mirror of
https://github.com/RPCSX/llvm.git
synced 2024-11-23 19:59:57 +00:00
67a2b82b14
Register coalescing can change the target of a RegPair hint to a physreg, we should not crash on this. This also slightly improved the way ARMBaseRegisterInfo::updateRegAllocHint() works. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@233987 91177308-0d34-0410-b5e6-96231b3b80d8
23 lines
757 B
LLVM
23 lines
757 B
LLVM
; RUN: llc -o - %s
|
|
; ARM target used to fail an assertion if RegPair{Odd|Even} hint pointed to a
|
|
; physreg.
|
|
target datalayout = "e-m:o-p:32:32-f64:32:64-v64:32:64-v128:32:128-a:0:32-n32-S32"
|
|
target triple = "thumbv7-apple-tvos8.3.0"
|
|
|
|
declare i8* @llvm.frameaddress(i32) #1
|
|
declare i8* @llvm.returnaddress(i32) #1
|
|
|
|
@somevar = global [2 x i32] [i32 0, i32 0]
|
|
|
|
define void @__ubsan_handle_shift_out_of_bounds() #0 {
|
|
entry:
|
|
%0 = tail call i8* @llvm.frameaddress(i32 0)
|
|
%1 = ptrtoint i8* %0 to i32
|
|
%2 = tail call i8* @llvm.returnaddress(i32 0)
|
|
%3 = ptrtoint i8* %2 to i32
|
|
%val0 = insertvalue [2 x i32] [i32 undef, i32 undef], i32 %3, 0
|
|
%val1 = insertvalue [2 x i32] %val0, i32 %1, 1
|
|
store [2 x i32] %val1, [2 x i32]* @somevar, align 8
|
|
ret void
|
|
}
|