[PATCH] i386/cpu: Rename AVX10_VNNI_INT to AVX10_V1_AUX

[PATCH] i386/cpu: Rename AVX10_VNNI_INT to AVX10_V1_AUX

[Hao Xiang]

Rename AVX10_VNNI_INT is to AVX10_V1_AUX to match the most recent version
of the AVX10.2 spec (revision 7.0) [*].  Intel renamed the CPUID flag to
apply to a variety of "convert" instructions as well as VNNI instructions.

The recent KVM patch has renamed AVX10_VNNI_INT to AVX10_V1_AUX. QEMU needs
to update the CPUID macro, feature property, AVX10.2 feature model and
Diamond Rapids CPU definition. Keep the CPUID location and bit unchanged.

[*] Intel® Advanced Vector Extensions 10.2 Architecture Specification 
(rev 7.0)

Signed-off-by: Hao Xiang <hao.xiang@linux.alibaba.com>
---
  target/i386/cpu.c | 6 +++---
  target/i386/cpu.h | 4 ++--
  2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/target/i386/cpu.c b/target/i386/cpu.c
index 9b9ed2d1e3..e7e35c3f5b 100644
--- a/target/i386/cpu.c
+++ b/target/i386/cpu.c
@@ -1366,7 +1366,7 @@ FeatureWordInfo feature_word_info[FEATURE_WORDS] = {
      [FEAT_24_1_ECX] = {
          .type = CPUID_FEATURE_WORD,
          .feat_names = {
-            [2] = "avx10-vnni-int",
+            [2] = "avx10-v1-aux",
          },
          .cpuid = {
              .eax = 0x24,
@@ -2387,7 +2387,7 @@ static const AVX10VersionDefinition 
builtin_avx10_defs[] = {
          .version = 2,
          .max_subleaf = 1,
          .features = (FeatureMask[]) {
-            { FEAT_24_1_ECX,         CPUID_24_1_ECX_AVX10_VNNI_INT },
+            { FEAT_24_1_ECX,         CPUID_24_1_ECX_AVX10_V1_AUX },
              { /* end of list */ }
          }
      },
@@ -5538,7 +5538,7 @@ static const X86CPUDefinition builtin_x86_defs[] = {
           * Though this bit will be set by avx_version=2, it's better to
           * explicitly enumerate this feature here.
           */
-        .features[FEAT_24_1_ECX] = CPUID_24_1_ECX_AVX10_VNNI_INT,
+        .features[FEAT_24_1_ECX] = CPUID_24_1_ECX_AVX10_V1_AUX,
          .features[FEAT_8000_0001_ECX] =
              CPUID_EXT3_LAHF_LM | CPUID_EXT3_ABM | 
CPUID_EXT3_3DNOWPREFETCH,
          .features[FEAT_8000_0001_EDX] =
diff --git a/target/i386/cpu.h b/target/i386/cpu.h
index 9f222a0c9f..6f7230031b 100644
--- a/target/i386/cpu.h
+++ b/target/i386/cpu.h
@@ -1126,8 +1126,8 @@ uint64_t x86_cpu_get_supported_feature_word(X86CPU 
*cpu, FeatureWord w);
                                           CPUID_24_0_EBX_AVX10_256 | \
                                           CPUID_24_0_EBX_AVX10_512)

-/* AVX10_VNNI_INT instruction */
-#define CPUID_24_1_ECX_AVX10_VNNI_INT         (1U << 2)
+/* AVX10_V1_AUX instruction */
+#define CPUID_24_1_ECX_AVX10_V1_AUX           (1U << 2)

  /*
   * New Conditional Instructions (NCIs), explicit New Data Destination 
(NDD)
-- 
2.51.1

Re: [PATCH] i386/cpu: Rename AVX10_VNNI_INT to AVX10_V1_AUX

[zhao1.liu]

Hi Xiang,

On Tue, Oct 27, 2026 at 04:30:27PM +0800, Hao Xiang wrote:
> Date: Tue, 27 Oct 2026 16:30:27 +0800
> From: Hao Xiang <hao.xiang@linux.alibaba.com>
> Subject: [PATCH] i386/cpu: Rename AVX10_VNNI_INT to AVX10_V1_AUX
> X-Mailer: [Alimail-Mailagent revision 74][W4_0.2.3][v5ForWebDing][Chrome]
> 
> Rename AVX10_VNNI_INT is to AVX10_V1_AUX to match the most recent version
> of the AVX10.2 spec (revision 7.0) [*].  Intel renamed the CPUID flag to
> apply to a variety of "convert" instructions as well as VNNI instructions.
> 
> The recent KVM patch has renamed AVX10_VNNI_INT to AVX10_V1_AUX. QEMU needs
> to update the CPUID macro, feature property, AVX10.2 feature model and
> Diamond Rapids CPU definition. Keep the CPUID location and bit unchanged.
> 
> [*] Intel® Advanced Vector Extensions 10.2 Architecture Specification 
> (rev 7.0)
> 
> Signed-off-by: Hao Xiang <hao.xiang@linux.alibaba.com>
> ---
>   target/i386/cpu.c | 6 +++---
>   target/i386/cpu.h | 4 ++--
>   2 files changed, 5 insertions(+), 5 deletions(-)

Thanks for this patch! I posted something similar before :)

https://lore.kernel.org/qemu-devel/20260930045237.784060-1-zhao1.liu@intel.com/

(There is one additional change: I add an alias for the vnni-int bit to avoid
 potential API compatibility issues.)

Regards,
Zhao