| View previous topic :: View next topic |
| Author |
Message |
dE_logics
Advocate
Joined: 02 Jan 2009
Posts: 2266
Location: $TERM
|
 Posted: Thu Sep 12, 2024 10:28 am Post subject: Determining -march=native output. |
 |
|
I've tried
| Code: | | gcc -### -march=native -E /usr/include/stdlib.h 2>&1 | grep "/usr/libexec/gcc/.*cc1" |
OR
| Code: | | gcc -march=native -E -v - </dev/null 2>&1 | grep cc1 |
(which both result in the same outputs)
I get the following --
| Code: | | /usr/libexec/gcc/x86_64-pc-linux-gnu/14/cc1 -E -quiet /usr/include/stdlib.h "-march=znver1" -mmmx -mpopcnt -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -mavx -mavx2 -msse4a -mno-fma4 -mno-xop -mfma -mno-avx512f -mbmi -mbmi2 -maes -mpclmul -mno-avx512vl -mno-avx512bw -mno-avx512dq -mno-avx512cd -mno-avx512vbmi -mno-avx512ifma -mno-avx512vpopcntdq -mno-avx512vbmi2 -mno-gfni -mno-vpclmulqdq -mno-avx512vnni -mno-avx512bitalg -mno-avx512bf16 -mno-avx512vp2intersect -mno-3dnow -madx -mabm -mno-cldemote -mclflushopt -mno-clwb -mclzero -mcx16 -mno-enqcmd -mf16c -mfsgsbase -mfxsr -mno-hle -msahf -mno-lwp -mlzcnt -mmovbe -mno-movdir64b -mno-movdiri -mmwaitx -mno-pconfig -mno-pku -mprfchw -mno-ptwrite -mno-rdpid -mrdrnd -mrdseed -mno-rtm -mno-serialize -mno-sgx -msha -mno-shstk -mno-tbm -mno-tsxldtrk -mno-vaes -mno-waitpkg -mno-wbnoinvd -mxsave -mxsavec -mxsaveopt -mxsaves -mno-amx-tile -mno-amx-int8 -mno-amx-bf16 -mno-uintr -mno-hreset -mno-kl -mno-widekl -mno-avxvnni -mno-avx512fp16 -mno-avxifma -mno-avxvnniint8 -mno-avxneconvert -mno-cmpccxadd -mno-amx-fp16 -mno-prefetchi -mno-raoint -mno-amx-complex -mno-avxvnniint16 -mno-sm3 -mno-sha512 -mno-sm4 -mno-apxf -mno-usermsr --param "l1-cache-size=32" --param "l1-cache-line-size=64" --param "l2-cache-size=512" "-mtune=znver1" -dumpbase stdlib.h -dumpbase-ext .h |
Based on this I've been trying to cross-compile the kernel with the following KCFLAGS --
| Code: | KCFLAGS="-march=znver1 -mmmx -mpopcnt -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -mavx -mavx2 -msse4a -mno-fma4 -mno-xop -mfma -mno-avx512f -mbmi -mbmi2 -maes -mpclmul -mno-avx512vl -mno-avx512bw -mno-avx512dq -mno-avx512cd -mno-avx512vbmi -mno-avx512ifma -mno-avx512vpopcntdq -mno-avx512vbmi2 -mno-gfni -mno-vpclmulqdq -mno-avx512vnni -mno-avx512bitalg -mno-avx512bf16 -mno-avx512vp2intersect -mno-3dnow -madx -mabm -mno-cldemote -mclflushopt -mno-clwb -mclzero -mcx16 -mno-enqcmd -mf16c -mfsgsbase -mfxsr -mno-hle -msahf -mno-lwp -mlzcnt -mmovbe -mno-movdir64b -mno-movdiri -mmwaitx -mno-pconfig -mno-pku -mprfchw -mno-ptwrite -mno-rdpid -mrdrnd -mrdseed -mno-rtm -mno-serialize -mno-sgx -msha -mno-shstk -mno-tbm -mno-tsxldtrk -mno-vaes -mno-waitpkg -mno-wbnoinvd -mxsave -mxsavec -mxsaveopt -mxsaves -mno-amx-tile -mno-amx-int8 -mno-amx-bf16 -mno-uintr -mno-hreset -mno-kl -mno-widekl -mno-avxvnni -mno-avx512fp16 -mno-avxifma -mno-avxvnniint8 -mno-avxneconvert -mno-cmpccxadd -mno-amx-fp16 -mno-prefetchi -mno-raoint -mno-amx-complex -mno-avxvnniint16 -mno-sm3 -mno-sha512 -mno-sm4 -mno-apxf -mno-usermsr --param l1-cache-size=32 --param l1-cache-line-size=64 --param l2-cache-size=512 -flto=1 -fuse-linker-plugin -ffat-lto-objects -O3"
|
However this does not work. After the kernel gets executed, the system reboots.
If I --
| Code: | | "-march=native -O3 -pipe -Wno-error -flto=1 -fuse-linker-plugin -ffat-lto-objects" |
It works. I fail to understand how. Can someone put some light on this?
_________________
My blog |
|
| Back to top |
|
 |
krumpf
Apprentice
Joined: 15 Jul 2018
Posts: 187
|
| Posted: Thu Sep 12, 2024 8:44 pm Post subject: |
|
|
To check native flags, I use the following :
| Code: | | gcc -march=native -Q --help=target |
Compiling the kernel with gcc using LTO is possible, but requires a patch (read here).
Afaik, many flags you've added seem useless, as march=native should already cover most of them. Check in kernel menuconfig submenu 'Processor type and features' (gentoo-sources package needs USE=experimental to display additional CPUs in the list, notably the intel/amd native)
_________________
Dragon Princess Music Games Heroes and villains |
|
| Back to top |
|
|
dE_logics
Advocate
Joined: 02 Jan 2009
Posts: 2266
Location: $TERM
|
| Posted: Fri Sep 13, 2024 3:43 am Post subject: |
|
|
I'm cross compiling. -march=native is not an option.
I was aware of gcc -march=native -Q --help=target but it seems it's output is different from the commands which I'm using. However it's more 'official' therefore I'm going to try this out. Also I'll try disabling LTO (although it's working without any patches).
Thanks for notifying about these.
_________________
My blog |
|
| Back to top |
|
|
dE_logics
Advocate
Joined: 02 Jan 2009
Posts: 2266
Location: $TERM
|
| Posted: Fri Sep 13, 2024 5:53 am Post subject: |
|
|
This results in a bigger and more epic fail.
./arch/x86/include/asm/preempt.h:27: Error: no such instruction: `movl %gs:DWORD PTR __preempt_count[rip],ecx'
_________________
My blog |
|
| Back to top |
|
|
dE_logics
Advocate
Joined: 02 Jan 2009
Posts: 2266
Location: $TERM
|
| Posted: Fri Sep 13, 2024 8:27 am Post subject: |
|
|
Turns out app-misc/resolve-march-native is the solution.
_________________
My blog |
|
| Back to top |
|
|
Hu
Administrator
Joined: 06 Mar 2007
Posts: 22617
|
| Posted: Fri Sep 13, 2024 2:13 pm Post subject: |
|
|
| From what architecture are you cross-compiling, and to what architecture? I would not expect -march=native to be meaningful in that situation, since it directs the compiler to build for the system on which it is running. |
|
| Back to top |
|
|
|
Kernel & Hardware
