Skip to contentSkip to conversation

x86/cpufeatures: Move the definition of X86_FEATURE_AMX_* to the word 18

PATCH In review Like Xu 4 participants 10 messages Jan 17, 2022 – Feb 8, 2022 x86

lore
Share & tools
lore

Like Xu

author
#0 lore raw

PATCH 1/1 x86/cpufeatures: Move the definition of X86_FEATURE_AMX_* to the word 18

From: Like Xu <[email protected]>

We have defined the word 18 for Intel-defined CPU features from CPUID level 0x00000007:0 (EDX). Let's move the definitions of X86_FEATURE_AMX_* to the right entry to prevent misinterpretation. No functional change intended.

---
 arch/x86/include/asm/cpufeatures.h | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
arch/x86/include/asm/cpufeatures.h +3 -3
index 6db4e2932b3d..5cd22090e53d 100644
@@ -299,9 +299,6 @@
 /* Intel-defined CPU features, CPUID level 0x00000007:1 (EAX), word 12 */
 #define X86_FEATURE_AVX_VNNI		(12*32+ 4) /* AVX VNNI instructions */
 #define X86_FEATURE_AVX512_BF16		(12*32+ 5) /* AVX512 BFLOAT16 instructions */
-#define X86_FEATURE_AMX_BF16		(18*32+22) /* AMX bf16 Support */
-#define X86_FEATURE_AMX_TILE		(18*32+24) /* AMX tile Support */
-#define X86_FEATURE_AMX_INT8		(18*32+25) /* AMX int8 Support */
 
 /* AMD-defined CPU features, CPUID level 0x80000008 (EBX), word 13 */
 #define X86_FEATURE_CLZERO		(13*32+ 0) /* CLZERO instruction */
@@ -390,7 +387,10 @@
 #define X86_FEATURE_TSXLDTRK		(18*32+16) /* TSX Suspend Load Address Tracking */
 #define X86_FEATURE_PCONFIG		(18*32+18) /* Intel PCONFIG */
 #define X86_FEATURE_ARCH_LBR		(18*32+19) /* Intel ARCH LBR */
+#define X86_FEATURE_AMX_BF16		(18*32+22) /* AMX bf16 Support */
 #define X86_FEATURE_AVX512_FP16		(18*32+23) /* AVX512 FP16 */
+#define X86_FEATURE_AMX_TILE		(18*32+24) /* AMX tile Support */
+#define X86_FEATURE_AMX_INT8		(18*32+25) /* AMX int8 Support */
 #define X86_FEATURE_SPEC_CTRL		(18*32+26) /* "" Speculation Control (IBRS + IBPB) */
 #define X86_FEATURE_INTEL_STIBP		(18*32+27) /* "" Single Thread Indirect Branch Predictors */
 #define X86_FEATURE_FLUSH_L1D		(18*32+28) /* Flush L1D cache */
Signature · 1 line
2.33.1
Reviewed-by on v1 1/1
#1 lore raw
Quoted 3 lines · Like Xu

We have defined the word 18 for Intel-defined CPU features from CPUID level 0x00000007:0 (EDX). Let's move the definitions of X86_FEATURE_AMX_* to the right entry to prevent misinterpretation. No functional change intended.

#2 lore raw
In reply to Like Xu

What tree is this against? I see BF16 and INT8 in some old versions of Chang's patches, but not current kernels. All I see right now in tip/master is:

Quoted 1 line

#define X86_FEATURE_AMX_TILE (18*32+24) /* AMX tile ...

It's still in the wrong spot, but the other two features aren't there.

Quoted 1 line

We have defined the word 18 for Intel-defined CPU features from CPUID level> 0x00000007:0 (EDX). Let's move the definitions of X86_FEATURE_AMX_* to

the> right entry to prevent misinterpretation. No functional change intended.
Please, no "we's" in changelogs. Don't say, "let's move". Just say:
"Move..."

The subject could probably also be trimmed a bit. Perhaps:

x86/cpu: Move AMX CPU feature defines to correct word location

#3 lore raw
Quoted 7 lines · Dave Hansen

What tree is this against? I see BF16 and INT8 in some old versions of Chang's patches, but not current kernels. All I see right now in tip/master is:

Quoted 1 line

#define X86_FEATURE_AMX_TILE (18*32+24) /* AMX tile ...

It's still in the wrong spot, but the other two features aren't there.

It was added for the KVM side of AMX (commit 690a757d610e, "kvm: x86: Add CPUID support for Intel AMX") and is in Linus's tree.

Paolo

Quoted 11 lines
Quoted 1 line

We have defined the word 18 for Intel-defined CPU features from CPUID level> 0x00000007:0 (EDX). Let's move the definitions of X86_FEATURE_AMX_* to

the> right entry to prevent misinterpretation. No functional change intended.
Please, no "we's" in changelogs. Don't say, "let's move". Just say:
"Move..."

The subject could probably also be trimmed a bit. Perhaps:

x86/cpu: Move AMX CPU feature defines to correct word location

author Reviewed-by Paolo Bonzini
#4 lore raw
Quoted 5 lines · Paolo Bonzini
Quoted 4 lines · Dave Hansen

What tree is this against?  I see BF16 and INT8 in some old versions of Chang's patches, but not current kernels.  All I see right now in tip/master is:

Quoted 20 lines
Quoted 3 lines
Quoted 1 line

#define X86_FEATURE_AMX_TILE            (18*32+24) /* AMX tile ...

It's still in the wrong spot, but the other two features aren't there.

It was added for the KVM side of AMX (commit 690a757d610e, "kvm: x86: Add CPUID support for Intel AMX") and is in Linus's tree.

Paolo

Quoted 10 lines
Quoted 1 line

We have defined the word 18 for Intel-defined CPU features from CPUID level> 0x00000007:0 (EDX). Let's move the definitions of X86_FEATURE_AMX_* to

the> right entry to prevent misinterpretation. No functional change intended.
Please, no "we's" in changelogs.  Don't say, "let's move".  Just say:
"Move..." >>
The subject could probably also be trimmed a bit.  Perhaps:

    x86/cpu: Move AMX CPU feature defines to correct word location

Thanks Dave and Paolo. Just for your convenience:

From 588c2221999c1f5860188a7cbaeb0d4f80c6d727 Mon Sep 17 00:00:00 2001
From: Like Xu <[email protected]>
Date: Mon, 17 Jan 2022 14:23:44 +0800
Subject: [PATCH v2] x86/cpufeatures: Move AMX CPU feature defines to correct
word location

From: Like Xu <[email protected]>

The word 18 for Intel-defined CPU features from CPUID level 0x00000007:0 (EDX) has been defined in the same file. Move the definitions of X86_FEATURE_AMX_* to the right entry to prevent misinterpretation. No functional change intended.

---
v1 -> v2 Changelog:
- Refine the commit message and subject; (Dave)

 arch/x86/include/asm/cpufeatures.h | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
arch/x86/include/asm/cpufeatures.h +3 -3
index 6db4e2932b3d..5cd22090e53d 100644
@@ -299,9 +299,6 @@
 /* Intel-defined CPU features, CPUID level 0x00000007:1 (EAX), word 12 */
 #define X86_FEATURE_AVX_VNNI		(12*32+ 4) /* AVX VNNI instructions */
 #define X86_FEATURE_AVX512_BF16		(12*32+ 5) /* AVX512 BFLOAT16 instructions */
-#define X86_FEATURE_AMX_BF16		(18*32+22) /* AMX bf16 Support */
-#define X86_FEATURE_AMX_TILE		(18*32+24) /* AMX tile Support */
-#define X86_FEATURE_AMX_INT8		(18*32+25) /* AMX int8 Support */
 
 /* AMD-defined CPU features, CPUID level 0x80000008 (EBX), word 13 */
 #define X86_FEATURE_CLZERO		(13*32+ 0) /* CLZERO instruction */
@@ -390,7 +387,10 @@
 #define X86_FEATURE_TSXLDTRK		(18*32+16) /* TSX Suspend Load Address Tracking */
 #define X86_FEATURE_PCONFIG		(18*32+18) /* Intel PCONFIG */
 #define X86_FEATURE_ARCH_LBR		(18*32+19) /* Intel ARCH LBR */
+#define X86_FEATURE_AMX_BF16		(18*32+22) /* AMX bf16 Support */
 #define X86_FEATURE_AVX512_FP16		(18*32+23) /* AVX512 FP16 */
+#define X86_FEATURE_AMX_TILE		(18*32+24) /* AMX tile Support */
+#define X86_FEATURE_AMX_INT8		(18*32+25) /* AMX int8 Support */
 #define X86_FEATURE_SPEC_CTRL		(18*32+26) /* "" Speculation Control (IBRS + 

IBPB) */
#define X86_FEATURE_INTEL_STIBP (18*32+27) /* "" Single Thread Indirect
Branch Predictors */
#define X86_FEATURE_FLUSH_L1D (18*32+28) /* Flush L1D cache */

Signature · 1 line
2.33.1
maintainer
#5 lore raw
Quoted 14 lines · Like Xu

Thanks Dave and Paolo. Just for your convenience:

From 588c2221999c1f5860188a7cbaeb0d4f80c6d727 Mon Sep 17 00:00:00 2001
From: Like Xu <[email protected]>
Date: Mon, 17 Jan 2022 14:23:44 +0800
Subject: [PATCH v2] x86/cpufeatures: Move AMX CPU feature defines to correct
word location

From: Like Xu <[email protected]>

The word 18 for Intel-defined CPU features from CPUID level 0x00000007:0 (EDX) has been defined in the same file. Move the definitions of X86_FEATURE_AMX_* to the right entry to prevent misinterpretation. No functional change intended.

Quoted 8 lines

---
v1 -> v2 Changelog:
- Refine the commit message and subject; (Dave)

 arch/x86/include/asm/cpufeatures.h | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
$ test-apply.sh /tmp/urgent.01
checking file arch/x86/include/asm/cpufeatures.h
Hunk #1 FAILED at 299.
patch: **** malformed patch at line 45: IBPB) */

Don't ever send patches from

User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:91.0) Gecko/20100101 Thunderbird/91.5.0

That thing mangles them.

See Documentation/process/email-clients.rst for further info.

Signature · 4 lines
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette
author
#6 lore raw
Quoted 21 lines · Borislav Petkov
Quoted 20 lines · Like Xu

Thanks Dave and Paolo. Just for your convenience:

From 588c2221999c1f5860188a7cbaeb0d4f80c6d727 Mon Sep 17 00:00:00 2001
From: Like Xu <[email protected]>
Date: Mon, 17 Jan 2022 14:23:44 +0800
Subject: [PATCH v2] x86/cpufeatures: Move AMX CPU feature defines to correct
word location

From: Like Xu <[email protected]>

The word 18 for Intel-defined CPU features from CPUID level 0x00000007:0 (EDX) has been defined in the same file. Move the definitions of X86_FEATURE_AMX_* to the right entry to prevent misinterpretation. No functional change intended.

---
v1 -> v2 Changelog:
- Refine the commit message and subject; (Dave)

Quoted 4 lines
Quoted 2 lines
  arch/x86/include/asm/cpufeatures.h | 6 +++---
  1 file changed, 3 insertions(+), 3 deletions(-)
$ test-apply.sh /tmp/urgent.01

Emm, are you willing to make "test-apply.sh" public (or have done) and let others benefit ?

Quoted 5 lines

checking file arch/x86/include/asm/cpufeatures.h
Hunk #1 FAILED at 299.
patch: **** malformed patch at line 45: IBPB) */

Don't ever send patches from

Got you.

Quoted 7 lines

User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:91.0) Gecko/20100101 Thunderbird/91.5.0

That thing mangles them.

See Documentation/process/email-clients.rst for further info.

Sorry for the late reply, and here's a new version from git-send-email:

https://lore.kernel.org/lkml/[email protected]/

maintainer
#7 lore raw
Quoted 2 lines · Like Xu

Emm, are you willing to make "test-apply.sh" public (or have done) and let others benefit ?

It is nothing special - just a simple script which does first

patch -p1 --dry-run -l -i [patch file]

to see whether it would even apply.

Quoted 1 line

Sorry for the late reply, and here's a new version from git-send-email:

Yeah, someone already took care of this:

https://git.kernel.org/tip/ae75fa54228162ecd65341f9780886f21f557cc4

Signature · 4 lines
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette
author
#8 lore raw
Quoted 8 lines · Borislav Petkov
Quoted 2 lines · Like Xu

Emm, are you willing to make "test-apply.sh" public (or have done) and let others benefit ?

It is nothing special - just a simple script which does first

patch -p1 --dry-run -l -i [patch file]

Quoted 8 lines

to see whether it would even apply.

Quoted 1 line

Sorry for the late reply, and here's a new version from git-send-email:

Yeah, someone already took care of this:

https://git.kernel.org/tip/ae75fa54228162ecd65341f9780886f21f557cc4

Uh, it's Jim and we have worked together on many KVM issues.

BTW, I'm not sure why we prefer:

#define X86_FEATURE_AVX512_FP16 (18*32+23) /* AVX512 FP16 */
#define X86_FEATURE_SPEC_CTRL (18*32+26) /* "" Speculation Control (IBRS + IBPB) */
+#define X86_FEATURE_AMX_BF16 (18*32+22) /* AMX bf16 Support */
+#define X86_FEATURE_AMX_TILE (18*32+24) /* AMX tile Support */
+#define X86_FEATURE_AMX_INT8 (18*32+25) /* AMX int8 Support */

rather than:

+#define X86_FEATURE_AMX_BF16 (18*32+22) /* AMX bf16 Support */
#define X86_FEATURE_AVX512_FP16 (18*32+23) /* AVX512 FP16 */
+#define X86_FEATURE_AMX_TILE (18*32+24) /* AMX tile Support */
+#define X86_FEATURE_AMX_INT8 (18*32+25) /* AMX int8 Support */

It would make more sense to put them in order, what do you think? Need v3 to sort it ?

Thanks,
Like Xu

maintainer
#9 lore raw
Quoted 1 line · Like Xu

It would make more sense to put them in order, what do you think?

Ah, good catch, thanks.

I'll edit the patch directly.

Signature · 4 lines
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette