Support for bindless mode in L0 - improvements

Related-To: NEO-6448

- add new IGC compilation flag when bindless mode used
- fix SBA programming of BindlessSurfaceStateSize -
always set maximum surface state count
- fix residency of global DSH heap on gen9 - gen12lp
in bindless mode
- add L0 aub test with bindless kernel - disabled
- partial fixes in OCL aub tests


Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
This commit is contained in:
Mateusz Hoppe
2021-11-30 22:59:19 +00:00
committed by Compute-Runtime-Automation
parent 55959d4d1d
commit 8b233f7f45
21 changed files with 352 additions and 19 deletions

View File

@@ -13,6 +13,14 @@ set(TEST_MODULES
${CMAKE_CURRENT_SOURCE_DIR}/test_modules/test_kernel.cl
)
set(TEST_KERNEL_BINDLESS_internal_options
"-cl-intel-use-bindless-mode -cl-intel-use-bindless-advanced-mode"
)
set(TEST_KERNEL_BINDLESS
${CMAKE_CURRENT_SOURCE_DIR}/test_modules/stateful_copy_buffer.cl
)
set(l0_test_kernels_outputs)
macro(macro_for_each_core_type)
foreach(PLATFORM_TYPE ${PLATFORM_TYPES})
@@ -20,12 +28,18 @@ macro(macro_for_each_core_type)
get_family_name_with_type(${CORE_TYPE} ${PLATFORM_TYPE})
string(TOLOWER ${PLATFORM_TYPE} PLATFORM_TYPE_LOWER)
set(PLATFORM_LOWER ${DEFAULT_SUPPORTED_${CORE_TYPE}_${PLATFORM_TYPE}_PLATFORM})
string(TOLOWER ${CORE_TYPE} CORE_TYPE_LOWER)
level_zero_generate_kernels(l0_test_kernel_outputs ${PLATFORM_LOWER} ${family_name_with_type} "-g" ${TEST_MODULES})
#skip Gen8 bindless kernel generation
if(NOT ("${CORE_TYPE_LOWER}" STREQUAL "gen8"))
level_zero_generate_kernels_with_internal_options(l0_bindless_test_kernel_outputs ${PLATFORM_LOWER} ${family_name_with_type} "bindless" "-g" ${TEST_KERNEL_BINDLESS_internal_options} ${TEST_KERNEL_BINDLESS})
endif()
endif()
endforeach()
endmacro()
apply_macro_for_each_core_type("TESTED")
add_custom_target(l0_common_test_kernels DEPENDS ${l0_test_kernel_outputs} copy_compiler_files)
add_custom_target(l0_common_test_kernels DEPENDS ${l0_test_kernel_outputs} ${l0_bindless_test_kernel_outputs} copy_compiler_files)
set_target_properties(l0_common_test_kernels PROPERTIES FOLDER ${TARGET_NAME_L0})
add_dependencies(prepare_test_kernels_for_l0 l0_common_test_kernels)

View File

@@ -49,3 +49,54 @@ function(level_zero_generate_kernels target_list platform_name suffix options)
set(${target_list} ${${target_list}} PARENT_SCOPE)
endfunction()
function(level_zero_generate_kernels_with_internal_options target_list platform_name suffix prefix options internal_options)
list(APPEND results copy_compiler_files)
set(outputdir "${TargetDir}/level_zero/${suffix}/test_files/${NEO_ARCH}/")
foreach(filepath ${ARGN})
get_filename_component(filename ${filepath} NAME)
get_filename_component(basename ${filepath} NAME_WE)
get_filename_component(workdir ${filepath} DIRECTORY)
set(outputpath_base "${outputdir}${prefix}_${basename}_${suffix}")
if(NOT NEO_DISABLE_BUILTINS_COMPILATION)
set(output_files
${outputpath_base}.bin
${outputpath_base}.gen
${outputpath_base}.spv
${outputpath_base}.dbg
)
set(output_name "-output" "${prefix}_${basename}")
string(CONCAT options \" ${options} \" )
string(CONCAT internal_options \" ${internal_options} \" )
add_custom_command(
COMMAND echo generate ${ocloc_cmd_prefix} -q -file ${filename} -device ${platform_name} -out_dir ${outputdir} ${output_name} -options ${options} -internal_options ${internal_options} , workdir is ${workdir}
OUTPUT ${output_files}
COMMAND ${ocloc_cmd_prefix} -q -file ${filename} -device ${platform_name} -out_dir ${outputdir} ${output_name} -options ${options} -internal_options ${internal_options}
WORKING_DIRECTORY ${workdir}
DEPENDS ${filepath} ocloc
)
list(APPEND ${target_list} ${output_files})
else()
foreach(_file_name "bin" "gen" "spv" "dbg")
set(_file_prebuilt "${NEO_SOURCE_DIR}/../neo_test_kernels/level_zero/${suffix}/test_files/${NEO_ARCH}/${prefix}_${basename}_${suffix}.${_file_name}")
add_custom_command(
OUTPUT ${outputpath_base}.${_file_name}
COMMAND ${CMAKE_COMMAND} -E make_directory ${outputdir}
COMMAND ${CMAKE_COMMAND} -E copy_if_different ${_file_prebuilt} ${outputdir}
)
list(APPEND ${target_list} ${outputpath_base}.${_file_name})
endforeach()
endif()
endforeach()
set(${target_list} ${${target_list}} PARENT_SCOPE)
endfunction()

View File

@@ -0,0 +1,14 @@
/*
* Copyright (C) 2021 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
*/
__kernel void StatefulCopyBuffer(
const __global uchar* src,
__global uchar* dst)
{
uint id = get_global_id(0);
dst[id] = src[id];
}