diff mbox

[3.8.y.z,extended,stable] Patch "x86, AVX-512: AVX-512 Feature Detection" has been added to staging queue

Message ID 1397777393-4492-1-git-send-email-kamal@canonical.com
State New
Headers show

Commit Message

Kamal Mostafa April 17, 2014, 11:29 p.m. UTC 
This is a note to let you know that I have just added a patch titled

    x86, AVX-512: AVX-512 Feature Detection

to the linux-3.8.y-queue branch of the 3.8.y.z extended stable tree 
which can be found at:

 http://kernel.ubuntu.com/git?p=ubuntu/linux.git;a=shortlog;h=refs/heads/linux-3.8.y-queue

This patch is scheduled to be released in version 3.8.13.22.

If you, or anyone else, feels it should not be added to this tree, please 
reply to this email.

For more information about the 3.8.y.z tree, see
https://wiki.ubuntu.com/Kernel/Dev/ExtendedStable

Thanks.
-Kamal

------

From 072af6571a063dd603521b6c947cf094b7d347a1 Mon Sep 17 00:00:00 2001
From: Fenghua Yu <fenghua.yu@intel.com>
Date: Thu, 20 Feb 2014 13:24:50 -0800
Subject: x86, AVX-512: AVX-512 Feature Detection

commit 8e5780fdeef7dc490b3f0b3a62704593721fa4f3 upstream.

AVX-512 is an extention of AVX2. Its spec can be found at:
http://download-software.intel.com/sites/default/files/managed/71/2e/319433-017.pdf

This patch detects AVX-512 features by CPUID.

Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Link: http://lkml.kernel.org/r/1392931491-33237-1-git-send-email-fenghua.yu@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Signed-off-by: Kamal Mostafa <kamal@canonical.com>
---
 arch/x86/include/asm/cpufeature.h | 4 ++++
 1 file changed, 4 insertions(+)

--
1.9.1

Comments

Fenghua Yu April 18, 2014, 2:13 a.m. UTC | #1 
Hi, Kamal,

Except the patch in your email, there is another patch:
https://lkml.org/lkml/2014/2/20/546

Please add this patch as well if not done yet.

Thanks.

-Fenghua

> -----Original Message-----
> From: Kamal Mostafa [mailto:kamal@canonical.com]
> Sent: Thursday, April 17, 2014 4:30 PM
> To: Yu, Fenghua
> Cc: H. Peter Anvin; Kamal Mostafa; kernel-team@lists.ubuntu.com
> Subject: [3.8.y.z extended stable] Patch "x86, AVX-512: AVX-512 Feature
> Detection" has been added to staging queue
> 
> This is a note to let you know that I have just added a patch titled
> 
>     x86, AVX-512: AVX-512 Feature Detection
> 
> to the linux-3.8.y-queue branch of the 3.8.y.z extended stable tree
> which can be found at:
> 
> 
> http://kernel.ubuntu.com/git?p=ubuntu/linux.git;a=shortlog;h=refs/heads
> /linux-3.8.y-queue
> 
> This patch is scheduled to be released in version 3.8.13.22.
> 
> If you, or anyone else, feels it should not be added to this tree,
> please
> reply to this email.
> 
> For more information about the 3.8.y.z tree, see
> https://wiki.ubuntu.com/Kernel/Dev/ExtendedStable
> 
> Thanks.
> -Kamal
> 
> ------
> 
> From 072af6571a063dd603521b6c947cf094b7d347a1 Mon Sep 17 00:00:00 2001
> From: Fenghua Yu <fenghua.yu@intel.com>
> Date: Thu, 20 Feb 2014 13:24:50 -0800
> Subject: x86, AVX-512: AVX-512 Feature Detection
> 
> commit 8e5780fdeef7dc490b3f0b3a62704593721fa4f3 upstream.
> 
> AVX-512 is an extention of AVX2. Its spec can be found at:
> http://download-
> software.intel.com/sites/default/files/managed/71/2e/319433-017.pdf
> 
> This patch detects AVX-512 features by CPUID.
> 
> Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
> Link: http://lkml.kernel.org/r/1392931491-33237-1-git-send-email-
> fenghua.yu@intel.com
> Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
> Signed-off-by: Kamal Mostafa <kamal@canonical.com>
> ---
>  arch/x86/include/asm/cpufeature.h | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/arch/x86/include/asm/cpufeature.h
> b/arch/x86/include/asm/cpufeature.h
> index d3f0db1..0480fb9 100644
> --- a/arch/x86/include/asm/cpufeature.h
> +++ b/arch/x86/include/asm/cpufeature.h
> @@ -211,9 +211,13 @@
>  #define X86_FEATURE_ERMS	(9*32+ 9) /* Enhanced REP MOVSB/STOSB */
>  #define X86_FEATURE_INVPCID	(9*32+10) /* Invalidate Processor Context
> ID */
>  #define X86_FEATURE_RTM		(9*32+11) /* Restricted
> Transactional Memory */
> +#define X86_FEATURE_AVX512F	(9*32+16) /* AVX-512 Foundation */
>  #define X86_FEATURE_RDSEED	(9*32+18) /* The RDSEED instruction */
>  #define X86_FEATURE_ADX		(9*32+19) /* The ADCX and ADOX
> instructions */
>  #define X86_FEATURE_SMAP	(9*32+20) /* Supervisor Mode Access
> Prevention */
> +#define X86_FEATURE_AVX512PF	(9*32+26) /* AVX-512 Prefetch */
> +#define X86_FEATURE_AVX512ER	(9*32+27) /* AVX-512 Exponential and
> Reciprocal */
> +#define X86_FEATURE_AVX512CD	(9*32+28) /* AVX-512 Conflict Detection
> */
> 
>  #if defined(__KERNEL__) && !defined(__ASSEMBLY__)
> 
> --
> 1.9.1
Kamal Mostafa April 18, 2014, 5:43 p.m. UTC | #2 
On Fri, 2014-04-18 at 02:13 +0000, Yu, Fenghua wrote:
> Hi, Kamal,
> 
> Except the patch in your email, there is another patch:
> https://lkml.org/lkml/2014/2/20/546
> 
> Please add this patch as well if not done yet.


Hi Fenghua-

I didn't pick up that additional patch:

  c2bc11f x86, AVX-512: Enable AVX-512 States Context Switch

because its only (?) purpose seemed to be to modify the definition of
XSTATE_LAZY, which doesn't exist in 3.8-stable[0].

Would just adding the additional XSTATE_ definitions be useful for
3.8-stable for some other reason?

Thanks for reviewing this!

 -Kamal

[0] http://kernel.ubuntu.com/git?p=ubuntu/linux.git;a=blob;f=arch/x86/include/asm/xsave.h;h=0415cdabb5a663d1684284525a3fe24b7b3cd359;hb=refs/heads/linux-3.8.y-queue


> > -----Original Message-----
> > From: Kamal Mostafa [mailto:kamal@canonical.com]
> > Sent: Thursday, April 17, 2014 4:30 PM
> > To: Yu, Fenghua
> > Cc: H. Peter Anvin; Kamal Mostafa; kernel-team@lists.ubuntu.com
> > Subject: [3.8.y.z extended stable] Patch "x86, AVX-512: AVX-512 Feature
> > Detection" has been added to staging queue
> > 
> > This is a note to let you know that I have just added a patch titled
> > 
> >     x86, AVX-512: AVX-512 Feature Detection
> > 
> > to the linux-3.8.y-queue branch of the 3.8.y.z extended stable tree
> > which can be found at:
> > 
> > 
> > http://kernel.ubuntu.com/git?p=ubuntu/linux.git;a=shortlog;h=refs/heads
> > /linux-3.8.y-queue
> > 
> > This patch is scheduled to be released in version 3.8.13.22.
> > 
> > If you, or anyone else, feels it should not be added to this tree,
> > please
> > reply to this email.
> > 
> > For more information about the 3.8.y.z tree, see
> > https://wiki.ubuntu.com/Kernel/Dev/ExtendedStable
> > 
> > Thanks.
> > -Kamal
> > 
> > ------
> > 
> > From 072af6571a063dd603521b6c947cf094b7d347a1 Mon Sep 17 00:00:00 2001
> > From: Fenghua Yu <fenghua.yu@intel.com>
> > Date: Thu, 20 Feb 2014 13:24:50 -0800
> > Subject: x86, AVX-512: AVX-512 Feature Detection
> > 
> > commit 8e5780fdeef7dc490b3f0b3a62704593721fa4f3 upstream.
> > 
> > AVX-512 is an extention of AVX2. Its spec can be found at:
> > http://download-
> > software.intel.com/sites/default/files/managed/71/2e/319433-017.pdf
> > 
> > This patch detects AVX-512 features by CPUID.
> > 
> > Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
> > Link: http://lkml.kernel.org/r/1392931491-33237-1-git-send-email-
> > fenghua.yu@intel.com
> > Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
> > Signed-off-by: Kamal Mostafa <kamal@canonical.com>
> > ---
> >  arch/x86/include/asm/cpufeature.h | 4 ++++
> >  1 file changed, 4 insertions(+)
> > 
> > diff --git a/arch/x86/include/asm/cpufeature.h
> > b/arch/x86/include/asm/cpufeature.h
> > index d3f0db1..0480fb9 100644
> > --- a/arch/x86/include/asm/cpufeature.h
> > +++ b/arch/x86/include/asm/cpufeature.h
> > @@ -211,9 +211,13 @@
> >  #define X86_FEATURE_ERMS	(9*32+ 9) /* Enhanced REP MOVSB/STOSB */
> >  #define X86_FEATURE_INVPCID	(9*32+10) /* Invalidate Processor Context
> > ID */
> >  #define X86_FEATURE_RTM		(9*32+11) /* Restricted
> > Transactional Memory */
> > +#define X86_FEATURE_AVX512F	(9*32+16) /* AVX-512 Foundation */
> >  #define X86_FEATURE_RDSEED	(9*32+18) /* The RDSEED instruction */
> >  #define X86_FEATURE_ADX		(9*32+19) /* The ADCX and ADOX
> > instructions */
> >  #define X86_FEATURE_SMAP	(9*32+20) /* Supervisor Mode Access
> > Prevention */
> > +#define X86_FEATURE_AVX512PF	(9*32+26) /* AVX-512 Prefetch */
> > +#define X86_FEATURE_AVX512ER	(9*32+27) /* AVX-512 Exponential and
> > Reciprocal */
> > +#define X86_FEATURE_AVX512CD	(9*32+28) /* AVX-512 Conflict Detection
> > */
> > 
> >  #if defined(__KERNEL__) && !defined(__ASSEMBLY__)
> > 
> > --
> > 1.9.1
>
Fenghua Yu April 22, 2014, 9:01 p.m. UTC | #3 
> From: Kamal Mostafa [mailto:kamal@canonical.com]
> Sent: Friday, April 18, 2014 10:44 AM
> To: Yu, Fenghua
> Cc: H. Peter Anvin; kernel-team@lists.ubuntu.com
> Subject: Re: [3.8.y.z extended stable] Patch "x86, AVX-512: AVX-512
> Feature Detection" has been added to staging queue
> 
> On Fri, 2014-04-18 at 02:13 +0000, Yu, Fenghua wrote:
> > Hi, Kamal,
> >
> > Except the patch in your email, there is another patch:
> > https://lkml.org/lkml/2014/2/20/546
> >
> > Please add this patch as well if not done yet.
> 
> 
> Hi Fenghua-
> 
> I didn't pick up that additional patch:

Actually this patch is important for AVX 512. Without this patch, AVX 512 applications can not run because their context switch will fail.

> 
>   c2bc11f x86, AVX-512: Enable AVX-512 States Context Switch
> 
> because its only (?) purpose seemed to be to modify the definition of
> XSTATE_LAZY, which doesn't exist in 3.8-stable[0].

The point is XCNTXT_MASK has XSTATE_LAZY. So the patch actually changes XCNTXT_MASK and thus enables AVX 512 context switch.

For 3.8 kernel, you can directly add the three AVX-512 bits to XCNTXT_MASK:

-#define XCNTXT_MASK (XSTATE_FP | XSTATE_SSE | XSTATE_YMM)
+#define XCNTXT_MASK (XSTATE_FP | XSTATE_SSE | XSTATE_YMM
+	| XSTATE_OPMASK | XSTATE_ZMM_Hi256 | XSTATE_Hi16_ZMM)

This should work to enable AVX 512 context switch.

> 
> Would just adding the additional XSTATE_ definitions be useful for
> 3.8-stable for some other reason?

If you want 3.8-stable to support AVX 512 feature, then this additional patch is a must-have patch. Otherwise, it's unnecessary to add this patch.

> 
> Thanks for reviewing this!
> 
>  -Kamal
>
Kamal Mostafa April 23, 2014, 5:22 p.m. UTC | #4 
On Tue, 2014-04-22 at 21:01 +0000, Yu, Fenghua wrote:
> > From: Kamal Mostafa [mailto:kamal@canonical.com]
> > Sent: Friday, April 18, 2014 10:44 AM
> > To: Yu, Fenghua
> > Cc: H. Peter Anvin; kernel-team@lists.ubuntu.com
> > Subject: Re: [3.8.y.z extended stable] Patch "x86, AVX-512: AVX-512
> > Feature Detection" has been added to staging queue
> > 
> > On Fri, 2014-04-18 at 02:13 +0000, Yu, Fenghua wrote:
> > > Hi, Kamal,
> > >
> > > Except the patch in your email, there is another patch:
> > > https://lkml.org/lkml/2014/2/20/546
> > >
> > > Please add this patch as well if not done yet.
> > 
> > 
> > Hi Fenghua-
> > 
> > I didn't pick up that additional patch:
> 
> Actually this patch is important for AVX 512. Without this patch, AVX 512 applications can not run because their context switch will fail.
> 
> > 
> >   c2bc11f x86, AVX-512: Enable AVX-512 States Context Switch
> > 
> > because its only (?) purpose seemed to be to modify the definition of
> > XSTATE_LAZY, which doesn't exist in 3.8-stable[0].
> 
> The point is XCNTXT_MASK has XSTATE_LAZY. So the patch actually changes XCNTXT_MASK and thus enables AVX 512 context switch.
> 
> For 3.8 kernel, you can directly add the three AVX-512 bits to XCNTXT_MASK:
> 
> -#define XCNTXT_MASK (XSTATE_FP | XSTATE_SSE | XSTATE_YMM)
> +#define XCNTXT_MASK (XSTATE_FP | XSTATE_SSE | XSTATE_YMM
> +	| XSTATE_OPMASK | XSTATE_ZMM_Hi256 | XSTATE_Hi16_ZMM)
> 
> This should work to enable AVX 512 context switch.


Very good Fenghua -- I'll apply that to XCNTXT_MASK.

Thanks for your help!

 -Kamal
diff mbox

Patch

diff --git a/arch/x86/include/asm/cpufeature.h b/arch/x86/include/asm/cpufeature.h
index d3f0db1..0480fb9 100644
--- a/arch/x86/include/asm/cpufeature.h
+++ b/arch/x86/include/asm/cpufeature.h
@@ -211,9 +211,13 @@ 
 #define X86_FEATURE_ERMS	(9*32+ 9) /* Enhanced REP MOVSB/STOSB */
 #define X86_FEATURE_INVPCID	(9*32+10) /* Invalidate Processor Context ID */
 #define X86_FEATURE_RTM		(9*32+11) /* Restricted Transactional Memory */
+#define X86_FEATURE_AVX512F	(9*32+16) /* AVX-512 Foundation */
 #define X86_FEATURE_RDSEED	(9*32+18) /* The RDSEED instruction */
 #define X86_FEATURE_ADX		(9*32+19) /* The ADCX and ADOX instructions */
 #define X86_FEATURE_SMAP	(9*32+20) /* Supervisor Mode Access Prevention */
+#define X86_FEATURE_AVX512PF	(9*32+26) /* AVX-512 Prefetch */
+#define X86_FEATURE_AVX512ER	(9*32+27) /* AVX-512 Exponential and Reciprocal */
+#define X86_FEATURE_AVX512CD	(9*32+28) /* AVX-512 Conflict Detection */

 #if defined(__KERNEL__) && !defined(__ASSEMBLY__)