Add mechanism to build CK for select data types, add Navi3x CI. (#790)

* allow building CK for specific data types

* add CI build and test stage on Naiv3x without some int8 instances

* add missing gemm fp16 instances

* add the changes to the missed cmake file

* add empty lines at end of source files

* Do not build quantization client example on navi3 in CI

* disable batched_gemm_multi_d_int8 instances with DTYPES

* disable device_conv2d_bwd_data_instance with DTYPES

* fix ckprofiler for conv_bwd_data for int8

* properly isolate the conv_bwd_data int8 instances

* remove empty line

[ROCm/composable_kernel commit: 189ea3b9aa]

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

@@ -1,84 +1,93 @@
-add_instance_library(device_gemm_instance
-   device_gemm_dl_f16_f16_f16_km_kn_mn_instance.cpp
-   device_gemm_dl_f16_f16_f16_km_kn_mn_irregular_instance.cpp
-   device_gemm_dl_f16_f16_f16_km_nk_mn_instance.cpp
-   device_gemm_dl_f16_f16_f16_km_nk_mn_irregular_instance.cpp
-   device_gemm_dl_f16_f16_f16_mk_kn_mn_instance.cpp
-   device_gemm_dl_f16_f16_f16_mk_kn_mn_irregular_instance.cpp
-   device_gemm_dl_f16_f16_f16_mk_nk_mn_instance.cpp
-   device_gemm_dl_f16_f16_f16_mk_nk_mn_irregular_instance.cpp
-   device_gemm_dl_f32_f32_f32_km_kn_mn_instance.cpp
-   device_gemm_dl_f32_f32_f32_km_nk_mn_instance.cpp
-   device_gemm_dl_f32_f32_f32_mk_kn_mn_instance.cpp
-   device_gemm_dl_f32_f32_f32_mk_nk_mn_instance.cpp
-   device_gemm_dl_i8_i8_i8_km_kn_mn_instance.cpp
-   device_gemm_dl_i8_i8_i8_km_kn_mn_irregular_instance.cpp
-   device_gemm_dl_i8_i8_i8_km_nk_mn_instance.cpp
-   device_gemm_dl_i8_i8_i8_km_nk_mn_irregular_instance.cpp
-   device_gemm_dl_i8_i8_i8_mk_kn_mn_instance.cpp
-   device_gemm_dl_i8_i8_i8_mk_kn_mn_irregular_instance.cpp
-   device_gemm_dl_i8_i8_i8_mk_nk_mn_instance.cpp
-   device_gemm_dl_i8_i8_i8_mk_nk_mn_irregular_instance.cpp
-   device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instance.cpp
-   device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_kn_mn_instance.cpp
-   device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_nk_mn_instance.cpp
-   device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instance.cpp
-   device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instance.cpp
-   device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp
-   device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instance.cpp
-   device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp
-   device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instance.cpp
-   device_gemm_xdl_c_shuffle_f32_f32_f32_km_kn_mn_instance.cpp
-   device_gemm_xdl_c_shuffle_f32_f32_f32_km_nk_mn_instance.cpp
-   device_gemm_xdl_c_shuffle_f32_f32_f32_mk_kn_mn_instance.cpp
-   device_gemm_xdl_c_shuffle_f32_f32_f32_mk_nk_mn_instance.cpp
-   device_gemm_xdl_c_shuffle_i8_i8_i8_km_kn_mn_instance.cpp
-   device_gemm_xdl_c_shuffle_i8_i8_i8_km_nk_mn_instance.cpp
-   device_gemm_xdl_c_shuffle_i8_i8_i8_mk_kn_mn_instance.cpp
-   device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instance.cpp
+set(GEMM_INSTANCES)
+if(DTYPES MATCHES "fp64" OR NOT DEFINED DTYPES)
+    list(APPEND GEMM_INSTANCES device_gemm_xdl_f64_f64_f64_mk_kn_mn_instance.cpp)
+    list(APPEND GEMM_INSTANCES device_gemm_xdl_f64_f64_f64_mk_nk_mn_instance.cpp)
+    list(APPEND GEMM_INSTANCES device_gemm_xdl_f64_f64_f64_km_kn_mn_instance.cpp)
+    list(APPEND GEMM_INSTANCES device_gemm_xdl_f64_f64_f64_km_nk_mn_instance.cpp)
+endif()
+if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f32_f32_f32_mk_kn_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f32_f32_f32_mk_nk_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f32_f32_f32_km_kn_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f32_f32_f32_km_nk_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_f32_f32_f32_mk_kn_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_f32_f32_f32_mk_nk_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_f32_f32_f32_km_kn_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_f32_f32_f32_km_nk_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_f32_f32_f32_mk_kn_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_f32_f32_f32_mk_nk_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_f32_f32_f32_km_kn_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_f32_f32_f32_km_nk_mn_instance.cpp)
+endif()
+if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_f16_f16_f16_mk_kn_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_f16_f16_f16_mk_kn_mn_irregular_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_f16_f16_f16_mk_nk_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_f16_f16_f16_mk_nk_mn_irregular_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_f16_f16_f16_km_kn_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_f16_f16_f16_km_kn_mn_irregular_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_f16_f16_f16_km_nk_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_f16_f16_f16_km_nk_mn_irregular_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/km_kn_mn_add_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/km_kn_mn_default_pipeline_v1_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/km_kn_mn_default_pipeline_v2_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/km_kn_mn_default_pipeline_v2_opt_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/km_kn_mn_interwave_pipeline_v1_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/km_kn_mn_irregular_default_pipeline_v1_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/km_kn_mn_irregular_default_pipeline_v2_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/km_kn_mn_irregular_interwave_pipeline_v1_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/km_nk_mn_add_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/km_nk_mn_default_pipeline_v1_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/km_nk_mn_default_pipeline_v2_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/km_nk_mn_default_pipeline_v2_opt_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/km_nk_mn_interwave_pipeline_v1_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/km_nk_mn_irregular_default_pipeline_v1_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/km_nk_mn_irregular_default_pipeline_v2_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/km_nk_mn_irregular_interwave_pipeline_v1_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/mk_kn_mn_add_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/mk_kn_mn_default_pipeline_v1_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/mk_kn_mn_default_pipeline_v2_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/mk_kn_mn_default_pipeline_v2_opt_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/mk_kn_mn_interwave_pipeline_v1_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/mk_kn_mn_irregular_default_pipeline_v1_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/mk_kn_mn_irregular_default_pipeline_v2_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/mk_kn_mn_irregular_interwave_pipeline_v1_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/mk_nk_mn_add_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/mk_nk_mn_default_pipeline_v1_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/mk_nk_mn_default_pipeline_v2_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/mk_nk_mn_default_pipeline_v2_opt_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/mk_nk_mn_interwave_pipeline_v1_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/mk_nk_mn_irregular_default_pipeline_v1_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/mk_nk_mn_irregular_default_pipeline_v2_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_f16_f16_f16/mk_nk_mn_irregular_interwave_pipeline_v1_instance.cpp)
+endif()
+if(DTYPES MATCHES "int8" OR NOT DEFINED DTYPES)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_i8_i8_i8_mk_kn_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_i8_i8_i8_mk_kn_mn_irregular_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_i8_i8_i8_mk_nk_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_i8_i8_i8_mk_nk_mn_irregular_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_i8_i8_i8_km_kn_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_i8_i8_i8_km_kn_mn_irregular_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_i8_i8_i8_km_nk_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_dl_i8_i8_i8_km_nk_mn_irregular_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_i8_i8_i8_mk_kn_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_i8_i8_i8_km_kn_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_i8_i8_i8_km_nk_mn_instance.cpp)
+endif()
+if(DTYPES MATCHES "bf16" OR NOT DEFINED DTYPES)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_kn_mn_instance.cpp)
+   list(APPEND GEMM_INSTANCES device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_nk_mn_instance.cpp)
+endif()
 
-   device_gemm_xdl_f16_f16_f16/km_kn_mn_add_instance.cpp
-   device_gemm_xdl_f16_f16_f16/km_kn_mn_default_pipeline_v1_instance.cpp
-   device_gemm_xdl_f16_f16_f16/km_kn_mn_default_pipeline_v2_instance.cpp
-   device_gemm_xdl_f16_f16_f16/km_kn_mn_default_pipeline_v2_opt_instance.cpp
-   device_gemm_xdl_f16_f16_f16/km_kn_mn_interwave_pipeline_v1_instance.cpp
-   device_gemm_xdl_f16_f16_f16/km_kn_mn_irregular_default_pipeline_v1_instance.cpp
-   device_gemm_xdl_f16_f16_f16/km_kn_mn_irregular_default_pipeline_v2_instance.cpp
-   device_gemm_xdl_f16_f16_f16/km_kn_mn_irregular_interwave_pipeline_v1_instance.cpp
-   device_gemm_xdl_f16_f16_f16/km_nk_mn_add_instance.cpp
-   device_gemm_xdl_f16_f16_f16/km_nk_mn_default_pipeline_v1_instance.cpp
-   device_gemm_xdl_f16_f16_f16/km_nk_mn_default_pipeline_v2_instance.cpp
-   device_gemm_xdl_f16_f16_f16/km_nk_mn_default_pipeline_v2_opt_instance.cpp
-   device_gemm_xdl_f16_f16_f16/km_nk_mn_interwave_pipeline_v1_instance.cpp
-   device_gemm_xdl_f16_f16_f16/km_nk_mn_irregular_default_pipeline_v1_instance.cpp
-   device_gemm_xdl_f16_f16_f16/km_nk_mn_irregular_default_pipeline_v2_instance.cpp
-   device_gemm_xdl_f16_f16_f16/km_nk_mn_irregular_interwave_pipeline_v1_instance.cpp
-   device_gemm_xdl_f16_f16_f16/mk_kn_mn_add_instance.cpp
-   device_gemm_xdl_f16_f16_f16/mk_kn_mn_default_pipeline_v1_instance.cpp
-   device_gemm_xdl_f16_f16_f16/mk_kn_mn_default_pipeline_v2_instance.cpp
-   device_gemm_xdl_f16_f16_f16/mk_kn_mn_default_pipeline_v2_opt_instance.cpp
-   device_gemm_xdl_f16_f16_f16/mk_kn_mn_interwave_pipeline_v1_instance.cpp
-   device_gemm_xdl_f16_f16_f16/mk_kn_mn_irregular_default_pipeline_v1_instance.cpp
-   device_gemm_xdl_f16_f16_f16/mk_kn_mn_irregular_default_pipeline_v2_instance.cpp
-   device_gemm_xdl_f16_f16_f16/mk_kn_mn_irregular_interwave_pipeline_v1_instance.cpp
-   device_gemm_xdl_f16_f16_f16/mk_nk_mn_add_instance.cpp
-   device_gemm_xdl_f16_f16_f16/mk_nk_mn_default_pipeline_v1_instance.cpp
-   device_gemm_xdl_f16_f16_f16/mk_nk_mn_default_pipeline_v2_instance.cpp
-   device_gemm_xdl_f16_f16_f16/mk_nk_mn_default_pipeline_v2_opt_instance.cpp
-   device_gemm_xdl_f16_f16_f16/mk_nk_mn_interwave_pipeline_v1_instance.cpp
-   device_gemm_xdl_f16_f16_f16/mk_nk_mn_irregular_default_pipeline_v1_instance.cpp
-   device_gemm_xdl_f16_f16_f16/mk_nk_mn_irregular_default_pipeline_v2_instance.cpp
-   device_gemm_xdl_f16_f16_f16/mk_nk_mn_irregular_interwave_pipeline_v1_instance.cpp
-
-   device_gemm_xdl_f32_f32_f32_km_kn_mn_instance.cpp
-   device_gemm_xdl_f32_f32_f32_km_nk_mn_instance.cpp
-   device_gemm_xdl_f32_f32_f32_mk_kn_mn_instance.cpp
-   device_gemm_xdl_f32_f32_f32_mk_nk_mn_instance.cpp
-   device_gemm_xdl_f64_f64_f64_km_kn_mn_instance.cpp
-   device_gemm_xdl_f64_f64_f64_km_nk_mn_instance.cpp
-   device_gemm_xdl_f64_f64_f64_mk_kn_mn_instance.cpp
-   device_gemm_xdl_f64_f64_f64_mk_nk_mn_instance.cpp
-)
+add_instance_library(device_gemm_instance ${GEMM_INSTANCES})
 
 set(ENABLE_PIPELINE_V2_OPT OFF)
 

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_i8_i8_i8_km_kn_mn_instance.cpp

@@ -8,7 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_dl.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
-
+#ifdef __int8__
 namespace ck {
 namespace tensor_operation {
 namespace device {
@@ -80,3 +80,4 @@ void add_device_gemm_dl_i8_i8_i8_km_kn_mn_instances(
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
+#endif

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_i8_i8_i8_km_kn_mn_irregular_instance.cpp

@@ -8,7 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_dl.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
-
+#ifdef __int8__
 namespace ck {
 namespace tensor_operation {
 namespace device {
@@ -78,3 +78,4 @@ void add_device_gemm_dl_i8_i8_i8_km_kn_mn_irregular_instances(
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
+#endif

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_i8_i8_i8_km_nk_mn_instance.cpp

@@ -8,7 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_dl.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
-
+#ifdef __int8__
 namespace ck {
 namespace tensor_operation {
 namespace device {
@@ -80,3 +80,4 @@ void add_device_gemm_dl_i8_i8_i8_km_nk_mn_instances(
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
+#endif

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_i8_i8_i8_km_nk_mn_irregular_instance.cpp

@@ -8,7 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_dl.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
-
+#ifdef __int8__
 namespace ck {
 namespace tensor_operation {
 namespace device {
@@ -78,3 +78,4 @@ void add_device_gemm_dl_i8_i8_i8_km_nk_mn_irregular_instances(
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
+#endif

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_i8_i8_i8_mk_kn_mn_instance.cpp

@@ -8,7 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_dl.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
-
+#ifdef __int8__
 namespace ck {
 namespace tensor_operation {
 namespace device {
@@ -80,3 +80,4 @@ void add_device_gemm_dl_i8_i8_i8_mk_kn_mn_instances(
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
+#endif

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_i8_i8_i8_mk_kn_mn_irregular_instance.cpp

@@ -8,7 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_dl.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
-
+#ifdef __int8__
 namespace ck {
 namespace tensor_operation {
 namespace device {
@@ -78,3 +78,4 @@ void add_device_gemm_dl_i8_i8_i8_mk_kn_mn_irregular_instances(
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
+#endif

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_i8_i8_i8_mk_nk_mn_instance.cpp

@@ -8,7 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_dl.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
-
+#ifdef __int8__
 namespace ck {
 namespace tensor_operation {
 namespace device {
@@ -80,3 +80,4 @@ void add_device_gemm_dl_i8_i8_i8_mk_nk_mn_instances(
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
+#endif

library/src/tensor_operation_instance/gpu/gemm/device_gemm_dl_i8_i8_i8_mk_nk_mn_irregular_instance.cpp

@@ -8,7 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_dl.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
-
+#ifdef __int8__
 namespace ck {
 namespace tensor_operation {
 namespace device {
@@ -78,3 +78,4 @@ void add_device_gemm_dl_i8_i8_i8_mk_nk_mn_irregular_instances(
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
+#endif

library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_i8_i8_i8_km_kn_mn_instance.cpp

@@ -8,7 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
-
+#ifdef __int8__
 namespace ck {
 namespace tensor_operation {
 namespace device {
@@ -66,3 +66,4 @@ void add_device_gemm_xdl_c_shuffle_i8_i8_i8_km_kn_mn_instances(
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
+#endif

library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_i8_i8_i8_km_nk_mn_instance.cpp

@@ -8,7 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
-
+#ifdef __int8__
 namespace ck {
 namespace tensor_operation {
 namespace device {
@@ -66,3 +66,4 @@ void add_device_gemm_xdl_c_shuffle_i8_i8_i8_km_nk_mn_instances(
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
+#endif

library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_i8_i8_i8_mk_kn_mn_instance.cpp

@@ -8,7 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
-
+#ifdef __int8__
 namespace ck {
 namespace tensor_operation {
 namespace device {
@@ -66,3 +66,4 @@ void add_device_gemm_xdl_c_shuffle_i8_i8_i8_mk_kn_mn_instances(
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
+#endif

library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instance.cpp

@@ -8,7 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
-
+#ifdef __int8__
 namespace ck {
 namespace tensor_operation {
 namespace device {
@@ -63,3 +63,4 @@ void add_device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instances(
 } // namespace device
 } // namespace tensor_operation
 } // namespace ck
+#endif