19380bae05
Csmith discovered a program that caused wrong code generation with -O0: When handling a SIGN_EXTEND in expandRxSBG(), RxSBG.BitSize may be less than the Input width (if a truncate was previously traversed), so maskMatters() should be called with a masked based on the width of the sign extend result instead. Review: Ulrich Weigand llvm-svn: 319892
25 lines
683 B
LLVM
25 lines
683 B
LLVM
; Test that a rosbg conversion involving a sign extend operation rotates with
|
|
; the right number of steps.
|
|
;
|
|
; RUN: llc < %s -mtriple=s390x-linux-gnu -O0 | FileCheck %s
|
|
|
|
@g_136 = external global i16, align 2
|
|
@g_999 = external global i32, align 4
|
|
|
|
; Function Attrs: nounwind
|
|
define void @main() {
|
|
%1 = load i32, i32* undef, align 4
|
|
store i16 -28141, i16* @g_136, align 2
|
|
%2 = load i32, i32* undef, align 4
|
|
%3 = xor i32 -28141, %2
|
|
%4 = xor i32 %1, %3
|
|
%5 = sext i32 %4 to i64
|
|
%6 = icmp sgt i64 0, %5
|
|
%7 = zext i1 %6 to i32
|
|
%8 = load i32, i32* @g_999, align 4
|
|
%9 = or i32 %8, %7
|
|
; CHECK: rosbg %r1, %r3, 63, 63, 33
|
|
store i32 %9, i32* @g_999, align 4
|
|
ret void
|
|
}
|