[Mlir-commits] [mlir] [mlir][tosa] Switch matmul_t output to bf16 from fp16 (PR #212970)

Jeremy Johnson llvmlistbot at llvm.org
Thu Jul 30 02:26:18 PDT 2026


https://github.com/jjohnson-arm created https://github.com/llvm/llvm-project/pull/212970

Note that this implements the TOSA specification change: https://github.com/arm/tosa-specification/pull/67

>From 2b894ad60e4bc4a76522f66d5ed6f2d6c02b280b Mon Sep 17 00:00:00 2001
From: Jeremy Johnson <jeremy.johnson at arm.com>
Date: Wed, 29 Jul 2026 15:15:32 +0100
Subject: [PATCH] [mlir][tosa] Switch matmul_t output to bf16 from fp16

Signed-off-by: Jeremy Johnson <jeremy.johnson at arm.com>
Change-Id: I568a53053215ca2312129b521671979cea948869
---
 .../Dialect/Tosa/IR/TosaComplianceData.h.inc  | 488 ++++++++++--------
 mlir/test/Dialect/Tosa/invalid_extension.mlir |   8 +-
 mlir/test/Dialect/Tosa/ops.mlir               |   6 +-
 .../tosa-validation-version-1p1-valid.mlir    |   6 +-
 4 files changed, 281 insertions(+), 227 deletions(-)

diff --git a/mlir/include/mlir/Dialect/Tosa/IR/TosaComplianceData.h.inc b/mlir/include/mlir/Dialect/Tosa/IR/TosaComplianceData.h.inc
index d08cd7453bea0..cf02420fb67ea 100644
--- a/mlir/include/mlir/Dialect/Tosa/IR/TosaComplianceData.h.inc
+++ b/mlir/include/mlir/Dialect/Tosa/IR/TosaComplianceData.h.inc
@@ -72,8 +72,6 @@ profileComplianceMap = {
          SpecificationVersion::V_1_1_DRAFT},
         {{fp32T, fp16T, fp32T, fp16T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{fp16T, fp16T, fp16T, fp16T, fp32T},
-         SpecificationVersion::V_1_1_DRAFT},
         {{fp32T, fp32T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}}}}},
     {"tosa.max_pool2d",
@@ -1682,13 +1680,7 @@ extensionComplianceMap = {
      {{{Extension::int16},
        {{{i16T, i16T, i16T, i16T, i48T}, SpecificationVersion::V_1_1_DRAFT}}},
       {{Extension::fp8e4m3},
-       {{{fp8e4m3T, fp8e4m3T, fp8e4m3T, fp8e4m3T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e4m3T, fp16T, fp8e4m3T, fp16T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp16T, fp8e4m3T, fp16T, fp8e4m3T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e4m3T, fp8e4m3T, fp8e4m3T, fp8e4m3T, fp32T},
+        {{{fp8e4m3T, fp8e4m3T, fp8e4m3T, fp8e4m3T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{fp8e4m3T, fp16T, fp8e4m3T, fp16T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
@@ -1698,94 +1690,108 @@ extensionComplianceMap = {
          SpecificationVersion::V_1_1_DRAFT},
         {{fp32T, fp8e4m3T, fp32T, fp8e4m3T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}}},
-      {{Extension::fp8e4m3, Extension::fp8e5m2},
-       {{{fp8e4m3T, fp8e5m2T, fp8e4m3T, fp8e5m2T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e5m2T, fp8e4m3T, fp8e5m2T, fp8e4m3T, fp16T},
+      {{Extension::fp8e4m3, Extension::fp8e5m2, Extension::bf16},
+       {{{fp8e4m3T, fp8e5m2T, fp8e4m3T, fp8e5m2T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e4m3T, fp8e5m2T, fp8e4m3T, fp8e5m2T, fp32T},
+        {{fp8e5m2T, fp8e4m3T, fp8e5m2T, fp8e4m3T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::fp8e4m3, Extension::fp8e5m2},
+       {{{fp8e4m3T, fp8e5m2T, fp8e4m3T, fp8e5m2T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{fp8e5m2T, fp8e4m3T, fp8e5m2T, fp8e4m3T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
       {{Extension::bf16, Extension::fp8e4m3},
-       {{{fp8e4m3T, bf16T, fp8e4m3T, bf16T, fp16T},
+       {{{fp8e4m3T, bf16T, fp8e4m3T, bf16T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bf16T, fp8e4m3T, bf16T, fp8e4m3T, fp16T},
+        {{bf16T, fp8e4m3T, bf16T, fp8e4m3T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
         {{fp8e4m3T, bf16T, fp8e4m3T, bf16T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bf16T, fp8e4m3T, bf16T, fp8e4m3T, fp32T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{fp8e4m3T, fp8e4m3T, fp8e4m3T, fp8e4m3T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{fp8e4m3T, fp16T, fp8e4m3T, fp16T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{fp16T, fp8e4m3T, fp16T, fp8e4m3T, bf16T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_fp8e4m3},
-       {{{fp8e4m3T, bs32_fp8ue8m0_fp8e4m3T, fp8e4m3T, fp32T, fp16T},
+      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_fp8e4m3, Extension::bf16},
+       {{{fp8e4m3T, bs32_fp8ue8m0_fp8e4m3T, fp8e4m3T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e4m3T, fp8e4m3T, fp32T, fp8e4m3T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e4m3T, bs32_fp8ue8m0_fp8e4m3T, fp8e4m3T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp8e4m3T, fp8e4m3T, fp32T, fp8e4m3T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_fp8e4m3},
+       {{{fp8e4m3T, bs32_fp8ue8m0_fp8e4m3T, fp8e4m3T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp8e4m3T, fp8e4m3T, fp32T, fp8e4m3T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_fp8e5m2},
-       {{{fp8e4m3T, bs32_fp8ue8m0_fp8e5m2T, fp8e4m3T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e5m2T, fp8e4m3T, fp32T, fp8e4m3T, fp16T},
+      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_fp8e5m2, Extension::bf16},
+       {{{fp8e4m3T, bs32_fp8ue8m0_fp8e5m2T, fp8e4m3T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e4m3T, bs32_fp8ue8m0_fp8e5m2T, fp8e4m3T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp8e5m2T, fp8e4m3T, fp32T, fp8e4m3T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_fp8e5m2},
+       {{{fp8e4m3T, bs32_fp8ue8m0_fp8e5m2T, fp8e4m3T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp8e5m2T, fp8e4m3T, fp32T, fp8e4m3T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_fp6e3m2},
-       {{{fp8e4m3T, bs32_fp8ue8m0_fp6e3m2T, fp8e4m3T, fp32T, fp16T},
+      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_fp6e3m2, Extension::bf16},
+       {{{fp8e4m3T, bs32_fp8ue8m0_fp6e3m2T, fp8e4m3T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e3m2T, fp8e4m3T, fp32T, fp8e4m3T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e4m3T, bs32_fp8ue8m0_fp6e3m2T, fp8e4m3T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp6e3m2T, fp8e4m3T, fp32T, fp8e4m3T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_fp6e3m2},
+       {{{fp8e4m3T, bs32_fp8ue8m0_fp6e3m2T, fp8e4m3T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp6e3m2T, fp8e4m3T, fp32T, fp8e4m3T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_fp6e2m3},
-       {{{fp8e4m3T, bs32_fp8ue8m0_fp6e2m3T, fp8e4m3T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e2m3T, fp8e4m3T, fp32T, fp8e4m3T, fp16T},
+      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_fp6e2m3, Extension::bf16},
+       {{{fp8e4m3T, bs32_fp8ue8m0_fp6e2m3T, fp8e4m3T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e4m3T, bs32_fp8ue8m0_fp6e2m3T, fp8e4m3T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp6e2m3T, fp8e4m3T, fp32T, fp8e4m3T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_fp6e2m3},
+       {{{fp8e4m3T, bs32_fp8ue8m0_fp6e2m3T, fp8e4m3T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp6e2m3T, fp8e4m3T, fp32T, fp8e4m3T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_fp4e2m1},
-       {{{fp8e4m3T, bs32_fp8ue8m0_fp4e2m1T, fp8e4m3T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp4e2m1T, fp8e4m3T, fp32T, fp8e4m3T, fp16T},
+      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_fp4e2m1, Extension::bf16},
+       {{{fp8e4m3T, bs32_fp8ue8m0_fp4e2m1T, fp8e4m3T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e4m3T, bs32_fp8ue8m0_fp4e2m1T, fp8e4m3T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp4e2m1T, fp8e4m3T, fp32T, fp8e4m3T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_fp4e2m1},
+       {{{fp8e4m3T, bs32_fp8ue8m0_fp4e2m1T, fp8e4m3T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp4e2m1T, fp8e4m3T, fp32T, fp8e4m3T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_int8},
-       {{{fp8e4m3T, bs32_fp8ue8m0_mxint8T, fp8e4m3T, fp32T, fp16T},
+      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_int8, Extension::bf16},
+       {{{fp8e4m3T, bs32_fp8ue8m0_mxint8T, fp8e4m3T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_mxint8T, fp8e4m3T, fp32T, fp8e4m3T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e4m3T, bs32_fp8ue8m0_mxint8T, fp8e4m3T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_mxint8T, fp8e4m3T, fp32T, fp8e4m3T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::fp8e4m3, Extension::mx_common, Extension::mx_int8},
+       {{{fp8e4m3T, bs32_fp8ue8m0_mxint8T, fp8e4m3T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_mxint8T, fp8e4m3T, fp32T, fp8e4m3T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
       {{Extension::fp8e5m2},
-       {{{fp8e5m2T, fp8e5m2T, fp8e5m2T, fp8e5m2T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e5m2T, fp16T, fp8e5m2T, fp16T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp16T, fp8e5m2T, fp16T, fp8e5m2T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e5m2T, fp8e5m2T, fp8e5m2T, fp8e5m2T, fp32T},
+       {{{fp8e5m2T, fp8e5m2T, fp8e5m2T, fp8e5m2T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{fp8e5m2T, fp16T, fp8e5m2T, fp16T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
@@ -1796,81 +1802,99 @@ extensionComplianceMap = {
         {{fp32T, fp8e5m2T, fp32T, fp8e5m2T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}}},
       {{Extension::bf16, Extension::fp8e5m2},
-       {{{fp8e5m2T, bf16T, fp8e5m2T, bf16T, fp16T},
+       {{{fp8e5m2T, bf16T, fp8e5m2T, bf16T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bf16T, fp8e5m2T, bf16T, fp8e5m2T, fp16T},
+        {{bf16T, fp8e5m2T, bf16T, fp8e5m2T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
         {{fp8e5m2T, bf16T, fp8e5m2T, bf16T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bf16T, fp8e5m2T, bf16T, fp8e5m2T, fp32T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{fp8e5m2T, fp8e5m2T, fp8e5m2T, fp8e5m2T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{fp8e5m2T, fp16T, fp8e5m2T, fp16T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{fp16T, fp8e5m2T, fp16T, fp8e5m2T, bf16T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_fp8e4m3},
-       {{{fp8e5m2T, bs32_fp8ue8m0_fp8e4m3T, fp8e5m2T, fp32T, fp16T},
+      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_fp8e4m3, Extension::bf16},
+       {{{fp8e5m2T, bs32_fp8ue8m0_fp8e4m3T, fp8e5m2T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e4m3T, fp8e5m2T, fp32T, fp8e5m2T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e5m2T, bs32_fp8ue8m0_fp8e4m3T, fp8e5m2T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp8e4m3T, fp8e5m2T, fp32T, fp8e5m2T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_fp8e4m3},
+       {{{fp8e5m2T, bs32_fp8ue8m0_fp8e4m3T, fp8e5m2T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp8e4m3T, fp8e5m2T, fp32T, fp8e5m2T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_fp8e5m2},
-       {{{fp8e5m2T, bs32_fp8ue8m0_fp8e5m2T, fp8e5m2T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e5m2T, fp8e5m2T, fp32T, fp8e5m2T, fp16T},
+      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_fp8e5m2, Extension::bf16},
+       {{{fp8e5m2T, bs32_fp8ue8m0_fp8e5m2T, fp8e5m2T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e5m2T, bs32_fp8ue8m0_fp8e5m2T, fp8e5m2T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp8e5m2T, fp8e5m2T, fp32T, fp8e5m2T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_fp8e5m2},
+       {{{fp8e5m2T, bs32_fp8ue8m0_fp8e5m2T, fp8e5m2T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp8e5m2T, fp8e5m2T, fp32T, fp8e5m2T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_fp6e3m2},
-       {{{fp8e5m2T, bs32_fp8ue8m0_fp6e3m2T, fp8e5m2T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e3m2T, fp8e5m2T, fp32T, fp8e5m2T, fp16T},
+      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_fp6e3m2, Extension::bf16},
+       {{{fp8e5m2T, bs32_fp8ue8m0_fp6e3m2T, fp8e5m2T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e5m2T, bs32_fp8ue8m0_fp6e3m2T, fp8e5m2T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp6e3m2T, fp8e5m2T, fp32T, fp8e5m2T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_fp6e3m2},
+       {{{fp8e5m2T, bs32_fp8ue8m0_fp6e3m2T, fp8e5m2T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp6e3m2T, fp8e5m2T, fp32T, fp8e5m2T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_fp6e2m3},
-       {{{fp8e5m2T, bs32_fp8ue8m0_fp6e2m3T, fp8e5m2T, fp32T, fp16T},
+      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_fp6e2m3, Extension::bf16},
+       {{{fp8e5m2T, bs32_fp8ue8m0_fp6e2m3T, fp8e5m2T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e2m3T, fp8e5m2T, fp32T, fp8e5m2T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e5m2T, bs32_fp8ue8m0_fp6e2m3T, fp8e5m2T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp6e2m3T, fp8e5m2T, fp32T, fp8e5m2T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_fp6e2m3},
+       {{{fp8e5m2T, bs32_fp8ue8m0_fp6e2m3T, fp8e5m2T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp6e2m3T, fp8e5m2T, fp32T, fp8e5m2T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_fp4e2m1},
-       {{{fp8e5m2T, bs32_fp8ue8m0_fp4e2m1T, fp8e5m2T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp4e2m1T, fp8e5m2T, fp32T, fp8e5m2T, fp16T},
+      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_fp4e2m1, Extension::bf16},
+       {{{fp8e5m2T, bs32_fp8ue8m0_fp4e2m1T, fp8e5m2T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e5m2T, bs32_fp8ue8m0_fp4e2m1T, fp8e5m2T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp4e2m1T, fp8e5m2T, fp32T, fp8e5m2T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_fp4e2m1},
+       {{{fp8e5m2T, bs32_fp8ue8m0_fp4e2m1T, fp8e5m2T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp4e2m1T, fp8e5m2T, fp32T, fp8e5m2T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_int8},
-       {{{fp8e5m2T, bs32_fp8ue8m0_mxint8T, fp8e5m2T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_mxint8T, fp8e5m2T, fp32T, fp8e5m2T, fp16T},
+      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_int8, Extension::bf16},
+       {{{fp8e5m2T, bs32_fp8ue8m0_mxint8T, fp8e5m2T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{fp8e5m2T, bs32_fp8ue8m0_mxint8T, fp8e5m2T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_mxint8T, fp8e5m2T, fp32T, fp8e5m2T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::fp8e5m2, Extension::mx_common, Extension::mx_int8},
+       {{{fp8e5m2T, bs32_fp8ue8m0_mxint8T, fp8e5m2T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_mxint8T, fp8e5m2T, fp32T, fp8e5m2T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
       {{Extension::bf16},
-       {{{fp16T, bf16T, fp16T, bf16T, fp16T},
+       {{{fp16T, bf16T, fp16T, bf16T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bf16T, fp16T, bf16T, fp16T, fp16T},
+        {{bf16T, fp16T, bf16T, fp16T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bf16T, bf16T, bf16T, bf16T, fp16T},
+        {{bf16T, bf16T, bf16T, bf16T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
         {{fp16T, bf16T, fp16T, bf16T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
@@ -1883,13 +1907,7 @@ extensionComplianceMap = {
         {{bf16T, bf16T, bf16T, bf16T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}}},
       {{Extension::mx_common, Extension::mx_fp8e4m3},
-       {{{fp16T, bs32_fp8ue8m0_fp8e4m3T, fp16T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e4m3T, fp16T, fp32T, fp16T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp16T, bs32_fp8ue8m0_fp8e4m3T, fp16T, fp32T, fp32T},
+       {{{fp16T, bs32_fp8ue8m0_fp8e4m3T, fp16T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{fp32T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
@@ -1901,13 +1919,7 @@ extensionComplianceMap = {
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
       {{Extension::mx_common, Extension::mx_fp8e5m2},
-       {{{fp16T, bs32_fp8ue8m0_fp8e5m2T, fp16T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e5m2T, fp16T, fp32T, fp16T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp16T, bs32_fp8ue8m0_fp8e5m2T, fp16T, fp32T, fp32T},
+       {{{fp16T, bs32_fp8ue8m0_fp8e5m2T, fp16T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{fp32T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
@@ -1919,13 +1931,7 @@ extensionComplianceMap = {
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
       {{Extension::mx_common, Extension::mx_fp6e3m2},
-       {{{fp16T, bs32_fp8ue8m0_fp6e3m2T, fp16T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e3m2T, fp16T, fp32T, fp16T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp16T, bs32_fp8ue8m0_fp6e3m2T, fp16T, fp32T, fp32T},
+       {{{fp16T, bs32_fp8ue8m0_fp6e3m2T, fp16T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{fp32T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
@@ -1937,13 +1943,7 @@ extensionComplianceMap = {
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
       {{Extension::mx_common, Extension::mx_fp6e2m3},
-       {{{fp16T, bs32_fp8ue8m0_fp6e2m3T, fp16T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e2m3T, fp16T, fp32T, fp16T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp16T, bs32_fp8ue8m0_fp6e2m3T, fp16T, fp32T, fp32T},
+       {{{fp16T, bs32_fp8ue8m0_fp6e2m3T, fp16T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{fp32T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
@@ -1955,13 +1955,7 @@ extensionComplianceMap = {
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
       {{Extension::mx_common, Extension::mx_fp4e2m1},
-       {{{fp16T, bs32_fp8ue8m0_fp4e2m1T, fp16T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp4e2m1T, fp16T, fp32T, fp16T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp16T, bs32_fp8ue8m0_fp4e2m1T, fp16T, fp32T, fp32T},
+       {{{fp16T, bs32_fp8ue8m0_fp4e2m1T, fp16T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{fp32T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
@@ -1973,13 +1967,7 @@ extensionComplianceMap = {
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
       {{Extension::mx_common, Extension::mx_int8},
-       {{{fp16T, bs32_fp8ue8m0_mxint8T, fp16T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_mxint8T, fp16T, fp32T, fp16T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_mxint8T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{fp16T, bs32_fp8ue8m0_mxint8T, fp16T, fp32T, fp32T},
+       {{{fp16T, bs32_fp8ue8m0_mxint8T, fp16T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{fp32T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
@@ -1991,211 +1979,277 @@ extensionComplianceMap = {
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
       {{Extension::bf16, Extension::mx_common, Extension::mx_fp8e4m3},
-       {{{bf16T, bs32_fp8ue8m0_fp8e4m3T, bf16T, fp32T, fp16T},
+       {{{bf16T, bs32_fp8ue8m0_fp8e4m3T, bf16T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e4m3T, bf16T, fp32T, bf16T, fp16T},
+        {{bs32_fp8ue8m0_fp8e4m3T, bf16T, fp32T, bf16T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bf16T, bs32_fp8ue8m0_fp8e4m3T, bf16T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp8e4m3T, bf16T, fp32T, bf16T, fp32T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{fp16T, bs32_fp8ue8m0_fp8e4m3T, fp16T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{bs32_fp8ue8m0_fp8e4m3T, fp16T, fp32T, fp16T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
       {{Extension::bf16, Extension::mx_common, Extension::mx_fp8e5m2},
-       {{{bf16T, bs32_fp8ue8m0_fp8e5m2T, bf16T, fp32T, fp16T},
+       {{{bf16T, bs32_fp8ue8m0_fp8e5m2T, bf16T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e5m2T, bf16T, fp32T, bf16T, fp16T},
+        {{bs32_fp8ue8m0_fp8e5m2T, bf16T, fp32T, bf16T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bf16T, bs32_fp8ue8m0_fp8e5m2T, bf16T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp8e5m2T, bf16T, fp32T, bf16T, fp32T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{fp16T, bs32_fp8ue8m0_fp8e5m2T, fp16T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{bs32_fp8ue8m0_fp8e5m2T, fp16T, fp32T, fp16T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
       {{Extension::bf16, Extension::mx_common, Extension::mx_fp6e3m2},
-       {{{bf16T, bs32_fp8ue8m0_fp6e3m2T, bf16T, fp32T, fp16T},
+       {{{bf16T, bs32_fp8ue8m0_fp6e3m2T, bf16T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e3m2T, bf16T, fp32T, bf16T, fp16T},
+        {{bs32_fp8ue8m0_fp6e3m2T, bf16T, fp32T, bf16T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bf16T, bs32_fp8ue8m0_fp6e3m2T, bf16T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp6e3m2T, bf16T, fp32T, bf16T, fp32T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{fp16T, bs32_fp8ue8m0_fp6e3m2T, fp16T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{bs32_fp8ue8m0_fp6e3m2T, fp16T, fp32T, fp16T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
       {{Extension::bf16, Extension::mx_common, Extension::mx_fp6e2m3},
-       {{{bf16T, bs32_fp8ue8m0_fp6e2m3T, bf16T, fp32T, fp16T},
+       {{{bf16T, bs32_fp8ue8m0_fp6e2m3T, bf16T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e2m3T, bf16T, fp32T, bf16T, fp16T},
+        {{bs32_fp8ue8m0_fp6e2m3T, bf16T, fp32T, bf16T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bf16T, bs32_fp8ue8m0_fp6e2m3T, bf16T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp6e2m3T, bf16T, fp32T, bf16T, fp32T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{fp16T, bs32_fp8ue8m0_fp6e2m3T, fp16T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{bs32_fp8ue8m0_fp6e2m3T, fp16T, fp32T, fp16T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
       {{Extension::bf16, Extension::mx_common, Extension::mx_fp4e2m1},
-       {{{bf16T, bs32_fp8ue8m0_fp4e2m1T, bf16T, fp32T, fp16T},
+       {{{bf16T, bs32_fp8ue8m0_fp4e2m1T, bf16T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp4e2m1T, bf16T, fp32T, bf16T, fp16T},
+        {{bs32_fp8ue8m0_fp4e2m1T, bf16T, fp32T, bf16T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bf16T, bs32_fp8ue8m0_fp4e2m1T, bf16T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp4e2m1T, bf16T, fp32T, bf16T, fp32T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{fp16T, bs32_fp8ue8m0_fp4e2m1T, fp16T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{bs32_fp8ue8m0_fp4e2m1T, fp16T, fp32T, fp16T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
       {{Extension::bf16, Extension::mx_common, Extension::mx_int8},
-       {{{bf16T, bs32_fp8ue8m0_mxint8T, bf16T, fp32T, fp16T},
+       {{{bf16T, bs32_fp8ue8m0_mxint8T, bf16T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_mxint8T, bf16T, fp32T, bf16T, fp16T},
+        {{bs32_fp8ue8m0_mxint8T, bf16T, fp32T, bf16T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bf16T, bs32_fp8ue8m0_mxint8T, bf16T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_mxint8T, bf16T, fp32T, bf16T, fp32T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{fp16T, bs32_fp8ue8m0_mxint8T, fp16T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{bs32_fp8ue8m0_mxint8T, fp16T, fp32T, fp16T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT},
+        {{bs32_fp8ue8m0_mxint8T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::mx_common, Extension::mx_fp8e4m3, Extension::mx_fp8e5m2},
-       {{{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, fp16T},
+      {{Extension::mx_common, Extension::mx_fp8e4m3, Extension::mx_fp8e5m2, Extension::bf16},
+       {{{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::mx_common, Extension::mx_fp8e4m3, Extension::mx_fp8e5m2},
+       {{{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::mx_common, Extension::mx_fp6e3m2, Extension::mx_fp8e4m3},
-       {{{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, fp16T},
+      {{Extension::mx_common, Extension::mx_fp6e3m2, Extension::mx_fp8e4m3, Extension::bf16},
+       {{{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::mx_common, Extension::mx_fp6e3m2, Extension::mx_fp8e4m3},
+       {{{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::mx_common, Extension::mx_fp6e2m3, Extension::mx_fp8e4m3},
-       {{{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, fp16T},
+      {{Extension::mx_common, Extension::mx_fp6e2m3, Extension::mx_fp8e4m3, Extension::bf16},
+       {{{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::mx_common, Extension::mx_fp6e2m3, Extension::mx_fp8e4m3},
+       {{{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::mx_common, Extension::mx_fp4e2m1, Extension::mx_fp8e4m3},
-       {{{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, fp16T},
+      {{Extension::mx_common, Extension::mx_fp4e2m1, Extension::mx_fp8e4m3, Extension::bf16},
+       {{{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::mx_common, Extension::mx_fp4e2m1, Extension::mx_fp8e4m3},
+       {{{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::mx_common, Extension::mx_fp8e4m3, Extension::mx_int8},
-       {{{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_mxint8T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, fp16T},
+      {{Extension::mx_common, Extension::mx_fp8e4m3, Extension::mx_int8, Extension::bf16},
+       {{{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_mxint8T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::mx_common, Extension::mx_fp8e4m3, Extension::mx_int8},
+       {{{bs32_fp8ue8m0_fp8e4m3T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_mxint8T, bs32_fp8ue8m0_fp8e4m3T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::mx_common, Extension::mx_fp6e3m2, Extension::mx_fp8e5m2},
-       {{{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, fp16T},
+      {{Extension::mx_common, Extension::mx_fp6e3m2, Extension::mx_fp8e5m2, Extension::bf16},
+       {{{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::mx_common, Extension::mx_fp6e3m2, Extension::mx_fp8e5m2},
+       {{{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::mx_common, Extension::mx_fp6e2m3, Extension::mx_fp8e5m2},
-       {{{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, fp16T},
+      {{Extension::mx_common, Extension::mx_fp6e2m3, Extension::mx_fp8e5m2, Extension::bf16},
+       {{{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::mx_common, Extension::mx_fp6e2m3, Extension::mx_fp8e5m2},
+       {{{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::mx_common, Extension::mx_fp4e2m1, Extension::mx_fp8e5m2},
-       {{{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, fp16T},
+      {{Extension::mx_common, Extension::mx_fp4e2m1, Extension::mx_fp8e5m2, Extension::bf16},
+       {{{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::mx_common, Extension::mx_fp4e2m1, Extension::mx_fp8e5m2},
+       {{{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::mx_common, Extension::mx_fp8e5m2, Extension::mx_int8},
-       {{{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_mxint8T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, fp16T},
+      {{Extension::mx_common, Extension::mx_fp8e5m2, Extension::mx_int8, Extension::bf16},
+       {{{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_mxint8T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::mx_common, Extension::mx_fp8e5m2, Extension::mx_int8},
+       {{{bs32_fp8ue8m0_fp8e5m2T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_mxint8T, bs32_fp8ue8m0_fp8e5m2T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::mx_common, Extension::mx_fp6e2m3, Extension::mx_fp6e3m2},
-       {{{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, fp16T},
+      {{Extension::mx_common, Extension::mx_fp6e2m3, Extension::mx_fp6e3m2, Extension::bf16},
+       {{{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::mx_common, Extension::mx_fp6e2m3, Extension::mx_fp6e3m2},
+       {{{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::mx_common, Extension::mx_fp4e2m1, Extension::mx_fp6e3m2},
-       {{{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, fp16T},
+      {{Extension::mx_common, Extension::mx_fp4e2m1, Extension::mx_fp6e3m2, Extension::bf16},
+       {{{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::mx_common, Extension::mx_fp4e2m1, Extension::mx_fp6e3m2},
+       {{{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::mx_common, Extension::mx_fp6e3m2, Extension::mx_int8},
-       {{{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_mxint8T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, fp16T},
+      {{Extension::mx_common, Extension::mx_fp6e3m2, Extension::mx_int8, Extension::bf16},
+       {{{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_mxint8T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::mx_common, Extension::mx_fp6e3m2, Extension::mx_int8},
+       {{{bs32_fp8ue8m0_fp6e3m2T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_mxint8T, bs32_fp8ue8m0_fp6e3m2T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::mx_common, Extension::mx_fp4e2m1, Extension::mx_fp6e2m3},
-       {{{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, fp16T},
+      {{Extension::mx_common, Extension::mx_fp4e2m1, Extension::mx_fp6e2m3, Extension::bf16},
+       {{{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::mx_common, Extension::mx_fp4e2m1, Extension::mx_fp6e2m3},
+       {{{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::mx_common, Extension::mx_fp6e2m3, Extension::mx_int8},
-       {{{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, fp16T},
+      {{Extension::mx_common, Extension::mx_fp6e2m3, Extension::mx_int8, Extension::bf16},
+       {{{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_mxint8T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_mxint8T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::mx_common, Extension::mx_fp6e2m3, Extension::mx_int8},
+       {{{bs32_fp8ue8m0_fp6e2m3T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_mxint8T, bs32_fp8ue8m0_fp6e2m3T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
        allOf},
-      {{Extension::mx_common, Extension::mx_fp4e2m1, Extension::mx_int8},
-       {{{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, fp16T},
-         SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_mxint8T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, fp16T},
+      {{Extension::mx_common, Extension::mx_fp4e2m1, Extension::mx_int8, Extension::bf16},
+       {{{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, bf16T},
          SpecificationVersion::V_1_1_DRAFT},
-        {{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, fp32T},
+        {{bs32_fp8ue8m0_mxint8T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, bf16T},
+         SpecificationVersion::V_1_1_DRAFT}},
+       allOf},
+      {{Extension::mx_common, Extension::mx_fp4e2m1, Extension::mx_int8},
+       {{{bs32_fp8ue8m0_fp4e2m1T, bs32_fp8ue8m0_mxint8T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT},
         {{bs32_fp8ue8m0_mxint8T, bs32_fp8ue8m0_fp4e2m1T, fp32T, fp32T, fp32T},
          SpecificationVersion::V_1_1_DRAFT}},
diff --git a/mlir/test/Dialect/Tosa/invalid_extension.mlir b/mlir/test/Dialect/Tosa/invalid_extension.mlir
index b7b0e19ba04ce..63fe0e6a2bdc9 100644
--- a/mlir/test/Dialect/Tosa/invalid_extension.mlir
+++ b/mlir/test/Dialect/Tosa/invalid_extension.mlir
@@ -378,12 +378,12 @@ func.func @test_cast_fp8_block_scaled(%arg0: tensor<4x32xf8E4M3FN>) -> tensor<4x
 
 // -----
 
-func.func @test_matmul_t_mixed_block_scaled_operands(%arg0: tensor<4x8x32xf16>, %arg1: tensor<4x16x32x!tosa.block_scaled<BLOCK_SHAPE_32:f8E8M0FNU:f6E2M3FN>>) -> tensor<4x8x16xf16> {
+func.func @test_matmul_t_mixed_block_scaled_operands(%arg0: tensor<4x8x32xf16>, %arg1: tensor<4x16x32x!tosa.block_scaled<BLOCK_SHAPE_32:f8E8M0FNU:f6E2M3FN>>) -> tensor<4x8x16xbf16> {
   %azp0 = "tosa.const"() <{values = dense<0.0> : tensor<1xf16>}> : () -> tensor<1xf16>
   %bzp0 = "tosa.const"() <{values = dense<0.0> : tensor<1xf32>}> : () -> tensor<1xf32>
-  // expected-error at +1 {{'tosa.matmul_t' op illegal: requires all of [mx_common, mx_fp6e2m3] profiles/extensions to be specified in the target environment}}
-  %0 = tosa.matmul_t %arg0, %arg1, %azp0, %bzp0 : (tensor<4x8x32xf16>, tensor<4x16x32x!tosa.block_scaled<BLOCK_SHAPE_32:f8E8M0FNU:f6E2M3FN>>, tensor<1xf16>, tensor<1xf32>) -> tensor<4x8x16xf16>
-  return %0 : tensor<4x8x16xf16>
+  // expected-error at +1 {{'tosa.matmul_t' op illegal: requires all of [bf16, mx_common, mx_fp6e2m3] profiles/extensions to be specified in the target environment}}
+  %0 = tosa.matmul_t %arg0, %arg1, %azp0, %bzp0 : (tensor<4x8x32xf16>, tensor<4x16x32x!tosa.block_scaled<BLOCK_SHAPE_32:f8E8M0FNU:f6E2M3FN>>, tensor<1xf16>, tensor<1xf32>) -> tensor<4x8x16xbf16>
+  return %0 : tensor<4x8x16xbf16>
 }
 
 // -----
diff --git a/mlir/test/Dialect/Tosa/ops.mlir b/mlir/test/Dialect/Tosa/ops.mlir
index fd134370327de..43aee928aee1f 100644
--- a/mlir/test/Dialect/Tosa/ops.mlir
+++ b/mlir/test/Dialect/Tosa/ops.mlir
@@ -288,11 +288,11 @@ func.func @test_matmul_t_bf16(%arg0: tensor<2x14x19xbf16>, %arg1: tensor<2x28x19
 
 // -----
 // CHECK-LABEL: test_matmul_t_fp8_mixed
-func.func @test_matmul_t_fp8_mixed(%arg0: tensor<2x14x19xf8E4M3FN>, %arg1: tensor<2x28x19xf8E5M2>) -> tensor<2x14x28xf16> {
+func.func @test_matmul_t_fp8_mixed(%arg0: tensor<2x14x19xf8E4M3FN>, %arg1: tensor<2x28x19xf8E5M2>) -> tensor<2x14x28xbf16> {
 %azp0 = "tosa.const"() <{values = dense<0.0> : tensor<1xf8E4M3FN>}> : () -> tensor<1xf8E4M3FN>
 %bzp0 = "tosa.const"() <{values = dense<0.0> : tensor<1xf8E5M2>}> : () -> tensor<1xf8E5M2>
-%0 = tosa.matmul_t %arg0, %arg1, %azp0, %bzp0 : (tensor<2x14x19xf8E4M3FN>, tensor<2x28x19xf8E5M2>, tensor<1xf8E4M3FN>, tensor<1xf8E5M2>)  -> tensor<2x14x28xf16>
-  return %0 : tensor<2x14x28xf16>
+%0 = tosa.matmul_t %arg0, %arg1, %azp0, %bzp0 : (tensor<2x14x19xf8E4M3FN>, tensor<2x28x19xf8E5M2>, tensor<1xf8E4M3FN>, tensor<1xf8E5M2>)  -> tensor<2x14x28xbf16>
+  return %0 : tensor<2x14x28xbf16>
 }
 
 // -----
diff --git a/mlir/test/Dialect/Tosa/tosa-validation-version-1p1-valid.mlir b/mlir/test/Dialect/Tosa/tosa-validation-version-1p1-valid.mlir
index 87a1c17ce1154..071db7a1ffc1b 100644
--- a/mlir/test/Dialect/Tosa/tosa-validation-version-1p1-valid.mlir
+++ b/mlir/test/Dialect/Tosa/tosa-validation-version-1p1-valid.mlir
@@ -53,11 +53,11 @@ func.func @test_matmul_t_with_block_scaled_inputs_type_fp4e2m1_fp6e2m3(%arg0: te
 // -----
 
 // CHECK-LABEL: test_matmul_t_f16_with_block_scaled_inputs_type_fp6e3m2
-func.func @test_matmul_t_f16_with_block_scaled_inputs_type_fp6e3m2(%arg0: tensor<4x8x32xf16>, %arg1: tensor<4x16x32x!tosa.block_scaled<BLOCK_SHAPE_32:f8E8M0FNU:f6E3M2FN>>) -> tensor<4x8x16xf16> {
+func.func @test_matmul_t_f16_with_block_scaled_inputs_type_fp6e3m2(%arg0: tensor<4x8x32xf16>, %arg1: tensor<4x16x32x!tosa.block_scaled<BLOCK_SHAPE_32:f8E8M0FNU:f6E3M2FN>>) -> tensor<4x8x16xbf16> {
   %azp0 = "tosa.const"() <{values = dense<0.0> : tensor<1xf16>}> : () -> tensor<1xf16>
   %bzp0 = "tosa.const"() <{values = dense<0.0> : tensor<1xf32>}> : () -> tensor<1xf32>
-  %0 = tosa.matmul_t %arg0, %arg1, %azp0, %bzp0 : (tensor<4x8x32xf16>, tensor<4x16x32x!tosa.block_scaled<BLOCK_SHAPE_32:f8E8M0FNU:f6E3M2FN>>, tensor<1xf16>, tensor<1xf32>) -> tensor<4x8x16xf16>
-  return %0 : tensor<4x8x16xf16>
+  %0 = tosa.matmul_t %arg0, %arg1, %azp0, %bzp0 : (tensor<4x8x32xf16>, tensor<4x16x32x!tosa.block_scaled<BLOCK_SHAPE_32:f8E8M0FNU:f6E3M2FN>>, tensor<1xf16>, tensor<1xf32>) -> tensor<4x8x16xbf16>
+  return %0 : tensor<4x8x16xbf16>
 }
 
 // -----



More information about the Mlir-commits mailing list