merge from develop and revisison for pr#881

29448ffd · Harisankar Sadasivan · 9223a5e2 · 8f84a012 · 29448ffd · 29448ffd
Commit 29448ffd authored Sep 08, 2023 by Harisankar Sadasivan
20 changed files
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/CMakeLists.txt
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/CMakeLists.txt
+if(DTYPES MATCHES "fp16" OR DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
 add_instance_library(device_batched_gemm_reduce_instance
    device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gmk_gkn_gmn_instance.cpp
    device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gmk_gnk_gmn_instance.cpp
    device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gkm_gkn_gmn_instance.cpp
    device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gkm_gnk_gmn_instance.cpp
 )
-
+endif()
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gkm_gkn_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gkm_gkn_gmn_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gkm_gnk_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gkm_gnk_gmn_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gmk_gkn_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gmk_gkn_gmn_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gmk_gnk_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_reduce/device_batched_gemm_reduce_xdl_cshuffle_f16_f16_f16_f32_f32_gmk_gnk_gmn_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm/CMakeLists.txt
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm/CMakeLists.txt
+if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
 add_instance_library(device_batched_gemm_softmax_gemm_instance
    device_batched_gemm_softmax_gemm_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
 )
-
+endif()
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm/device_batched_gemm_softmax_gemm_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm/device_batched_gemm_softmax_gemm_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/CMakeLists.txt
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/CMakeLists.txt
-add_instance_library(device_batched_gemm_softmax_gemm_permute_instance
-    device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
-    device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp
-    device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
-    device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp
-)
+set(DEVICE_BATCHED_GEMM_SOFTMAX_GEMM_PERMUTE_INSTANCES)
+if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
+    list(APPEND DEVICE_BATCHED_GEMM_SOFTMAX_GEMM_PERMUTE_INSTANCES device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp)
+    list(APPEND DEVICE_BATCHED_GEMM_SOFTMAX_GEMM_PERMUTE_INSTANCES device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp)
+endif()
+if(DTYPES MATCHES "bf16" OR NOT DEFINED DTYPES)
+    list(APPEND DEVICE_BATCHED_GEMM_SOFTMAX_GEMM_PERMUTE_INSTANCES device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp)
+    list(APPEND DEVICE_BATCHED_GEMM_SOFTMAX_GEMM_PERMUTE_INSTANCES device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp)
+endif()
+add_instance_library(device_batched_gemm_softmax_gemm_permute_instance ${DEVICE_BATCHED_GEMM_SOFTMAX_GEMM_PERMUTE_INSTANCES})

--- a/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/batchnorm/CMakeLists.txt
+++ b/library/src/tensor_operation_instance/gpu/batchnorm/CMakeLists.txt
--- a/library/src/tensor_operation_instance/gpu/batchnorm/device_batchnorm_backward_bf16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batchnorm/device_batchnorm_backward_bf16_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/batchnorm/device_batchnorm_backward_f16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batchnorm/device_batchnorm_backward_f16_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/batchnorm/device_batchnorm_backward_f32_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batchnorm/device_batchnorm_backward_f32_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/batchnorm/device_batchnorm_backward_f64_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batchnorm/device_batchnorm_backward_f64_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/batchnorm/device_batchnorm_forward_bf16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batchnorm/device_batchnorm_forward_bf16_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/batchnorm/device_batchnorm_forward_f16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batchnorm/device_batchnorm_forward_f16_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/batchnorm/device_batchnorm_forward_f32_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batchnorm/device_batchnorm_forward_f32_instance.cpp