AMX was introduced in QEMU commit 1f16764f7d4515bfd5e4ae0aae814fa280a7d0c8
and following commits.
---
src/cpu_map/sync_qemu_i386.py | 3 +++
src/cpu_map/x86_features.xml | 9 +++++++++
2 files changed, 12 insertions(+)
diff --git a/src/cpu_map/sync_qemu_i386.py b/src/cpu_map/sync_qemu_i386.py
index 4dd9f3b84d..6a46f87cff 100755
--- a/src/cpu_map/sync_qemu_i386.py
+++ b/src/cpu_map/sync_qemu_i386.py
@@ -72,6 +72,9 @@ def translate_feature(name):
"CPUID_7_0_EDX_SPEC_CTRL": "spec-ctrl",
"CPUID_7_0_EDX_SPEC_CTRL_SSBD": "ssbd",
"CPUID_7_0_EDX_STIBP": "stibp",
+ "CPUID_7_0_EDX_AMX_BF16": "amx-bf16",
+ "CPUID_7_0_EDX_AMX_TILE": "amx-tile",
+ "CPUID_7_0_EDX_AMX_INT8": "amx-int8",
"CPUID_7_1_EAX_AVX512_BF16": "avx512-bf16",
"CPUID_7_1_EAX_AVX_VNNI": "avx-vnni",
"CPUID_8000_0008_EBX_AMD_SSBD": "amd-ssbd",
diff --git a/src/cpu_map/x86_features.xml b/src/cpu_map/x86_features.xml
index 4cf3ff0804..102d39f626 100644
--- a/src/cpu_map/x86_features.xml
+++ b/src/cpu_map/x86_features.xml
@@ -347,6 +347,15 @@
<feature name='pconfig'>
<cpuid eax_in='0x07' ecx_in='0x00' edx='0x00040000'/>
</feature>
+ <feature name='amx-bf16'>
+ <cpuid eax_in='0x07' ecx_in='0x00' edx='0x00400000'/>
+ </feature>
+ <feature name='amx-tile'>
+ <cpuid eax_in='0x07' ecx_in='0x00' edx='0x01000000'/>
+ </feature>
+ <feature name='amx-int8'>
+ <cpuid eax_in='0x07' ecx_in='0x00' edx='0x02000000'/>
+ </feature>
<feature name='spec-ctrl'>
<cpuid eax_in='0x07' ecx_in='0x00' edx='0x04000000'/>
</feature>
--
2.34.1
On Tue, Sep 13, 2022 at 12:30:31PM -0700, Lin Yang wrote: >AMX was introduced in QEMU commit 1f16764f7d4515bfd5e4ae0aae814fa280a7d0c8 >and following commits. Some of them not pushed yet I believe: https://lists.gnu.org/archive/html/qemu-devel/2022-08/msg01936.html Otherwise the patch looks good (I'm not sure what's the planned usage of the python script, but this patch just adds features that'd be missing in the future, so I think it's fine). Reviewed-by: Martin Kletzander <mkletzan@redhat.com> >--- > src/cpu_map/sync_qemu_i386.py | 3 +++ > src/cpu_map/x86_features.xml | 9 +++++++++ > 2 files changed, 12 insertions(+) > >diff --git a/src/cpu_map/sync_qemu_i386.py b/src/cpu_map/sync_qemu_i386.py >index 4dd9f3b84d..6a46f87cff 100755 >--- a/src/cpu_map/sync_qemu_i386.py >+++ b/src/cpu_map/sync_qemu_i386.py >@@ -72,6 +72,9 @@ def translate_feature(name): > "CPUID_7_0_EDX_SPEC_CTRL": "spec-ctrl", > "CPUID_7_0_EDX_SPEC_CTRL_SSBD": "ssbd", > "CPUID_7_0_EDX_STIBP": "stibp", >+ "CPUID_7_0_EDX_AMX_BF16": "amx-bf16", >+ "CPUID_7_0_EDX_AMX_TILE": "amx-tile", >+ "CPUID_7_0_EDX_AMX_INT8": "amx-int8", > "CPUID_7_1_EAX_AVX512_BF16": "avx512-bf16", > "CPUID_7_1_EAX_AVX_VNNI": "avx-vnni", > "CPUID_8000_0008_EBX_AMD_SSBD": "amd-ssbd", >diff --git a/src/cpu_map/x86_features.xml b/src/cpu_map/x86_features.xml >index 4cf3ff0804..102d39f626 100644 >--- a/src/cpu_map/x86_features.xml >+++ b/src/cpu_map/x86_features.xml >@@ -347,6 +347,15 @@ > <feature name='pconfig'> > <cpuid eax_in='0x07' ecx_in='0x00' edx='0x00040000'/> > </feature> >+ <feature name='amx-bf16'> >+ <cpuid eax_in='0x07' ecx_in='0x00' edx='0x00400000'/> >+ </feature> >+ <feature name='amx-tile'> >+ <cpuid eax_in='0x07' ecx_in='0x00' edx='0x01000000'/> >+ </feature> >+ <feature name='amx-int8'> >+ <cpuid eax_in='0x07' ecx_in='0x00' edx='0x02000000'/> >+ </feature> > <feature name='spec-ctrl'> > <cpuid eax_in='0x07' ecx_in='0x00' edx='0x04000000'/> > </feature> >-- >2.34.1 >
On Thu, Sep 22, 2022 at 04:35:51PM +0200, Martin Kletzander wrote: >On Tue, Sep 13, 2022 at 12:30:31PM -0700, Lin Yang wrote: >>AMX was introduced in QEMU commit 1f16764f7d4515bfd5e4ae0aae814fa280a7d0c8 >>and following commits. > >Some of them not pushed yet I believe: > >https://lists.gnu.org/archive/html/qemu-devel/2022-08/msg01936.html > >Otherwise the patch looks good (I'm not sure what's the planned usage of >the python script, but this patch just adds features that'd be missing >in the future, so I think it's fine). > >Reviewed-by: Martin Kletzander <mkletzan@redhat.com> > I still can't see the other commits in QEMU mentioning the other features, but I can't see how those missing could cause an issue, so in the end I removed the "and following commits". We also require a Signed-off-by for contributions as written in our contribution guidelines: https://libvirt.org/hacking.html If you say I can add one for you with the current e-mail address there's no need to resend this patch and I can add it before pushing. Thanks for the patch.
On Thu, Sep 22, 2022 at 04:35:51PM +0200, Martin Kletzander wrote: > >On Tue, Sep 13, 2022 at 12:30:31PM -0700, Lin Yang wrote: > >>AMX was introduced in QEMU commit 1f16764f7d4515bfd5e4ae0aae814fa280a7d0c8 > >>and following commits. > > > >Some of them not pushed yet I believe: > > > >https://lists.gnu.org/archive/html/qemu-devel/2022-08/msg01936.html > > > >Otherwise the patch looks good (I'm not sure what's the planned usage of > >the python script, but this patch just adds features that'd be missing > >in the future, so I think it's fine). > > > >Reviewed-by: Martin Kletzander <mkletzan@redhat.com> > > > > I still can't see the other commits in QEMU mentioning the other > features, but I can't see how those missing could cause an issue, so in > the end I removed the "and following commits". > > We also require a Signed-off-by for contributions as written in our > contribution guidelines: > > https://libvirt.org/hacking.html > > If you say I can add one for you with the current e-mail address there's > no need to resend this patch and I can add it before pushing. > > Thanks for the patch. Totally okay to me. Please sign with “Lin Yang <lin.a.yang@intel.com>". Thank you so much for reviewing this patch. Thanks, Lin.
On Fri, Nov 04, 2022 at 05:11:10AM +0000, Yang, Lin A wrote: >On Thu, Sep 22, 2022 at 04:35:51PM +0200, Martin Kletzander wrote: >> >On Tue, Sep 13, 2022 at 12:30:31PM -0700, Lin Yang wrote: >> >>AMX was introduced in QEMU commit 1f16764f7d4515bfd5e4ae0aae814fa280a7d0c8 >> >>and following commits. >> > >> >Some of them not pushed yet I believe: >> > >> >https://lists.gnu.org/archive/html/qemu-devel/2022-08/msg01936.html >> > >> >Otherwise the patch looks good (I'm not sure what's the planned usage of >> >the python script, but this patch just adds features that'd be missing >> >in the future, so I think it's fine). >> > >> >Reviewed-by: Martin Kletzander <mkletzan@redhat.com> >> > >> >> I still can't see the other commits in QEMU mentioning the other >> features, but I can't see how those missing could cause an issue, so in >> the end I removed the "and following commits". >> >> We also require a Signed-off-by for contributions as written in our >> contribution guidelines: >> >> https://libvirt.org/hacking.html >> >> If you say I can add one for you with the current e-mail address there's >> no need to resend this patch and I can add it before pushing. >> >> Thanks for the patch. > >Totally okay to me. Please sign with “Lin Yang <lin.a.yang@intel.com>". >Thank you so much for reviewing this patch. > Thanks, pushed. >Thanks, >Lin.
© 2016 - 2024 Red Hat, Inc.