[CK_TILE] Add PagedAttention kernels

example/ck_tile/01_fmha/CMakeLists.txt

@@ -1,7 +1,27 @@
-# generate a list of kernels, but not actually emit files at config stage
+# validate user-specified fmha_fwd API list
+set(FMHA_FWD_KNOWN_APIS "fwd;fwd_splitkv;fwd_appendkv")
+set(FMHA_FWD_ENABLE_APIS "fwd" CACHE STRING
+    "semicolon-separated list of APIs to generate (${FMHA_FWD_KNOWN_APIS}) & link, or \"all\".")
+if(FMHA_FWD_ENABLE_APIS STREQUAL "all")
+  set(FMHA_FWD_ENABLE_APIS ${FMHA_FWD_KNOWN_APIS})
+endif()
+
+foreach(api ${FMHA_FWD_ENABLE_APIS})
+  if(NOT "${api}" IN_LIST FMHA_FWD_KNOWN_APIS)
+    message(FATAL_ERROR "${api} isn't a known api: ${FMHA_FWD_KNOWN_APIS}.")
+  endif()
+endforeach()
+
+# "fwd" is a must-have api for the fmha_fwd example, add it if not specified
+if(NOT "fwd" IN_LIST FMHA_FWD_ENABLE_APIS)
+  list(APPEND FMHA_FWD_ENABLE_APIS "fwd")
+endif()
+
+string(REPLACE ";" "," FMHA_FWD_APIS "${FMHA_FWD_ENABLE_APIS}")
+# generate a list of kernels, but not actually emit files at config sta
 execute_process(
   COMMAND ${Python3_EXECUTABLE} ${CMAKE_CURRENT_LIST_DIR}/generate.py
-  --api fwd,fwd_splitkv --list_blobs ${CMAKE_CURRENT_BINARY_DIR}/fwd_blob_list.txt
+  --api ${FMHA_FWD_APIS} --list_blobs ${CMAKE_CURRENT_BINARY_DIR}/fwd_blob_list.txt
 )
 
 execute_process(
@@ -17,7 +37,7 @@ file(STRINGS ${CMAKE_CURRENT_BINARY_DIR}/bwd_blob_list.txt FMHA_BWD_GEN_BLOBS)
 add_custom_command(
   OUTPUT ${FMHA_FWD_GEN_BLOBS}
   COMMAND ${Python3_EXECUTABLE} ${CMAKE_CURRENT_LIST_DIR}/generate.py
-  --api fwd,fwd_splitkv --output_dir ${CMAKE_CURRENT_BINARY_DIR}
+  --api ${FMHA_FWD_APIS} --output_dir ${CMAKE_CURRENT_BINARY_DIR}
 )
 
 add_custom_command(
@@ -60,6 +80,20 @@ else()
 endif()
 list(APPEND EXAMPLE_FMHA_BWD_COMPILE_OPTIONS -Wno-undefined-func-template -fgpu-flush-denormals-to-zero)
 
+# conditionally enable call to the fwd_splitkv API in fmha_fwd example
+if("fwd_splitkv" IN_LIST FMHA_FWD_ENABLE_APIS)
+  list(APPEND EXAMPLE_FMHA_FWD_COMPILE_OPTIONS -DCK_TILE_FMHA_FWD_SPLITKV_API=1)
+else()
+  list(APPEND EXAMPLE_FMHA_FWD_COMPILE_OPTIONS -DCK_TILE_FMHA_FWD_SPLITKV_API=0)
+endif()
+
+# conditionally enable call to the fwd_appendkv API in fmha_fwd example
+if("fwd_appendkv" IN_LIST FMHA_FWD_ENABLE_APIS)
+  list(APPEND EXAMPLE_FMHA_FWD_COMPILE_OPTIONS -DCK_TILE_FMHA_FWD_APPENDKV_API=1)
+else()
+  list(APPEND EXAMPLE_FMHA_FWD_COMPILE_OPTIONS -DCK_TILE_FMHA_FWD_APPENDKV_API=0)
+endif()
+
 # Allow comparing floating points directly in order to check sentinel values
 list(APPEND EXAMPLE_FMHA_FWD_COMPILE_OPTIONS -Wno-float-equal)
 list(APPEND EXAMPLE_FMHA_BWD_COMPILE_OPTIONS -Wno-float-equal)

example/ck_tile/01_fmha/codegen/cpp_symbol_map.py

@@ -82,6 +82,18 @@ def get_mask_check_map(mask : str):
     "dropout_wg16_storerandval" : "t.has_dropout == true && t.is_store_randval == true",
 }
 
+ROPE_MAP = {
+    "no" : "ck_tile::RotaryEmbeddingEnum::NONE",
+    "inter"  : "ck_tile::RotaryEmbeddingEnum::INTERLEAVED",
+    "half" : "ck_tile::RotaryEmbeddingEnum::HALF_ROTATED"
+}
+
+ROPE_CHECK_MAP = {
+    "no"    : "rope_enum::none",
+    "inter" : "rope_enum::interleaved",
+    "half"  : "rope_enum::half_rotated"
+}
+
 MODE_MAP = {
     "batch" : "false",
     "group" : "true"
@@ -105,4 +117,4 @@ def get_mask_check_map(mask : str):
 BOOL_MAP = {
     "t" : "true",
     "f" : "false"
-}
+}