[llvm-bugs] [Bug 52320] New: [BranchFolding] Merging blocks with identical PATCHPOINTs
via llvm-bugs
llvm-bugs at lists.llvm.org
Tue Oct 26 02:27:28 PDT 2021
https://bugs.llvm.org/show_bug.cgi?id=52320
Bug ID: 52320
Summary: [BranchFolding] Merging blocks with identical
PATCHPOINTs
Product: libraries
Version: trunk
Hardware: PC
OS: Linux
Status: NEW
Severity: normal
Priority: P
Component: Common Code Generator Code
Assignee: unassignedbugs at nondot.org
Reporter: paulsson at linux.vnet.ibm.com
CC: llvm-bugs at lists.llvm.org
Created attachment 25392
--> https://bugs.llvm.org/attachment.cgi?id=25392&action=edit
testcase
I suspect that there is a bug in branch-folding as it merges two different
patchpoints into one:
define void @fun(i32 %Arg) {
%c = icmp eq i32 %Arg, 0
br i1 %c, label %patch0, label %patch1
patch0:
call void (i64, i32, i8*, i32, ...) @llvm.experimental.patchpoint.void(i64 0,
i32 14, i8* null, i32 0)
br label %exit
patch1:
call void (i64, i32, i8*, i32, ...) @llvm.experimental.patchpoint.void(i64 0,
i32 14, i8* null, i32 0)
br label %exit
exit:
ret void
}
declare void @llvm.experimental.patchpoint.void(i64, i32, i8*, i32, ...)
llc -mtriple=s390x-linux-gnu -O3 -o - ./tc_patchpoint_branchfolder.ll
CFG optimizer will see that these are identical and therefore remove the
conditional branch:
# *** IR Dump After Control Flow Optimizer (branch-folder) ***:
bb.0 (%ir-block.0):
STMG killed $r14d, killed $r15d, $r15d, 112
$r15d = AGHI $r15d(tied-def 0), -160, implicit-def dead $cc
CHI killed renamable $r2l, 0, implicit-def $cc
PATCHPOINT 0, 14, 0, 0, 0, <regmask $f8d $f9d $f10d $f11d $f12d $f13d $f14d
$f15d $f8q $f9q $f12q $f13q $f8s $f9s $f10s $f11s $f12s $f13s $f14s $f15s $r6d
$r7d $r8d $r9d $r10d $r11d $r12d $r13d $r14d $r15d $r6h $r7h $r8h and 22
more...>, implicit-def dead early-clobber $r0d, implicit-def dead early-clobber
$r1d, implicit-def dead early-clobber $r14d
$r14d, $r15d = LMG $r15d, 272
Return
# End machine code for function fun
IIUC, PATCHPOINTs can not be considered identical as they will be patched
during runtime and therefore are conceptually always different..?
--
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/20211026/1ec3a865/attachment.html>
More information about the llvm-bugs
mailing list