Add Inductor benchmark for scaled Blackwell persistent + TMA

Differential Revision: D82699362

Pull Request resolved: #453

Author: jananisriram

tritonbench/operators/fp8_gemm/fp8_gemm.py

@@ -189,6 +189,27 @@ def blackwell_persistent_tma_fp8_gemm(self, a, b, scale_a, scale_b):
                 self.extra_args.scaling_rowwise,
             )
 
+        @register_benchmark(enabled=True)
+        def blackwell_pt2_fp8_gemm(self, a, b, scale_a, scale_b):
+            torch._dynamo.reset()
+            with inductor_config.patch(
+                max_autotune=True,
+                max_autotune_gemm_backends="TRITON",
+                autotune_fallback_to_aten=False,
+            ):
+                f = lambda a, b: torch._scaled_mm(
+                    a,
+                    b,
+                    scale_a,
+                    scale_b,
+                    use_fast_accum=True,
+                    out_dtype=self._get_dtype(),
+                )
+                compiled = torch.compile(f, dynamic=False)
+                compiled(a, b)
+
+            return lambda: compiled(a, b)
+
     @register_benchmark()
     def triton_fp8_gemm(self, a, b, scale_a, scale_b):
         return lambda: tutorial_matmul(a, b)