Add description for profiler operations

profiler/src/profile_batched_gemm.cpp

@@ -197,4 +197,4 @@ int profile_batched_gemm(int argc, char* argv[])
     }
 }
 
-REGISTER_PROFILER_OPERATION("batched_gemm", profile_batched_gemm)
+REGISTER_PROFILER_OPERATION("batched_gemm", "Batched GEMM", profile_batched_gemm);

profiler/src/profile_batched_gemm_add_relu_gemm_add.cpp

@@ -111,7 +111,7 @@ int profile_batched_gemm_add_relu_gemm_add(int argc, char* argv[])
     else
     {
         printf("arg1: tensor operation (batched_gemm_add_relu_gemm_add: "
-               "Batched_GEMM+Add+Relu+Gemm+Add)\n");
+               "Batched GEMM+Add+Relu+GEMM+Add)\n");
         printf("arg2: data type (1: fp16)\n");
         printf("arg3: matrix layout (0: Relu(A0[m, k] * B0[n, k] + D0[m, n]) * B1[n, o] + D1[m, o] "
                "= E1[m, o]; 1: Relu(A0[m, k] * B0[n, k] + D0[m, n]) * B1[o, n] + D1[m, o] = "
@@ -209,4 +209,6 @@ int profile_batched_gemm_add_relu_gemm_add(int argc, char* argv[])
     return 0;
 }
 
-REGISTER_PROFILER_OPERATION("batched_gemm_add_relu_gemm_add", profile_batched_gemm_add_relu_gemm_add)
+REGISTER_PROFILER_OPERATION("batched_gemm_add_relu_gemm_add",
+                            "Batched GEMM+Add+Relu+GEMM+Add",
+                            profile_batched_gemm_add_relu_gemm_add);

profiler/src/profile_batched_gemm_gemm.cpp

@@ -102,7 +102,7 @@ int profile_batched_gemm_gemm(int argc, char* argv[])
     }
     else
     {
-        printf("arg1: tensor operation (batched_gemm_gemm: Batched_GEMM+Gemm)\n");
+        printf("arg1: tensor operation (batched_gemm_gemm: Batched GEMM+GEMM)\n");
         printf("arg2: data type (1: fp16)\n");
         printf("arg3: matrix layout (0: Relu(A0[m, k] * B0[n, k] + D0[m, n]) * B1[n, o] + D1[m, o] "
                "= E1[m, o];  1: Relu(A0[m, k] * B0[n, k] + D0[m, n]) * B1[o, n] + D1[m, o] = E1[m, "
@@ -181,4 +181,4 @@ int profile_batched_gemm_gemm(int argc, char* argv[])
     return 0;
 }
 
-REGISTER_PROFILER_OPERATION("batched_gemm_gemm", profile_batched_gemm_gemm)
+REGISTER_PROFILER_OPERATION("batched_gemm_gemm", "Batched GEMM+GEMM", profile_batched_gemm_gemm);

profiler/src/profile_batched_gemm_reduce.cpp

@@ -27,7 +27,7 @@ int profile_batched_gemm_reduce(int argc, char* argv[])
 
     if(argc != 15)
     {
-        printf("arg1: tensor operation (batched_gemm_reduce: BatchedGEMM+Reduce)\n");
+        printf("arg1: tensor operation (batched_gemm_reduce: Batched GEMM+Reduce)\n");
         printf("arg2: data type (0: fp32; 1: fp16)\n");
         printf("arg3: matrix layout (0: A[m, k] * B[k, n] = C[m, n];\n");
         printf("                     1: A[m, k] * B[n, k] = C[m, n];\n");
@@ -153,4 +153,6 @@ int profile_batched_gemm_reduce(int argc, char* argv[])
     return 0;
 }
 
-REGISTER_PROFILER_OPERATION("batched_gemm_reduce", profile_batched_gemm_reduce)
+REGISTER_PROFILER_OPERATION("batched_gemm_reduce",
+                            "Batched GEMM+Reduce",
+                            profile_batched_gemm_reduce);

profiler/src/profile_conv_bwd_data.cpp

@@ -184,4 +184,4 @@ int profile_conv_bwd_data(int argc, char* argv[])
     return 1;
 }
 
-REGISTER_PROFILER_OPERATION("conv_bwd_data", profile_conv_bwd_data)
+REGISTER_PROFILER_OPERATION("conv_bwd_data", "Convolution Backward Data", profile_conv_bwd_data);

profiler/src/profile_conv_fwd.cpp

@@ -186,4 +186,4 @@ int profile_conv_fwd(int argc, char* argv[])
     return 1;
 }
 
-REGISTER_PROFILER_OPERATION("conv_fwd", profile_conv_fwd)
+REGISTER_PROFILER_OPERATION("conv_fwd", "Convolution Forward", profile_conv_fwd);

profiler/src/profile_conv_fwd_bias_relu.cpp

@@ -37,7 +37,7 @@ int profile_conv_fwd_bias_relu(int argc, char* argv[])
 {
     if(argc != 25)
     {
-        printf("arg1: tensor operation (conv_fwd_bias_relu: ForwardConvolution+Bias+ReLu)\n");
+        printf("arg1: tensor operation (conv_fwd_bias_relu: Convolution Forward+Bias+ReLU)\n");
         printf("arg2: data type (0: fp32; 1: fp16)\n");
         printf("arg3: input tensor layout (0: NCHW; 1: NHWC)\n");
         printf("arg4: weight tensor layout (0: KCYX; 1: KYXC)\n");
@@ -116,4 +116,6 @@ int profile_conv_fwd_bias_relu(int argc, char* argv[])
     return 0;
 }
 
-REGISTER_PROFILER_OPERATION("conv_fwd_bias_relu", profile_conv_fwd_bias_relu)
+REGISTER_PROFILER_OPERATION("conv_fwd_bias_relu",
+                            "Convolution Forward+Bias+ReLU",
+                            profile_conv_fwd_bias_relu);

profiler/src/profile_conv_fwd_bias_relu_add.cpp

@@ -38,7 +38,7 @@ int profile_conv_fwd_bias_relu_add(int argc, char* argv[])
     if(argc != 25)
     {
         printf(
-            "arg1: tensor operation (conv_fwd_bias_relu_add: ForwardConvolution+Bias+ReLu+Add)\n");
+            "arg1: tensor operation (conv_fwd_bias_relu_add: Convolution Forward+Bias+ReLU+Add)\n");
         printf("arg2: data type (0: fp32; 1: fp16)\n");
         printf("arg3: input tensor layout (0: NCHW; 1: NHWC)\n");
         printf("arg4: weight tensor layout (0: KCYX; 1: KYXC)\n");
@@ -117,4 +117,6 @@ int profile_conv_fwd_bias_relu_add(int argc, char* argv[])
     return 0;
 }
 
-REGISTER_PROFILER_OPERATION("conv_fwd_bias_relu_add", profile_conv_fwd_bias_relu_add)
+REGISTER_PROFILER_OPERATION("conv_fwd_bias_relu_add",
+                            "Convolution Forward+Bias+ReLU+Add",
+                            profile_conv_fwd_bias_relu_add);

profiler/src/profile_gemm.cpp

@@ -186,4 +186,4 @@ int profile_gemm(int argc, char* argv[])
     }
 }
 
-REGISTER_PROFILER_OPERATION("gemm", profile_gemm)
+REGISTER_PROFILER_OPERATION("gemm", "GEMM", profile_gemm);

profiler/src/profile_gemm_add_add_fastgelu.cpp

@@ -152,4 +152,6 @@ int profile_gemm_add_add_fastgelu(int argc, char* argv[])
     }
 }
 
-REGISTER_PROFILER_OPERATION("gemm_add_add_fastgelu", profile_gemm_add_add_fastgelu)
+REGISTER_PROFILER_OPERATION("gemm_add_add_fastgelu",
+                            "GEMM+Add+Add+FastGeLU",
+                            profile_gemm_add_add_fastgelu);

profiler/src/profile_gemm_bias_add_reduce.cpp

@@ -27,7 +27,7 @@ int profile_gemm_bias_add_reduce(int argc, char* argv[])
 
     if(!(argc == 14 || argc == 15))
     {
-        printf("arg1: tensor operation (gemm: GEMM+bias+add+Reduce)\n");
+        printf("arg1: tensor operation (gemm_bias_add_reduce: GEMM+Bias+Add+Reduce)\n");
         printf("arg2: data type (0: fp32; 1: fp16)\n");
         printf("arg3: matrix layout (0: A[m, k] * B[k, n] = C[m, n];\n");
         printf("                     1: A[m, k] * B[n, k] = C[m, n];\n");
@@ -161,4 +161,6 @@ int profile_gemm_bias_add_reduce(int argc, char* argv[])
     return 0;
 }
 
-REGISTER_PROFILER_OPERATION("gemm_bias_add_reduce", profile_gemm_bias_add_reduce)
+REGISTER_PROFILER_OPERATION("gemm_bias_add_reduce",
+                            "GEMM+Bias+Add+Reduce",
+                            profile_gemm_bias_add_reduce);

profiler/src/profile_gemm_bilinear.cpp

@@ -146,4 +146,4 @@ int profile_gemm_bilinear(int argc, char* argv[])
     }
 }
 
-REGISTER_PROFILER_OPERATION("gemm_bilinear", profile_gemm_bilinear)
+REGISTER_PROFILER_OPERATION("gemm_bilinear", "GEMM+Bilinear", profile_gemm_bilinear);

profiler/src/profile_gemm_reduce.cpp

@@ -27,7 +27,7 @@ int profile_gemm_reduce(int argc, char* argv[])
 
     if(!(argc == 14 || argc == 15))
     {
-        printf("arg1: tensor operation (gemm: GEMM+Reduce)\n");
+        printf("arg1: tensor operation (gemm_reduce: GEMM+Reduce)\n");
         printf("arg2: data type (0: fp32; 1: fp16)\n");
         printf("arg3: matrix layout (0: A[m, k] * B[k, n] = C[m, n];\n");
         printf("                     1: A[m, k] * B[n, k] = C[m, n];\n");
@@ -148,4 +148,4 @@ int profile_gemm_reduce(int argc, char* argv[])
     return 0;
 }
 
-REGISTER_PROFILER_OPERATION("gemm_reduce", profile_gemm_reduce)
+REGISTER_PROFILER_OPERATION("gemm_reduce", "GEMM+Reduce", profile_gemm_reduce);

profiler/src/profile_gemm_splitk.cpp

@@ -148,4 +148,4 @@ int profile_gemm_splitk(int argc, char* argv[])
     }
 }
 
-REGISTER_PROFILER_OPERATION("gemm_splitk", profile_gemm_splitk)
+REGISTER_PROFILER_OPERATION("gemm_splitk", "Split-K GEMM", profile_gemm_splitk);

profiler/src/profile_grouped_conv_bwd_weight.cpp

@@ -26,20 +26,21 @@ enum struct ConvDataType
 
 static void print_helper_msg()
 {
-    std::cout << "arg1: tensor operation (conv_bwd_weight: Convolution Backward Weight\n"
-              << "arg2: data type (0: Input fp32, Weight fp32, Output fp32\n"
-              << "                 1: Input fp16, Weight fp16, Output fp16\n"
-              << "                 2: Input bf16, Weight fp32, Output bf16)\n"
-              << "arg3: tensor layout (0: Input[G, N, C, Hi, Wi], Weight[G, K, C, Y, X], Output[G, "
-                 "N, K, Ho, Wo]\n"
-              << "                     1: Input[G, N, Hi, Wi, C], Weight[G, K, Y, X, C], Output[G, "
-                 "N, Ho, Wo, K]\n"
-              << "arg4: verification (0: no, 1: yes)\n"
-              << "arg5: initialization (0: no init, 1: integer value, 2: decimal value)\n"
-              << "arg6: print tensor value (0: no; 1: yes)\n"
-              << "arg7: time kernel (0: no, 1: yes)\n"
-              << ck::utils::conv::get_conv_param_parser_helper_msg() << " SplitK\n"
-              << std::endl;
+    std::cout
+        << "arg1: tensor operation (grouped_conv_bwd_weight: Grouped Convolution Backward Weight\n"
+        << "arg2: data type (0: Input fp32, Weight fp32, Output fp32\n"
+        << "                 1: Input fp16, Weight fp16, Output fp16\n"
+        << "                 2: Input bf16, Weight fp32, Output bf16)\n"
+        << "arg3: tensor layout (0: Input[G, N, C, Hi, Wi], Weight[G, K, C, Y, X], Output[G, "
+           "N, K, Ho, Wo]\n"
+        << "                     1: Input[G, N, Hi, Wi, C], Weight[G, K, Y, X, C], Output[G, "
+           "N, Ho, Wo, K]\n"
+        << "arg4: verification (0: no, 1: yes)\n"
+        << "arg5: initialization (0: no init, 1: integer value, 2: decimal value)\n"
+        << "arg6: print tensor value (0: no; 1: yes)\n"
+        << "arg7: time kernel (0: no, 1: yes)\n"
+        << ck::utils::conv::get_conv_param_parser_helper_msg() << " SplitK\n"
+        << std::endl;
 }
 
 } // namespace
@@ -176,4 +177,6 @@ int profile_grouped_conv_bwd_weight(int argc, char* argv[])
     return 1;
 }
 
-REGISTER_PROFILER_OPERATION("grouped_conv_bwd_weight", profile_grouped_conv_bwd_weight)
+REGISTER_PROFILER_OPERATION("grouped_conv_bwd_weight",
+                            "Grouped Convolution Backward Weight",
+                            profile_grouped_conv_bwd_weight);

profiler/src/profile_grouped_conv_fwd.cpp

@@ -254,4 +254,6 @@ int profile_grouped_conv_fwd(int argc, char* argv[])
     return 1;
 }
 
-REGISTER_PROFILER_OPERATION("grouped_conv_fwd", profile_grouped_conv_fwd)
+REGISTER_PROFILER_OPERATION("grouped_conv_fwd",
+                            "Grouped Convolution Forward",
+                            profile_grouped_conv_fwd);

profiler/src/profile_grouped_gemm.cpp

@@ -163,4 +163,4 @@ int profile_grouped_gemm(int argc, char* argv[])
     return 0;
 }
 
-REGISTER_PROFILER_OPERATION("grouped_gemm", profile_grouped_gemm)
+REGISTER_PROFILER_OPERATION("grouped_gemm", "Grouped GEMM", profile_grouped_gemm);

profiler/src/profile_groupnorm.cpp

@@ -46,7 +46,7 @@ struct GroupnormArgParser
 
 void print_help_groupnorm()
 {
-    std::cout << "arg1: tensor operation (groupnorm: Group normalization)\n"
+    std::cout << "arg1: tensor operation (groupnorm: Group Normalization)\n"
               << "arg2: data type (0: fp16; 1: fp32)\n"
               << "arg3: verification (0: no; 1: yes)\n"
               << "arg4: initialization (0: no init; 1: integer value; 2: decimal value)\n"
@@ -106,4 +106,4 @@ int profile_groupnorm(int argc, char* argv[])
     return 0;
 }
 
-REGISTER_PROFILER_OPERATION("groupnorm", profile_groupnorm)
+REGISTER_PROFILER_OPERATION("groupnorm", "Group Normalization", profile_groupnorm);

profiler/src/profile_layernorm.cpp

@@ -98,4 +98,4 @@ int profile_layernorm(int argc, char* argv[])
     return 0;
 }
 
-REGISTER_PROFILER_OPERATION("layernorm", profile_layernorm)
+REGISTER_PROFILER_OPERATION("layernorm", "Layer Normalization", profile_layernorm);

profiler/src/profile_reduce.cpp

@@ -431,4 +431,4 @@ int profile_reduce(int argc, char* argv[])
     return (0);
 };
 
-REGISTER_PROFILER_OPERATION("reduce", profile_reduce)
+REGISTER_PROFILER_OPERATION("reduce", "Reduce", profile_reduce);