Skip to contentSkip to conversation

KVM: x86: Remove AMX-TF32 enumeration

PATCH Superseded Chang S. Bae 4 participants 7 messages Jul 8 – Jul 20

Subsystems (3)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 KVM: x86: Remove AMX-TF32 enumeration

AMX-TF32 enumeration was added based on the Intel documentation [1]. The recent update removed the feature, indicating that the feature is no longer planned for hardware implementation.

[1] Intel Architecture Instruction Set Extensions and Future Features

---
Note: The feature first appeared in the September 2025 revision of the document and was removed in the June 2026 revision.

---
 arch/x86/kvm/cpuid.c         | 1 -
 arch/x86/kvm/reverse_cpuid.h | 1 -
 2 files changed, 2 deletions(-)
arch/x86/kvm/cpuid.c +0 -1
index 591d2294acd7..759a1c557080 100644
@@ -1082,7 +1082,6 @@ void kvm_initialize_cpu_caps(void)
 		F(AMX_COMPLEX_ALIAS),
 		F(AMX_FP16_ALIAS),
 		F(AMX_FP8),
-		F(AMX_TF32),
 		F(AMX_AVX512),
 		F(AMX_MOVRS),
 	);
arch/x86/kvm/reverse_cpuid.h +0 -1
index 2ad25781cefb..1bb13903c5b2 100644
@@ -56,7 +56,6 @@
 #define X86_FEATURE_AMX_COMPLEX_ALIAS	KVM_X86_FEATURE(CPUID_1E_1_EAX, 2)
 #define X86_FEATURE_AMX_FP16_ALIAS	KVM_X86_FEATURE(CPUID_1E_1_EAX, 3)
 #define X86_FEATURE_AMX_FP8		KVM_X86_FEATURE(CPUID_1E_1_EAX, 4)
-#define X86_FEATURE_AMX_TF32		KVM_X86_FEATURE(CPUID_1E_1_EAX, 6)
 #define X86_FEATURE_AMX_AVX512		KVM_X86_FEATURE(CPUID_1E_1_EAX, 7)
 #define X86_FEATURE_AMX_MOVRS		KVM_X86_FEATURE(CPUID_1E_1_EAX, 8)
 
Signature · 1 line
2.53.0
#1 lore raw
Quoted 3 lines · Chang S. Bae

AMX-TF32 enumeration was added based on the Intel documentation [1]. The recent update removed the feature, indicating that the feature is no longer planned for hardware implementation.

Patch looks good to me.

BTW, I think the bit cannot be assigned for new other usage in the future and becomes reserved forever, right?

Quoted 37 lines

[1] Intel Architecture Instruction Set Extensions and Future Features

---
Note: The feature first appeared in the September 2025 revision of the document and was removed in the June 2026 revision.

---
  arch/x86/kvm/cpuid.c         | 1 -
  arch/x86/kvm/reverse_cpuid.h | 1 -
  2 files changed, 2 deletions(-)
arch/x86/kvm/cpuid.c +0 -1
index 591d2294acd7..759a1c557080 100644
@@ -1082,7 +1082,6 @@ void kvm_initialize_cpu_caps(void)
  		F(AMX_COMPLEX_ALIAS),
  		F(AMX_FP16_ALIAS),
  		F(AMX_FP8),
-		F(AMX_TF32),
  		F(AMX_AVX512),
  		F(AMX_MOVRS),
  	);
arch/x86/kvm/reverse_cpuid.h +0 -1
index 2ad25781cefb..1bb13903c5b2 100644
@@ -56,7 +56,6 @@
  #define X86_FEATURE_AMX_COMPLEX_ALIAS	KVM_X86_FEATURE(CPUID_1E_1_EAX, 2)
  #define X86_FEATURE_AMX_FP16_ALIAS	KVM_X86_FEATURE(CPUID_1E_1_EAX, 3)
  #define X86_FEATURE_AMX_FP8		KVM_X86_FEATURE(CPUID_1E_1_EAX, 4)
-#define X86_FEATURE_AMX_TF32		KVM_X86_FEATURE(CPUID_1E_1_EAX, 6)
  #define X86_FEATURE_AMX_AVX512		KVM_X86_FEATURE(CPUID_1E_1_EAX, 7)
  #define X86_FEATURE_AMX_MOVRS		KVM_X86_FEATURE(CPUID_1E_1_EAX, 8)
 
#2 lore raw
Quoted 2 lines · Xiaoyao Li

BTW, I think the bit cannot be assigned for new other usage in the future and becomes reserved forever, right?

Reserving the bit forever looks contrary to the purpose of this patch. Right now the bit is already taken. And shipping processors will report that CPUID bit as zero. So this removal intentionally makes the software into the "it was never defined before" state. Whether the bit will eventually be repurposed is what I don't know. But that being said, I think it should be also backported to @stable.

Thanks,
Chang

Reviewed-by on v1 1/1
#3 lore raw
In reply to Chang S. Bae
Quoted 14 lines · Chang S. Bae

Date: Wed, 8 Jul 2026 21:01:18 +0000
From: "Chang S. Bae" <[email protected]>
Subject: [PATCH] KVM: x86: Remove AMX-TF32 enumeration
X-Mailer: git-send-email 2.51.0

AMX-TF32 enumeration was added based on the Intel documentation [1]. The recent update removed the feature, indicating that the feature is no longer planned for hardware implementation.

[1] Intel Architecture Instruction Set Extensions and Future Features

Quoted 7 lines

---
Note: The feature first appeared in the September 2025 revision of the document and was removed in the June 2026 revision.

---
 arch/x86/kvm/cpuid.c         | 1 -
 arch/x86/kvm/reverse_cpuid.h | 1 -
 2 files changed, 2 deletions(-)

LGTM,

Thanks,
Zhao

Reviewed-by on v1 1/1
#5 lore raw
In reply to Chang S. Bae
Quoted 9 lines · Chang S. Bae

AMX-TF32 enumeration was added based on the Intel documentation [1]. The recent update removed the feature, indicating that the feature is no longer planned for hardware implementation.

[1] Intel Architecture Instruction Set Extensions and Future Features

One nit below.

Quoted 3 lines

---
Note: The feature first appeared in the September 2025 revision of the document and was removed in the June 2026 revision.

IMHO, the note is informative.
Is it better to move it to the body of the changelog?

Quoted 29 lines
---
 arch/x86/kvm/cpuid.c         | 1 -
 arch/x86/kvm/reverse_cpuid.h | 1 -
 2 files changed, 2 deletions(-)
arch/x86/kvm/cpuid.c +0 -1
index 591d2294acd7..759a1c557080 100644
@@ -1082,7 +1082,6 @@ void kvm_initialize_cpu_caps(void)
 		F(AMX_COMPLEX_ALIAS),
 		F(AMX_FP16_ALIAS),
 		F(AMX_FP8),
-		F(AMX_TF32),
 		F(AMX_AVX512),
 		F(AMX_MOVRS),
 	);
arch/x86/kvm/reverse_cpuid.h +0 -1
index 2ad25781cefb..1bb13903c5b2 100644
@@ -56,7 +56,6 @@
 #define X86_FEATURE_AMX_COMPLEX_ALIAS	KVM_X86_FEATURE(CPUID_1E_1_EAX, 2)
 #define X86_FEATURE_AMX_FP16_ALIAS	KVM_X86_FEATURE(CPUID_1E_1_EAX, 3)
 #define X86_FEATURE_AMX_FP8		KVM_X86_FEATURE(CPUID_1E_1_EAX, 4)
-#define X86_FEATURE_AMX_TF32		KVM_X86_FEATURE(CPUID_1E_1_EAX, 6)
 #define X86_FEATURE_AMX_AVX512		KVM_X86_FEATURE(CPUID_1E_1_EAX, 7)
 #define X86_FEATURE_AMX_MOVRS		KVM_X86_FEATURE(CPUID_1E_1_EAX, 8)
 
#6 lore raw
Quoted 3 lines · Binbin Wu

On 7/9/2026 5:01 AM, Chang S. Bae wrote:

Thanks!

Quoted 9 lines

One nit below.

Quoted 3 lines

---
Note: The feature first appeared in the September 2025 revision of the document and was removed in the June 2026 revision.

IMHO, the note is informative.
Is it better to move it to the body of the changelog?

Will do that.

Thanks,
Chang