Code clean-up (#1285)

* code clean-up

* remove the profiling output samples

include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_dl.hpp

@@ -552,7 +552,7 @@ struct DeviceGemmMultipleD_Dl : public DeviceGemmMultipleD<ALayout,
     static bool IsSupportedArgument(const Argument& arg)
     {
         if(ck::get_device_name() == "gfx906" || ck::is_xdl_supported() ||
-           ck::is_navi2_supported() || ck::is_navi3_supported())
+           ck::is_gfx103_supported() || ck::is_gfx11_supported())
         {
             return GridwiseGemm::CheckValidity(
                 arg.a_grid_desc_k0_m_k1_, arg.b_grid_desc_k0_n_k1_, arg.e_grid_desc_m_n_);

include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_wmma_cshuffle.hpp

@@ -515,7 +515,7 @@ struct DeviceGemmMultipleD_Wmma_CShuffle : public DeviceGemmMultipleD<ALayout,
 
     static bool IsSupportedArgument(const Argument& arg)
     {
-        if(ck::is_navi3_supported())
+        if(ck::is_gfx11_supported())
         {
             if constexpr(!(is_same_v<AccDataType, float> || is_same_v<AccDataType, int32_t>))
             {

include/ck/tensor_operation/gpu/device/impl/device_gemm_wmma.hpp

@@ -443,7 +443,7 @@ struct DeviceGemmWmma_CShuffle : public DeviceGemm<ALayout,
 
     static bool IsSupportedArgument(const Argument& arg)
     {
-        if(ck::is_navi3_supported())
+        if(ck::is_gfx11_supported())
         {
             if constexpr(!(is_same_v<AccDataType, float> || is_same_v<AccDataType, ck::half_t> ||
                            is_same_v<AccDataType, int32_t>))

include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_data_multiple_d_wmma_cshuffle.hpp

@@ -629,7 +629,7 @@ struct DeviceGroupedConvBwdDataMultipleD_Wmma_CShuffle
     static bool IsSupportedArgument(const Argument& arg)
     {
         // check device
-        if(ck::is_navi3_supported())
+        if(ck::is_gfx11_supported())
         {
             if constexpr(!(is_same_v<AccDataType, float> || is_same_v<AccDataType, int32_t>))
             {

include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_wmma_cshuffle.hpp

@@ -692,7 +692,7 @@ struct DeviceGroupedConvBwdWeight_Wmma_CShuffle
     static bool IsSupportedArgument(const Argument& arg)
     {
         // check device
-        if(ck::is_navi3_supported())
+        if(ck::is_gfx11_supported())
         {
             if constexpr(!(is_same_v<AccDataType, float> || is_same_v<AccDataType, int32_t>))
             {

include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_dl_multiple_d_nhwc_kyxc_nhwk.hpp

@@ -666,7 +666,7 @@ struct DeviceGroupedConvFwdDlMultipleD_NHWC_KYXC_NHWK
 
         // check device
         if(!(ck::get_device_name() == "gfx906" || ck::is_xdl_supported() ||
-             ck::is_navi2_supported() || ck::is_navi3_supported()))
+             ck::is_gfx103_supported() || ck::is_gfx11_supported()))
         {
             return false;
         }

include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_dl_nhwc_kyxc_nhwk.hpp

@@ -601,8 +601,8 @@ struct DeviceGroupedConvFwdDl_NHWC_KYXC_NHWK : public DeviceGroupedConvFwd<NDimS
         namespace ctc = tensor_layout::convolution;
 
         // check device
-        if(!(ck::get_device_name() == "gfx906" || ck::is_navi2_supported() ||
-             ck::is_navi3_supported()))
+        if(!(ck::get_device_name() == "gfx906" || ck::is_gfx103_supported() ||
+             ck::is_gfx11_supported()))
         {
             return false;
         }

include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_multiple_d_wmma_cshuffle.hpp

@@ -581,7 +581,7 @@ struct DeviceGroupedConvFwdMultipleD_Wmma_CShuffle
         namespace ctc = tensor_layout::convolution;
 
         // check device
-        if(ck::is_navi3_supported())
+        if(ck::is_gfx11_supported())
         {
             if constexpr(!(is_same_v<AccDataType, float> || is_same_v<AccDataType, int32_t>))
             {

include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_dl.hpp

@@ -673,7 +673,7 @@ struct DeviceGroupedGemmMultipleD_Dl : public DeviceGroupedGemm<ALayout,
         }
 
         if(ck::get_device_name() == "gfx906" || ck::is_xdl_supported() ||
-           ck::is_navi2_supported() || ck::is_navi3_supported())
+           ck::is_gfx103_supported() || ck::is_gfx11_supported())
         {
             for(std::size_t i = 0; i < arg.gemm_desc_kernel_arg_.size(); i++)
             {

include/ck/tensor_operation/gpu/device/impl/device_grouped_query_attention_forward_wmma.hpp

@@ -596,7 +596,7 @@ struct DeviceGroupedQueryAttentionForward_Wmma
 
     static bool IsSupportedArgument(const RawArg& arg)
     {
-        if(ck::is_navi3_supported())
+        if(ck::is_gfx11_supported())
         {
             if constexpr(!(is_same_v<Acc0DataType, float> || is_same_v<Acc0DataType, int32_t>))
             {
@@ -958,7 +958,7 @@ struct DeviceGroupedQueryAttentionForward_Wmma
 #if 0
     static bool IsSupportedArgument(const Argument& arg)
     {
-        if(ck::is_navi3_supported())
+        if(ck::is_gfx11_supported())
         {
             if constexpr(!(is_same_v<Acc0DataType, float> || is_same_v<Acc0DataType, int32_t>))
             {

include/ck/tensor_operation/gpu/device/impl/device_multi_query_attention_forward_wmma.hpp

@@ -594,7 +594,7 @@ struct DeviceMultiQueryAttentionForward_Wmma
 
     static bool IsSupportedArgument(const RawArg& arg)
     {
-        if(ck::is_navi3_supported())
+        if(ck::is_gfx11_supported())
         {
             if constexpr(!(is_same_v<Acc0DataType, float> || is_same_v<Acc0DataType, int32_t>))
             {
@@ -950,7 +950,7 @@ struct DeviceMultiQueryAttentionForward_Wmma
 #if 0
     static bool IsSupportedArgument(const Argument& arg)
     {
-        if(ck::is_navi3_supported())
+        if(ck::is_gfx11_supported())
         {
             if constexpr(!(is_same_v<Acc0DataType, float> || is_same_v<Acc0DataType, int32_t>))
             {

include/ck/tensor_operation/gpu/grid/block_to_ctile_map.hpp

@@ -260,7 +260,7 @@ struct BlockToCTileMap_M00_N0_M01Adapt : BlockToCTileMap_M00_N0_M01Adapt<MPerBlo
 };
 
 // Grouped Rows of column-vectors WGP mapping
-// Optimized for MI300-like multipe-die chip
+// Optimized for gfx94x-like multipe-die chip
 
 template <index_t GroupNum, index_t MPerBlock, index_t NPerBlock>
 struct BlockToCTileMap_Grouped_M00_N0_M01Adapt

include/ck/tensor_operation/gpu/warp/wmma_gemm.hpp

@@ -95,7 +95,7 @@ struct wmma_type<WmmaInstr::wmma_f32_16x16x16_f16,
 
     // Wave mode dependent propety
     static constexpr index_t wave_size = Number<WaveSize>{};
-    // * Fixed in Navi3x, Will be wave mode dependent on Navi4x
+    // * Fixed on gfx11, Will be wave mode dependent for future architectures
     static constexpr index_t num_src_a_vgprs_per_wave = m_per_wmma * src_a_data_size / 4;
     static constexpr index_t num_src_b_vgprs_per_wave = n_per_wmma * src_b_data_size / 4;
     // * num_acc_vgprs_per_wave alone M direction

include/ck/utility/amd_xdlops.hpp

@@ -4,7 +4,7 @@
 #pragma once
 
 namespace ck {
-// Define the common macro for MI300 models
+// Define the common macro for gfx94x models
 #if defined(__gfx940__) || defined(__gfx941__) || defined(__gfx942__)
 #define __gfx94__
 #endif

include/ck/utility/type_convert.hpp

@@ -8,7 +8,7 @@
 #include "ck/utility/random_gen.hpp"
 
 namespace ck {
-// Define the common macro for MI300 models
+// Define the common macro for gfx94x models
 #if defined(__gfx940__) || defined(__gfx941__) || defined(__gfx942__)
 #define __gfx94__
 #endif

profiler/README.md

@@ -13,15 +13,6 @@
 ./bin/ckProfiler      gemm         1       1       1     1    0       5  3840 4096 4096     4096    4096    4096
 ```
 
-Result (MI100 @ 1087Mhz, 133.5TFlops peak FP16)
-```bash
-a_m_k: dim 2, lengths {3840, 4096}, strides {4096, 1}
-b_k_n: dim 2, lengths {4096, 4096}, strides {1, 4096}
-c_m_n: dim 2, lengths {3840, 4096}, strides {4096, 1}
-....
-Best Perf: 1.1933 ms, 107.977 TFlops, 79.0848 GB/s
-```
-
 ## Profile 2D forward convolution kernels
 ```bash
 #arg1: tensor operation (conv=Convolution)
@@ -37,15 +28,6 @@ Best Perf: 1.1933 ms, 107.977 TFlops, 79.0848 GB/s
  ################          op datatype  in_layout   wei_layout  out_layout  verify  init  log  repeat  N__ K___ C___ Y X Hi__ Wi__ Strides Dilations LeftPads RightPads
  ./bin/ckProfiler  conv2d_fwd        1          1            1           1       1     1    0       5  128  256  192 3 3   71   71     2 2       1 1      1 1       1 1
 ```
-Result (MI100 @ 1087Mhz, 133.5TFlops peak FP16)
-
-```bash
-in_n_c_hi_wi: dim 4, lengths {128, 192, 71, 71}, strides {967872, 1, 13632, 192}
-wei_k_c_y_x: dim 4, lengths {256, 192, 3, 3}, strides {1728, 1, 576, 192}
-out_n_k_ho_wo: dim 4, lengths {128, 256, 36, 36}, strides {331776, 1, 9216, 256}
-....
-Best Perf: 1.42509 ms, 102.988 TFlops, 234.086 GB/s
-```
 
 ## Profile contraction kernels
 ```bash
@@ -71,16 +53,6 @@ Best Perf: 1.42509 ms, 102.988 TFlops, 234.086 GB/s
 ./bin/ckProfiler contraction_bilinear         0                 0        2      1       0     0    0     1    1.0   1.0 128 128 128 128 128 128
 ```
 
-Result (MI100)
-```bash
-a_m_k: dim 4, lengths {128, 128, 128, 128}, strides {2097152, 16384, 128, 1}
-b_k_n: dim 4, lengths {128, 128, 128, 128}, strides {128, 1, 2097152, 16384}
-d_m_n: dim 4, lengths {128, 128, 128, 128}, strides {2097152, 16384, 128, 1}
-e_m_n: dim 4, lengths {128, 128, 128, 128}, strides {2097152, 16384, 128, 1}
-....
-Best Perf: 211.405 ms, 41.6077 TFlops, 15.2372 GB/s
-```
-
 ## Profile batched gemm multiple D kernels
 ```bash
 #arg1: tensor operation (batched_gemm_multi_d=Batched GEMM multi D);
@@ -99,14 +71,6 @@ Best Perf: 211.405 ms, 41.6077 TFlops, 15.2372 GB/s
 ./bin/ckProfiler batched_gemm_multi_d         0       1       0     0    0     1 4096 4096 4096    4096    4096    4096     16777216     16777216     16777216         16
 ```
 
-Result (Radeon RX 6800 XT)
-```bash
-arg.a_grid_desc_k0_m0_m1_k1_{2048, 4096, 2}
-arg.b_grid_desc_k0_n0_n1_k1_{2048, 4096, 2}
-arg.e_grid_desc_m_n_{ 4096, 4096}
-....
-Best Perf: 58.0306 ms, 37.8942 TFlops, 27.7545 GB/s
-```
 ## Profile grouped convolution backward data kernels
 ```bash
 # arg1: tensor operation (grouped_conv_bwd_data: Grouped Convolution Backward Data)
@@ -134,20 +98,6 @@ Best Perf: 58.0306 ms, 37.8942 TFlops, 27.7545 GB/s
 
 ```
 
-Result (MI100, FP16, GNHWC_GKYXC_GNHWK)
-
-```bash
-out: dim 5, lengths {32, 4, 192, 28, 28}, strides {602112, 150528, 1, 5376, 192}
-wei: dim 5, lengths {32, 192, 192, 3, 3}, strides {331776, 1728, 1, 576, 192}
-in: dim 5, lengths {32, 4, 192, 28, 28}, strides {602112, 150528, 1, 5376, 192}
-....
-Best configuration parameters:
-name: DeviceGroupedConvBwdDataMultipleD_Xdl_CShuffle_v1<256, 128, 256, 32, 8, 2, Default, 32, 32, 2, 4, 8, 4, 1, 1>
-avg_time: 0.768321
-tflops: 86.6679
-GB/s: 127.947
-```
-
 ## Profile grouped convolution backward weight kernels
 ```bash
 # arg1: tensor operation (grouped_conv_bwd_weight: Grouped Convolution Backward Weight)
@@ -179,19 +129,6 @@ GB/s: 127.947
 
 ```
 
-Result (MI100, FP16, GNHWC_GKYXC_GNHWK)
-
-```bash
-input: dim 5, lengths {32, 512, 1024, 28, 28}, strides {411041792, 802816, 1, 28672, 1024}
-weight: dim 5, lengths {32, 512, 1024, 3, 3}, strides {4718592, 9216, 1, 3072, 1024}
-output: dim 5, lengths {32, 512, 512, 26, 26}, strides {177209344, 346112, 1, 13312, 512}
-....
-Best configuration parameters:
-name: DeviceGroupedConvBwdWeight_Xdl_CShuffle<256, 256, 128, 4, Default, 8, 4, 2, 8, 4, 8, 2, 1, 1, 8>
-avg_time: 68.5216
-tflops: 95.337
-GB/s: 69.2301
-```
 Note: This kernel use atomic add, this will cause output buffer to be accumulated multiple times, causing verification failure. To work around it, do not use CK's own timer and do verification at the same time.
 
 ## Profile image to column/column to image kernels
@@ -224,17 +161,6 @@ Note: This kernel use atomic add, this will cause output buffer to be accumulate
 
 ```
 
-Result (MI210, FP32, NHWC)
-
-```bash
-input: dim 5, lengths {1, 256, 512, 28, 28}, strides {102760448, 401408, 1, 14336, 512}
-output: dim 2, lengths {173056, 4608}, strides {4608, 1}
-....
-Best configuration parameters:
-name: DeviceImageToColumn<128, 32, 64, 4>
-avg_time: 3.12326
-GB/s: 2042.59
-```
 Note: Column to image kernel adds to the output memory, this will cause output buffer to be accumulated multiple times, causing verification failure. To work around it, do not use CK's own timer and do verification at the same time.
 
 ## Profile Permute scale kernels
@@ -254,12 +180,3 @@ Note: Column to image kernel adds to the output memory, this will cause output b
 ################            op datatype  verify  init  log  time  dim0 dim1 dim2 in_stride0 in_stride1 in_stride2 out_stride0 out_stride1 out_stride2
 ./bin/ckProfiler permute_scale        0       1     1    0     1    64   64   64       4096         64          1           1          64        4096
 ```
-
-Result (MI100, FP32)
-
-```bash
-A: dim 3, lengths {64, 64, 64}, strides {4096, 64, 1}
-B: dim 3, lengths {64, 64, 64}, strides {1, 64, 4096}
-....
-Best perf = 0.0146878 ms, 142.782 GB/s, DeviceElementwiseNormalizationImpl<3, 2>
-```

script/test_convnd_fwd.sh

@@ -65,7 +65,7 @@ set -- "${POSITIONAL[@]}" # restore positional parameters
 # NUMACTL="numactl --cpunodebind=1 --membind=1"
 NUMACTL=
 # ENV_CONF=
-GPU=mi100
+GPU=gfx908
 PROF_ITER_COUNT=10000
 LOG_DIR_PATH=../log/${LOG_DIR}
 set -x

test/grouped_convnd_bwd_weight/test_grouped_convnd_bwd_weight.cpp

@@ -55,14 +55,14 @@ class TestGroupedConvndBwdWeight : public ::testing::Test
             }
         }
 
-        if(ck::is_navi3_supported())
+        if(ck::is_gfx11_supported())
         {
-            // on navi3x only support for 3d is implemented
+            // on gfx11 only support for 3d is implemented
             if constexpr(NDimSpatial{} != 3)
             {
                 return true;
             }
-            // on navi3x only support for i8 and fp16 is implemented
+            // on gfx11 only support for i8 and fp16 is implemented
             if constexpr(!((std::is_same_v<InDataType, int8_t> &&
                             std::is_same_v<WeiDataType, int8_t> &&
                             std::is_same_v<OutDataType, int8_t>) ||
@@ -80,7 +80,7 @@ class TestGroupedConvndBwdWeight : public ::testing::Test
         }
         else
         {
-            // support for i8 is only implemented on navi3x
+            // support for i8 is only implemented on gfx11
             if constexpr(std::is_same_v<InDataType, int8_t> &&
                          std::is_same_v<WeiDataType, int8_t> && std::is_same_v<OutDataType, int8_t>)
             {