On 2020-02-24 4:54 a.m., Christophe Lyon wrote:
Hi,
On Sun, 23 Feb 2020 at 22:26, Vladimir Makarov <vmaka...@redhat.com> wrote:
The following patch is for
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93564
The patch was successfully bootstrapped on x86-64 and benchmarked on
SPEC2000.
It seems this patch causes regression on some arm cores (seen on
cortex-a15 and cortex-57).
When configuring
--target arm-none-linux-gnueabihf
--with-cpu cortex-a57
--with-fpu crypto-neon-fp-armv8
gcc.target/arm/unaligned-memcpy-2.c: ldrd found 1 times
FAIL: gcc.target/arm/unaligned-memcpy-2.c scan-assembler-times ldrd 0
gcc.target/arm/unaligned-memcpy-2.c: strd found 2 times
FAIL: gcc.target/arm/unaligned-memcpy-2.c scan-assembler-times strd 1
This one regresses only on cortex-a15:
gcc.target/arm/unaligned-memcpy-3.c: ldrd found 2 times
FAIL: gcc.target/arm/unaligned-memcpy-3.c scan-assembler-times ldrd 1
gcc.target/arm/unaligned-memcpy-3.c: strd found 1 times
FAIL: gcc.target/arm/unaligned-memcpy-3.c scan-assembler-times strd 0
I've just submitted a patch which I hope will solve some of the failures.
