf1d265092f6f..3a44de6de51ff7 100644
--- a/llvm/lib/CodeGen/TargetOptionsImpl.cpp
+++ b/llvm/lib/CodeGen/TargetOptionsImpl.cpp
@@ -40,6 +40,19 @@ bool TargetOptions::DisableFramePointerElim(const MachineFunction &MF) const {
   llvm_unreachable("unknown frame pointer flag");
 }
 
+/// EnableLeafFramePointerElim - This returns true if leaf frame pointer elimination
+/// optimization should be disabled for the given machine function.
+bool TargetOptions::EnableLeafFramePointerElim(const MachineFunction &MF) const {
+  const Function &F = MF.getFunction();
+
+  if (!F.hasFnAttribute("frame-pointer"))
+    return false;
+  StringRef FP = F.getFnAttribute("frame-pointer").getValueAsString();
+  if (FP == "all")
+    return true;
+  return false;
+}
+
 bool TargetOptions::FramePointerIsReserved(const MachineFunction &MF) const {
   // Check to see if the target want to forcibly keep frame pointer.
   if (MF.getSubtarget().getFrameLowering()->keepFramePointer(MF))
diff --git a/llvm/lib/Target/ARM/ARMFrameLowering.cpp b/llvm/lib/Target/ARM/ARMFrameLowering.cpp
index 40354f99559896..473490badb7d12 100644
--- a/llvm/lib/Target/ARM/ARMFrameLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMFrameLowering.cpp
@@ -2271,6 +2271,7 @@ void ARMFrameLowering::determineCalleeSaves(MachineFunction &MF,
   // is spilled in the order specified by getCalleeSavedRegs() to make it easier
   // to combine multiple loads / stores.
   bool CanEliminateFrame = !(requiresAAPCSFrameRecord(MF) && hasFP(MF));
+  bool CanEliminateLeafFrame = !MF.getTarget().Options.EnableLeafFramePointerElim(MF);
   bool CS1Spilled = false;
   bool LRSpilled = false;
   unsigned NumGPRSpills = 0;
@@ -2513,7 +2514,7 @@ void ARMFrameLowering::determineCalleeSaves(MachineFunction &MF,
                     << "; EstimatedFPStack: " << MaxFixedOffset - MaxFPOffset
                     << "; BigFrameOffsets: " << BigFrameOffsets << "\n");
   if (BigFrameOffsets ||
-      !CanEliminateFrame || RegInfo->cannotEliminateFrame(MF)) {
+      !CanEliminateFrame || RegInfo->cannotEliminateFrame(MF) || !CanEliminateLeafFrame) {
     AFI->setHasStackFrame(true);
 
     if (HasFP) {
diff --git a/llvm/test/CodeGen/ARM/arm-shrink-wrapping.ll b/llvm/test/CodeGen/ARM/arm-shrink-wrapping.ll
index aa79e4156dac11..b6adc995091cea 100644
--- a/llvm/test/CodeGen/ARM/arm-shrink-wrapping.ll
+++ b/llvm/test/CodeGen/ARM/arm-shrink-wrapping.ll
@@ -1732,7 +1732,7 @@ if.end:
 ; Another infinite loop test this time with two nested infinite loop.
 ; infiniteloop3
 ; bx lr
-define void @infiniteloop3() "frame-pointer"="all" {
+define void @infiniteloop3() "frame-pointer"="none" {
 ; ARM-LABEL: infiniteloop3:
 ; ARM:       @ %bb.0: @ %entry
 ; ARM-NEXT:    mov r0, #0
diff --git a/llvm/test/CodeGen/ARM/call-tc.ll b/llvm/test/CodeGen/ARM/call-tc.ll
index 18d83bdc03e22f..9c70bac0322fe1 100644
--- a/llvm/test/CodeGen/ARM/call-tc.ll
+++ b/llvm/test/CodeGen/ARM/call-tc.ll
@@ -17,7 +17,7 @@ define void @t1() "frame-pointer"="all" {
         ret void
 }
 
-define void @t2() "frame-pointer"="all" {
+define void @t2() "frame-pointer"="none" {
 ; CHECKV6-LABEL: t2:
 ; CHECKV6: bx r0
 ; CHECKT2D-LABEL: t2:
@@ -102,7 +102,7 @@ bb:
 
 ; Make sure codegenprep is duplicating ret instructions to enable tail calls.
 ; rdar://11140249
-define i32 @t8(i32 %x) nounwind ssp "frame-pointer"="all" {
+define i32 @t8(i32 %x) nounwind ssp "frame-pointer"="none" {
 entry:
 ; CHECKT2D-LABEL: t8:
 ; CHECKT2D-NOT: push
diff --git a/llvm/test/CodeGen/ARM/debug-frame.ll b/llvm/test/CodeGen/ARM/debug-frame.ll
index faeafdf45dc392..72e7cfcab487a7 100644
--- a/llvm/test/CodeGen/ARM/debug-frame.ll
+++ b/llvm/test/CodeGen/ARM/debug-frame.ll
@@ -526,7 +526,7 @@ entry:
 ; Test 4
 ;-------------------------------------------------------------------------------
 
-define void @test4() nounwind {
+define void @test4() nounwind "frame-pointer"="none" {
 entry:
   ret void
 }
diff --git a/llvm/test/CodeGen/ARM/ehabi.ll b/llvm/test/CodeGen/ARM/ehabi.ll
index fea497076030f1..d1a4e9a6bccad0 100644
--- a/llvm/test/CodeGen/ARM/ehabi.ll
+++ b/llvm/test/CodeGen/ARM/ehabi.ll
@@ -575,7 +575,7 @@ entry:
 ; Test 4
 ;-------------------------------------------------------------------------------
 
-define void @test4() nounwind {
+define void @test4() nounwind "frame-pointer"="none" {
 entry:
   ret void
 }
diff --git a/llvm/test/CodeGen/ARM/frame-chain.ll b/llvm/test/CodeGen/ARM/frame-chain.ll
index e37213e4aaf8b8..7b722cd5fcef24 100644
--- a/llvm/test/CodeGen/ARM/frame-chain.ll
+++ b/llvm/test/CodeGen/ARM/frame-chain.ll
@@ -10,11 +10,14 @@
 define dso_local noundef i32 @leaf(i32 noundef %0) {
 ; LEAF-FP-LABEL: leaf:
 ; LEAF-FP:       @ %bb.0:
-; LEAF-FP-NEXT:    .pad #4
-; LEAF-FP-NEXT:    sub sp, sp, #4
-; LEAF-FP-NEXT:    str r0, [sp]
+; LEAF-FP-NEXT:    .save {r11, lr}
+; LEAF-FP-NEXT:    push {r11, lr}
+; LEAF-FP-NEXT:    .setfp r11, sp
+; LEAF-FP-NEXT:    mov r11, sp
+; LEAF-FP-NEXT:    push {r0}
 ; LEAF-FP-NEXT:    add r0, r0, #4
-; LEAF-FP-NEXT:    add sp, sp, #4
+; LEAF-FP-NEXT:    mov sp, r11
+; LEAF-FP-NEXT:    pop {r11, lr}
 ; LEAF-FP-NEXT:    mov pc, lr
 ;
 ; LEAF-FP-AAPCS-LABEL: leaf:
diff --git a/llvm/test/CodeGen/ARM/ifcvt5.ll b/llvm/test/CodeGen/ARM/ifcvt5.ll
index dc9a3400b691ac..30a92eb34989a6 100644
--- a/llvm/test/CodeGen/ARM/ifcvt5.ll
+++ b/llvm/test/CodeGen/ARM/ifcvt5.ll
@@ -5,7 +5,7 @@
 
 @x = external global ptr		; <ptr> [#uses=1]
 
-define void @foo(i32 %a) "frame-pointer"="all" {
+define void @foo(i32 %a) "frame-pointer"="none" {
 ; A8-LABEL: foo:
 ; A8:       @ %bb.0: @ %entry
 ; A8-NEXT:    movw r1, :lower16:(L_x$non_lazy_ptr-(LPC0_0+8))
diff --git a/llvm/test/CodeGen/ARM/ldrd.ll b/llvm/test/CodeGen/ARM/ldrd.ll
index cf5c2dfe5ef60b..3cf10f0e64b4d1 100644
--- a/llvm/test/CodeGen/ARM/ldrd.ll
+++ b/llvm/test/CodeGen/ARM/ldrd.ll
@@ -168,7 +168,7 @@ define void @ldrd_postupdate_inc(ptr %p0) "frame-pointer"="all" {
 ; NORMAL: strd r1, r2, [r0], #-8
 ; CONSERVATIVE-NOT: strd
 ; CHECK: bx lr
-define ptr @strd_postupdate_dec(ptr %p0, i32 %v0, i32 %v1) "frame-pointer"="all" {
+define ptr @strd_postupdate_dec(ptr %p0, i32 %v0, i32 %v1) "frame-pointer"="none" {
   %p0.1 = getelementptr i32, ptr %p0, i32 1
   store i32 %v0, ptr %p0
   store i32 %v1, ptr %p0.1
@@ -180,7 +180,7 @@ define ptr @strd_postupdate_dec(ptr %p0, i32 %v0, i32 %v1) "frame-pointer"="all"
 ; NORMAL: strd r1, r2, [r0], #8
 ; CONSERVATIVE-NOT: strd
 ; CHECK: bx lr
-define ptr @strd_postupdate_inc(ptr %p0, i32 %v0, i32 %v1) "frame-pointer"="all" {
+define ptr @strd_postupdate_inc(ptr %p0, i32 %v0, i32 %v1) "frame-pointer"="none" {
   %p0.1 = getelementptr i32, ptr %p0, i32 1
   store i32 %v0, ptr %p0
   store i32 %v1, ptr %p0.1
diff --git a/llvm/test/CodeGen/ARM/stack-size-section.ll b/llvm/test/CodeGen/ARM/stack-size-section.ll
index fb23e358d856ee..8272389719a691 100644
--- a/llvm/test/CodeGen/ARM/stack-size-section.ll
+++ b/llvm/test/CodeGen/ARM/stack-size-section.ll
@@ -29,4 +29,4 @@ define void @dynalloc(i32 %N) #0 {
   ret void
 }
 
-attributes #0 = { "frame-pointer"="all" }
+attributes #0 = { "frame-pointer"="none" }
diff --git a/llvm/test/CodeGen/ARM/v7k-abi-align.ll b/llvm/test/CodeGen/ARM/v7k-abi-align.ll
index 20c7aea5dcbe6b..b27c4354f432a1 100644
--- a/llvm/test/CodeGen/ARM/v7k-abi-align.ll
+++ b/llvm/test/CodeGen/ARM/v7k-abi-align.ll
@@ -117,7 +117,7 @@ define void @test_dpr_unwind_align_no_dprs() "frame-pointer"="all" {
 
 ; 128-bit vectors should use 128-bit (i.e. correctly aligned) slots on
 ; the stack.
-define <4 x float> @test_v128_stack_pass([8 x double], float, <4 x float> %in) "frame-pointer"="all" {
+define <4 x float> @test_v128_stack_pass([8 x double], float, <4 x float> %in) "frame-pointer"="none" {
 ; CHECK-LABEL: test_v128_stack_pass:
 ; CHECK: add r[[ADDR:[0-9]+]], sp, #16
 ; CHECK: vld1.64 {d0, d1}, [r[[ADDR]]:128]
@@ -140,7 +140,7 @@ define void @test_v128_stack_pass_varargs(<4 x float> %in) "frame-pointer"="all"
 
 ; To be compatible with AAPCS's va_start model (store r0-r3 at incoming SP, give
 ; a single pointer), 64-bit quantities must be pass
-define i64 @test_64bit_gpr_align(i32, i64 %r2_r3, i32 %sp) "frame-pointer"="all" {
+define i64 @test_64bit_gpr_align(i32, i64 %r2_r3, i32 %sp) "frame-pointer"="none" {
 ; CHECK-LABEL: test_64bit_gpr_align:
 ; CHECK: ldr [[RHS:r[0-9]+]], [sp]
 ; CHECK: adds r0, [[RHS]], r2
diff --git a/llvm/test/CodeGen/Thumb/frame-chain.ll b/llvm/test/CodeGen/Thumb/frame-chain.ll
index eb62ce09caf1be..e68fc626be9819 100644
--- a/llvm/test/CodeGen/Thumb/frame-chain.ll
+++ b/llvm/test/CodeGen/Thumb/frame-chain.ll
@@ -8,12 +8,16 @@
 define dso_local noundef i32 @leaf(i32 noundef %0) {
 ; LEAF-FP-LABEL: leaf:
 ; LEAF-FP:       @ %bb.0:
-; LEAF-FP-NEXT:    .pad #4
-; LEAF-FP-NEXT:    sub sp, #4
-; LEAF-FP-NEXT:    str r0, [sp]
-; LEAF-FP-NEXT:    adds r0, r0, #4
-; LEAF-FP-NEXT:    add sp, #4
-; LEAF-FP-NEXT:    bx lr
+; LEAF-FP-NEXT:    .save	{r7, lr}
+; LEAF-FP-NEXT:    push	{r7, lr}
+; LEAF-FP-NEXT:	   .setfp	r7, sp
+; LEAF-FP-NEXT:	   add	r7, sp, #0
+; LEAF-FP-NEXT:	   .pad	#4
+; LEAF-FP-NEXT:	   sub	sp, #4
+; LEAF-FP-NEXT:	   str	r0, [sp]
+; LEAF-FP-NEXT:	   adds	r0, r0, #4
+; LEAF-FP-NEXT:	   add	sp, #4
+; LEAF-FP-NEXT:	   pop	{r7, pc}
 ;
 ; LEAF-FP-AAPCS-LABEL: leaf:
 ; LEAF-FP-AAPCS:       @ %bb.0:
diff --git a/llvm/test/CodeGen/Thumb2/frame-pointer.ll b/llvm/test/CodeGen/Thumb2/frame-pointer.ll
index ae3c1c8a50e2b4..85c919a50d88c1 100644
--- a/llvm/test/CodeGen/Thumb2/frame-pointer.ll
+++ b/llvm/test/CodeGen/Thumb2/frame-pointer.ll
@@ -14,7 +14,7 @@ define void @leaf() {
 
 ; Leaf function, frame pointer is requested but we don't need any stack frame,
 ; so don't create a frame pointer.
-define void @leaf_nofpelim() "frame-pointer"="all" {
+define void @leaf_nofpelim() "frame-pointer"="none" {
 ; CHECK-LABEL: leaf_nofpelim:
 ; CHECK-NOT: push
 ; CHECK-NOT: sp
diff --git a/llvm/test/CodeGen/Thumb2/frameless.ll b/llvm/test/CodeGen/Thumb2/frameless.ll
index 01e0414de37d93..44914136b1f839 100644
--- a/llvm/test/CodeGen/Thumb2/frameless.ll
+++ b/llvm/test/CodeGen/Thumb2/frameless.ll
@@ -1,5 +1,5 @@
-; RUN: llc < %s -mtriple=thumbv7-apple-darwin -frame-pointer=all | not grep mov
-; RUN: llc < %s -mtriple=thumbv7-linux -frame-pointer=all | not grep mov
+; RUN: llc < %s -mtriple=thumbv7-apple-darwin -frame-pointer=none | not grep mov
+; RUN: llc < %s -mtriple=thumbv7-linux -frame-pointer=none | not grep mov
 
 define void @t() nounwind readnone {
   ret void
diff --git a/llvm/test/CodeGen/Thumb2/frameless2.ll b/llvm/test/CodeGen/Thumb2/frameless2.ll
index 4750527ae555cd..4848deaf8a1e4c 100644
--- a/llvm/test/CodeGen/Thumb2/frameless2.ll
+++ b/llvm/test/CodeGen/Thumb2/frameless2.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -mtriple=thumbv7-apple-darwin -frame-pointer=all | not grep r7
+; RUN: llc < %s -mtriple=thumbv7-apple-darwin -frame-pointer=none | not grep r7
 
 %struct.noise3 = type { [3 x [17 x i32]] }
 %struct.noiseguard = type { i32, i32, i32 }
diff --git a/llvm/test/CodeGen/Thumb2/machine-licm.ll b/llvm/test/CodeGen/Thumb2/machine-licm.ll
index 5a2ec9280de770..a2f379f7b54384 100644
--- a/llvm/test/CodeGen/Thumb2/machine-licm.ll
+++ b/llvm/test/CodeGen/Thumb2/machine-licm.ll
@@ -1,5 +1,5 @@
-; RUN: llc < %s -mtriple=thumbv7-apple-darwin -mcpu=cortex-a8 -relocation-model=dynamic-no-pic -frame-pointer=all | FileCheck %s
-; RUN: llc < %s -mtriple=thumbv7-apple-darwin -mcpu=cortex-a8 -relocation-model=pic -frame-pointer=all | FileCheck %s --check-prefix=PIC
+; RUN: llc < %s -mtriple=thumbv7-apple-darwin -mcpu=cortex-a8 -relocation-model=dynamic-no-pic -frame-pointer=none | FileCheck %s
+; RUN: llc < %s -mtriple=thumbv7-apple-darwin -mcpu=cortex-a8 -relocation-model=pic -frame-pointer=none | FileCheck %s --check-prefix=PIC
 ; rdar://7353541
 ; rdar://7354376
 
diff --git a/llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/arm_generated_funcs.ll b/llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/arm_generated_funcs.ll
index bae66d456f89a5..174cca4fab0982 100644
--- a/llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/arm_generated_funcs.ll
+++ b/llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/arm_generated_funcs.ll
@@ -60,4 +60,4 @@ define dso_local i32 @main() #0 {
   ret i32 0
 }
 
-attributes #0 = { noredzone nounwind ssp uwtable "frame-pointer"="all" }
+attributes #0 = { noredzone nounwind ssp uwtable "frame-pointer"="none" }
diff --git a/llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/arm_generated_funcs.ll.generated.expected b/llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/arm_generated_funcs.ll.generated.expected
index de5571f6436154..2dfb725f556655 100644
--- a/llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/arm_generated_funcs.ll.generated.expected
+++ b/llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/arm_generated_funcs.ll.generated.expected
@@ -61,7 +61,7 @@ define dso_local i32 @main() #0 {
   ret i32 0
 }
 
-attributes #0 = { noredzone nounwind ssp uwtable "frame-pointer"="all" }
+attributes #0 = { noredzone nounwind ssp uwtable "frame-pointer"="none" }
 ; CHECK-LABEL: check_boundaries:
 ; CHECK:       @ %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #20
diff --git a/llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/arm_generated_funcs.ll.nogenerated.expected b/llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/arm_generated_funcs.ll.nogenerated.expected
index 4f623384ade602..85d3389cdaaf92 100644
--- a/llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/arm_generated_funcs.ll.nogenerated.expected
+++ b/llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/arm_generated_funcs.ll.nogenerated.expected
@@ -121,4 +121,4 @@ define dso_local i32 @main() #0 {
   ret i32 0
 }
 
-attributes #0 = { noredzone nounwind ssp uwtable "frame-pointer"="all" }
+attributes #0 = { noredzone nounwind ssp uwtable "frame-pointer"="none" }

``````````

</details>


https://github.com/llvm/llvm-project/pull/109628

More information about the llvm-commits mailing list