x86/cpufeatures: Add AVX512 Bit Matrix Multiply (BMM) and Bit reversal support

[PATCH] x86/cpufeatures: Add AVX512 Bit Matrix Multiply (BMM) and Bit reversal support

Posted by Nikunj A Dadhania 2 days, 13 hours ago

Add support for AVX512 Bit Matrix Multiply (BMM) and Bit Reversal
instructions, a feature that enables bit matrix multiply operations and
bit reversal, which is exposed via CPUID leaf 0x80000021_EAX[23].

Expose the support to guests when available by including it in the CPUID
leaf 0x80000021_EAX feature list.

While at it, reorder PREFETCHI to match the bit position order in CPUID
leaf 0x80000021_EAX for better organization.

Signed-off-by: Nikunj A Dadhania <nikunj@amd.com>
---

AMD64 Bit Matrix Multiply and Bit Reversal Instructions
Publication #69192 Revision: 1.00
Issue Date: January 2026

https://docs.amd.com/v/u/en-US/69192-PUB
---
 arch/x86/include/asm/cpufeatures.h | 1 +
 arch/x86/kvm/cpuid.c               | 3 ++-
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/arch/x86/include/asm/cpufeatures.h b/arch/x86/include/asm/cpufeatures.h
index c3b53beb1300..2f1583c4bdc0 100644
--- a/arch/x86/include/asm/cpufeatures.h
+++ b/arch/x86/include/asm/cpufeatures.h
@@ -472,6 +472,7 @@
 #define X86_FEATURE_GP_ON_USER_CPUID	(20*32+17) /* User CPUID faulting */
 
 #define X86_FEATURE_PREFETCHI		(20*32+20) /* Prefetch Data/Instruction to Cache Level */
+#define X86_FEATURE_AVX512_BMM		(20*32+23) /* AVX512 Bit Matrix Multiply instructions */
 #define X86_FEATURE_SBPB		(20*32+27) /* Selective Branch Prediction Barrier */
 #define X86_FEATURE_IBPB_BRTYPE		(20*32+28) /* MSR_PRED_CMD[IBPB] flushes all branch type predictions */
 #define X86_FEATURE_SRSO_NO		(20*32+29) /* CPU is not affected by SRSO */
diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
index 88a5426674a1..b36e8f10f509 100644
--- a/arch/x86/kvm/cpuid.c
+++ b/arch/x86/kvm/cpuid.c
@@ -1218,11 +1218,12 @@ void kvm_set_cpu_caps(void)
 		F(NULL_SEL_CLR_BASE),
 		/* UpperAddressIgnore */
 		F(AUTOIBRS),
-		F(PREFETCHI),
 		EMULATED_F(NO_SMM_CTL_MSR),
 		/* PrefetchCtlMsr */
 		/* GpOnUserCpuid */
 		/* EPSF */
+		F(PREFETCHI),
+		F(AVX512_BMM),
 		SYNTHESIZED_F(SBPB),
 		SYNTHESIZED_F(IBPB_BRTYPE),
 		SYNTHESIZED_F(SRSO_NO),

base-commit: e89f0e9a0a007e8c3afb8ecd739c0b3255422b00
-- 
2.48.1

Re: [PATCH] x86/cpufeatures: Add AVX512 Bit Matrix Multiply (BMM) and Bit reversal support

Posted by Sean Christopherson 2 days, 3 hours ago

KVM: x86: Advertise AVX512 Bit Matrix Multiply (BMM) to userspace

Because the primary focus of the change is quite clearly to add KVM support,
not to simply define the feature flag.

On Thu, Feb 05, 2026, Nikunj A Dadhania wrote:
> Add support for AVX512 Bit Matrix Multiply (BMM) and Bit Reversal
> instructions, a feature that enables bit matrix multiply operations and
> bit reversal, which is exposed via CPUID leaf 0x80000021_EAX[23].
> 
> Expose the support to guests when available by including it in the CPUID

Advertise to userspace.  The VMM decides whether or not to enumerate features to
guests.

> leaf 0x80000021_EAX feature list.
> 
> While at it, reorder PREFETCHI to match the bit position order in CPUID
> leaf 0x80000021_EAX for better organization.
> 
> Signed-off-by: Nikunj A Dadhania <nikunj@amd.com>
> ---
> 
> AMD64 Bit Matrix Multiply and Bit Reversal Instructions
> Publication #69192 Revision: 1.00
> Issue Date: January 2026
> 
> https://docs.amd.com/v/u/en-US/69192-PUB
> ---
>  arch/x86/include/asm/cpufeatures.h | 1 +
>  arch/x86/kvm/cpuid.c               | 3 ++-
>  2 files changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/x86/include/asm/cpufeatures.h b/arch/x86/include/asm/cpufeatures.h
> index c3b53beb1300..2f1583c4bdc0 100644
> --- a/arch/x86/include/asm/cpufeatures.h
> +++ b/arch/x86/include/asm/cpufeatures.h
> @@ -472,6 +472,7 @@
>  #define X86_FEATURE_GP_ON_USER_CPUID	(20*32+17) /* User CPUID faulting */
>  
>  #define X86_FEATURE_PREFETCHI		(20*32+20) /* Prefetch Data/Instruction to Cache Level */
> +#define X86_FEATURE_AVX512_BMM		(20*32+23) /* AVX512 Bit Matrix Multiply instructions */
>  #define X86_FEATURE_SBPB		(20*32+27) /* Selective Branch Prediction Barrier */
>  #define X86_FEATURE_IBPB_BRTYPE		(20*32+28) /* MSR_PRED_CMD[IBPB] flushes all branch type predictions */
>  #define X86_FEATURE_SRSO_NO		(20*32+29) /* CPU is not affected by SRSO */
> diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
> index 88a5426674a1..b36e8f10f509 100644
> --- a/arch/x86/kvm/cpuid.c
> +++ b/arch/x86/kvm/cpuid.c
> @@ -1218,11 +1218,12 @@ void kvm_set_cpu_caps(void)
>  		F(NULL_SEL_CLR_BASE),
>  		/* UpperAddressIgnore */
>  		F(AUTOIBRS),
> -		F(PREFETCHI),
>  		EMULATED_F(NO_SMM_CTL_MSR),
>  		/* PrefetchCtlMsr */
>  		/* GpOnUserCpuid */
>  		/* EPSF */
> +		F(PREFETCHI),
> +		F(AVX512_BMM),
>  		SYNTHESIZED_F(SBPB),
>  		SYNTHESIZED_F(IBPB_BRTYPE),
>  		SYNTHESIZED_F(SRSO_NO),
> 
> base-commit: e89f0e9a0a007e8c3afb8ecd739c0b3255422b00
> -- 
> 2.48.1
>

Re: [PATCH] x86/cpufeatures: Add AVX512 Bit Matrix Multiply (BMM) and Bit reversal support

Posted by Nikunj A. Dadhania 1 day, 14 hours ago


On 2/5/2026 8:40 PM, Sean Christopherson wrote:
> KVM: x86: Advertise AVX512 Bit Matrix Multiply (BMM) to userspace
> 
> Because the primary focus of the change is quite clearly to add KVM support,
> not to simply define the feature flag.

Sure, will update.

> 
> On Thu, Feb 05, 2026, Nikunj A Dadhania wrote:
>> Add support for AVX512 Bit Matrix Multiply (BMM) and Bit Reversal
>> instructions, a feature that enables bit matrix multiply operations and
>> bit reversal, which is exposed via CPUID leaf 0x80000021_EAX[23].
>>
>> Expose the support to guests when available by including it in the CPUID
> 
> Advertise to userspace.  The VMM decides whether or not to enumerate features to
> guests.

Ack.

Regards
Nikunj

> 
>> leaf 0x80000021_EAX feature list.
>>
>> While at it, reorder PREFETCHI to match the bit position order in CPUID
>> leaf 0x80000021_EAX for better organization.
>>
>> Signed-off-by: Nikunj A Dadhania <nikunj@amd.com>
>> ---
>>
>> AMD64 Bit Matrix Multiply and Bit Reversal Instructions
>> Publication #69192 Revision: 1.00
>> Issue Date: January 2026
>>
>> https://docs.amd.com/v/u/en-US/69192-PUB
>> ---
>>  arch/x86/include/asm/cpufeatures.h | 1 +
>>  arch/x86/kvm/cpuid.c               | 3 ++-
>>  2 files changed, 3 insertions(+), 1 deletion(-)
>>
>> diff --git a/arch/x86/include/asm/cpufeatures.h b/arch/x86/include/asm/cpufeatures.h
>> index c3b53beb1300..2f1583c4bdc0 100644
>> --- a/arch/x86/include/asm/cpufeatures.h
>> +++ b/arch/x86/include/asm/cpufeatures.h
>> @@ -472,6 +472,7 @@
>>  #define X86_FEATURE_GP_ON_USER_CPUID	(20*32+17) /* User CPUID faulting */
>>  
>>  #define X86_FEATURE_PREFETCHI		(20*32+20) /* Prefetch Data/Instruction to Cache Level */
>> +#define X86_FEATURE_AVX512_BMM		(20*32+23) /* AVX512 Bit Matrix Multiply instructions */
>>  #define X86_FEATURE_SBPB		(20*32+27) /* Selective Branch Prediction Barrier */
>>  #define X86_FEATURE_IBPB_BRTYPE		(20*32+28) /* MSR_PRED_CMD[IBPB] flushes all branch type predictions */
>>  #define X86_FEATURE_SRSO_NO		(20*32+29) /* CPU is not affected by SRSO */
>> diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
>> index 88a5426674a1..b36e8f10f509 100644
>> --- a/arch/x86/kvm/cpuid.c
>> +++ b/arch/x86/kvm/cpuid.c
>> @@ -1218,11 +1218,12 @@ void kvm_set_cpu_caps(void)
>>  		F(NULL_SEL_CLR_BASE),
>>  		/* UpperAddressIgnore */
>>  		F(AUTOIBRS),
>> -		F(PREFETCHI),
>>  		EMULATED_F(NO_SMM_CTL_MSR),
>>  		/* PrefetchCtlMsr */
>>  		/* GpOnUserCpuid */
>>  		/* EPSF */
>> +		F(PREFETCHI),
>> +		F(AVX512_BMM),
>>  		SYNTHESIZED_F(SBPB),
>>  		SYNTHESIZED_F(IBPB_BRTYPE),
>>  		SYNTHESIZED_F(SRSO_NO),
>>
>> base-commit: e89f0e9a0a007e8c3afb8ecd739c0b3255422b00
>> -- 
>> 2.48.1
>>