* A lot of additional testsuite failures when bootstrapping with "--with-arch=native --with-tune=native" on AMD Threadripper
@ 2023-04-12 9:02 Rainer Emrich
2023-04-12 10:57 ` Richard Biener
0 siblings, 1 reply; 3+ messages in thread
From: Rainer Emrich @ 2023-04-12 9:02 UTC (permalink / raw)
To: gcc Mailing List
[-- Attachment #1.1.1: Type: text/plain, Size: 2341 bytes --]
I did some optimization testing by bootstrapping with "--with-arch=native --with-tune=native"
on an AMD Threadripper system, that's "znver2".
I got a lot of additional testsuite failures especially in the jit testsuite.
The log shows a lot of failures of the following form:
libgccjit.so: error: bad value 'native' for '-march=' switch
libgccjit.so: note: valid arguments to '-march=' switch are: nocona core2 nehalem corei7 westmere sandybridge corei7-avx ivybridge core-avx-i haswell core-avx2 broadwell skylake skylake-avx512 cannonlake icelake-client rocketlake icelake-server cascadelake tigerlake cooperlake sapphirerapids emeraldrapids alderlake raptorlake meteorlake graniterapids bonnell atom silvermont slm goldmont goldmont-plus tremont sierraforest grandridge knl knm x86-64 x86-64-v2 x86-64-v3 x86-64-v4 eden-x2 nano nano-1000 nano-2000 nano-3000 nano-x2 eden-x4 nano-x4 lujiazui k8 k8-sse3 opteron opteron-sse3 athlon64 athlon64-sse3 athlon-fx amdfam10 barcelona bdver1 bdver2 bdver3 bdver4 znver1 znver2 znver3 znver4 btver1 btver2 native
DISCARDED exp6 : libgccjit.so: error: bad value 'native' for '-march=' switch
DISCARDED exp6 : libgccjit.so: note: valid arguments to '-march=' switch are: nocona core2 nehalem corei7 westmere sandybridge corei7-avx ivybridge core-avx-i haswell core-avx2 broadwell skylake skylake-avx512 cannonlake icelake-client rocketlake icelake-server cascadelake tigerlake cooperlake sapphirerapids emeraldrapids alderlake raptorlake meteorlake graniterapids bonnell atom silvermont slm goldmont goldmont-plus tremont sierraforest grandridge knl knm x86-64 x86-64-v2 x86-64-v3 x86-64-v4 eden-x2 nano nano-1000 nano-2000 nano-3000 nano-x2 eden-x4 nano-x4 lujiazui k8 k8-sse3 opteron opteron-sse3 athlon64 athlon64-sse3 athlon-fx amdfam10 barcelona bdver1 bdver2 bdver3 bdver4 znver1 znver2 znver3 znver4 btver1 btver2 native
Assembler messages:
Error: can't open /tmp/libgccjit-uacKEw/fake.s for reading: No such file or directory
DISCARDED exp6 : Assembler messages:
DISCARDED exp6 : Error: can't open /tmp/libgccjit-uacKEw/fake.s for reading: No such file or directory
libgccjit.so: error: error invoking gcc driver
DISCARDED exp6 : libgccjit.so: error: error invoking gcc driver
So, the "-march" switch in libgccjit.so is wrong!
I can open a PR if neccessary.
Rainer
[-- Attachment #1.1.2: OpenPGP public key --]
[-- Type: application/pgp-keys, Size: 8721 bytes --]
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 840 bytes --]
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: A lot of additional testsuite failures when bootstrapping with "--with-arch=native --with-tune=native" on AMD Threadripper
2023-04-12 9:02 A lot of additional testsuite failures when bootstrapping with "--with-arch=native --with-tune=native" on AMD Threadripper Rainer Emrich
@ 2023-04-12 10:57 ` Richard Biener
2023-04-12 11:23 ` Rainer Emrich
0 siblings, 1 reply; 3+ messages in thread
From: Richard Biener @ 2023-04-12 10:57 UTC (permalink / raw)
To: Rainer Emrich; +Cc: gcc Mailing List
On Wed, Apr 12, 2023 at 11:03 AM Rainer Emrich
<rainer@emrich-ebersheim.de> wrote:
>
> I did some optimization testing by bootstrapping with "--with-arch=native --with-tune=native"
> on an AMD Threadripper system, that's "znver2".
I'm not sure that's really supported, it appears to be not resolved at
configure time at least.
> I got a lot of additional testsuite failures especially in the jit testsuite.
Are you doing an LTO bootstrap? Please quote the full configure and
make commands.
Does it work when you don't enable jit?
> The log shows a lot of failures of the following form:
>
> libgccjit.so: error: bad value 'native' for '-march=' switch
> libgccjit.so: note: valid arguments to '-march=' switch are: nocona core2 nehalem corei7 westmere sandybridge corei7-avx ivybridge core-avx-i haswell core-avx2 broadwell skylake skylake-avx512 cannonlake icelake-client rocketlake icelake-server cascadelake tigerlake cooperlake sapphirerapids emeraldrapids alderlake raptorlake meteorlake graniterapids bonnell atom silvermont slm goldmont goldmont-plus tremont sierraforest grandridge knl knm x86-64 x86-64-v2 x86-64-v3 x86-64-v4 eden-x2 nano nano-1000 nano-2000 nano-3000 nano-x2 eden-x4 nano-x4 lujiazui k8 k8-sse3 opteron opteron-sse3 athlon64 athlon64-sse3 athlon-fx amdfam10 barcelona bdver1 bdver2 bdver3 bdver4 znver1 znver2 znver3 znver4 btver1 btver2 native
> DISCARDED exp6 : libgccjit.so: error: bad value 'native' for '-march=' switch
> DISCARDED exp6 : libgccjit.so: note: valid arguments to '-march=' switch are: nocona core2 nehalem corei7 westmere sandybridge corei7-avx ivybridge core-avx-i haswell core-avx2 broadwell skylake skylake-avx512 cannonlake icelake-client rocketlake icelake-server cascadelake tigerlake cooperlake sapphirerapids emeraldrapids alderlake raptorlake meteorlake graniterapids bonnell atom silvermont slm goldmont goldmont-plus tremont sierraforest grandridge knl knm x86-64 x86-64-v2 x86-64-v3 x86-64-v4 eden-x2 nano nano-1000 nano-2000 nano-3000 nano-x2 eden-x4 nano-x4 lujiazui k8 k8-sse3 opteron opteron-sse3 athlon64 athlon64-sse3 athlon-fx amdfam10 barcelona bdver1 bdver2 bdver3 bdver4 znver1 znver2 znver3 znver4 btver1 btver2 native
> Assembler messages:
> Error: can't open /tmp/libgccjit-uacKEw/fake.s for reading: No such file or directory
> DISCARDED exp6 : Assembler messages:
> DISCARDED exp6 : Error: can't open /tmp/libgccjit-uacKEw/fake.s for reading: No such file or directory
> libgccjit.so: error: error invoking gcc driver
> DISCARDED exp6 : libgccjit.so: error: error invoking gcc driver
>
> So, the "-march" switch in libgccjit.so is wrong!
>
> I can open a PR if neccessary.
>
> Rainer
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: A lot of additional testsuite failures when bootstrapping with "--with-arch=native --with-tune=native" on AMD Threadripper
2023-04-12 10:57 ` Richard Biener
@ 2023-04-12 11:23 ` Rainer Emrich
0 siblings, 0 replies; 3+ messages in thread
From: Rainer Emrich @ 2023-04-12 11:23 UTC (permalink / raw)
To: Richard Biener; +Cc: gcc Mailing List
[-- Attachment #1.1.1: Type: text/plain, Size: 5054 bytes --]
Am 12.04.2023 um 12:57 schrieb Richard Biener:
> On Wed, Apr 12, 2023 at 11:03 AM Rainer Emrich
> <rainer@emrich-ebersheim.de> wrote:
>>
>> I did some optimization testing by bootstrapping with "--with-arch=native --with-tune=native"
>> on an AMD Threadripper system, that's "znver2".
>
> I'm not sure that's really supported, it appears to be not resolved at
> configure time at least.
The bootstrap is successful.
>
>> I got a lot of additional testsuite failures especially in the jit testsuite.
>
> Are you doing an LTO bootstrap? Please quote the full configure and
> make commands.
That doesn't matter, I did both normal bootstrap and LTO bootstrap. The
example failure is from normal bootstrap:
Compiler version: 13.0.1 20230406 (experimental) [master
r13-7112-gd2a499a9881] (GCC)
Platform: x86_64-pc-linux-gnu
configure flags:
--prefix=/opt/devel/gnu/gcc/Linux/x86_64-pc-linux-gnu/Ubuntu_20.04/gcc-13.0.1-test
--with-gnu-as
--with-as=/opt/devel/gnu/gcc/Linux/x86_64-pc-linux-gnu/Ubuntu_20.04/gcc-13.0.1-test/bin/as
--with-gnu-ld
--with-ld=/opt/devel/gnu/gcc/Linux/x86_64-pc-linux-gnu/Ubuntu_20.04/gcc-13.0.1-test/bin/ld
--build=x86_64-pc-linux-gnu --enable-threads=posix --enable-shared
--enable-__cxa_atexit --enable-plugin --enable-languages=all
--with-gmp-include=/opt/devel/gnu/gcc/Linux/x86_64-pc-linux-gnu/Ubuntu_20.04/gcc-13.0.1-test/include
--with-gmp-lib=/opt/devel/gnu/gcc/Linux/x86_64-pc-linux-gnu/Ubuntu_20.04/gcc-13.0.1-test/lib64
--with-mpfr-include=/opt/devel/gnu/gcc/Linux/x86_64-pc-linux-gnu/Ubuntu_20.04/gcc-13.0.1-test/include
--with-mpfr-lib=/opt/devel/gnu/gcc/Linux/x86_64-pc-linux-gnu/Ubuntu_20.04/gcc-13.0.1-test/lib64
--with-mpc-include=/opt/devel/gnu/gcc/Linux/x86_64-pc-linux-gnu/Ubuntu_20.04/gcc-13.0.1-test/include
--with-mpc-lib=/opt/devel/gnu/gcc/Linux/x86_64-pc-linux-gnu/Ubuntu_20.04/gcc-13.0.1-test/lib64
--with-isl-include=/opt/devel/gnu/gcc/Linux/x86_64-pc-linux-gnu/Ubuntu_20.04/gcc-13.0.1-test/include
--with-isl-lib=/opt/devel/gnu/gcc/Linux/x86_64-pc-linux-gnu/Ubuntu_20.04/gcc-13.0.1-test/lib64
--with-local-prefix=/opt/devel/tec/devel/Linux/x86_64-pc-linux-gnu/Ubuntu_20.04/gcc-13.0.1-test
--enable-checking=release --disable-werror --disable-multilib
--with-sphinx-build=/home/rainer/sphinx-gcc/bin/sphinx-build
--enable-host-shared --with-arch=native --with-tune=native 20 15:54:19
UTC 2023 x86_64 x86_64 x86_64 GNU/Linux
The make command is:
"make -j 16"
> Does it work when you don't enable jit?As I said above, the bootstrap is successful. There are some additional
testsuite failures, but not many, only in the jit testsuite I get more
than 500 additional failures.
Additional info, even combined LTO profiled bootstrap is working, with
all languages enabled: c,ada,c++,d,fortran,go,jit,lto,m2,objc,obj-c++,rust
This is on Ubuntu 22.04.
> >> The log shows a lot of failures of the following form:
>>
>> libgccjit.so: error: bad value 'native' for '-march=' switch
>> libgccjit.so: note: valid arguments to '-march=' switch are: nocona core2 nehalem corei7 westmere sandybridge corei7-avx ivybridge core-avx-i haswell core-avx2 broadwell skylake skylake-avx512 cannonlake icelake-client rocketlake icelake-server cascadelake tigerlake cooperlake sapphirerapids emeraldrapids alderlake raptorlake meteorlake graniterapids bonnell atom silvermont slm goldmont goldmont-plus tremont sierraforest grandridge knl knm x86-64 x86-64-v2 x86-64-v3 x86-64-v4 eden-x2 nano nano-1000 nano-2000 nano-3000 nano-x2 eden-x4 nano-x4 lujiazui k8 k8-sse3 opteron opteron-sse3 athlon64 athlon64-sse3 athlon-fx amdfam10 barcelona bdver1 bdver2 bdver3 bdver4 znver1 znver2 znver3 znver4 btver1 btver2 native
>> DISCARDED exp6 : libgccjit.so: error: bad value 'native' for '-march=' switch
>> DISCARDED exp6 : libgccjit.so: note: valid arguments to '-march=' switch are: nocona core2 nehalem corei7 westmere sandybridge corei7-avx ivybridge core-avx-i haswell core-avx2 broadwell skylake skylake-avx512 cannonlake icelake-client rocketlake icelake-server cascadelake tigerlake cooperlake sapphirerapids emeraldrapids alderlake raptorlake meteorlake graniterapids bonnell atom silvermont slm goldmont goldmont-plus tremont sierraforest grandridge knl knm x86-64 x86-64-v2 x86-64-v3 x86-64-v4 eden-x2 nano nano-1000 nano-2000 nano-3000 nano-x2 eden-x4 nano-x4 lujiazui k8 k8-sse3 opteron opteron-sse3 athlon64 athlon64-sse3 athlon-fx amdfam10 barcelona bdver1 bdver2 bdver3 bdver4 znver1 znver2 znver3 znver4 btver1 btver2 native
>> Assembler messages:
>> Error: can't open /tmp/libgccjit-uacKEw/fake.s for reading: No such file or directory
>> DISCARDED exp6 : Assembler messages:
>> DISCARDED exp6 : Error: can't open /tmp/libgccjit-uacKEw/fake.s for reading: No such file or directory
>> libgccjit.so: error: error invoking gcc driver
>> DISCARDED exp6 : libgccjit.so: error: error invoking gcc driver
>>
>> So, the "-march" switch in libgccjit.so is wrong!
>>
>> I can open a PR if neccessary.
>>
>> Rainer
[-- Attachment #1.1.2: OpenPGP public key --]
[-- Type: application/pgp-keys, Size: 8721 bytes --]
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 840 bytes --]
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2023-04-12 11:23 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-04-12 9:02 A lot of additional testsuite failures when bootstrapping with "--with-arch=native --with-tune=native" on AMD Threadripper Rainer Emrich
2023-04-12 10:57 ` Richard Biener
2023-04-12 11:23 ` Rainer Emrich
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).