https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116623
Bug ID: 116623
Summary: [15 regression] regressions on arm-linux-gnueabihf
since r15-1619-g3b9b8d6cfdf593
Product: gcc
Version: 15.0
Status: UNCONFIRMED
Keywords: testsuite-fail
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: thiago.bauermann at linaro dot org
CC: jskumari at gcc dot gnu.org
Target Milestone: ---
Target: arm
Since commit r15-1619-g3b9b8d6cfdf593 we see these regressions on
armv8l-linux-gnueabihf:
Running gcc:gcc.target/arm/arm.exp ...
FAIL: gcc.target/arm/bfloat16_scalar_1_1.c check-function-bodies bfloat_mov_mr
FAIL: gcc.target/arm/bfloat16_scalar_1_2.c check-function-bodies bfloat_mov_mr
FAIL: gcc.target/arm/bfloat16_scalar_2_1.c check-function-bodies bfloat_mov_mr
FAIL: gcc.target/arm/bfloat16_scalar_2_2.c check-function-bodies bfloat_mov_mr
FAIL: gcc.target/arm/bfloat16_scalar_3_1.c check-function-bodies bfloat_mov_mr
FAIL: gcc.target/arm/bfloat16_scalar_3_2.c check-function-bodies bfloat_mov_mr
Running gcc:gcc.target/arm/mve/mve.exp ...
FAIL: gcc.target/arm/mve/dlstp-compile-asm-2.c check-function-bodies test5
FAIL: gcc.target/arm/mve/dlstp-compile-asm-2.c check-function-bodies test8
The bfloat16_scalar_*.c failures are all about the same code change:
body: .*\tldrh r4, \[.*\] @ __bf16
.*\tbx lr
against: push {r4}
sub sp, sp, #12
#foo
ldrh r3, [sp, #6] @ __bf16
mov r4, r3 @ __bf16
#foo
add sp, sp, #12
ldr r4, [sp], #4
bx lr
Diff from before and after assembly file:
--- before/bfloat16_scalar_1_1.s 2024-09-05 20:19:02.000000000 -0300
+++ after/bfloat16_scalar_1_1.s 2024-09-05 19:57:59.000000000 -0300
@@ -193,7 +193,8 @@
@ 0 "" 2
.thumb
.syntax unified
- ldrh r4, [sp, #6] @ __bf16
+ ldrh r3, [sp, #6] @ __bf16
+ mov r4, r3 @ __bf16
.syntax unified
@ 117
"/home/thiago.bauermann/src/gcc/gcc/testsuite/gcc.target/arm/bfloat16_scalar_1_1.c"
1
#foo
The gcc.target/arm/mve/dlstp-compile-asm-2.c failures are a bit more involved:
body: .*\tdlstp.8 lr, r[0-9]+
.*\tvldrb.8 q[0-9]+, \[r1\]
\tvldrb.8 q[0-9]+, \[r2\]
.*\tvadd.i8 (q[0-9]+), q[0-9]+, q[0-9]+
.*\tvstrb.8 \1, \[r2\]
\tvstrb.8 \1, \[r3\]
\tletp lr, .*
.*
against: push {r4, r5, r6, lr}
ldr ip, [sp, #16]
cmp ip, #0
ble .L37
mov r6, r3
sub r3, ip, #16
add r5, r2, ip
add r4, r1, ip
add r0, r0, ip
dlstp.8 lr, ip
sub r2, r4, ip
sub r1, r0, ip
vldrb.8 q3, [r1]
vldrb.8 q2, [r2]
sub r2, r5, ip
mov ip, r3
subs r3, r3, #16
vadd.i8 q3, q3, q2
vstrb.8 q3, [r2]
vstrb.8 q3, [r6]
letp lr, .L39
pop {r4, r5, r6, pc}
FAIL: gcc.target/arm/mve/dlstp-compile-asm-2.c check-function-bodies test5
body: .*\tdlstp.32 lr, r3
\tvldrw.32 q[0-9]+, \[r0\], #16
\tvctp.32 r4
\tvpst
\tvldrwt.32 q[0-9]+, \[r1\], #16
.*\tvadd.i32 (q[0-9]+), q[0-9]+, q[0-9]+
\tvstrw.32 \1, \[r2\], #16
\tletp lr, .*
.*
against: push {lr}
ldr ip, [sp, #4]
cmp r3, #0
ble .L59
dlstp.32 lr, r3
vldrw.32 q3, [r0], #16
vctp.32 ip
vpst
vldrwt.32 q2, [r1], #16
add ip, ip, #1
vadd.i32 q3, q3, q2
vstrw.32 q3, [r2], #16
letp lr, .L61
ldr pc, [sp], #4
FAIL: gcc.target/arm/mve/dlstp-compile-asm-2.c check-function-bodies test8
Diff from before and after assembly file:
--- before/dlstp-compile-asm-2.s 2024-09-05 20:14:36.000000000 -0300
+++ after/dlstp-compile-asm-2.s 2024-09-05 20:10:07.000000000 -0300
@@ -140,25 +140,26 @@
@ args = 4, pretend = 0, frame = 0
@ frame_needed = 0, uses_anonymous_args = 0
push {r4, r5, r6, lr}
- ldr r4, [sp, #16]
- cmp r4, #0
+ ldr ip, [sp, #16]
+ cmp ip, #0
ble .L37
- sub ip, r4, #16
- adds r6, r2, r4
- adds r5, r1, r4
- add r0, r0, r4
- dlstp.8 lr, r4
+ mov r6, r3
+ sub r3, ip, #16
+ add r5, r2, ip
+ add r4, r1, ip
+ add r0, r0, ip
+ dlstp.8 lr, ip
.L39:
- subs r2, r5, r4
- subs r1, r0, r4
+ sub r2, r4, ip
+ sub r1, r0, ip
vldrb.8 q3, [r1]
vldrb.8 q2, [r2]
- subs r2, r6, r4
- mov r4, ip
- sub ip, ip, #16
+ sub r2, r5, ip
+ mov ip, r3
+ subs r3, r3, #16
vadd.i8 q3, q3, q2
vstrb.8 q3, [r2]
- vstrb.8 q3, [r3]
+ vstrb.8 q3, [r6]
letp lr, .L39
.L37:
pop {r4, r5, r6, pc}
@@ -227,22 +228,22 @@
test8:
@ args = 4, pretend = 0, frame = 0
@ frame_needed = 0, uses_anonymous_args = 0
- push {r4, lr}
- ldr r4, [sp, #8]
+ push {lr}
+ ldr ip, [sp, #4]
cmp r3, #0
ble .L59
dlstp.32 lr, r3
.L61:
vldrw.32 q3, [r0], #16
- vctp.32 r4
+ vctp.32 ip
vpst
vldrwt.32 q2, [r1], #16
- adds r4, r4, #1
+ add ip, ip, #1
vadd.i32 q3, q3, q2
vstrw.32 q3, [r2], #16
letp lr, .L61
.L59:
- pop {r4, pc}
+ ldr pc, [sp], #4
.size test8, .-test8
.align 1
.p2align 2,,3
Configure flags:
~/src/gcc//configure" \
SHELL=/bin/bash \
--with-gnu-as \
--with-gnu-ld \
--enable-lto \
--enable-shared \
--without-included-gettext \
--enable-nls \
--with-system-zlib \
--disable-sjlj-exceptions \
--enable-gnu-unique-object \
--enable-linker-build-id \
--disable-libstdcxx-pch \
--enable-c99 \
--enable-clocale=gnu \
--enable-libstdcxx-debug \
--enable-long-long \
--with-isl=no \
--disable-multilib \
--with-float=hard \
--with-fpu=neon-fp-armv8 \
--with-mode=thumb \
--with-arch=armv8-a \
--enable-threads=posix \
--enable-multiarch \
--enable-libstdcxx-time=yes \
--enable-gnu-indirect-function \
--enable-checking=yes \
--disable-bootstrap \
--enable-languages=c,c++,fortran,lto \
--prefix=/tmp/gcc-native-arm-linux-gnueabihf \
--build=arm-linux-gnueabihf \
--host=arm-linux-gnueabihf \
--target=arm-linux-gnueabihf
We also see regressions caused by this commit in other testcases on
aarch64-linux-gnu and yet others on arm-none-eabi, but I'm filing separate
bugzillas for those.
