On 07/10/2024 10:15, Christophe Lyon wrote:
On Mon, 7 Oct 2024 at 11:04, Torbjorn SVENSSON
<torbjorn.svens...@foss.st.com> wrote:
On 2024-10-07 10:53, Andre Vieira (lists) wrote:
Hi Torbjorn,
2. All other the test cases in the list above: These need to be
adapted to the change introduced in r15-3606-g7d6c6a0d15c to have the
proper arch.
I've sent a patch that should fix these "regressions" in https://
gcc.gnu.org/pipermail/gcc-patches/2024-October/664611.html.
I presume you are using -march=armv8.1-m.main+mve.fp+fp.dp for these
rather than -mcpu? If I do:
RUNTESTFLAGS="--target_board=<...>/-mcpu=cortex-m55/-mfloat-abi=hard
arm.exp=vseleqdf.c" then it works just fine for me as the -mcpu=unset
does it work, but the -march=armv8.1-m.main+mve.fp+fp.dp does fail. I'll
talk to Richard E about this one.
For these results, I did as before. This is the command line that was
used for gcc.target/arm/vseleqdf.c:
.../bin/arm-none-eabi-gcc .../gcc/testsuite/gcc.target/arm/vseleqdf.c
-mthumb -march=armv8.1-m.main+mve+pacbti -mcpu=cortex-m85
-mfloat-abi=hard -mfpu=fpv5-d16 -fdiagnostics-plain-output -O2
-mcpu=cortex-a57 -mfpu=fp-armv8 -mfloat-abi=softfp -ffat-lto-objects
-fno-ident -S -o vseleqdf.s
@Torbjorn:
Where are all these options coming from though? There are some
'conflicts' here. If you are passing all of these to RUNTESTFLAGS then
... please don't.
For instance:
-march=armv8.1-m.main+mve+pacbti -mcpu=cortex-m85
These are contradictory:
Use either -march or -mcpu.
-mcpu=cortex-m85 is equivalent to:
-march=armv8.1-m.main+pacbti+mve.fp+fp.dp -mtune=cortex-m85
Your -march only enables integer mve it also doesn't enable any FP,
cortex-m85 only seems to allow for double precision FP, we dont' seem to
have a single FP configuration in GCC.
I also see a -mfpu=fpv5-d16, if you are passing that, then please don't.
You should be passing -mfpu=auto if anything when compiling for MVE
cores as there is no -mfpu= option that enables MVE and we don't
recommend using both architecture extensions and -mfpu options.
Having said all this, I do now realize vseleq.f64 is available for
armv8.1-m.main, so we should also enable that. I'm having a look!
Right: I was mainly mentioning the bootstrap problem in case it takes
time to come to an agreement for this patch. (bootstrap is easy to
fix)
(@Andre I did manually start a precommit CI bootstrap build for this
patch, but it fails to apply because we keep the baseline as before
bootstrap is broken)
Yeah, hadn't had the chance to look at the fail log. But I'll commit the
fix as obvious today, regardless of the issues above.
