Add datatypes for roofline profiling (#642)

Rebuild of rocm-amdgpu-bench roofline binaries for MI200/MI300 systems with rocm6.
Added datatype options to roofline feature.

---------

Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>

CHANGELOG.md

@@ -17,6 +17,9 @@ Full documentation for ROCm Compute Profiler is available at [https://rocm.docs.
   * --roofline-data-type / -R option added to specify which datatypes the user wants to capture in the roofline PDF plot outputs
   * Default is FP32, but user can specify as many types as desired to overlay on the same plot output
 
+* Additional datatypes for roofline profiling
+  * Now supports FP8, FP16, BF16, FP32, FP64, I8, I32, I64 (dependent on gpu architecture)
+
 ### Changed
 
 * Change normal_unit default to per_kernel

src/argparser.py

@@ -372,7 +372,7 @@ Examples:
         "-R",
         "--roofline-data-type",
         required=False,
-        choices=["FP8", "FP16", "BF16", "FP32", "FP64", "I8"],
+        choices=["FP8", "FP16", "BF16", "FP32", "FP64", "I8", "I32", "I64"],
         metavar="",
         nargs="+",
         type=str,

src/roofline.py

@@ -319,26 +319,29 @@ class Roofline:
             )
 
         # Plot peak MFMA ceiling
-        fig.add_trace(
+        if dtype in MFMA_DATATYPES:
-            go.Scatter(
+            fig.add_trace(
-                x=self.__ceiling_data["mfma"][0],
+                go.Scatter(
-                y=self.__ceiling_data["mfma"][1],
+                    x=self.__ceiling_data["mfma"][0],
-                name="Peak MFMA-{}".format(dtype),
+                    y=self.__ceiling_data["mfma"][1],
-                mode=plot_mode,
+                    name="Peak MFMA-{}".format(dtype),
-                hovertemplate="<b>%{text}</b>",
+                    mode=plot_mode,
-                text=[
+                    hovertemplate="<b>%{text}</b>",
-                    (
+                    text=[
-                        None
+                        (
-                        if self.__run_parameters["is_standalone"]
+                            None
-                        else "{} G{}/s".format(
+                            if self.__run_parameters["is_standalone"]
+                            else "{} G{}/s".format(
+                                to_int(self.__ceiling_data["mfma"][2]), ops_flops
+                            )
+                        ),
+                        "{} G{}/s".format(
                             to_int(self.__ceiling_data["mfma"][2]), ops_flops
-                        )
+                        ),
-                    ),
+                    ],
-                    "{} G{}/s".format(to_int(self.__ceiling_data["mfma"][2]), ops_flops),
+                    textposition="top left",
-                ],
+                )
-                textposition="top left",
             )
-        )
         #######################
         # Plot Application AI
         #######################

src/utils/roofline_calc.py

@@ -44,13 +44,13 @@ FONT_WEIGHT = "bold"
 # SUPPORTED_DATATYPES table is based on datatype support in rocm-amdgpu-bench repository
 # Indicates which datatypes per gpu arch can be generated by the roofline binary
 SUPPORTED_DATATYPES = {
-    "gfx90a": ["FP16", "BF16", "FP32", "FP64", "I8"],  # Unsupported: F8
+    "gfx90a": ["FP16", "BF16", "FP32", "FP64", "I8", "I32", "I64"],  # Unsupported: F8
-    "gfx940": ["FP8", "FP16", "FP32", "FP64"],  # Unsupported: BF16, I8
+    "gfx940": ["FP8", "FP16", "BF16", "FP32", "FP64", "I8", "I32", "I64"],  # Unsupported:
-    "gfx941": ["FP8", "FP16", "FP32", "FP64"],  # Unsupported: BF16, I8
+    "gfx941": ["FP8", "FP16", "BF16", "FP32", "FP64", "I8", "I32", "I64"],  # Unsupported:
-    "gfx942": ["FP8", "FP16", "FP32", "FP64"],  # Unsupported: BF16, I8
+    "gfx942": ["FP8", "FP16", "BF16", "FP32", "FP64", "I8", "I32", "I64"],  # Unsupported:
 }
 
-PEAK_OPS_DATATYPES = ["FP8", "FP32", "FP64"]
+PEAK_OPS_DATATYPES = ["FP8", "FP16", "BF16", "FP32", "FP64", "I8", "I32", "I64"]
 MFMA_DATATYPES = ["FP8", "FP16", "BF16", "FP32", "FP64", "I8"]
 
 TOP_N = 10
@@ -177,7 +177,7 @@ def calc_ceilings(roofline_parameters, dtype, benchmark_data):
         graphPoints["valu"].append(peakOps)
 
     # Plot MFMA roof
-    if x1_mfma != -1 or (dtype in MFMA_DATATYPES):  # assert that mfma has been assigned
+    if x1_mfma != -1 and (dtype in MFMA_DATATYPES):  # assert that mfma has been assigned
         x0_mfma = XMAX
         if x2_mfma < x0_mfma:
             x0_mfma = x2_mfma