[CompactUnwind] Fix register encoding logic

Fix a compact unwind encoding logic bug which would try to encode
more callee saved registers than it should, leading to early bail out
in the encoding logic and abusive use of DWARF frame mode unnecessarily.

Also remove no-compact-unwind.ll which was testing the wrong thing
based on this bug and move it to valid 'compact unwind' tests. Added
other few more tests too.

llvm-svn: 223676
This commit is contained in:
Bruno Cardoso Lopes 2014-12-08 18:18:32 +00:00
parent d669b5b6fa
commit 915b66faf0
4 changed files with 152 additions and 74 deletions

View File

@ -682,7 +682,7 @@ private:
// 4 3
// 5 3
//
for (unsigned i = 0; i != CU_NUM_SAVED_REGS; ++i) {
for (unsigned i = 0; i < RegCount; ++i) {
int CUReg = getCompactUnwindRegNum(SavedRegs[i]);
if (CUReg == -1) return ~0U;
SavedRegs[i] = CUReg;

View File

@ -1,12 +1,20 @@
; RUN: llc < %s -disable-fp-elim -mtriple x86_64-apple-darwin11 -mcpu corei7 | FileCheck -check-prefix=ASM %s
; RUN: llc < %s -disable-fp-elim -mtriple x86_64-apple-darwin11 -mcpu corei7 -filetype=obj -o - \
; RUN: | llvm-objdump -triple x86_64-apple-darwin11 -s - \
; RUN: | llvm-objdump -triple x86_64-apple-darwin11 -unwind-info - \
; RUN: | FileCheck -check-prefix=CU %s
; RUN: llc < %s -disable-fp-elim -mtriple x86_64-apple-darwin11 -mcpu corei7 \
; RUN: | llvm-mc -triple x86_64-apple-darwin11 -filetype=obj -o - \
; RUN: | llvm-objdump -triple x86_64-apple-darwin11 -s - \
; RUN: | llvm-objdump -triple x86_64-apple-darwin11 -unwind-info - \
; RUN: | FileCheck -check-prefix=FROM-ASM %s
; RUN: llc < %s -mtriple x86_64-apple-macosx10.8.0 -mcpu corei7 -filetype=obj -o - \
; RUN: | llvm-objdump -triple x86_64-apple-macosx10.8.0 -unwind-info - \
; RUN: | FileCheck -check-prefix=NOFP-CU %s
; RUN: llc < %s -mtriple x86_64-apple-darwin11 -mcpu corei7 \
; RUN: | llvm-mc -triple x86_64-apple-darwin11 -filetype=obj -o - \
; RUN: | llvm-objdump -triple x86_64-apple-darwin11 -unwind-info - \
; RUN: | FileCheck -check-prefix=NOFP-FROM-ASM %s
%ty = type { i8* }
@gv = external global i32
@ -17,15 +25,19 @@
; Even though we can't encode %rax into the compact unwind, We still want to be
; able to generate a compact unwind encoding in this particular case.
; CU: Contents of section __compact_unwind:
; CU-NEXT: 0020 00000000 00000000 1e000000 01000101
; CU-NEXT: 0030 00000000 00000000 00000000 00000000
; CU: Contents of __compact_unwind section:
; CU-NEXT: Entry at offset 0x0:
; CU-NEXT: start: 0x0 _test0
; CU-NEXT: length: 0x1e
; CU-NEXT: compact encoding: 0x01010001
; FROM-ASM: Contents of section __compact_unwind:
; FROM-ASM-NEXT: 0020 00000000 00000000 1e000000 01000101
; FROM-ASM-NEXT: 0030 00000000 00000000 00000000 00000000
; FROM-ASM: Contents of __compact_unwind section:
; FROM-ASM-NEXT: Entry at offset 0x0:
; FROM-ASM-NEXT: start: 0x0 _test0
; FROM-ASM-NEXT: length: 0x1e
; FROM-ASM-NEXT: compact encoding: 0x01010001
define i8* @foo(i64 %size) {
define i8* @test0(i64 %size) {
%addr = alloca i64, align 8
%tmp20 = load i32* @gv, align 4
%tmp21 = call i32 @bar()
@ -39,3 +51,61 @@ define i8* @foo(i64 %size) {
}
declare i32 @bar()
%"struct.dyld::MappedRanges" = type { [400 x %struct.anon], %"struct.dyld::MappedRanges"* }
%struct.anon = type { %class.ImageLoader*, i64, i64 }
%class.ImageLoader = type { i32 (...)**, i8*, i8*, i32, i64, i64, i32, i32, %"struct.ImageLoader::recursive_lock"*, i16, i16, [4 x i8] }
%"struct.ImageLoader::recursive_lock" = type { i32, i32 }
@G1 = external hidden global %"struct.dyld::MappedRanges", align 8
declare void @OSMemoryBarrier() optsize
; Test the code below uses UNWIND_X86_64_MODE_STACK_IMMD compact unwind
; encoding.
; NOFP-CU: Entry at offset 0x20:
; NOFP-CU-NEXT: start: 0x1d _test1
; NOFP-CU-NEXT: length: 0x42
; NOFP-CU-NEXT: compact encoding: 0x02040c0a
; NOFP-FROM-ASM: Entry at offset 0x20:
; NOFP-FROM-ASM-NEXT: start: 0x1d _test1
; NOFP-FROM-ASM-NEXT: length: 0x42
; NOFP-FROM-ASM-NEXT: compact encoding: 0x02040c0a
define void @test1(%class.ImageLoader* %image) optsize ssp uwtable {
entry:
br label %for.cond1.preheader
for.cond1.preheader: ; preds = %for.inc10, %entry
%p.019 = phi %"struct.dyld::MappedRanges"* [ @G1, %entry ], [ %1, %for.inc10 ]
br label %for.body3
for.body3: ; preds = %for.inc, %for.cond1.preheader
%indvars.iv = phi i64 [ 0, %for.cond1.preheader ], [ %indvars.iv.next, %for.inc ]
%image4 = getelementptr inbounds %"struct.dyld::MappedRanges"* %p.019, i64 0, i32 0, i64 %indvars.iv, i32 0
%0 = load %class.ImageLoader** %image4, align 8
%cmp5 = icmp eq %class.ImageLoader* %0, %image
br i1 %cmp5, label %if.then, label %for.inc
if.then: ; preds = %for.body3
tail call void @OSMemoryBarrier() optsize
store %class.ImageLoader* null, %class.ImageLoader** %image4, align 8
br label %for.inc
for.inc: ; preds = %if.then, %for.body3
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp eq i32 %lftr.wideiv, 400
br i1 %exitcond, label %for.inc10, label %for.body3
for.inc10: ; preds = %for.inc
%next = getelementptr inbounds %"struct.dyld::MappedRanges"* %p.019, i64 0, i32 1
%1 = load %"struct.dyld::MappedRanges"** %next, align 8
%cmp = icmp eq %"struct.dyld::MappedRanges"* %1, null
br i1 %cmp, label %for.end11, label %for.cond1.preheader
for.end11: ; preds = %for.inc10
ret void
}

View File

@ -1,64 +0,0 @@
; RUN: llc < %s -mtriple x86_64-apple-macosx10.8.0 -mcpu corei7 -filetype=obj -o - \
; RUN: | llvm-objdump -triple x86_64-apple-macosx10.8.0 -s - \
; RUN: | FileCheck -check-prefix=CU %s
; RUN: llc < %s -mtriple x86_64-apple-darwin11 -mcpu corei7 \
; RUN: | llvm-mc -triple x86_64-apple-darwin11 -filetype=obj -o - \
; RUN: | llvm-objdump -triple x86_64-apple-darwin11 -s - \
; RUN: | FileCheck -check-prefix=FROM-ASM %s
%"struct.dyld::MappedRanges" = type { [400 x %struct.anon], %"struct.dyld::MappedRanges"* }
%struct.anon = type { %class.ImageLoader*, i64, i64 }
%class.ImageLoader = type { i32 (...)**, i8*, i8*, i32, i64, i64, i32, i32, %"struct.ImageLoader::recursive_lock"*, i16, i16, [4 x i8] }
%"struct.ImageLoader::recursive_lock" = type { i32, i32 }
@G1 = external hidden global %"struct.dyld::MappedRanges", align 8
declare void @OSMemoryBarrier() optsize
; This compact unwind encoding indicates that we could not generate correct
; compact unwind encodings for this function. This then defaults to using the
; DWARF EH frame.
; CU: Contents of section __compact_unwind:
; CU-NEXT: 0048 00000000 00000000 42000000 00000004
; CU-NEXT: 0058 00000000 00000000 00000000 00000000
; FROM-ASM: Contents of section __compact_unwind:
; FROM-ASM-NEXT: 0048 00000000 00000000 42000000 00000004
; FROM-ASM-NEXT: 0058 00000000 00000000 00000000 00000000
define void @func(%class.ImageLoader* %image) optsize ssp uwtable {
entry:
br label %for.cond1.preheader
for.cond1.preheader: ; preds = %for.inc10, %entry
%p.019 = phi %"struct.dyld::MappedRanges"* [ @G1, %entry ], [ %1, %for.inc10 ]
br label %for.body3
for.body3: ; preds = %for.inc, %for.cond1.preheader
%indvars.iv = phi i64 [ 0, %for.cond1.preheader ], [ %indvars.iv.next, %for.inc ]
%image4 = getelementptr inbounds %"struct.dyld::MappedRanges"* %p.019, i64 0, i32 0, i64 %indvars.iv, i32 0
%0 = load %class.ImageLoader** %image4, align 8
%cmp5 = icmp eq %class.ImageLoader* %0, %image
br i1 %cmp5, label %if.then, label %for.inc
if.then: ; preds = %for.body3
tail call void @OSMemoryBarrier() optsize
store %class.ImageLoader* null, %class.ImageLoader** %image4, align 8
br label %for.inc
for.inc: ; preds = %if.then, %for.body3
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp eq i32 %lftr.wideiv, 400
br i1 %exitcond, label %for.inc10, label %for.body3
for.inc10: ; preds = %for.inc
%next = getelementptr inbounds %"struct.dyld::MappedRanges"* %p.019, i64 0, i32 1
%1 = load %"struct.dyld::MappedRanges"** %next, align 8
%cmp = icmp eq %"struct.dyld::MappedRanges"* %1, null
br i1 %cmp, label %for.end11, label %for.cond1.preheader
for.end11: ; preds = %for.inc10
ret void
}

View File

@ -0,0 +1,72 @@
# RUN: llvm-mc -filetype=obj -triple=x86_64-apple-darwin10.0 %s | llvm-objdump -unwind-info - | FileCheck %s
.section __TEXT,__text,regular,pure_instructions
.macosx_version_min 10, 10
# Check that we emit compact-unwind info with UNWIND_X86_MODE_STACK_IND encoding
# CHECK: Contents of __compact_unwind section:
# CHECK-NEXT: Entry at offset 0x0:
# CHECK-NEXT: start: 0x0 _test0
# CHECK-NEXT: length: 0x15
# CHECK-NEXT: compact encoding: 0x03056804
.globl _test0
_test0: ## @test0
.cfi_startproc
## BB#0: ## %entry
pushq %rbp
Ltmp0:
.cfi_def_cfa_offset 16
pushq %rbx
Ltmp1:
.cfi_def_cfa_offset 24
subq $14408, %rsp ## imm = 0x3848
Ltmp2:
.cfi_def_cfa_offset 14432
Ltmp3:
.cfi_offset %rbx, -24
Ltmp4:
.cfi_offset %rbp, -16
xorl %eax, %eax
addq $14408, %rsp ## imm = 0x3848
popq %rbx
popq %rbp
retq
.cfi_endproc
# Check that we emit compact-unwind info with UNWIND_X86_MODE_STACK_IMMD encoding
# CHECK: Entry at offset 0x20:
# CHECK-NEXT: start: 0x15 _test1
# CHECK-NEXT: length: 0x15
# CHECK-NEXT: compact encoding: 0x02360804
.globl _test1
_test1: ## @test1
.cfi_startproc
## BB#0: ## %entry
pushq %rbp
Ltmp10:
.cfi_def_cfa_offset 16
pushq %rbx
Ltmp11:
.cfi_def_cfa_offset 24
subq $408, %rsp ## imm = 0x198
Ltmp12:
.cfi_def_cfa_offset 432
Ltmp13:
.cfi_offset %rbx, -24
Ltmp14:
.cfi_offset %rbp, -16
xorl %eax, %eax
addq $408, %rsp ## imm = 0x198
popq %rbx
popq %rbp
retq
.cfi_endproc
.section __TEXT,__cstring,cstring_literals
L_.str: ## @.str
.asciz "%d\n"
.subsections_via_symbols