Merge branch 'amd-develop' into amd-master

library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_default_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_default_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Col,
+                                                      Row,
+                                                      F16,
+                                                      F16,
+                                                      F16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_instances<Interwave,
+                                                                             GemmDefault>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_kpadding_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_kpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Col,
+                                                      Row,
+                                                      F16,
+                                                      F16,
+                                                      F16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_instances<Interwave,
+                                                                             GemmKPadding>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_mnkpadding_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_mnkpadding_instances(
+    std::vector<std::unique_ptr<DeviceGemm_Streamk_V2<Row,
+                                                      Col,
+                                                      Row,
+                                                      F16,
+                                                      F16,
+                                                      F16,
+                                                      PassThrough,
+                                                      PassThrough,
+                                                      PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_instances<Interwave,
+                                                                             GemmMNKPadding>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_convinvscale/CMakeLists.txt

+# ONLY XDL_KERNELS
+set(GROUPED_CONV3D_FWD_CONVINVSCALE
+   xdl/device_grouped_conv3d_fwd_xdl_convinvscale_ndhwgc_gkzyxc_ndhwgk_f8_instance.cpp)
+
+add_instance_library(device_grouped_conv3d_fwd_convinvscale_instance ${GROUPED_CONV3D_FWD_CONVINVSCALE})

library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_convinvscale/xdl/device_grouped_conv3d_fwd_xdl_convinvscale_ndhwgc_gkzyxc_ndhwgk_f8_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_outelementop_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using ConvInvscale = ck::tensor_operation::element_wise::ConvInvscale;
+
+void add_device_grouped_conv3d_fwd_xdl_convinvscale_ndhwgc_gkzyxc_ndhwgk_f8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                ck::Tuple<>,
+                                                                NDHWGK,
+                                                                F8,
+                                                                F8,
+                                                                ck::Tuple<>,
+                                                                F8,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                ConvInvscale,
+                                                                F8,
+                                                                F8>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_fwd_xdl_outelementop_f8_instances<3,
+                                                              NDHWGC,
+                                                              GKZYXC,
+                                                              ck::Tuple<>,
+                                                              NDHWGK,
+                                                              ConvFwdDefault,
+                                                              ConvInvscale>{});
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_fwd_xdl_outelementop_f8_instances<3,
+                                                              NDHWGC,
+                                                              GKZYXC,
+                                                              ck::Tuple<>,
+                                                              NDHWGK,
+                                                              ConvFwd1x1P0,
+                                                              ConvInvscale>{});
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_fwd_xdl_outelementop_f8_instances<3,
+                                                              NDHWGC,
+                                                              GKZYXC,
+                                                              ck::Tuple<>,
+                                                              NDHWGK,
+                                                              ConvFwd1x1S1P0,
+                                                              ConvInvscale>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_convscale/CMakeLists.txt

 # ONLY XDL_KERNELS
 set(GROUPED_CONV3D_FWD_CONVSCALE
-   xdl/device_grouped_conv3d_fwd_xdl_convscale_ndhwgc_gkzyxc_ndhwgk_f8_instance.cpp)
+   xdl/device_grouped_conv3d_fwd_xdl_convscale_ndhwgc_gkzyxc_ndhwgk_f8_instance.cpp
+   xdl/device_grouped_conv3d_fwd_xdl_convscale_ndhwgc_gkzyxc_ndhwgk_bf8_instance.cpp
+   xdl/device_grouped_conv3d_fwd_xdl_convscale_ndhwgc_gkzyxc_ndhwgk_f8_bf8_instance.cpp)
 
 add_instance_library(device_grouped_conv3d_fwd_convscale_instance ${GROUPED_CONV3D_FWD_CONVSCALE})

library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_convscale/xdl/device_grouped_conv3d_fwd_xdl_convscale_ndhwgc_gkzyxc_ndhwgk_bf8_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_outelementop_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using ConvScale = ck::tensor_operation::element_wise::ConvScale;
+
+void add_device_grouped_conv3d_fwd_xdl_convscale_ndhwgc_gkzyxc_ndhwgk_bf8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                ck::Tuple<>,
+                                                                NDHWGK,
+                                                                BF8,
+                                                                BF8,
+                                                                ck::Tuple<>,
+                                                                F8,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                ConvScale,
+                                                                BF8>>>& instances)
+{
+
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_fwd_xdl_outelementop_bf8_instances<3,
+                                                               NDHWGC,
+                                                               GKZYXC,
+                                                               ck::Tuple<>,
+                                                               NDHWGK,
+                                                               ConvFwdDefault,
+                                                               ConvScale>{});
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_fwd_xdl_outelementop_bf8_instances<3,
+                                                               NDHWGC,
+                                                               GKZYXC,
+                                                               ck::Tuple<>,
+                                                               NDHWGK,
+                                                               ConvFwd1x1P0,
+                                                               ConvScale>{});
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_fwd_xdl_outelementop_bf8_instances<3,
+                                                               NDHWGC,
+                                                               GKZYXC,
+                                                               ck::Tuple<>,
+                                                               NDHWGK,
+                                                               ConvFwd1x1S1P0,
+                                                               ConvScale>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_convscale/xdl/device_grouped_conv3d_fwd_xdl_convscale_ndhwgc_gkzyxc_ndhwgk_f8_bf8_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_outelementop_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using ConvScale = ck::tensor_operation::element_wise::ConvScale;
+
+void add_device_grouped_conv3d_fwd_xdl_convscale_ndhwgc_gkzyxc_ndhwgk_f8_bf8_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                ck::Tuple<>,
+                                                                NDHWGK,
+                                                                F8,
+                                                                BF8,
+                                                                ck::Tuple<>,
+                                                                F8,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                ConvScale,
+                                                                F8,
+                                                                BF8>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_fwd_xdl_outelementop_f8_bf8_instances<3,
+                                                                  NDHWGC,
+                                                                  GKZYXC,
+                                                                  ck::Tuple<>,
+                                                                  NDHWGK,
+                                                                  ConvFwdDefault,
+                                                                  ConvScale>{});
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_fwd_xdl_outelementop_f8_bf8_instances<3,
+                                                                  NDHWGC,
+                                                                  GKZYXC,
+                                                                  ck::Tuple<>,
+                                                                  NDHWGK,
+                                                                  ConvFwd1x1P0,
+                                                                  ConvScale>{});
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_fwd_xdl_outelementop_f8_bf8_instances<3,
+                                                                  NDHWGC,
+                                                                  GKZYXC,
+                                                                  ck::Tuple<>,
+                                                                  NDHWGK,
+                                                                  ConvFwd1x1S1P0,
+                                                                  ConvScale>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/CMakeLists.txt

@@ -5,8 +5,22 @@ set(GROUPED_GEMM_TILE_LOOP_INSTANCES)
 list(APPEND GROUPED_GEMM_TILE_LOOP_INSTANCES 
 	device_grouped_gemm_xdl_tile_loop_f16_f16_f16_mk_kn_mn_instance.cpp
 	device_grouped_gemm_xdl_tile_loop_f16_f16_f16_mk_nk_mn_instance.cpp
-
-	device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_instance.cpp
+	device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp
+	device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_default_instance.cpp
+	device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_kpadding_instance.cpp
+	device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp
+	device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_mnpadding_instance.cpp
+	device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_mem_v1_default_instance.cpp
+	device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_mem_v1_kpadding_instance.cpp
+	device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_mem_v1_mnkpadding_instance.cpp
+	device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_mem_v1_mnpadding_instance.cpp
+	device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_mem_v2_default_instance.cpp
+	device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_mem_v2_kpadding_instance.cpp
+	device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_mem_v2_mnkpadding_instance.cpp
+	device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_mem_v2_mnpadding_instance.cpp
+	device_grouped_gemm_xdl_tile_loop_multiply_bias_bf16_i8_bf16_mk_kn_mn_instance.cpp
+	device_grouped_gemm_xdl_tile_loop_multiply_bias_fastgelu_bf16_i8_bf16_mk_kn_mn_instance.cpp
+	device_grouped_gemm_xdl_tile_loop_multiply_fastgelu_bf16_i8_bf16_mk_kn_mn_instance.cpp
 	)
 
 add_instance_library(device_grouped_gemm_tile_loop_instance ${GROUPED_GEMM_TILE_LOOP_INSTANCES})

library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_f16_f16_f16_mk_kn_mn_instance.cpp

@@ -38,16 +38,16 @@ using device_grouped_gemm_xdl_tile_loop_f16_f16_f16_mk_kn_mn_irregular_tile_inst
         //###########################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
         //###########################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
         //###########################################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   2,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   2,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 4>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   2,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   2,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 4>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,           S<8>>
     // clang-format on
     >;
 

library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_f16_f16_f16_mk_nk_mn_instance.cpp

@@ -37,19 +37,19 @@ using device_grouped_gemm_xdl_tile_loop_f16_f16_f16_mk_nk_mn_irregular_tile_inst
         //###########################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
         //###########################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
         //###########################################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   256,    64,   8,   8,   32,   32,    2,    4,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   128,    64,   8,   8,   32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    64,   8,   8,   32,   32,    2,    1,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    64,   8,   8,   32,   32,    1,    2,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,   128,    64,   8,   8,   32,   32,    4,    2,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    64,   8,   8,   32,   32,    2,    2,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    64,   8,   8,   32,   32,    2,    2,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    32,    64,   8,   8,   32,   32,    2,    1,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   128,    64,   8,   8,   32,   32,    1,    2,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   256,    64,   8,   8,   32,   32,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    64,    64,   8,   8,   32,   32,    2,    2,     S<8,  8, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8,  8, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    32,    64,   8,   8,   32,   32,    2,    1,     S<8,  8, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8,  8, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    32,    64,    64,   8,   8,   32,   32,    1,    2,     S<8,  8, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8,  8, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8>
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   256,    64,   8,   8,   32,   32,    2,    4,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   128,    64,   8,   8,   32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,    64,    64,   8,   8,   32,   32,    2,    1,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,    64,   128,    64,   8,   8,   32,   32,    1,    2,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,   128,    64,   8,   8,   32,   32,    4,    2,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    64,    64,   8,   8,   32,   32,    2,    2,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    64,   128,    64,   8,   8,   32,   32,    2,    2,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,   128,    32,    64,   8,   8,   32,   32,    2,    1,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 4>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   128,    64,   8,   8,   32,   32,    1,    2,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   128,    32,   256,    64,   8,   8,   32,   32,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    64,    64,   8,   8,   32,   32,    2,    2,     S<8,  8, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8,  8, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    64,    32,    64,   8,   8,   32,   32,    2,    1,     S<8,  8, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8,  8, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,           S<8>>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Col,    DsLayout,    Row,   F16,   F16,     F32,      F32,  DsDataType,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,    64,    32,    64,    64,   8,   8,   32,   32,    1,    2,     S<8,  8, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8,  8, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,           S<8>>
     // clang-format on
     >;
 

library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include <cstdlib>
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_multiple_d_xdl_cshuffle_tile_loop.hpp"
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+using BF16 = ck::bhalf_t;
+using I8   = int8_t;
+using F32  = float;
+
+using Row = ck::tensor_layout::gemm::RowMajor;
+using Col = ck::tensor_layout::gemm::ColumnMajor;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using PassThrough         = ck::tensor_operation::element_wise::PassThrough;
+using Multiply            = ck::tensor_operation::element_wise::Multiply;
+using MultiplyAddFastGelu = ck::tensor_operation::element_wise::MultiplyAddFastGelu;
+using MultiplyFastGelu    = ck::tensor_operation::element_wise::MultiplyFastGelu;
+using MultiplyAdd         = ck::tensor_operation::element_wise::MultiplyAdd;
+
+static constexpr auto GemmDefault    = GemmSpecialization::Default;
+static constexpr auto GemmKPadding   = GemmSpecialization::KPadding;
+static constexpr auto GemmMNPadding  = GemmSpecialization::MNPadding;
+static constexpr auto GemmMNKPadding = GemmSpecialization::MNKPadding;
+
+static constexpr auto Intrawave = BlockGemmPipelineScheduler::Intrawave;
+static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
+
+template <typename DsLayout,
+          typename DsDataType,
+          typename CDEElementwiseOp,
+          GemmSpecialization GemmSpec = GemmMNKPadding>
+using device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_comp_instances = std::tuple<
+    // clang-format off
+        //###########################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|                C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
+        //###########################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise|      Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
+        //###########################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|        Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //###########################################|       |       |            |       |      |      |        |         |            |      |            |            |                 |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |   S<C,D0...,D_N| 
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   256,   256,    32,   8,   4,   32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   128,   128,    64,   8,   4,   32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,    S<16, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   256,   256,    32,   8,   4,   32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   256,   256,    32,   8,   4,   32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   224,   256,    64,   8,   4,   16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,    S<16, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,         0,           1,           2,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   128,   128,    64,   8,   4,   32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,    S<16, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   128,   256,    32,   8,   4,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   128,   128,    64,   8,   4,   32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,    S<16, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,8>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v1>
+    // clang-format on
+    >;
+
+template <typename DsLayout,
+          typename DsDataType,
+          typename CDEElementwiseOp,
+          GemmSpecialization GemmSpec                 = GemmMNKPadding,
+          BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave>
+using device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_mem_instances =
+    std::tuple<
+        // clang-format off
+        //###########################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|                C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
+        //###########################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise|      Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
+        //###########################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|        Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //###########################################|       |       |            |       |      |      |        |         |            |      |            |            |                 |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |   S<C,D0...,D_N| 
+        // Latency friendly
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,    64,    16,    16,   256,   8,   4,   16,   16,    1,    1,     S<32, 2, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<64, 1, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 4>,        S<4,4,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   128,    16,    32,   256,   8,   4,   16,   16,    1,    1,     S<32, 4, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<64, 2, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 8>,        S<4,4,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
+        // Memory friendly
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,    64,    16,    16,   256,   8,   4,   16,   16,    1,    1,     S<32, 2, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<64, 1, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 4>,        S<4,4,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   128,    16,    32,   256,   8,   4,   16,   16,    1,    1,     S<32, 4, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<64, 2, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 8>,        S<4,4,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   128,    16,    64,   128,   8,   4,   16,   16,    1,    2,     S<16, 8, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<32, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 8>,        S<4,4,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   128,    32,    64,   128,   8,   4,   32,   32,    1,    1,     S<16, 8, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<32, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 8>,        S<8,8,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   128,    16,   128,    64,   8,   4,   16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 8>,        S<4,4,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   128,    32,   128,    64,   8,   4,   32,   32,    1,    2,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 8>,        S<8,8,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,    16,   256,    64,   8,   4,   16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,    S<16, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,              S<1, 16, 1, 16>,        S<4,4,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,    32,   256,    64,   8,   4,   32,   32,    1,    2,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,    S<16, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,              S<1, 16, 1, 16>,        S<8,8,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>
+        // clang-format on
+        >;
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_default_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_default_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemmTileLoop<Row,
+                                                          Row,
+                                                          ck::Tuple<Row>,
+                                                          Row,
+                                                          BF16,
+                                                          I8,
+                                                          ck::Tuple<BF16>,
+                                                          BF16,
+                                                          PassThrough,
+                                                          PassThrough,
+                                                          Multiply>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_comp_instances<ck::Tuple<Row>,
+                                                                               ck::Tuple<BF16>,
+                                                                               Multiply,
+                                                                               GemmDefault>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_kpadding_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_kpadding_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemmTileLoop<Row,
+                                                          Row,
+                                                          ck::Tuple<Row>,
+                                                          Row,
+                                                          BF16,
+                                                          I8,
+                                                          ck::Tuple<BF16>,
+                                                          BF16,
+                                                          PassThrough,
+                                                          PassThrough,
+                                                          Multiply>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_comp_instances<ck::Tuple<Row>,
+                                                                               ck::Tuple<BF16>,
+                                                                               Multiply,
+                                                                               GemmKPadding>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_mnkpadding_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemmTileLoop<Row,
+                                                          Row,
+                                                          ck::Tuple<Row>,
+                                                          Row,
+                                                          BF16,
+                                                          I8,
+                                                          ck::Tuple<BF16>,
+                                                          BF16,
+                                                          PassThrough,
+                                                          PassThrough,
+                                                          Multiply>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_comp_instances<ck::Tuple<Row>,
+                                                                               ck::Tuple<BF16>,
+                                                                               Multiply,
+                                                                               GemmMNKPadding>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_mnpadding_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_mnpadding_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemmTileLoop<Row,
+                                                          Row,
+                                                          ck::Tuple<Row>,
+                                                          Row,
+                                                          BF16,
+                                                          I8,
+                                                          ck::Tuple<BF16>,
+                                                          BF16,
+                                                          PassThrough,
+                                                          PassThrough,
+                                                          Multiply>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_comp_instances<ck::Tuple<Row>,
+                                                                               ck::Tuple<BF16>,
+                                                                               Multiply,
+                                                                               GemmMNPadding>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_instance.cpp

@@ -31,51 +31,63 @@ using MultiplyAddFastGelu = ck::tensor_operation::element_wise::MultiplyAddFastG
 using MultiplyFastGelu    = ck::tensor_operation::element_wise::MultiplyFastGelu;
 using MultiplyAdd         = ck::tensor_operation::element_wise::MultiplyAdd;
 
-static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
-
-template <typename DsLayout, typename DsDataType, typename CDEElementwiseOp>
-using device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_irregular_tile_instances = std::tuple<
-// clang-format off
+static constexpr auto GemmDefault    = GemmSpecialization::Default;
+static constexpr auto GemmKPadding   = GemmSpecialization::KPadding;
+static constexpr auto GemmMNPadding  = GemmSpecialization::MNPadding;
+static constexpr auto GemmMNKPadding = GemmSpecialization::MNKPadding;
+
+static constexpr auto Intrawave = BlockGemmPipelineScheduler::Intrawave;
+static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
+
+template <typename DsLayout,
+          typename DsDataType,
+          typename CDEElementwiseOp,
+          GemmSpecialization GemmSpec = GemmMNKPadding>
+using device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_comp_instances = std::tuple<
+    // clang-format off
         //###########################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|                C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
         //###########################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise|      Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
         //###########################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|        Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
-        //###########################################|       |       |            |       |      |      |        |         |            |      |            |            |                 |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-#if 1
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   2,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   2,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 4>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 4>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   2,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,              8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>
-#endif
-#if 0
-	//comp
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   256,    32,   8,   4,  32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8>,
-
-	//latency
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,    64,    16,   16,   256,   8,   4,  16,   16,    1,    1,     S<32, 2, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<64, 1, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 4>,               4>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   128,    16,   32,   256,   8,   4,  16,   16,    1,    1,     S<32, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<64, 2, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4>,
-
-	//mem
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,    64,    16,   16,   256,   8,   4,  16,   16,    1,    1,     S<32, 2, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<64, 1, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 4>,               4>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   128,    16,   32,   256,   8,   4,  16,   16,    1,    1,     S<32, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<64, 2, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   128,    16,   64,   128,   8,   4,  16,   16,    1,    2,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<32, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   128,    32,   64,   128,   8,   4,  32,   32,    1,    1,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<32, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   128,    16,  128,    64,   8,   4,  16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   128,    32,  128,    64,   8,   4,  32,   32,    1,    2,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               8>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   256,    16,  256,    64,   8,   4,  16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 16>,              4>,
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<    Row,    Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp, GemmMNKPadding,        1,   256,    32,  256,    64,   8,   4,  32,   32,    1,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 16>,              8>
-#endif
+        //###########################################|       |       |            |       |      |      |        |         |            |      |            |            |                 |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |   S<C,D0...,D_N| 
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   256,   256,    32,   8,   4,   32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   128,   128,    64,   8,   4,   32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,    S<16, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   256,   256,    32,   8,   4,   32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   256,   256,    32,   8,   4,   32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   224,   256,    64,   8,   4,   16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,    S<16, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,         0,           1,           2,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   128,   128,    64,   8,   4,   32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,    S<16, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   128,   256,    32,   8,   4,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   128,   128,    64,   8,   4,   32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,    S<16, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v1>
+
     // clang-format on
     >;
 
+template <typename DsLayout,
+          typename DsDataType,
+          typename CDEElementwiseOp,
+          GemmSpecialization GemmSpec                 = GemmMNKPadding,
+          BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave>
+using device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_mem_instances =
+    std::tuple<
+        // clang-format off
+        //###########################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|                C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
+        //###########################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise|      Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
+        //###########################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|        Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //###########################################|       |       |            |       |      |      |        |         |            |      |            |            |                 |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |   S<C,D0...,D_N| 
+        // Latency friendly
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,    64,    16,    16,   256,   8,   4,   16,   16,    1,    1,     S<32, 2, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<64, 1, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 4>,        S<4,4,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   128,    16,    32,   256,   8,   4,   16,   16,    1,    1,     S<32, 4, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<64, 2, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 8>,        S<4,4,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
+        // Memory friendly
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,    64,    16,    16,   256,   8,   4,   16,   16,    1,    1,     S<32, 2, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<64, 1, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 4>,        S<4,4,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   128,    16,    32,   256,   8,   4,   16,   16,    1,    1,     S<32, 4, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<64, 2, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 8>,        S<4,4,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   128,    16,    64,   128,   8,   4,   16,   16,    1,    2,     S<16, 8, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<32, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 8>,        S<4,4,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   128,    32,    64,   128,   8,   4,   32,   32,    1,    1,     S<16, 8, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<32, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 8>,        S<8,8,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   128,    16,   128,    64,   8,   4,   16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 8>,        S<4,4,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   128,    32,   128,    64,   8,   4,   32,   32,    1,    2,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 8>,        S<8,8,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,    16,   256,    64,   8,   4,   16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,    S<16, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,              S<1, 16, 1, 16>,        S<4,4,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,    32,   256,    64,   8,   4,   32,   32,    1,    2,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,    S<16, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,              S<1, 16, 1, 16>,        S<8,8,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>
+        // clang-format on
+        >;
+
 void add_device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_instances(
     std::vector<std::unique_ptr<DeviceGroupedGemmTileLoop<Row,
                                                           Row,
@@ -89,33 +101,89 @@ void add_device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_instan
                                                           PassThrough,
                                                           Multiply>>>& instances)
 {
+    // comp
     add_device_operation_instances(
         instances,
-        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_irregular_tile_instances<
-            ck::Tuple<Row>,
-            ck::Tuple<BF16>,
-            Multiply>{});
-}
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_comp_instances<ck::Tuple<Row>,
+                                                                               ck::Tuple<BF16>,
+                                                                               Multiply,
+                                                                               GemmDefault>{});
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_comp_instances<ck::Tuple<Row>,
+                                                                               ck::Tuple<BF16>,
+                                                                               Multiply,
+                                                                               GemmMNKPadding>{});
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_comp_instances<ck::Tuple<Row>,
+                                                                               ck::Tuple<BF16>,
+                                                                               Multiply,
+                                                                               GemmMNPadding>{});
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_comp_instances<ck::Tuple<Row>,
+                                                                               ck::Tuple<BF16>,
+                                                                               Multiply,
+                                                                               GemmKPadding>{});
+    // mem
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_mem_instances<ck::Tuple<Row>,
+                                                                              ck::Tuple<BF16>,
+                                                                              Multiply,
+                                                                              GemmDefault,
+                                                                              Intrawave>{});
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_mem_instances<ck::Tuple<Row>,
+                                                                              ck::Tuple<BF16>,
+                                                                              Multiply,
+                                                                              GemmMNKPadding,
+                                                                              Intrawave>{});
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_mem_instances<ck::Tuple<Row>,
+                                                                              ck::Tuple<BF16>,
+                                                                              Multiply,
+                                                                              GemmMNPadding,
+                                                                              Intrawave>{});
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_mem_instances<ck::Tuple<Row>,
+                                                                              ck::Tuple<BF16>,
+                                                                              Multiply,
+                                                                              GemmKPadding,
+                                                                              Intrawave>{});
 
-void add_device_grouped_gemm_xdl_tile_loop_multiply_bias_bf16_i8_bf16_mk_kn_mn_instances(
-    std::vector<std::unique_ptr<DeviceGroupedGemmTileLoop<Row,
-                                                          Row,
-                                                          ck::Tuple<Row, Row>,
-                                                          Row,
-                                                          BF16,
-                                                          I8,
-                                                          ck::Tuple<BF16, BF16>,
-                                                          BF16,
-                                                          PassThrough,
-                                                          PassThrough,
-                                                          MultiplyAdd>>>& instances)
-{
     add_device_operation_instances(
         instances,
-        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_irregular_tile_instances<
-            ck::Tuple<Row, Row>,
-            ck::Tuple<BF16, BF16>,
-            MultiplyAdd>{});
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_mem_instances<ck::Tuple<Row>,
+                                                                              ck::Tuple<BF16>,
+                                                                              Multiply,
+                                                                              GemmDefault,
+                                                                              Interwave>{});
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_mem_instances<ck::Tuple<Row>,
+                                                                              ck::Tuple<BF16>,
+                                                                              Multiply,
+                                                                              GemmMNKPadding,
+                                                                              Interwave>{});
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_mem_instances<ck::Tuple<Row>,
+                                                                              ck::Tuple<BF16>,
+                                                                              Multiply,
+                                                                              GemmMNPadding,
+                                                                              Interwave>{});
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_mem_instances<ck::Tuple<Row>,
+                                                                              ck::Tuple<BF16>,
+                                                                              Multiply,
+                                                                              GemmKPadding,
+                                                                              Interwave>{});
 }
 
 void add_device_grouped_gemm_xdl_tile_loop_multiply_bias_fastgelu_bf16_i8_bf16_mk_kn_mn_instances(

library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_mem_v1_default_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_mem_v1_default_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemmTileLoop<Row,
+                                                          Row,
+                                                          ck::Tuple<Row>,
+                                                          Row,
+                                                          BF16,
+                                                          I8,
+                                                          ck::Tuple<BF16>,
+                                                          BF16,
+                                                          PassThrough,
+                                                          PassThrough,
+                                                          Multiply>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_mem_instances<ck::Tuple<Row>,
+                                                                              ck::Tuple<BF16>,
+                                                                              Multiply,
+                                                                              GemmDefault,
+                                                                              Intrawave>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_mem_v1_kpadding_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_mem_v1_kpadding_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemmTileLoop<Row,
+                                                          Row,
+                                                          ck::Tuple<Row>,
+                                                          Row,
+                                                          BF16,
+                                                          I8,
+                                                          ck::Tuple<BF16>,
+                                                          BF16,
+                                                          PassThrough,
+                                                          PassThrough,
+                                                          Multiply>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_mem_instances<ck::Tuple<Row>,
+                                                                              ck::Tuple<BF16>,
+                                                                              Multiply,
+                                                                              GemmKPadding,
+                                                                              Intrawave>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck

library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_mem_v1_mnkpadding_instance.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_mem_v1_mnkpadding_instances(
+    std::vector<std::unique_ptr<DeviceGroupedGemmTileLoop<Row,
+                                                          Row,
+                                                          ck::Tuple<Row>,
+                                                          Row,
+                                                          BF16,
+                                                          I8,
+                                                          ck::Tuple<BF16>,
+                                                          BF16,
+                                                          PassThrough,
+                                                          PassThrough,
+                                                          Multiply>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_mem_instances<ck::Tuple<Row>,
+                                                                              ck::Tuple<BF16>,
+                                                                              Multiply,
+                                                                              GemmMNKPadding,
+                                                                              Intrawave>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck