tcg/optimize: Fix sign_mask for logical right-shift

The 'sign' computation is attempting to locate the sign bit that has
been repeated, so that we can test if that bit is known zero.  That
computation can be zero if there are no known sign repetitions.

Cc: qemu-stable@nongnu.org
Fixes: 93a967fbb5 ("tcg/optimize: Propagate sign info for shifting")
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2248
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
master
Richard Henderson 2024-03-26 11:21:38 -10:00
parent 889cd5a8e2
commit 2911e9b95f
3 changed files with 30 additions and 1 deletions

View File

@ -2376,7 +2376,7 @@ static bool fold_shift(OptContext *ctx, TCGOp *op)
* will not reduced the number of input sign repetitions.
*/
sign = (s_mask & -s_mask) >> 1;
if (!(z_mask & sign)) {
if (sign && !(z_mask & sign)) {
ctx->s_mask = s_mask;
}
break;

View File

@ -10,6 +10,7 @@ VPATH += $(AARCH64_SRC)
# Base architecture tests
AARCH64_TESTS=fcvt pcalign-a64 lse2-fault
AARCH64_TESTS += test-2248
fcvt: LDFLAGS+=-lm

View File

@ -0,0 +1,28 @@
/* SPDX-License-Identifier: GPL-2.0-or-later */
/* See https://gitlab.com/qemu-project/qemu/-/issues/2248 */
#include <assert.h>
__attribute__((noinline))
long test(long x, long y, long sh)
{
long r;
asm("cmp %1, %2\n\t"
"cset x12, lt\n\t"
"and w11, w12, #0xff\n\t"
"cmp w11, #0\n\t"
"csetm x14, ne\n\t"
"lsr x13, x14, %3\n\t"
"sxtb %0, w13"
: "=r"(r)
: "r"(x), "r"(y), "r"(sh)
: "x11", "x12", "x13", "x14");
return r;
}
int main()
{
long r = test(0, 1, 2);
assert(r == -1);
return 0;
}