[LLVMbugs] [Bug 18399] New: Thumb1: Folding SP Update incorrectly stop at modified pop
bugzilla-daemon at llvm.org
bugzilla-daemon at llvm.org
Mon Jan 6 11:22:14 PST 2014
http://llvm.org/bugs/show_bug.cgi?id=18399
Bug ID: 18399
Summary: Thumb1: Folding SP Update incorrectly stop at modified
pop
Product: libraries
Version: trunk
Hardware: PC
OS: Linux
Status: NEW
Severity: normal
Priority: P
Component: Backend: ARM
Assignee: unassignedbugs at nondot.org
Reporter: zhaoshiz at codeaurora.org
CC: llvmbugs at cs.uiuc.edu, t.p.northover at gmail.com
Classification: Unclassified
In Thumb1, we check on whether registers are callee-saved when deciding to
pass by a pop instruction or not. But folding SP update adds caller-saved
registers to the pop, rendering the above mechanism invalid.
If tryFoldSPUpdateIntoPushPop() returns true, we can remember the modified pop
instruction. Later when moving past by callee-saved register restoration, we
can simply skip the remembered pop.
Tim, do you want me fix it or you have other plan?
$ cat test_case.ll
declare void @bar(i8*)
define void @test_varsize(...) minsize {
%var = alloca i8, i32 8
call void @bar(i8* %var)
ret void
}
$ llc -mtriple=thumbv5-unkown-unkown -disable-fp-elim test_case.ll -o -
.syntax unified
.eabi_attribute 6, 3
.eabi_attribute 8, 1
.eabi_attribute 9, 1
.eabi_attribute 20, 1
.eabi_attribute 21, 1
.eabi_attribute 23, 3
.eabi_attribute 24, 1
.eabi_attribute 25, 1
.file "test_case.ll"
.text
.globl test_varsize
.align 1
.type test_varsize,%function
.code 16 @ @test_varsize
.thumb_func
test_varsize:
@ BB#0:
sub sp, #16
push {r5, r6, r7, lr}
add r7, sp, #8
str r3, [sp, #28]
str r2, [sp, #24]
str r1, [sp, #20]
str r0, [sp, #16]
mov r0, sp
bl bar
pop {r3}
add sp, #16
bx r3
bx lr
.Ltmp0:
.size test_varsize, .Ltmp0-test_varsize
--
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20140106/9ac690fc/attachment.html>
More information about the llvm-bugs
mailing list