ahmed/OpenCL-CTS
1
0
Fork 0
mirror of https://github.com/KhronosGroup/OpenCL-CTS.git synced 2026-03-20 14:39:01 +00:00
Code Packages Projects Releases Wiki Activity

Synchronise with Khronos-private Gitlab branch

Browse Source 
The maintenance of the conformance tests is moving to Github.

This commit contains all the changes that have been done in
Gitlab since the first public release of the conformance tests.

Signed-off-by: Kevin Petit <kevin.petit@arm.com>
This commit is contained in:
Kevin Petit
2019-02-20 16:36:05 +00:00
committed by Kévin Petit
parent 95196e7fb4
commit d8733efc0f
576 changed files with 212486 additions and 191776 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
test_conformance/api/CMakeLists.txt
View File

@@ -14,6 +14,7 @@ add_executable(conformance_test_api
test_platform.cpp
test_retain.cpp
test_device_min_data_type_align_size_alignment.cpp
test_queue_properties.cpp
test_mem_objects.cpp
test_bool.c
test_null_buffer_arg.c

54
test_conformance/api/Jamfile
View File

@@ -1,27 +1,27 @@
project
: requirements
<toolset>gcc:<cflags>-xc++
<toolset>msvc:<cflags>"/TP"
;
exe test_api
: main.c
test_api_min_max.c
test_binary.cpp
test_create_kernels.c
test_create_context_from_type.cpp
test_kernel_arg_changes.cpp
test_kernel_arg_multi_setup.cpp
test_kernels.c
test_native_kernel.cpp
test_queries.cpp
test_retain_program.c
test_platform.cpp
;
install dist
: test_api #test.lst
: <variant>debug:<location>$(DIST)/debug/tests/test_conformance/api
<variant>release:<location>$(DIST)/release/tests/test_conformance/api
;
project
: requirements
<toolset>gcc:<cflags>-xc++
<toolset>msvc:<cflags>"/TP"
;
exe test_api
: main.c
test_api_min_max.c
test_binary.cpp
test_create_kernels.c
test_create_context_from_type.cpp
test_kernel_arg_changes.cpp
test_kernel_arg_multi_setup.cpp
test_kernels.c
test_native_kernel.cpp
test_queries.cpp
test_retain_program.c
test_platform.cpp
;
install dist
: test_api #test.lst
: <variant>debug:<location>$(DIST)/debug/tests/test_conformance/api
<variant>release:<location>$(DIST)/release/tests/test_conformance/api
;

122
test_conformance/api/Makefile
View File

@@ -1,61 +1,61 @@
ifdef BUILD_WITH_ATF
ATF = -framework ATF
USE_ATF = -DUSE_ATF
endif
SRCS = main.c \
test_retain_program.c \
test_queries.cpp \
test_create_kernels.c \
test_kernels.c \
test_kernel_arg_info.c \
test_api_min_max.c \
test_kernel_arg_changes.cpp \
test_kernel_arg_multi_setup.cpp \
test_binary.cpp \
test_native_kernel.cpp \
test_create_context_from_type.cpp \
test_platform.cpp \
test_retain.cpp \
test_device_min_data_type_align_size_alignment.cpp \
test_mem_objects.cpp \
test_bool.c \
test_null_buffer_arg.c \
test_mem_object_info.cpp \
../../test_common/harness/errorHelpers.c \
../../test_common/harness/threadTesting.c \
../../test_common/harness/testHarness.c \
../../test_common/harness/imageHelpers.cpp \
../../test_common/harness/kernelHelpers.c \
../../test_common/harness/typeWrappers.cpp \
../../test_common/harness/mt19937.c \
../../test_common/harness/conversions.c
DEFINES = DONT_TEST_GARBAGE_POINTERS
SOURCES = $(abspath $(SRCS))
LIBPATH += -L/System/Library/Frameworks/OpenCL.framework/Libraries
LIBPATH += -L.
HEADERS =
TARGET = test_api
INCLUDE =
COMPILERFLAGS = -c -Wall -g -Wshorten-64-to-32
CC = c++
CFLAGS = $(COMPILERFLAGS) ${RC_CFLAGS} ${USE_ATF} $(DEFINES:%=-D%) $(INCLUDE)
CXXFLAGS = $(COMPILERFLAGS) ${RC_CFLAGS} ${USE_ATF} $(DEFINES:%=-D%) $(INCLUDE)
LIBRARIES = -framework OpenCL -framework OpenGL -framework GLUT -framework AppKit ${ATF}
OBJECTS := ${SOURCES:.c=.o}
OBJECTS := ${OBJECTS:.cpp=.o}
TARGETOBJECT =
all: $(TARGET)
$(TARGET): $(OBJECTS)
$(CC) $(RC_CFLAGS) $(OBJECTS) -o $@ $(LIBPATH) $(LIBRARIES)
clean:
rm -f $(TARGET) $(OBJECTS)
.DEFAULT:
@echo The target \"$@\" does not exist in Makefile.
ifdef BUILD_WITH_ATF
ATF = -framework ATF
USE_ATF = -DUSE_ATF
endif
SRCS = main.c \
test_retain_program.c \
test_queries.cpp \
test_create_kernels.c \
test_kernels.c \
test_kernel_arg_info.c \
test_api_min_max.c \
test_kernel_arg_changes.cpp \
test_kernel_arg_multi_setup.cpp \
test_binary.cpp \
test_native_kernel.cpp \
test_create_context_from_type.cpp \
test_platform.cpp \
test_retain.cpp \
test_device_min_data_type_align_size_alignment.cpp \
test_mem_objects.cpp \
test_bool.c \
test_null_buffer_arg.c \
test_mem_object_info.cpp \
../../test_common/harness/errorHelpers.c \
../../test_common/harness/threadTesting.c \
../../test_common/harness/testHarness.c \
../../test_common/harness/imageHelpers.cpp \
../../test_common/harness/kernelHelpers.c \
../../test_common/harness/typeWrappers.cpp \
../../test_common/harness/mt19937.c \
../../test_common/harness/conversions.c
DEFINES = DONT_TEST_GARBAGE_POINTERS
SOURCES = $(abspath $(SRCS))
LIBPATH += -L/System/Library/Frameworks/OpenCL.framework/Libraries
LIBPATH += -L.
HEADERS =
TARGET = test_api
INCLUDE =
COMPILERFLAGS = -c -Wall -g -Wshorten-64-to-32
CC = c++
CFLAGS = $(COMPILERFLAGS) ${RC_CFLAGS} ${USE_ATF} $(DEFINES:%=-D%) $(INCLUDE)
CXXFLAGS = $(COMPILERFLAGS) ${RC_CFLAGS} ${USE_ATF} $(DEFINES:%=-D%) $(INCLUDE)
LIBRARIES = -framework OpenCL -framework OpenGL -framework GLUT -framework AppKit ${ATF}
OBJECTS := ${SOURCES:.c=.o}
OBJECTS := ${OBJECTS:.cpp=.o}
TARGETOBJECT =
all: $(TARGET)
$(TARGET): $(OBJECTS)
$(CC) $(RC_CFLAGS) $(OBJECTS) -o $@ $(LIBPATH) $(LIBRARIES)
clean:
rm -f $(TARGET) $(OBJECTS)
.DEFAULT:
@echo The target \"$@\" does not exist in Makefile.

431
test_conformance/api/main.c
View File

@@ -1,215 +1,216 @@
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include <stdio.h>
#include <stdlib.h>
#if !defined(_WIN32)
#include <stdbool.h>
#endif
#include <math.h>
#include <string.h>
#include "procs.h"
#include "../../test_common/harness/testHarness.h"
#if !defined(_WIN32)
#include <unistd.h>
#endif
basefn basefn_list[] = {
test_get_platform_info,
test_get_sampler_info,
test_get_command_queue_info,
test_get_context_info,
test_get_device_info,
test_enqueue_task,
test_binary_get,
test_program_binary_create,
test_kernel_required_group_size,
test_release_kernel_order,
test_release_during_execute,
test_load_single_kernel,
test_load_two_kernels,
test_load_two_kernels_in_one,
test_load_two_kernels_manually,
test_get_program_info_kernel_names,
test_get_kernel_arg_info,
test_create_kernels_in_program,
test_get_kernel_info,
test_execute_kernel_local_sizes,
test_set_kernel_arg_by_index,
test_set_kernel_arg_constant,
test_set_kernel_arg_struct_array,
test_kernel_global_constant,
test_min_max_thread_dimensions,
test_min_max_work_items_sizes,
test_min_max_work_group_size,
test_min_max_read_image_args,
test_min_max_write_image_args,
test_min_max_mem_alloc_size,
test_min_max_image_2d_width,
test_min_max_image_2d_height,
test_min_max_image_3d_width,
test_min_max_image_3d_height,
test_min_max_image_3d_depth,
test_min_max_image_array_size,
test_min_max_image_buffer_size,
test_min_max_parameter_size,
test_min_max_samplers,
test_min_max_constant_buffer_size,
test_min_max_constant_args,
test_min_max_compute_units,
test_min_max_address_bits,
test_min_max_single_fp_config,
test_min_max_double_fp_config,
test_min_max_local_mem_size,
test_min_max_kernel_preferred_work_group_size_multiple,
test_min_max_execution_capabilities,
test_min_max_queue_properties,
test_min_max_device_version,
test_min_max_language_version,
test_kernel_arg_changes,
test_kernel_arg_multi_setup_random,
test_native_kernel,
test_create_context_from_type,
test_platform_extensions,
test_get_platform_ids,
test_for_bool_type,
test_repeated_setup_cleanup,
test_retain_queue_single,
test_retain_queue_multiple,
test_retain_mem_object_single,
test_retain_mem_object_multiple,
test_min_data_type_align_size_alignment,
test_mem_object_destructor_callback,
test_null_buffer_arg,
test_get_buffer_info,
test_get_image2d_info,
test_get_image3d_info,
test_get_image1d_info,
test_get_image1d_array_info,
test_get_image2d_array_info,
};
const char *basefn_names[] = {
"get_platform_info",
"get_sampler_info",
"get_command_queue_info",
"get_context_info",
"get_device_info",
"enqueue_task",
"binary_get",
"binary_create",
"kernel_required_group_size",
"release_kernel_order",
"release_during_execute",
"load_single_kernel",
"load_two_kernels",
"load_two_kernels_in_one",
"load_two_kernels_manually",
"get_program_info_kernel_names",
"get_kernel_arg_info",
"create_kernels_in_program",
"get_kernel_info",
"execute_kernel_local_sizes",
"set_kernel_arg_by_index",
"set_kernel_arg_constant",
"set_kernel_arg_struct_array",
"kernel_global_constant",
"min_max_thread_dimensions",
"min_max_work_items_sizes",
"min_max_work_group_size",
"min_max_read_image_args",
"min_max_write_image_args",
"min_max_mem_alloc_size",
"min_max_image_2d_width",
"min_max_image_2d_height",
"min_max_image_3d_width",
"min_max_image_3d_height",
"min_max_image_3d_depth",
"min_max_image_array_size",
"min_max_image_buffer_size",
"min_max_parameter_size",
"min_max_samplers",
"min_max_constant_buffer_size",
"min_max_constant_args",
"min_max_compute_units",
"min_max_address_bits",
"min_max_single_fp_config",
"min_max_double_fp_config",
"min_max_local_mem_size",
"min_max_kernel_preferred_work_group_size_multiple",
"min_max_execution_capabilities",
"min_max_queue_properties",
"min_max_device_version",
"min_max_language_version",
"kernel_arg_changes",
"kernel_arg_multi_setup_random",
"native_kernel",
"create_context_from_type",
"platform_extensions",
"get_platform_ids",
"bool_type",
"repeated_setup_cleanup",
"retain_queue_single",
"retain_queue_multiple",
"retain_mem_object_single",
"retain_mem_object_multiple",
"min_data_type_align_size_alignment",
"mem_object_destructor_callback",
"null_buffer_arg",
"get_buffer_info",
"get_image2d_info",
"get_image3d_info",
"get_image1d_info",
"get_image1d_array_info",
"get_image2d_array_info",
"all",
};
ct_assert((sizeof(basefn_names) / sizeof(basefn_names[0]) - 1) == (sizeof(basefn_list) / sizeof(basefn_list[0])));
int num_fns = sizeof(basefn_names) / sizeof(char *);
int main(int argc, const char *argv[])
{
return runTestHarness( argc, argv, num_fns, basefn_list, basefn_names, false, false, 0 );
}
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include <stdio.h>
#include <stdlib.h>
#if !defined(_WIN32)
#include <stdbool.h>
#endif
#include <math.h>
#include <string.h>
#include "procs.h"
#include "../../test_common/harness/testHarness.h"
#if !defined(_WIN32)
#include <unistd.h>
#endif
basefn basefn_list[] = {
test_get_platform_info,
test_get_sampler_info,
test_get_command_queue_info,
test_get_context_info,
test_get_device_info,
test_enqueue_task,
test_binary_get,
test_program_binary_create,
test_kernel_required_group_size,
test_release_kernel_order,
test_release_during_execute,
test_load_single_kernel,
test_load_two_kernels,
test_load_two_kernels_in_one,
test_load_two_kernels_manually,
test_get_program_info_kernel_names,
test_get_kernel_arg_info,
test_create_kernels_in_program,
test_get_kernel_info,
test_execute_kernel_local_sizes,
test_set_kernel_arg_by_index,
test_set_kernel_arg_constant,
test_set_kernel_arg_struct_array,
test_kernel_global_constant,
test_min_max_thread_dimensions,
test_min_max_work_items_sizes,
test_min_max_work_group_size,
test_min_max_read_image_args,
test_min_max_write_image_args,
test_min_max_mem_alloc_size,
test_min_max_image_2d_width,
test_min_max_image_2d_height,
test_min_max_image_3d_width,
test_min_max_image_3d_height,
test_min_max_image_3d_depth,
test_min_max_image_array_size,
test_min_max_image_buffer_size,
test_min_max_parameter_size,
test_min_max_samplers,
test_min_max_constant_buffer_size,
test_min_max_constant_args,
test_min_max_compute_units,
test_min_max_address_bits,
test_min_max_single_fp_config,
test_min_max_double_fp_config,
test_min_max_local_mem_size,
test_min_max_kernel_preferred_work_group_size_multiple,
test_min_max_execution_capabilities,
test_min_max_queue_properties,
test_min_max_device_version,
test_min_max_language_version,
test_kernel_arg_changes,
test_kernel_arg_multi_setup_random,
test_native_kernel,
test_create_context_from_type,
test_platform_extensions,
test_get_platform_ids,
test_for_bool_type,
test_repeated_setup_cleanup,
test_retain_queue_single,
test_retain_queue_multiple,
test_retain_mem_object_single,
test_retain_mem_object_multiple,
test_min_data_type_align_size_alignment,
test_mem_object_destructor_callback,
test_null_buffer_arg,
test_get_buffer_info,
test_get_image2d_info,
test_get_image3d_info,
test_get_image1d_info,
test_get_image1d_array_info,
test_get_image2d_array_info,
test_queue_properties,
};
const char *basefn_names[] = {
"get_platform_info",
"get_sampler_info",
"get_command_queue_info",
"get_context_info",
"get_device_info",
"enqueue_task",
"binary_get",
"binary_create",
"kernel_required_group_size",
"release_kernel_order",
"release_during_execute",
"load_single_kernel",
"load_two_kernels",
"load_two_kernels_in_one",
"load_two_kernels_manually",
"get_program_info_kernel_names",
"get_kernel_arg_info",
"create_kernels_in_program",
"get_kernel_info",
"execute_kernel_local_sizes",
"set_kernel_arg_by_index",
"set_kernel_arg_constant",
"set_kernel_arg_struct_array",
"kernel_global_constant",
"min_max_thread_dimensions",
"min_max_work_items_sizes",
"min_max_work_group_size",
"min_max_read_image_args",
"min_max_write_image_args",
"min_max_mem_alloc_size",
"min_max_image_2d_width",
"min_max_image_2d_height",
"min_max_image_3d_width",
"min_max_image_3d_height",
"min_max_image_3d_depth",
"min_max_image_array_size",
"min_max_image_buffer_size",
"min_max_parameter_size",
"min_max_samplers",
"min_max_constant_buffer_size",
"min_max_constant_args",
"min_max_compute_units",
"min_max_address_bits",
"min_max_single_fp_config",
"min_max_double_fp_config",
"min_max_local_mem_size",
"min_max_kernel_preferred_work_group_size_multiple",
"min_max_execution_capabilities",
"min_max_queue_properties",
"min_max_device_version",
"min_max_language_version",
"kernel_arg_changes",
"kernel_arg_multi_setup_random",
"native_kernel",
"create_context_from_type",
"platform_extensions",
"get_platform_ids",
"bool_type",
"repeated_setup_cleanup",
"retain_queue_single",
"retain_queue_multiple",
"retain_mem_object_single",
"retain_mem_object_multiple",
"min_data_type_align_size_alignment",
"mem_object_destructor_callback",
"null_buffer_arg",
"get_buffer_info",
"get_image2d_info",
"get_image3d_info",
"get_image1d_info",
"get_image1d_array_info",
"get_image2d_array_info",
"queue_properties",
"all",
};
ct_assert((sizeof(basefn_names) / sizeof(basefn_names[0]) - 1) == (sizeof(basefn_list) / sizeof(basefn_list[0])));
int num_fns = sizeof(basefn_names) / sizeof(char *);
int main(int argc, const char *argv[])
{
return runTestHarness( argc, argv, num_fns, basefn_list, basefn_names, false, false, 0 );
}

217
test_conformance/api/procs.h
View File

@@ -1,108 +1,109 @@
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "../../test_common/harness/errorHelpers.h"
#include "../../test_common/harness/kernelHelpers.h"
#include "../../test_common/harness/typeWrappers.h"
#include "../../test_common/harness/clImageHelper.h"
#include "../../test_common/harness/imageHelpers.h"
extern float calculate_ulperror(float a, float b);
extern int test_load_single_kernel(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_load_two_kernels(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_load_two_kernels_in_one(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_load_two_kernels_manually(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_get_program_info_kernel_names( cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_create_kernels_in_program(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_enqueue_task(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_repeated_setup_cleanup(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_for_bool_type(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_platform_extensions(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_get_platform_info(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_get_sampler_info(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_get_command_queue_info(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_get_context_info(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_get_device_info(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_kernel_required_group_size(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_binary_get(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_program_binary_create(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_release_kernel_order(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_release_during_execute(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_get_kernel_info(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_execute_kernel_local_sizes(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_set_kernel_arg_by_index(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_set_kernel_arg_struct(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_set_kernel_arg_constant(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_set_kernel_arg_struct_array(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_kernel_global_constant(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_thread_dimensions(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_work_items_sizes(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_work_group_size(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_read_image_args(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_write_image_args(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_mem_alloc_size(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_image_2d_width(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_image_2d_height(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_image_3d_width(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_image_3d_height(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_image_3d_depth(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_image_array_size(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_image_buffer_size(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_parameter_size(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_samplers(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_constant_buffer_size(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_constant_args(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_compute_units(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_address_bits(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_single_fp_config(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_double_fp_config(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_local_mem_size(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_kernel_preferred_work_group_size_multiple(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_execution_capabilities(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_queue_properties(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_device_version(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_language_version(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_native_kernel(cl_device_id device, cl_context context, cl_command_queue queue, int n_elems );
extern int test_create_context_from_type(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_get_platform_ids(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_kernel_arg_changes(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_kernel_arg_multi_setup_random(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_retain_queue_single(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_retain_queue_multiple(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_retain_mem_object_single(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_retain_mem_object_multiple(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_data_type_align_size_alignment(cl_device_id device, cl_context context, cl_command_queue queue, int n_elems );
extern int test_mem_object_destructor_callback(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_null_buffer_arg( cl_device_id device_id, cl_context context, cl_command_queue queue, int num_elements );
extern int test_get_buffer_info( cl_device_id deviceID, cl_context context, cl_command_queue ignoreQueue, int num_elements );
extern int test_get_image2d_info( cl_device_id deviceID, cl_context context, cl_command_queue ignoreQueue, int num_elements );
extern int test_get_image3d_info( cl_device_id deviceID, cl_context context, cl_command_queue ignoreQueue, int num_elements );
extern int test_get_image1d_info( cl_device_id deviceID, cl_context context, cl_command_queue ignoreQueue, int num_elements );
extern int test_get_image1d_array_info( cl_device_id deviceID, cl_context context, cl_command_queue ignoreQueue, int num_elements );
extern int test_get_image2d_array_info( cl_device_id deviceID, cl_context context, cl_command_queue ignoreQueue, int num_elements );
extern int test_get_kernel_arg_info( cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements );
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "../../test_common/harness/errorHelpers.h"
#include "../../test_common/harness/kernelHelpers.h"
#include "../../test_common/harness/typeWrappers.h"
#include "../../test_common/harness/clImageHelper.h"
#include "../../test_common/harness/imageHelpers.h"
extern float calculate_ulperror(float a, float b);
extern int test_load_single_kernel(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_load_two_kernels(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_load_two_kernels_in_one(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_load_two_kernels_manually(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_get_program_info_kernel_names( cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_create_kernels_in_program(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_enqueue_task(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_repeated_setup_cleanup(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_for_bool_type(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_platform_extensions(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_get_platform_info(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_get_sampler_info(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_get_command_queue_info(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_get_context_info(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_get_device_info(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_kernel_required_group_size(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_binary_get(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_program_binary_create(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_release_kernel_order(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_release_during_execute(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_get_kernel_info(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_execute_kernel_local_sizes(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_set_kernel_arg_by_index(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_set_kernel_arg_struct(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_set_kernel_arg_constant(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_set_kernel_arg_struct_array(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_kernel_global_constant(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_thread_dimensions(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_work_items_sizes(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_work_group_size(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_read_image_args(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_write_image_args(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_mem_alloc_size(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_image_2d_width(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_image_2d_height(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_image_3d_width(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_image_3d_height(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_image_3d_depth(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_image_array_size(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_image_buffer_size(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_parameter_size(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_samplers(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_constant_buffer_size(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_constant_args(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_compute_units(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_address_bits(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_single_fp_config(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_double_fp_config(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_local_mem_size(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_kernel_preferred_work_group_size_multiple(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_execution_capabilities(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_queue_properties(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_device_version(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_max_language_version(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_native_kernel(cl_device_id device, cl_context context, cl_command_queue queue, int n_elems );
extern int test_create_context_from_type(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_get_platform_ids(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_kernel_arg_changes(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_kernel_arg_multi_setup_random(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_retain_queue_single(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_retain_queue_multiple(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_retain_mem_object_single(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_retain_mem_object_multiple(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_min_data_type_align_size_alignment(cl_device_id device, cl_context context, cl_command_queue queue, int n_elems );
extern int test_mem_object_destructor_callback(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
extern int test_null_buffer_arg( cl_device_id device_id, cl_context context, cl_command_queue queue, int num_elements );
extern int test_get_buffer_info( cl_device_id deviceID, cl_context context, cl_command_queue ignoreQueue, int num_elements );
extern int test_get_image2d_info( cl_device_id deviceID, cl_context context, cl_command_queue ignoreQueue, int num_elements );
extern int test_get_image3d_info( cl_device_id deviceID, cl_context context, cl_command_queue ignoreQueue, int num_elements );
extern int test_get_image1d_info( cl_device_id deviceID, cl_context context, cl_command_queue ignoreQueue, int num_elements );
extern int test_get_image1d_array_info( cl_device_id deviceID, cl_context context, cl_command_queue ignoreQueue, int num_elements );
extern int test_get_image2d_array_info( cl_device_id deviceID, cl_context context, cl_command_queue ignoreQueue, int num_elements );
extern int test_get_kernel_arg_info( cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements );
extern int test_queue_properties( cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements );

72
test_conformance/api/testBase.h
View File

@@ -1,36 +1,36 @@
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#ifndef _testBase_h
#define _testBase_h
#include <stdio.h>
#include <stdlib.h>
#include <math.h>
#include <string.h>
#if !defined(_WIN32)
#include <stdbool.h>
#endif
#include <sys/types.h>
#include <sys/stat.h>
#include "procs.h"
#endif // _testBase_h
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#ifndef _testBase_h
#define _testBase_h
#include <stdio.h>
#include <stdlib.h>
#include <math.h>
#include <string.h>
#if !defined(_WIN32)
#include <stdbool.h>
#endif
#include <sys/types.h>
#include <sys/stat.h>
#include "procs.h"
#endif // _testBase_h

4066
test_conformance/api/test_api_min_max.c
View File

File diff suppressed because it is too large Load Diff

472
test_conformance/api/test_binary.cpp
View File

@@ -1,236 +1,236 @@
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
static const char *sample_binary_kernel_source[] = {
"__kernel void sample_test(__global float *src, __global int *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (int)src[tid] + 1;\n"
"\n"
"}\n" };
int test_binary_get(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
int error;
clProgramWrapper program;
size_t binarySize;
program = clCreateProgramWithSource( context, 1, sample_binary_kernel_source, NULL, &error );
test_error( error, "Unable to create program from source" );
// Build so we have a binary to get
error = clBuildProgram( program, 1, &deviceID, NULL, NULL, NULL );
test_error( error, "Unable to build test program" );
// Get the size of the resulting binary (only one device)
error = clGetProgramInfo( program, CL_PROGRAM_BINARY_SIZES, sizeof( binarySize ), &binarySize, NULL );
test_error( error, "Unable to get binary size" );
// Sanity check
if( binarySize == 0 )
{
log_error( "ERROR: Binary size of program is zero\n" );
return -1;
}
// Create a buffer and get the actual binary
unsigned char *binary;
binary = (unsigned char*)malloc(sizeof(unsigned char)*binarySize);
unsigned char *buffers[ 1 ] = { binary };
// Do another sanity check here first
size_t size;
error = clGetProgramInfo( program, CL_PROGRAM_BINARIES, 0, NULL, &size );
test_error( error, "Unable to get expected size of binaries array" );
if( size != sizeof( buffers ) )
{
log_error( "ERROR: Expected size of binaries array in clGetProgramInfo is incorrect (should be %d, got %d)\n", (int)sizeof( buffers ), (int)size );
free(binary);
return -1;
}
error = clGetProgramInfo( program, CL_PROGRAM_BINARIES, sizeof( buffers ), &buffers, NULL );
test_error( error, "Unable to get program binary" );
// No way to verify the binary is correct, so just be good with that
free(binary);
return 0;
}
int test_program_binary_create(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
/* To test this in a self-contained fashion, we have to create a program with
source, then get the binary, then use that binary to reload the program, and then verify */
int error;
clProgramWrapper program, program_from_binary;
size_t binarySize;
program = clCreateProgramWithSource( context, 1, sample_binary_kernel_source, NULL, &error );
test_error( error, "Unable to create program from source" );
// Build so we have a binary to get
error = clBuildProgram( program, 1, &deviceID, NULL, NULL, NULL );
test_error( error, "Unable to build test program" );
// Get the size of the resulting binary (only one device)
error = clGetProgramInfo( program, CL_PROGRAM_BINARY_SIZES, sizeof( binarySize ), &binarySize, NULL );
test_error( error, "Unable to get binary size" );
// Sanity check
if( binarySize == 0 )
{
log_error( "ERROR: Binary size of program is zero\n" );
return -1;
}
// Create a buffer and get the actual binary
unsigned char *binary;
binary = (unsigned char*)malloc(sizeof(unsigned char)*binarySize);
const unsigned char *buffers[ 1 ] = { binary };
error = clGetProgramInfo( program, CL_PROGRAM_BINARIES, sizeof( buffers ), &buffers, NULL );
test_error( error, "Unable to get program binary" );
cl_int loadErrors[ 1 ];
program_from_binary = clCreateProgramWithBinary( context, 1, &deviceID, &binarySize, buffers, loadErrors, &error );
test_error( error, "Unable to load valid program binary" );
test_error( loadErrors[ 0 ], "Unable to load valid device binary into program" );
error = clBuildProgram( program_from_binary, 1, &deviceID, NULL, NULL, NULL );
test_error( error, "Unable to build binary program" );
// Now get the binary one more time and verify it loaded the right binary
unsigned char *binary2;
binary2 = (unsigned char*)malloc(sizeof(unsigned char)*binarySize);
buffers[ 0 ] = binary2;
error = clGetProgramInfo( program_from_binary, CL_PROGRAM_BINARIES, sizeof( buffers ), &buffers, NULL );
test_error( error, "Unable to get program binary second time" );
if( memcmp( binary, binary2, binarySize ) != 0 )
{
log_error( "ERROR: Program binary is different when loaded from binary!\n" );
free(binary2);
free(binary);
return -1;
}
// Try again, this time without passing the status ptr in, to make sure we still
// get a valid binary
clProgramWrapper programWithoutStatus = clCreateProgramWithBinary( context, 1, &deviceID, &binarySize, buffers, NULL, &error );
test_error( error, "Unable to load valid program binary when binary_status pointer is NULL" );
error = clBuildProgram( programWithoutStatus, 1, &deviceID, NULL, NULL, NULL );
test_error( error, "Unable to build binary program" );
// Now get the binary one more time and verify it loaded the right binary
unsigned char *binary3;
binary3 = (unsigned char*)malloc(sizeof(unsigned char)*binarySize);
buffers[ 0 ] = binary3;
error = clGetProgramInfo( program_from_binary, CL_PROGRAM_BINARIES, sizeof( buffers ), &buffers, NULL );
test_error( error, "Unable to get program binary second time" );
if( memcmp( binary, binary3, binarySize ) != 0 )
{
log_error( "ERROR: Program binary is different when status pointer is NULL!\n" );
free(binary3);
free(binary2);
free(binary);
return -1;
}
free(binary3);
// Now execute them both to see that they both do the same thing.
clMemWrapper in, out, out_binary;
clKernelWrapper kernel, kernel_binary;
cl_int *out_data, *out_data_binary;
cl_float *in_data;
size_t size_to_run = 1000;
// Allocate some data
in_data = (cl_float*)malloc(sizeof(cl_float)*size_to_run);
out_data = (cl_int*)malloc(sizeof(cl_int)*size_to_run);
out_data_binary = (cl_int*)malloc(sizeof(cl_int)*size_to_run);
memset(out_data, 0, sizeof(cl_int)*size_to_run);
memset(out_data_binary, 0, sizeof(cl_int)*size_to_run);
for (size_t i=0; i<size_to_run; i++)
in_data[i] = (cl_float)i;
// Create the buffers
in = clCreateBuffer(context, CL_MEM_COPY_HOST_PTR, sizeof(cl_float)*size_to_run, in_data, &error);
test_error( error, "clCreateBuffer failed");
out = clCreateBuffer(context, CL_MEM_COPY_HOST_PTR, sizeof(cl_int)*size_to_run, out_data, &error);
test_error( error, "clCreateBuffer failed");
out_binary = clCreateBuffer(context, CL_MEM_COPY_HOST_PTR, sizeof(cl_int)*size_to_run, out_data_binary, &error);
test_error( error, "clCreateBuffer failed");
// Create the kernels
kernel = clCreateKernel(program, "sample_test", &error);
test_error( error, "clCreateKernel failed");
kernel_binary = clCreateKernel(program_from_binary, "sample_test", &error);
test_error( error, "clCreateKernel from binary failed");
// Set the arguments
error = clSetKernelArg(kernel, 0, sizeof(in), &in);
test_error( error, "clSetKernelArg failed");
error = clSetKernelArg(kernel, 1, sizeof(out), &out);
test_error( error, "clSetKernelArg failed");
error = clSetKernelArg(kernel_binary, 0, sizeof(in), &in);
test_error( error, "clSetKernelArg failed");
error = clSetKernelArg(kernel_binary, 1, sizeof(out_binary), &out_binary);
test_error( error, "clSetKernelArg failed");
// Execute the kernels
error = clEnqueueNDRangeKernel(queue, kernel, 1, NULL, &size_to_run, NULL, 0, NULL, NULL);
test_error( error, "clEnqueueNDRangeKernel failed");
error = clEnqueueNDRangeKernel(queue, kernel_binary, 1, NULL, &size_to_run, NULL, 0, NULL, NULL);
test_error( error, "clEnqueueNDRangeKernel for binary kernel failed");
// Finish up
error = clFinish(queue);
test_error( error, "clFinish failed");
// Get the results back
error = clEnqueueReadBuffer(queue, out, CL_TRUE, 0, sizeof(cl_int)*size_to_run, out_data, 0, NULL, NULL);
test_error( error, "clEnqueueReadBuffer failed");
error = clEnqueueReadBuffer(queue, out_binary, CL_TRUE, 0, sizeof(cl_int)*size_to_run, out_data_binary, 0, NULL, NULL);
test_error( error, "clEnqueueReadBuffer failed");
// Compare the results
if( memcmp( out_data, out_data_binary, sizeof(cl_int)*size_to_run ) != 0 )
{
log_error( "ERROR: Results from executing binary and regular kernel differ.\n" );
free(binary2);
free(binary);
return -1;
}
// All done!
free(in_data);
free(out_data);
free(out_data_binary);
free(binary2);
free(binary);
return 0;
}
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
static const char *sample_binary_kernel_source[] = {
"__kernel void sample_test(__global float *src, __global int *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (int)src[tid] + 1;\n"
"\n"
"}\n" };
int test_binary_get(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
int error;
clProgramWrapper program;
size_t binarySize;
program = clCreateProgramWithSource( context, 1, sample_binary_kernel_source, NULL, &error );
test_error( error, "Unable to create program from source" );
// Build so we have a binary to get
error = clBuildProgram( program, 1, &deviceID, NULL, NULL, NULL );
test_error( error, "Unable to build test program" );
// Get the size of the resulting binary (only one device)
error = clGetProgramInfo( program, CL_PROGRAM_BINARY_SIZES, sizeof( binarySize ), &binarySize, NULL );
test_error( error, "Unable to get binary size" );
// Sanity check
if( binarySize == 0 )
{
log_error( "ERROR: Binary size of program is zero\n" );
return -1;
}
// Create a buffer and get the actual binary
unsigned char *binary;
binary = (unsigned char*)malloc(sizeof(unsigned char)*binarySize);
unsigned char *buffers[ 1 ] = { binary };
// Do another sanity check here first
size_t size;
error = clGetProgramInfo( program, CL_PROGRAM_BINARIES, 0, NULL, &size );
test_error( error, "Unable to get expected size of binaries array" );
if( size != sizeof( buffers ) )
{
log_error( "ERROR: Expected size of binaries array in clGetProgramInfo is incorrect (should be %d, got %d)\n", (int)sizeof( buffers ), (int)size );
free(binary);
return -1;
}
error = clGetProgramInfo( program, CL_PROGRAM_BINARIES, sizeof( buffers ), &buffers, NULL );
test_error( error, "Unable to get program binary" );
// No way to verify the binary is correct, so just be good with that
free(binary);
return 0;
}
int test_program_binary_create(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
/* To test this in a self-contained fashion, we have to create a program with
source, then get the binary, then use that binary to reload the program, and then verify */
int error;
clProgramWrapper program, program_from_binary;
size_t binarySize;
program = clCreateProgramWithSource( context, 1, sample_binary_kernel_source, NULL, &error );
test_error( error, "Unable to create program from source" );
// Build so we have a binary to get
error = clBuildProgram( program, 1, &deviceID, NULL, NULL, NULL );
test_error( error, "Unable to build test program" );
// Get the size of the resulting binary (only one device)
error = clGetProgramInfo( program, CL_PROGRAM_BINARY_SIZES, sizeof( binarySize ), &binarySize, NULL );
test_error( error, "Unable to get binary size" );
// Sanity check
if( binarySize == 0 )
{
log_error( "ERROR: Binary size of program is zero\n" );
return -1;
}
// Create a buffer and get the actual binary
unsigned char *binary;
binary = (unsigned char*)malloc(sizeof(unsigned char)*binarySize);
const unsigned char *buffers[ 1 ] = { binary };
error = clGetProgramInfo( program, CL_PROGRAM_BINARIES, sizeof( buffers ), &buffers, NULL );
test_error( error, "Unable to get program binary" );
cl_int loadErrors[ 1 ];
program_from_binary = clCreateProgramWithBinary( context, 1, &deviceID, &binarySize, buffers, loadErrors, &error );
test_error( error, "Unable to load valid program binary" );
test_error( loadErrors[ 0 ], "Unable to load valid device binary into program" );
error = clBuildProgram( program_from_binary, 1, &deviceID, NULL, NULL, NULL );
test_error( error, "Unable to build binary program" );
// Now get the binary one more time and verify it loaded the right binary
unsigned char *binary2;
binary2 = (unsigned char*)malloc(sizeof(unsigned char)*binarySize);
buffers[ 0 ] = binary2;
error = clGetProgramInfo( program_from_binary, CL_PROGRAM_BINARIES, sizeof( buffers ), &buffers, NULL );
test_error( error, "Unable to get program binary second time" );
if( memcmp( binary, binary2, binarySize ) != 0 )
{
log_error( "ERROR: Program binary is different when loaded from binary!\n" );
free(binary2);
free(binary);
return -1;
}
// Try again, this time without passing the status ptr in, to make sure we still
// get a valid binary
clProgramWrapper programWithoutStatus = clCreateProgramWithBinary( context, 1, &deviceID, &binarySize, buffers, NULL, &error );
test_error( error, "Unable to load valid program binary when binary_status pointer is NULL" );
error = clBuildProgram( programWithoutStatus, 1, &deviceID, NULL, NULL, NULL );
test_error( error, "Unable to build binary program" );
// Now get the binary one more time and verify it loaded the right binary
unsigned char *binary3;
binary3 = (unsigned char*)malloc(sizeof(unsigned char)*binarySize);
buffers[ 0 ] = binary3;
error = clGetProgramInfo( program_from_binary, CL_PROGRAM_BINARIES, sizeof( buffers ), &buffers, NULL );
test_error( error, "Unable to get program binary second time" );
if( memcmp( binary, binary3, binarySize ) != 0 )
{
log_error( "ERROR: Program binary is different when status pointer is NULL!\n" );
free(binary3);
free(binary2);
free(binary);
return -1;
}
free(binary3);
// Now execute them both to see that they both do the same thing.
clMemWrapper in, out, out_binary;
clKernelWrapper kernel, kernel_binary;
cl_int *out_data, *out_data_binary;
cl_float *in_data;
size_t size_to_run = 1000;
// Allocate some data
in_data = (cl_float*)malloc(sizeof(cl_float)*size_to_run);
out_data = (cl_int*)malloc(sizeof(cl_int)*size_to_run);
out_data_binary = (cl_int*)malloc(sizeof(cl_int)*size_to_run);
memset(out_data, 0, sizeof(cl_int)*size_to_run);
memset(out_data_binary, 0, sizeof(cl_int)*size_to_run);
for (size_t i=0; i<size_to_run; i++)
in_data[i] = (cl_float)i;
// Create the buffers
in = clCreateBuffer(context, CL_MEM_COPY_HOST_PTR, sizeof(cl_float)*size_to_run, in_data, &error);
test_error( error, "clCreateBuffer failed");
out = clCreateBuffer(context, CL_MEM_COPY_HOST_PTR, sizeof(cl_int)*size_to_run, out_data, &error);
test_error( error, "clCreateBuffer failed");
out_binary = clCreateBuffer(context, CL_MEM_COPY_HOST_PTR, sizeof(cl_int)*size_to_run, out_data_binary, &error);
test_error( error, "clCreateBuffer failed");
// Create the kernels
kernel = clCreateKernel(program, "sample_test", &error);
test_error( error, "clCreateKernel failed");
kernel_binary = clCreateKernel(program_from_binary, "sample_test", &error);
test_error( error, "clCreateKernel from binary failed");
// Set the arguments
error = clSetKernelArg(kernel, 0, sizeof(in), &in);
test_error( error, "clSetKernelArg failed");
error = clSetKernelArg(kernel, 1, sizeof(out), &out);
test_error( error, "clSetKernelArg failed");
error = clSetKernelArg(kernel_binary, 0, sizeof(in), &in);
test_error( error, "clSetKernelArg failed");
error = clSetKernelArg(kernel_binary, 1, sizeof(out_binary), &out_binary);
test_error( error, "clSetKernelArg failed");
// Execute the kernels
error = clEnqueueNDRangeKernel(queue, kernel, 1, NULL, &size_to_run, NULL, 0, NULL, NULL);
test_error( error, "clEnqueueNDRangeKernel failed");
error = clEnqueueNDRangeKernel(queue, kernel_binary, 1, NULL, &size_to_run, NULL, 0, NULL, NULL);
test_error( error, "clEnqueueNDRangeKernel for binary kernel failed");
// Finish up
error = clFinish(queue);
test_error( error, "clFinish failed");
// Get the results back
error = clEnqueueReadBuffer(queue, out, CL_TRUE, 0, sizeof(cl_int)*size_to_run, out_data, 0, NULL, NULL);
test_error( error, "clEnqueueReadBuffer failed");
error = clEnqueueReadBuffer(queue, out_binary, CL_TRUE, 0, sizeof(cl_int)*size_to_run, out_data_binary, 0, NULL, NULL);
test_error( error, "clEnqueueReadBuffer failed");
// Compare the results
if( memcmp( out_data, out_data_binary, sizeof(cl_int)*size_to_run ) != 0 )
{
log_error( "ERROR: Results from executing binary and regular kernel differ.\n" );
free(binary2);
free(binary);
return -1;
}
// All done!
free(in_data);
free(out_data);
free(out_data_binary);
free(binary2);
free(binary);
return 0;
}

104
test_conformance/api/test_bool.c
View File

@@ -1,52 +1,52 @@
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
#include "../../test_common/harness/testHarness.h"
const char *kernel_with_bool[] = {
"__kernel void kernel_with_bool(__global float *src, __global int *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" bool myBool = (src[tid] < 0.5f) && (src[tid] > -0.5f);\n"
" if(myBool)\n"
" {\n"
" dst[tid] = (int)src[tid];\n"
" }\n"
" else\n"
" {\n"
" dst[tid] = 0;\n"
" }\n"
"\n"
"}\n"
};
int test_for_bool_type(cl_device_id deviceID, cl_context context,
cl_command_queue queue, int num_elements)
{
cl_program program;
cl_kernel kernel;
int err = create_single_kernel_helper(context,
&program,
&kernel,
1, kernel_with_bool,
"kernel_with_bool" );
return err;
}
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
#include "../../test_common/harness/testHarness.h"
const char *kernel_with_bool[] = {
"__kernel void kernel_with_bool(__global float *src, __global int *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" bool myBool = (src[tid] < 0.5f) && (src[tid] > -0.5f);\n"
" if(myBool)\n"
" {\n"
" dst[tid] = (int)src[tid];\n"
" }\n"
" else\n"
" {\n"
" dst[tid] = 0;\n"
" }\n"
"\n"
"}\n"
};
int test_for_bool_type(cl_device_id deviceID, cl_context context,
cl_command_queue queue, int num_elements)
{
cl_program program;
cl_kernel kernel;
int err = create_single_kernel_helper(context,
&program,
&kernel,
1, kernel_with_bool,
"kernel_with_bool" );
return err;
}

270
test_conformance/api/test_create_context_from_type.cpp
View File

@@ -1,135 +1,135 @@
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
#ifndef _WIN32
#include <unistd.h>
#endif
#include "../../test_common/harness/conversions.h"
extern cl_uint gRandomSeed;
void CL_CALLBACK notify_callback(const char *errinfo, const void *private_info, size_t cb, void *user_data)
{
log_info( "%s\n", errinfo );
}
int test_create_context_from_type(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
int error;
clProgramWrapper program;
clKernelWrapper kernel;
clMemWrapper streams[2];
clContextWrapper context_to_test;
clCommandQueueWrapper queue_to_test;
size_t threads[1], localThreads[1];
cl_float inputData[10];
cl_int outputData[10];
int i;
RandomSeed seed( gRandomSeed );
const char *sample_single_test_kernel[] = {
"__kernel void sample_test(__global float *src, __global int *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (int)src[tid];\n"
"\n"
"}\n" };
cl_device_type type;
error = clGetDeviceInfo(deviceID, CL_DEVICE_TYPE, sizeof(type), &type, NULL);
test_error(error, "clGetDeviceInfo for CL_DEVICE_TYPE failed\n");
cl_platform_id platform;
error = clGetDeviceInfo(deviceID, CL_DEVICE_PLATFORM, sizeof(platform), &platform, NULL);
test_error(error, "clGetDeviceInfo for CL_DEVICE_PLATFORM failed\n");
cl_context_properties properties[3] = {
(cl_context_properties)CL_CONTEXT_PLATFORM,
(cl_context_properties)platform,
NULL
};
context_to_test = clCreateContextFromType(properties, type, notify_callback, NULL, &error);
test_error(error, "clCreateContextFromType failed");
if (context_to_test == NULL) {
log_error("clCreateContextFromType returned NULL, but error was CL_SUCCESS.");
return -1;
}
queue_to_test = clCreateCommandQueue(context_to_test, deviceID, NULL, &error);
test_error(error, "clCreateCommandQueue failed");
if (queue_to_test == NULL) {
log_error("clCreateCommandQueue returned NULL, but error was CL_SUCCESS.");
return -1;
}
/* Create a kernel to test with */
if( create_single_kernel_helper( context_to_test, &program, &kernel, 1, sample_single_test_kernel, "sample_test" ) != 0 )
{
return -1;
}
/* Create some I/O streams */
streams[0] = clCreateBuffer(context_to_test, (cl_mem_flags)(CL_MEM_READ_WRITE), sizeof(cl_float) * 10, NULL, &error);
test_error( error, "Creating test array failed" );
streams[1] = clCreateBuffer(context_to_test, (cl_mem_flags)(CL_MEM_READ_WRITE), sizeof(cl_int) * 10, NULL, &error);
test_error( error, "Creating test array failed" );
/* Write some test data */
memset( outputData, 0, sizeof( outputData ) );
for (i=0; i<10; i++)
inputData[i] = get_random_float(-(float) 0x7fffffff, (float) 0x7fffffff, seed);
error = clEnqueueWriteBuffer(queue_to_test, streams[0], CL_TRUE, 0, sizeof(cl_float)*10, (void *)inputData, 0, NULL, NULL);
test_error( error, "Unable to set testing kernel data" );
/* Test setting the arguments by index manually */
error = clSetKernelArg(kernel, 1, sizeof( streams[1] ), &streams[1]);
test_error( error, "Unable to set indexed kernel arguments" );
error = clSetKernelArg(kernel, 0, sizeof( streams[0] ), &streams[0]);
test_error( error, "Unable to set indexed kernel arguments" );
/* Test running the kernel and verifying it */
threads[0] = (size_t)10;
error = get_max_common_work_group_size( context_to_test, kernel, threads[0], &localThreads[0] );
test_error( error, "Unable to get work group size to use" );
error = clEnqueueNDRangeKernel( queue_to_test, kernel, 1, NULL, threads, localThreads, 0, NULL, NULL );
test_error( error, "Kernel execution failed" );
error = clEnqueueReadBuffer( queue_to_test, streams[1], CL_TRUE, 0, sizeof(cl_int)*10, (void *)outputData, 0, NULL, NULL );
test_error( error, "Unable to get result data" );
for (i=0; i<10; i++)
{
if (outputData[i] != (int)inputData[i])
{
log_error( "ERROR: Data did not verify on first pass!\n" );
return -1;
}
}
return 0;
}
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
#ifndef _WIN32
#include <unistd.h>
#endif
#include "../../test_common/harness/conversions.h"
extern cl_uint gRandomSeed;
void CL_CALLBACK notify_callback(const char *errinfo, const void *private_info, size_t cb, void *user_data)
{
log_info( "%s\n", errinfo );
}
int test_create_context_from_type(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
int error;
clProgramWrapper program;
clKernelWrapper kernel;
clMemWrapper streams[2];
clContextWrapper context_to_test;
clCommandQueueWrapper queue_to_test;
size_t threads[1], localThreads[1];
cl_float inputData[10];
cl_int outputData[10];
int i;
RandomSeed seed( gRandomSeed );
const char *sample_single_test_kernel[] = {
"__kernel void sample_test(__global float *src, __global int *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (int)src[tid];\n"
"\n"
"}\n" };
cl_device_type type;
error = clGetDeviceInfo(deviceID, CL_DEVICE_TYPE, sizeof(type), &type, NULL);
test_error(error, "clGetDeviceInfo for CL_DEVICE_TYPE failed\n");
cl_platform_id platform;
error = clGetDeviceInfo(deviceID, CL_DEVICE_PLATFORM, sizeof(platform), &platform, NULL);
test_error(error, "clGetDeviceInfo for CL_DEVICE_PLATFORM failed\n");
cl_context_properties properties[3] = {
(cl_context_properties)CL_CONTEXT_PLATFORM,
(cl_context_properties)platform,
NULL
};
context_to_test = clCreateContextFromType(properties, type, notify_callback, NULL, &error);
test_error(error, "clCreateContextFromType failed");
if (context_to_test == NULL) {
log_error("clCreateContextFromType returned NULL, but error was CL_SUCCESS.");
return -1;
}
queue_to_test = clCreateCommandQueue(context_to_test, deviceID, NULL, &error);
test_error(error, "clCreateCommandQueue failed");
if (queue_to_test == NULL) {
log_error("clCreateCommandQueue returned NULL, but error was CL_SUCCESS.");
return -1;
}
/* Create a kernel to test with */
if( create_single_kernel_helper( context_to_test, &program, &kernel, 1, sample_single_test_kernel, "sample_test" ) != 0 )
{
return -1;
}
/* Create some I/O streams */
streams[0] = clCreateBuffer(context_to_test, (cl_mem_flags)(CL_MEM_READ_WRITE), sizeof(cl_float) * 10, NULL, &error);
test_error( error, "Creating test array failed" );
streams[1] = clCreateBuffer(context_to_test, (cl_mem_flags)(CL_MEM_READ_WRITE), sizeof(cl_int) * 10, NULL, &error);
test_error( error, "Creating test array failed" );
/* Write some test data */
memset( outputData, 0, sizeof( outputData ) );
for (i=0; i<10; i++)
inputData[i] = get_random_float(-(float) 0x7fffffff, (float) 0x7fffffff, seed);
error = clEnqueueWriteBuffer(queue_to_test, streams[0], CL_TRUE, 0, sizeof(cl_float)*10, (void *)inputData, 0, NULL, NULL);
test_error( error, "Unable to set testing kernel data" );
/* Test setting the arguments by index manually */
error = clSetKernelArg(kernel, 1, sizeof( streams[1] ), &streams[1]);
test_error( error, "Unable to set indexed kernel arguments" );
error = clSetKernelArg(kernel, 0, sizeof( streams[0] ), &streams[0]);
test_error( error, "Unable to set indexed kernel arguments" );
/* Test running the kernel and verifying it */
threads[0] = (size_t)10;
error = get_max_common_work_group_size( context_to_test, kernel, threads[0], &localThreads[0] );
test_error( error, "Unable to get work group size to use" );
error = clEnqueueNDRangeKernel( queue_to_test, kernel, 1, NULL, threads, localThreads, 0, NULL, NULL );
test_error( error, "Kernel execution failed" );
error = clEnqueueReadBuffer( queue_to_test, streams[1], CL_TRUE, 0, sizeof(cl_int)*10, (void *)outputData, 0, NULL, NULL );
test_error( error, "Unable to get result data" );
for (i=0; i<10; i++)
{
if (outputData[i] != (int)inputData[i])
{
log_error( "ERROR: Data did not verify on first pass!\n" );
return -1;
}
}
return 0;
}

1286
test_conformance/api/test_create_kernels.c
View File

File diff suppressed because it is too large Load Diff

120
test_conformance/api/test_device_min_data_type_align_size_alignment.cpp
View File

@@ -1,60 +1,60 @@
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
#include "../../test_common/harness/testHarness.h"
#ifndef _WIN32
#include <unistd.h>
#endif
int IsAPowerOfTwo( unsigned long x )
{
return 0 == (x & (x-1));
}
int test_min_data_type_align_size_alignment(cl_device_id device, cl_context context, cl_command_queue queue, int n_elems )
{
cl_uint min_alignment;
if (gHasLong)
min_alignment = sizeof(cl_long)*16;
else
min_alignment = sizeof(cl_int)*16;
int error = 0;
cl_uint alignment;
error = clGetDeviceInfo(device, CL_DEVICE_MEM_BASE_ADDR_ALIGN, sizeof(alignment), &alignment, NULL);
test_error(error, "clGetDeviceInfo for CL_DEVICE_MEM_BASE_ADDR_ALIGN failed");
log_info("Device reported CL_DEVICE_MEM_BASE_ADDR_ALIGN = %lu bits.\n", (unsigned long)alignment);
// Verify the size is large enough
if (alignment < min_alignment*8) {
log_error("ERROR: alignment too small. Minimum alignment for %s16 is %lu bits, device reported %lu bits.",
(gHasLong) ? "long" : "int",
(unsigned long)(min_alignment*8), (unsigned long)alignment);
return -1;
}
// Verify the size is a power of two
if (!IsAPowerOfTwo((unsigned long)alignment)) {
log_error("ERROR: alignment is not a power of two.\n");
return -1;
}
return 0;
}
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
#include "../../test_common/harness/testHarness.h"
#ifndef _WIN32
#include <unistd.h>
#endif
int IsAPowerOfTwo( unsigned long x )
{
return 0 == (x & (x-1));
}
int test_min_data_type_align_size_alignment(cl_device_id device, cl_context context, cl_command_queue queue, int n_elems )
{
cl_uint min_alignment;
if (gHasLong)
min_alignment = sizeof(cl_long)*16;
else
min_alignment = sizeof(cl_int)*16;
int error = 0;
cl_uint alignment;
error = clGetDeviceInfo(device, CL_DEVICE_MEM_BASE_ADDR_ALIGN, sizeof(alignment), &alignment, NULL);
test_error(error, "clGetDeviceInfo for CL_DEVICE_MEM_BASE_ADDR_ALIGN failed");
log_info("Device reported CL_DEVICE_MEM_BASE_ADDR_ALIGN = %lu bits.\n", (unsigned long)alignment);
// Verify the size is large enough
if (alignment < min_alignment*8) {
log_error("ERROR: alignment too small. Minimum alignment for %s16 is %lu bits, device reported %lu bits.",
(gHasLong) ? "long" : "int",
(unsigned long)(min_alignment*8), (unsigned long)alignment);
return -1;
}
// Verify the size is a power of two
if (!IsAPowerOfTwo((unsigned long)alignment)) {
log_error("ERROR: alignment is not a power of two.\n");
return -1;
}
return 0;
}

282
test_conformance/api/test_kernel_arg_changes.cpp
View File

@@ -1,141 +1,141 @@
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
extern "C" { extern cl_uint gRandomSeed;}
// This test is designed to stress changing kernel arguments between execute calls (that are asynchronous and thus
// potentially overlapping) to make sure each kernel gets the right arguments
// Note: put a delay loop in the kernel to make sure we have time to queue the next kernel before this one finishes
const char *inspect_image_kernel_source[] = {
"__kernel void sample_test(read_only image2d_t src, __global int *outDimensions )\n"
"{\n"
" int tid = get_global_id(0), i;\n"
" for( i = 0; i < 100000; i++ ); \n"
" outDimensions[tid * 2] = get_image_width(src) * tid;\n"
" outDimensions[tid * 2 + 1] = get_image_height(src) * tid;\n"
"\n"
"}\n" };
#define NUM_TRIES 100
#define NUM_THREADS 2048
int test_kernel_arg_changes(cl_device_id device, cl_context context, cl_command_queue queue, int num_elements)
{
clProgramWrapper program;
clKernelWrapper kernel;
int error, i;
clMemWrapper images[ NUM_TRIES ];
size_t sizes[ NUM_TRIES ][ 2 ];
clMemWrapper results[ NUM_TRIES ];
cl_image_format imageFormat;
size_t maxWidth, maxHeight;
size_t threads[1], localThreads[1];
cl_int resultArray[ NUM_THREADS * 2 ];
char errStr[ 128 ];
RandomSeed seed( gRandomSeed );
PASSIVE_REQUIRE_IMAGE_SUPPORT( device )
// Just get any ol format to test with
error = get_8_bit_image_format( context, CL_MEM_OBJECT_IMAGE2D, CL_MEM_READ_WRITE, 0, &imageFormat );
test_error( error, "Unable to obtain suitable image format to test with!" );
// Create our testing kernel
error = create_single_kernel_helper( context, &program, &kernel, 1, inspect_image_kernel_source, "sample_test" );
test_error( error, "Unable to create testing kernel" );
// Get max dimensions for each of our images
error = clGetDeviceInfo( device, CL_DEVICE_IMAGE2D_MAX_WIDTH, sizeof( maxWidth ), &maxWidth, NULL );
error |= clGetDeviceInfo( device, CL_DEVICE_IMAGE2D_MAX_HEIGHT, sizeof( maxHeight ), &maxHeight, NULL );
test_error( error, "Unable to get max image dimensions for device" );
// Get the number of threads we'll be able to run
threads[0] = NUM_THREADS;
error = get_max_common_work_group_size( context, kernel, threads[0], &localThreads[0] );
test_error( error, "Unable to get work group size for kernel" );
// Create a variety of images and output arrays
for( i = 0; i < NUM_TRIES; i++ )
{
sizes[ i ][ 0 ] = genrand_int32(seed) % (maxWidth/32) + 1;
sizes[ i ][ 1 ] = genrand_int32(seed) % (maxHeight/32) + 1;
images[ i ] = create_image_2d( context, (cl_mem_flags)(CL_MEM_READ_ONLY),
&imageFormat, sizes[ i ][ 0], sizes[ i ][ 1 ], 0, NULL, &error );
if( images[i] == NULL )
{
log_error("Failed to create image %d of size %d x %d (%s).\n", i, (int)sizes[i][0], (int)sizes[i][1], IGetErrorString( error ));
return -1;
}
results[ i ] = clCreateBuffer( context, (cl_mem_flags)(CL_MEM_READ_WRITE), sizeof( cl_int ) * threads[0] * 2, NULL, &error );
if( results[i] == NULL)
{
log_error("Failed to create array %d of size %d.\n", i, (int)threads[0]*2);
return -1;
}
}
// Start setting arguments and executing kernels
for( i = 0; i < NUM_TRIES; i++ )
{
// Set the arguments for this try
error = clSetKernelArg( kernel, 0, sizeof( cl_mem ), &images[ i ] );
sprintf( errStr, "Unable to set argument 0 for kernel try %d", i );
test_error( error, errStr );
error = clSetKernelArg( kernel, 1, sizeof( cl_mem ), &results[ i ] );
sprintf( errStr, "Unable to set argument 1 for kernel try %d", i );
test_error( error, errStr );
// Queue up execution
error = clEnqueueNDRangeKernel( queue, kernel, 1, NULL, threads, localThreads, 0, NULL, NULL );
sprintf( errStr, "Unable to execute kernel try %d", i );
test_error( error, errStr );
}
// Read the results back out, one at a time, and verify
for( i = 0; i < NUM_TRIES; i++ )
{
error = clEnqueueReadBuffer( queue, results[ i ], CL_TRUE, 0, sizeof( cl_int ) * threads[0] * 2, resultArray, 0, NULL, NULL );
sprintf( errStr, "Unable to read results for kernel try %d", i );
test_error( error, errStr );
// Verify. Each entry should be n * the (width/height) of image i
for( int j = 0; j < NUM_THREADS; j++ )
{
if( resultArray[ j * 2 + 0 ] != (int)sizes[ i ][ 0 ] * j )
{
log_error( "ERROR: Verficiation for kernel try %d, sample %d FAILED, expected a width of %d, got %d\n",
i, j, (int)sizes[ i ][ 0 ] * j, resultArray[ j * 2 + 0 ] );
return -1;
}
if( resultArray[ j * 2 + 1 ] != (int)sizes[ i ][ 1 ] * j )
{
log_error( "ERROR: Verficiation for kernel try %d, sample %d FAILED, expected a height of %d, got %d\n",
i, j, (int)sizes[ i ][ 1 ] * j, resultArray[ j * 2 + 1 ] );
return -1;
}
}
}
// If we got here, everything verified successfully
return 0;
}
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
extern "C" { extern cl_uint gRandomSeed;}
// This test is designed to stress changing kernel arguments between execute calls (that are asynchronous and thus
// potentially overlapping) to make sure each kernel gets the right arguments
// Note: put a delay loop in the kernel to make sure we have time to queue the next kernel before this one finishes
const char *inspect_image_kernel_source[] = {
"__kernel void sample_test(read_only image2d_t src, __global int *outDimensions )\n"
"{\n"
" int tid = get_global_id(0), i;\n"
" for( i = 0; i < 100000; i++ ); \n"
" outDimensions[tid * 2] = get_image_width(src) * tid;\n"
" outDimensions[tid * 2 + 1] = get_image_height(src) * tid;\n"
"\n"
"}\n" };
#define NUM_TRIES 100
#define NUM_THREADS 2048
int test_kernel_arg_changes(cl_device_id device, cl_context context, cl_command_queue queue, int num_elements)
{
clProgramWrapper program;
clKernelWrapper kernel;
int error, i;
clMemWrapper images[ NUM_TRIES ];
size_t sizes[ NUM_TRIES ][ 2 ];
clMemWrapper results[ NUM_TRIES ];
cl_image_format imageFormat;
size_t maxWidth, maxHeight;
size_t threads[1], localThreads[1];
cl_int resultArray[ NUM_THREADS * 2 ];
char errStr[ 128 ];
RandomSeed seed( gRandomSeed );
PASSIVE_REQUIRE_IMAGE_SUPPORT( device )
// Just get any ol format to test with
error = get_8_bit_image_format( context, CL_MEM_OBJECT_IMAGE2D, CL_MEM_READ_WRITE, 0, &imageFormat );
test_error( error, "Unable to obtain suitable image format to test with!" );
// Create our testing kernel
error = create_single_kernel_helper( context, &program, &kernel, 1, inspect_image_kernel_source, "sample_test" );
test_error( error, "Unable to create testing kernel" );
// Get max dimensions for each of our images
error = clGetDeviceInfo( device, CL_DEVICE_IMAGE2D_MAX_WIDTH, sizeof( maxWidth ), &maxWidth, NULL );
error |= clGetDeviceInfo( device, CL_DEVICE_IMAGE2D_MAX_HEIGHT, sizeof( maxHeight ), &maxHeight, NULL );
test_error( error, "Unable to get max image dimensions for device" );
// Get the number of threads we'll be able to run
threads[0] = NUM_THREADS;
error = get_max_common_work_group_size( context, kernel, threads[0], &localThreads[0] );
test_error( error, "Unable to get work group size for kernel" );
// Create a variety of images and output arrays
for( i = 0; i < NUM_TRIES; i++ )
{
sizes[ i ][ 0 ] = genrand_int32(seed) % (maxWidth/32) + 1;
sizes[ i ][ 1 ] = genrand_int32(seed) % (maxHeight/32) + 1;
images[ i ] = create_image_2d( context, (cl_mem_flags)(CL_MEM_READ_ONLY),
&imageFormat, sizes[ i ][ 0], sizes[ i ][ 1 ], 0, NULL, &error );
if( images[i] == NULL )
{
log_error("Failed to create image %d of size %d x %d (%s).\n", i, (int)sizes[i][0], (int)sizes[i][1], IGetErrorString( error ));
return -1;
}
results[ i ] = clCreateBuffer( context, (cl_mem_flags)(CL_MEM_READ_WRITE), sizeof( cl_int ) * threads[0] * 2, NULL, &error );
if( results[i] == NULL)
{
log_error("Failed to create array %d of size %d.\n", i, (int)threads[0]*2);
return -1;
}
}
// Start setting arguments and executing kernels
for( i = 0; i < NUM_TRIES; i++ )
{
// Set the arguments for this try
error = clSetKernelArg( kernel, 0, sizeof( cl_mem ), &images[ i ] );
sprintf( errStr, "Unable to set argument 0 for kernel try %d", i );
test_error( error, errStr );
error = clSetKernelArg( kernel, 1, sizeof( cl_mem ), &results[ i ] );
sprintf( errStr, "Unable to set argument 1 for kernel try %d", i );
test_error( error, errStr );
// Queue up execution
error = clEnqueueNDRangeKernel( queue, kernel, 1, NULL, threads, localThreads, 0, NULL, NULL );
sprintf( errStr, "Unable to execute kernel try %d", i );
test_error( error, errStr );
}
// Read the results back out, one at a time, and verify
for( i = 0; i < NUM_TRIES; i++ )
{
error = clEnqueueReadBuffer( queue, results[ i ], CL_TRUE, 0, sizeof( cl_int ) * threads[0] * 2, resultArray, 0, NULL, NULL );
sprintf( errStr, "Unable to read results for kernel try %d", i );
test_error( error, errStr );
// Verify. Each entry should be n * the (width/height) of image i
for( int j = 0; j < NUM_THREADS; j++ )
{
if( resultArray[ j * 2 + 0 ] != (int)sizes[ i ][ 0 ] * j )
{
log_error( "ERROR: Verficiation for kernel try %d, sample %d FAILED, expected a width of %d, got %d\n",
i, j, (int)sizes[ i ][ 0 ] * j, resultArray[ j * 2 + 0 ] );
return -1;
}
if( resultArray[ j * 2 + 1 ] != (int)sizes[ i ][ 1 ] * j )
{
log_error( "ERROR: Verficiation for kernel try %d, sample %d FAILED, expected a height of %d, got %d\n",
i, j, (int)sizes[ i ][ 1 ] * j, resultArray[ j * 2 + 1 ] );
return -1;
}
}
}
// If we got here, everything verified successfully
return 0;
}

11167
test_conformance/api/test_kernel_arg_info.c
View File

File diff suppressed because it is too large Load Diff

554
test_conformance/api/test_kernel_arg_multi_setup.cpp
View File

@@ -1,277 +1,277 @@
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
#include "../../test_common/harness/conversions.h"
// This test is designed to stress passing multiple vector parameters to kernels and verifying access between them all
const char *multi_arg_kernel_source_pattern =
"__kernel void sample_test(__global %s *src1, __global %s *src2, __global %s *src3, __global %s *dst1, __global %s *dst2, __global %s *dst3 )\n"
"{\n"
" int tid = get_global_id(0);\n"
" dst1[tid] = src1[tid];\n"
" dst2[tid] = src2[tid];\n"
" dst3[tid] = src3[tid];\n"
"}\n";
extern cl_uint gRandomSeed;
#define MAX_ERROR_TOLERANCE 0.0005f
int test_multi_arg_set(cl_device_id device, cl_context context, cl_command_queue queue,
ExplicitType vec1Type, int vec1Size,
ExplicitType vec2Type, int vec2Size,
ExplicitType vec3Type, int vec3Size, MTdata d)
{
clProgramWrapper program;
clKernelWrapper kernel;
int error, i, j;
clMemWrapper streams[ 6 ];
size_t threads[1], localThreads[1];
char programSrc[ 10248 ], vec1Name[ 64 ], vec2Name[ 64 ], vec3Name[ 64 ];
char sizeNames[][ 4 ] = { "", "2", "3", "4", "", "", "", "8" };
const char *ptr;
void *initData[3], *resultData[3];
// Create the program source
sprintf( vec1Name, "%s%s", get_explicit_type_name( vec1Type ), sizeNames[ vec1Size - 1 ] );
sprintf( vec2Name, "%s%s", get_explicit_type_name( vec2Type ), sizeNames[ vec2Size - 1 ] );
sprintf( vec3Name, "%s%s", get_explicit_type_name( vec3Type ), sizeNames[ vec3Size - 1 ] );
sprintf( programSrc, multi_arg_kernel_source_pattern,
vec1Name, vec2Name, vec3Name, vec1Name, vec2Name, vec3Name,
vec1Size, vec1Size, vec2Size, vec2Size, vec3Size, vec3Size );
ptr = programSrc;
// Create our testing kernel
error = create_single_kernel_helper( context, &program, &kernel, 1, &ptr, "sample_test" );
test_error( error, "Unable to create testing kernel" );
// Get thread dimensions
threads[0] = 1024;
error = get_max_common_work_group_size( context, kernel, threads[0], &localThreads[0] );
test_error( error, "Unable to get work group size for kernel" );
// Create input streams
initData[ 0 ] = create_random_data( vec1Type, d, (unsigned int)threads[ 0 ] * vec1Size );
streams[ 0 ] = clCreateBuffer( context, (cl_mem_flags)( CL_MEM_COPY_HOST_PTR ), get_explicit_type_size( vec1Type ) * threads[0] * vec1Size, initData[ 0 ], &error );
test_error( error, "Unable to create testing stream" );
initData[ 1 ] = create_random_data( vec2Type, d, (unsigned int)threads[ 0 ] * vec2Size );
streams[ 1 ] = clCreateBuffer( context, (cl_mem_flags)( CL_MEM_COPY_HOST_PTR ), get_explicit_type_size( vec2Type ) * threads[0] * vec2Size, initData[ 1 ], &error );
test_error( error, "Unable to create testing stream" );
initData[ 2 ] = create_random_data( vec3Type, d, (unsigned int)threads[ 0 ] * vec3Size );
streams[ 2 ] = clCreateBuffer( context, (cl_mem_flags)( CL_MEM_COPY_HOST_PTR ), get_explicit_type_size( vec3Type ) * threads[0] * vec3Size, initData[ 2 ], &error );
test_error( error, "Unable to create testing stream" );
streams[ 3 ] = clCreateBuffer( context, (cl_mem_flags)(CL_MEM_READ_WRITE), get_explicit_type_size( vec1Type ) * threads[0] * vec1Size, NULL, &error );
test_error( error, "Unable to create testing stream" );
streams[ 4 ] = clCreateBuffer( context, (cl_mem_flags)(CL_MEM_READ_WRITE), get_explicit_type_size( vec2Type ) * threads[0] * vec2Size, NULL, &error );
test_error( error, "Unable to create testing stream" );
streams[ 5 ] = clCreateBuffer( context, (cl_mem_flags)(CL_MEM_READ_WRITE), get_explicit_type_size( vec3Type ) * threads[0] * vec3Size, NULL, &error );
test_error( error, "Unable to create testing stream" );
// Set the arguments
error = 0;
for( i = 0; i < 6; i++ )
error |= clSetKernelArg( kernel, i, sizeof( cl_mem ), &streams[ i ] );
test_error( error, "Unable to set arguments for kernel" );
// Execute!
error = clEnqueueNDRangeKernel( queue, kernel, 1, NULL, threads, localThreads, 0, NULL, NULL );
test_error( error, "Unable to execute kernel" );
// Read results
resultData[0] = malloc( get_explicit_type_size( vec1Type ) * vec1Size * threads[0] );
resultData[1] = malloc( get_explicit_type_size( vec2Type ) * vec2Size * threads[0] );
resultData[2] = malloc( get_explicit_type_size( vec3Type ) * vec3Size * threads[0] );
error = clEnqueueReadBuffer( queue, streams[ 3 ], CL_TRUE, 0, get_explicit_type_size( vec1Type ) * vec1Size * threads[ 0 ], resultData[0], 0, NULL, NULL );
error |= clEnqueueReadBuffer( queue, streams[ 4 ], CL_TRUE, 0, get_explicit_type_size( vec2Type ) * vec2Size * threads[ 0 ], resultData[1], 0, NULL, NULL );
error |= clEnqueueReadBuffer( queue, streams[ 5 ], CL_TRUE, 0, get_explicit_type_size( vec3Type ) * vec3Size * threads[ 0 ], resultData[2], 0, NULL, NULL );
test_error( error, "Unable to read result stream" );
// Verify
char *ptr1 = (char *)initData[ 0 ], *ptr2 = (char *)resultData[ 0 ];
size_t span = get_explicit_type_size( vec1Type );
for( i = 0; i < (int)threads[0]; i++ )
{
for( j = 0; j < vec1Size; j++ )
{
if( memcmp( ptr1 + span * j , ptr2 + span * j, span ) != 0 )
{
log_error( "ERROR: Value did not validate for component %d of item %d of stream 0!\n", j, i );
free( initData[ 0 ] );
free( initData[ 1 ] );
free( initData[ 2 ] );
free( resultData[ 0 ] );
free( resultData[ 1 ] );
free( resultData[ 2 ] );
return -1;
}
}
ptr1 += span * vec1Size;
ptr2 += span * vec1Size;
}
ptr1 = (char *)initData[ 1 ];
ptr2 = (char *)resultData[ 1 ];
span = get_explicit_type_size( vec2Type );
for( i = 0; i < (int)threads[0]; i++ )
{
for( j = 0; j < vec2Size; j++ )
{
if( memcmp( ptr1 + span * j , ptr2 + span * j, span ) != 0 )
{
log_error( "ERROR: Value did not validate for component %d of item %d of stream 1!\n", j, i );
free( initData[ 0 ] );
free( initData[ 1 ] );
free( initData[ 2 ] );
free( resultData[ 0 ] );
free( resultData[ 1 ] );
free( resultData[ 2 ] );
return -1;
}
}
ptr1 += span * vec2Size;
ptr2 += span * vec2Size;
}
ptr1 = (char *)initData[ 2 ];
ptr2 = (char *)resultData[ 2 ];
span = get_explicit_type_size( vec3Type );
for( i = 0; i < (int)threads[0]; i++ )
{
for( j = 0; j < vec3Size; j++ )
{
if( memcmp( ptr1 + span * j , ptr2 + span * j, span ) != 0 )
{
log_error( "ERROR: Value did not validate for component %d of item %d of stream 2!\n", j, i );
free( initData[ 0 ] );
free( initData[ 1 ] );
free( initData[ 2 ] );
free( resultData[ 0 ] );
free( resultData[ 1 ] );
free( resultData[ 2 ] );
return -1;
}
}
ptr1 += span * vec3Size;
ptr2 += span * vec3Size;
}
// If we got here, everything verified successfully
free( initData[ 0 ] );
free( initData[ 1 ] );
free( initData[ 2 ] );
free( resultData[ 0 ] );
free( resultData[ 1 ] );
free( resultData[ 2 ] );
return 0;
}
int test_kernel_arg_multi_setup_exhaustive(cl_device_id device, cl_context context, cl_command_queue queue, int num_elements)
{
// Loop through every combination of input and output types
ExplicitType types[] = { kChar, kShort, kInt, kFloat, kNumExplicitTypes };
int type1, type2, type3;
int size1, size2, size3;
RandomSeed seed( gRandomSeed );
log_info( "\n" ); // for formatting
for( type1 = 0; types[ type1 ] != kNumExplicitTypes; type1++ )
{
for( type2 = 0; types[ type2 ] != kNumExplicitTypes; type2++ )
{
for( type3 = 0; types[ type3 ] != kNumExplicitTypes; type3++ )
{
log_info( "\n\ttesting %s, %s, %s...", get_explicit_type_name( types[ type1 ] ), get_explicit_type_name( types[ type2 ] ), get_explicit_type_name( types[ type3 ] ) );
// Loop through every combination of vector size
for( size1 = 2; size1 <= 8; size1 <<= 1 )
{
for( size2 = 2; size2 <= 8; size2 <<= 1 )
{
for( size3 = 2; size3 <= 8; size3 <<= 1 )
{
log_info(".");
fflush( stdout);
if( test_multi_arg_set( device, context, queue,
types[ type1 ], size1,
types[ type2 ], size2,
types[ type3 ], size3, seed ) )
return -1;
}
}
}
}
}
}
log_info( "\n" );
return 0;
}
int test_kernel_arg_multi_setup_random(cl_device_id device, cl_context context, cl_command_queue queue, int num_elements)
{
// Loop through a selection of combinations
ExplicitType types[] = { kChar, kShort, kInt, kFloat, kNumExplicitTypes };
int type1, type2, type3;
int size1, size2, size3;
RandomSeed seed( gRandomSeed );
num_elements = 3*3*3*4;
log_info( "Testing %d random configurations\n", num_elements );
// Loop through every combination of vector size
for( size1 = 2; size1 <= 8; size1 <<= 1 )
{
for( size2 = 2; size2 <= 8; size2 <<= 1 )
{
for( size3 = 2; size3 <= 8; size3 <<= 1 )
{
// Loop through 4 type combinations for each size combination
int n;
for (n=0; n<4; n++) {
type1 = (int)get_random_float(0,4, seed);
type2 = (int)get_random_float(0,4, seed);
type3 = (int)get_random_float(0,4, seed);
log_info( "\ttesting %s%d, %s%d, %s%d...\n",
get_explicit_type_name( types[ type1 ] ), size1,
get_explicit_type_name( types[ type2 ] ), size2,
get_explicit_type_name( types[ type3 ] ), size3 );
if( test_multi_arg_set( device, context, queue,
types[ type1 ], size1,
types[ type2 ], size2,
types[ type3 ], size3, seed ) )
return -1;
}
}
}
}
return 0;
}
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
#include "../../test_common/harness/conversions.h"
// This test is designed to stress passing multiple vector parameters to kernels and verifying access between them all
const char *multi_arg_kernel_source_pattern =
"__kernel void sample_test(__global %s *src1, __global %s *src2, __global %s *src3, __global %s *dst1, __global %s *dst2, __global %s *dst3 )\n"
"{\n"
" int tid = get_global_id(0);\n"
" dst1[tid] = src1[tid];\n"
" dst2[tid] = src2[tid];\n"
" dst3[tid] = src3[tid];\n"
"}\n";
extern cl_uint gRandomSeed;
#define MAX_ERROR_TOLERANCE 0.0005f
int test_multi_arg_set(cl_device_id device, cl_context context, cl_command_queue queue,
ExplicitType vec1Type, int vec1Size,
ExplicitType vec2Type, int vec2Size,
ExplicitType vec3Type, int vec3Size, MTdata d)
{
clProgramWrapper program;
clKernelWrapper kernel;
int error, i, j;
clMemWrapper streams[ 6 ];
size_t threads[1], localThreads[1];
char programSrc[ 10248 ], vec1Name[ 64 ], vec2Name[ 64 ], vec3Name[ 64 ];
char sizeNames[][ 4 ] = { "", "2", "3", "4", "", "", "", "8" };
const char *ptr;
void *initData[3], *resultData[3];
// Create the program source
sprintf( vec1Name, "%s%s", get_explicit_type_name( vec1Type ), sizeNames[ vec1Size - 1 ] );
sprintf( vec2Name, "%s%s", get_explicit_type_name( vec2Type ), sizeNames[ vec2Size - 1 ] );
sprintf( vec3Name, "%s%s", get_explicit_type_name( vec3Type ), sizeNames[ vec3Size - 1 ] );
sprintf( programSrc, multi_arg_kernel_source_pattern,
vec1Name, vec2Name, vec3Name, vec1Name, vec2Name, vec3Name,
vec1Size, vec1Size, vec2Size, vec2Size, vec3Size, vec3Size );
ptr = programSrc;
// Create our testing kernel
error = create_single_kernel_helper( context, &program, &kernel, 1, &ptr, "sample_test" );
test_error( error, "Unable to create testing kernel" );
// Get thread dimensions
threads[0] = 1024;
error = get_max_common_work_group_size( context, kernel, threads[0], &localThreads[0] );
test_error( error, "Unable to get work group size for kernel" );
// Create input streams
initData[ 0 ] = create_random_data( vec1Type, d, (unsigned int)threads[ 0 ] * vec1Size );
streams[ 0 ] = clCreateBuffer( context, (cl_mem_flags)( CL_MEM_COPY_HOST_PTR ), get_explicit_type_size( vec1Type ) * threads[0] * vec1Size, initData[ 0 ], &error );
test_error( error, "Unable to create testing stream" );
initData[ 1 ] = create_random_data( vec2Type, d, (unsigned int)threads[ 0 ] * vec2Size );
streams[ 1 ] = clCreateBuffer( context, (cl_mem_flags)( CL_MEM_COPY_HOST_PTR ), get_explicit_type_size( vec2Type ) * threads[0] * vec2Size, initData[ 1 ], &error );
test_error( error, "Unable to create testing stream" );
initData[ 2 ] = create_random_data( vec3Type, d, (unsigned int)threads[ 0 ] * vec3Size );
streams[ 2 ] = clCreateBuffer( context, (cl_mem_flags)( CL_MEM_COPY_HOST_PTR ), get_explicit_type_size( vec3Type ) * threads[0] * vec3Size, initData[ 2 ], &error );
test_error( error, "Unable to create testing stream" );
streams[ 3 ] = clCreateBuffer( context, (cl_mem_flags)(CL_MEM_READ_WRITE), get_explicit_type_size( vec1Type ) * threads[0] * vec1Size, NULL, &error );
test_error( error, "Unable to create testing stream" );
streams[ 4 ] = clCreateBuffer( context, (cl_mem_flags)(CL_MEM_READ_WRITE), get_explicit_type_size( vec2Type ) * threads[0] * vec2Size, NULL, &error );
test_error( error, "Unable to create testing stream" );
streams[ 5 ] = clCreateBuffer( context, (cl_mem_flags)(CL_MEM_READ_WRITE), get_explicit_type_size( vec3Type ) * threads[0] * vec3Size, NULL, &error );
test_error( error, "Unable to create testing stream" );
// Set the arguments
error = 0;
for( i = 0; i < 6; i++ )
error |= clSetKernelArg( kernel, i, sizeof( cl_mem ), &streams[ i ] );
test_error( error, "Unable to set arguments for kernel" );
// Execute!
error = clEnqueueNDRangeKernel( queue, kernel, 1, NULL, threads, localThreads, 0, NULL, NULL );
test_error( error, "Unable to execute kernel" );
// Read results
resultData[0] = malloc( get_explicit_type_size( vec1Type ) * vec1Size * threads[0] );
resultData[1] = malloc( get_explicit_type_size( vec2Type ) * vec2Size * threads[0] );
resultData[2] = malloc( get_explicit_type_size( vec3Type ) * vec3Size * threads[0] );
error = clEnqueueReadBuffer( queue, streams[ 3 ], CL_TRUE, 0, get_explicit_type_size( vec1Type ) * vec1Size * threads[ 0 ], resultData[0], 0, NULL, NULL );
error |= clEnqueueReadBuffer( queue, streams[ 4 ], CL_TRUE, 0, get_explicit_type_size( vec2Type ) * vec2Size * threads[ 0 ], resultData[1], 0, NULL, NULL );
error |= clEnqueueReadBuffer( queue, streams[ 5 ], CL_TRUE, 0, get_explicit_type_size( vec3Type ) * vec3Size * threads[ 0 ], resultData[2], 0, NULL, NULL );
test_error( error, "Unable to read result stream" );
// Verify
char *ptr1 = (char *)initData[ 0 ], *ptr2 = (char *)resultData[ 0 ];
size_t span = get_explicit_type_size( vec1Type );
for( i = 0; i < (int)threads[0]; i++ )
{
for( j = 0; j < vec1Size; j++ )
{
if( memcmp( ptr1 + span * j , ptr2 + span * j, span ) != 0 )
{
log_error( "ERROR: Value did not validate for component %d of item %d of stream 0!\n", j, i );
free( initData[ 0 ] );
free( initData[ 1 ] );
free( initData[ 2 ] );
free( resultData[ 0 ] );
free( resultData[ 1 ] );
free( resultData[ 2 ] );
return -1;
}
}
ptr1 += span * vec1Size;
ptr2 += span * vec1Size;
}
ptr1 = (char *)initData[ 1 ];
ptr2 = (char *)resultData[ 1 ];
span = get_explicit_type_size( vec2Type );
for( i = 0; i < (int)threads[0]; i++ )
{
for( j = 0; j < vec2Size; j++ )
{
if( memcmp( ptr1 + span * j , ptr2 + span * j, span ) != 0 )
{
log_error( "ERROR: Value did not validate for component %d of item %d of stream 1!\n", j, i );
free( initData[ 0 ] );
free( initData[ 1 ] );
free( initData[ 2 ] );
free( resultData[ 0 ] );
free( resultData[ 1 ] );
free( resultData[ 2 ] );
return -1;
}
}
ptr1 += span * vec2Size;
ptr2 += span * vec2Size;
}
ptr1 = (char *)initData[ 2 ];
ptr2 = (char *)resultData[ 2 ];
span = get_explicit_type_size( vec3Type );
for( i = 0; i < (int)threads[0]; i++ )
{
for( j = 0; j < vec3Size; j++ )
{
if( memcmp( ptr1 + span * j , ptr2 + span * j, span ) != 0 )
{
log_error( "ERROR: Value did not validate for component %d of item %d of stream 2!\n", j, i );
free( initData[ 0 ] );
free( initData[ 1 ] );
free( initData[ 2 ] );
free( resultData[ 0 ] );
free( resultData[ 1 ] );
free( resultData[ 2 ] );
return -1;
}
}
ptr1 += span * vec3Size;
ptr2 += span * vec3Size;
}
// If we got here, everything verified successfully
free( initData[ 0 ] );
free( initData[ 1 ] );
free( initData[ 2 ] );
free( resultData[ 0 ] );
free( resultData[ 1 ] );
free( resultData[ 2 ] );
return 0;
}
int test_kernel_arg_multi_setup_exhaustive(cl_device_id device, cl_context context, cl_command_queue queue, int num_elements)
{
// Loop through every combination of input and output types
ExplicitType types[] = { kChar, kShort, kInt, kFloat, kNumExplicitTypes };
int type1, type2, type3;
int size1, size2, size3;
RandomSeed seed( gRandomSeed );
log_info( "\n" ); // for formatting
for( type1 = 0; types[ type1 ] != kNumExplicitTypes; type1++ )
{
for( type2 = 0; types[ type2 ] != kNumExplicitTypes; type2++ )
{
for( type3 = 0; types[ type3 ] != kNumExplicitTypes; type3++ )
{
log_info( "\n\ttesting %s, %s, %s...", get_explicit_type_name( types[ type1 ] ), get_explicit_type_name( types[ type2 ] ), get_explicit_type_name( types[ type3 ] ) );
// Loop through every combination of vector size
for( size1 = 2; size1 <= 8; size1 <<= 1 )
{
for( size2 = 2; size2 <= 8; size2 <<= 1 )
{
for( size3 = 2; size3 <= 8; size3 <<= 1 )
{
log_info(".");
fflush( stdout);
if( test_multi_arg_set( device, context, queue,
types[ type1 ], size1,
types[ type2 ], size2,
types[ type3 ], size3, seed ) )
return -1;
}
}
}
}
}
}
log_info( "\n" );
return 0;
}
int test_kernel_arg_multi_setup_random(cl_device_id device, cl_context context, cl_command_queue queue, int num_elements)
{
// Loop through a selection of combinations
ExplicitType types[] = { kChar, kShort, kInt, kFloat, kNumExplicitTypes };
int type1, type2, type3;
int size1, size2, size3;
RandomSeed seed( gRandomSeed );
num_elements = 3*3*3*4;
log_info( "Testing %d random configurations\n", num_elements );
// Loop through every combination of vector size
for( size1 = 2; size1 <= 8; size1 <<= 1 )
{
for( size2 = 2; size2 <= 8; size2 <<= 1 )
{
for( size3 = 2; size3 <= 8; size3 <<= 1 )
{
// Loop through 4 type combinations for each size combination
int n;
for (n=0; n<4; n++) {
type1 = (int)get_random_float(0,4, seed);
type2 = (int)get_random_float(0,4, seed);
type3 = (int)get_random_float(0,4, seed);
log_info( "\ttesting %s%d, %s%d, %s%d...\n",
get_explicit_type_name( types[ type1 ] ), size1,
get_explicit_type_name( types[ type2 ] ), size2,
get_explicit_type_name( types[ type3 ] ), size3 );
if( test_multi_arg_set( device, context, queue,
types[ type1 ], size1,
types[ type2 ], size2,
types[ type3 ], size3, seed ) )
return -1;
}
}
}
}
return 0;
}

1408
test_conformance/api/test_kernels.c
View File

File diff suppressed because it is too large Load Diff

1502
test_conformance/api/test_mem_object_info.cpp
View File

File diff suppressed because it is too large Load Diff

216
test_conformance/api/test_mem_objects.cpp
View File

@@ -1,108 +1,108 @@
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
static volatile cl_int sDestructorIndex;
void CL_CALLBACK mem_destructor_callback( cl_mem memObject, void * userData )
{
int * userPtr = (int *)userData;
// ordering of callbacks is guaranteed, meaning we don't need to do atomic operation here
*userPtr = ++sDestructorIndex;
}
#ifndef ABS
#define ABS( x ) ( ( x < 0 ) ? -x : x )
#endif
int test_mem_object_destructor_callback_single( clMemWrapper &memObject )
{
cl_int error;
int i;
// Set up some variables to catch the order in which callbacks are called
volatile int callbackOrders[ 3 ] = { 0, 0, 0 };
sDestructorIndex = 0;
// Set up the callbacks
error = clSetMemObjectDestructorCallback( memObject, mem_destructor_callback, (void*) &callbackOrders[ 0 ] );
test_error( error, "Unable to set destructor callback" );
error = clSetMemObjectDestructorCallback( memObject, mem_destructor_callback, (void*) &callbackOrders[ 1 ] );
test_error( error, "Unable to set destructor callback" );
error = clSetMemObjectDestructorCallback( memObject, mem_destructor_callback, (void*) &callbackOrders[ 2 ] );
test_error( error, "Unable to set destructor callback" );
// Now release the buffer, which SHOULD call the callbacks
error = clReleaseMemObject( memObject );
test_error( error, "Unable to release test buffer" );
// Note: since we manually released the mem wrapper, we need to set it to NULL to prevent a double-release
memObject = NULL;
// At this point, all three callbacks should have already been called
int numErrors = 0;
for( i = 0; i < 3; i++ )
{
// Spin waiting for the release to finish. If you don't call the mem_destructor_callback, you will not
// pass the test. bugzilla 6316
while( 0 == callbackOrders[i] )
{}
if( ABS( callbackOrders[ i ] ) != 3-i )
{
log_error( "\tERROR: Callback %d was called in the wrong order! (Was called order %d, should have been order %d)\n",
i+1, ABS( callbackOrders[ i ] ), i );
numErrors++;
}
}
return ( numErrors > 0 ) ? -1 : 0;
}
int test_mem_object_destructor_callback(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
clMemWrapper testBuffer, testImage;
cl_int error;
// Create a buffer and an image to test callbacks against
testBuffer = clCreateBuffer( context, CL_MEM_READ_WRITE, 1024, NULL, &error );
test_error( error, "Unable to create testing buffer" );
if( test_mem_object_destructor_callback_single( testBuffer ) != 0 )
{
log_error( "ERROR: Destructor callbacks for buffer object FAILED\n" );
return -1;
}
if( checkForImageSupport( deviceID ) == 0 )
{
cl_image_format imageFormat = { CL_RGBA, CL_SIGNED_INT8 };
testImage = create_image_2d( context, CL_MEM_READ_ONLY, &imageFormat, 16, 16, 0, NULL, &error );
test_error( error, "Unable to create testing image" );
if( test_mem_object_destructor_callback_single( testImage ) != 0 )
{
log_error( "ERROR: Destructor callbacks for image object FAILED\n" );
return -1;
}
}
return 0;
}
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
static volatile cl_int sDestructorIndex;
void CL_CALLBACK mem_destructor_callback( cl_mem memObject, void * userData )
{
int * userPtr = (int *)userData;
// ordering of callbacks is guaranteed, meaning we don't need to do atomic operation here
*userPtr = ++sDestructorIndex;
}
#ifndef ABS
#define ABS( x ) ( ( x < 0 ) ? -x : x )
#endif
int test_mem_object_destructor_callback_single( clMemWrapper &memObject )
{
cl_int error;
int i;
// Set up some variables to catch the order in which callbacks are called
volatile int callbackOrders[ 3 ] = { 0, 0, 0 };
sDestructorIndex = 0;
// Set up the callbacks
error = clSetMemObjectDestructorCallback( memObject, mem_destructor_callback, (void*) &callbackOrders[ 0 ] );
test_error( error, "Unable to set destructor callback" );
error = clSetMemObjectDestructorCallback( memObject, mem_destructor_callback, (void*) &callbackOrders[ 1 ] );
test_error( error, "Unable to set destructor callback" );
error = clSetMemObjectDestructorCallback( memObject, mem_destructor_callback, (void*) &callbackOrders[ 2 ] );
test_error( error, "Unable to set destructor callback" );
// Now release the buffer, which SHOULD call the callbacks
error = clReleaseMemObject( memObject );
test_error( error, "Unable to release test buffer" );
// Note: since we manually released the mem wrapper, we need to set it to NULL to prevent a double-release
memObject = NULL;
// At this point, all three callbacks should have already been called
int numErrors = 0;
for( i = 0; i < 3; i++ )
{
// Spin waiting for the release to finish. If you don't call the mem_destructor_callback, you will not
// pass the test. bugzilla 6316
while( 0 == callbackOrders[i] )
{}
if( ABS( callbackOrders[ i ] ) != 3-i )
{
log_error( "\tERROR: Callback %d was called in the wrong order! (Was called order %d, should have been order %d)\n",
i+1, ABS( callbackOrders[ i ] ), i );
numErrors++;
}
}
return ( numErrors > 0 ) ? -1 : 0;
}
int test_mem_object_destructor_callback(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
clMemWrapper testBuffer, testImage;
cl_int error;
// Create a buffer and an image to test callbacks against
testBuffer = clCreateBuffer( context, CL_MEM_READ_WRITE, 1024, NULL, &error );
test_error( error, "Unable to create testing buffer" );
if( test_mem_object_destructor_callback_single( testBuffer ) != 0 )
{
log_error( "ERROR: Destructor callbacks for buffer object FAILED\n" );
return -1;
}
if( checkForImageSupport( deviceID ) == 0 )
{
cl_image_format imageFormat = { CL_RGBA, CL_SIGNED_INT8 };
testImage = create_image_2d( context, CL_MEM_READ_ONLY, &imageFormat, 16, 16, 0, NULL, &error );
test_error( error, "Unable to create testing image" );
if( test_mem_object_destructor_callback_single( testImage ) != 0 )
{
log_error( "ERROR: Destructor callbacks for image object FAILED\n" );
return -1;
}
}
return 0;
}

242
test_conformance/api/test_native_kernel.cpp
View File

@@ -1,121 +1,121 @@
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
#ifndef _WIN32
#include <unistd.h>
#endif
#include "../../test_common/harness/conversions.h"
extern cl_uint gRandomSeed;
static void CL_CALLBACK test_native_kernel_fn( void *userData )
{
struct arg_struct {
cl_int * source;
cl_int * dest;
cl_int count;
} *args = (arg_struct *)userData;
for( cl_int i = 0; i < args->count; i++ )
args->dest[ i ] = args->source[ i ];
}
int test_native_kernel(cl_device_id device, cl_context context, cl_command_queue queue, int n_elems )
{
int error;
RandomSeed seed( gRandomSeed );
// Check if we support native kernels
cl_device_exec_capabilities capabilities;
error = clGetDeviceInfo(device, CL_DEVICE_EXECUTION_CAPABILITIES, sizeof(capabilities), &capabilities, NULL);
if (!(capabilities & CL_EXEC_NATIVE_KERNEL)) {
log_info("Device does not support CL_EXEC_NATIVE_KERNEL.\n");
return 0;
}
clMemWrapper streams[ 2 ];
#if !(defined (_WIN32) && defined (_MSC_VER))
cl_int inBuffer[ n_elems ], outBuffer[ n_elems ];
#else
cl_int* inBuffer = (cl_int *)_malloca( n_elems * sizeof(cl_int) );
cl_int* outBuffer = (cl_int *)_malloca( n_elems * sizeof(cl_int) );
#endif
clEventWrapper finishEvent;
struct arg_struct
{
cl_mem inputStream;
cl_mem outputStream;
cl_int count;
} args;
// Create some input values
generate_random_data( kInt, n_elems, seed, inBuffer );
// Create I/O streams
streams[ 0 ] = clCreateBuffer( context, CL_MEM_COPY_HOST_PTR, n_elems * sizeof(cl_int), inBuffer, &error );
test_error( error, "Unable to create I/O stream" );
streams[ 1 ] = clCreateBuffer( context, 0, n_elems * sizeof(cl_int), NULL, &error );
test_error( error, "Unable to create I/O stream" );
// Set up the arrays to call with
args.inputStream = streams[ 0 ];
args.outputStream = streams[ 1 ];
args.count = n_elems;
void * memLocs[ 2 ] = { &args.inputStream, &args.outputStream };
// Run the kernel
error = clEnqueueNativeKernel( queue, test_native_kernel_fn,
&args, sizeof( args ),
2, &streams[ 0 ],
(const void **)memLocs,
0, NULL, &finishEvent );
test_error( error, "Unable to queue native kernel" );
// Finish and wait for the kernel to complete
error = clFinish( queue );
test_error(error, "clFinish failed");
error = clWaitForEvents( 1, &finishEvent );
test_error(error, "clWaitForEvents failed");
// Now read the results and verify
error = clEnqueueReadBuffer( queue, streams[ 1 ], CL_TRUE, 0, n_elems * sizeof(cl_int), outBuffer, 0, NULL, NULL );
test_error( error, "Unable to read results" );
for( int i = 0; i < n_elems; i++ )
{
if( inBuffer[ i ] != outBuffer[ i ] )
{
log_error( "ERROR: Data sample %d for native kernel did not validate (expected %d, got %d)\n",
i, (int)inBuffer[ i ], (int)outBuffer[ i ] );
return 1;
}
}
return 0;
}
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
#ifndef _WIN32
#include <unistd.h>
#endif
#include "../../test_common/harness/conversions.h"
extern cl_uint gRandomSeed;
static void CL_CALLBACK test_native_kernel_fn( void *userData )
{
struct arg_struct {
cl_int * source;
cl_int * dest;
cl_int count;
} *args = (arg_struct *)userData;
for( cl_int i = 0; i < args->count; i++ )
args->dest[ i ] = args->source[ i ];
}
int test_native_kernel(cl_device_id device, cl_context context, cl_command_queue queue, int n_elems )
{
int error;
RandomSeed seed( gRandomSeed );
// Check if we support native kernels
cl_device_exec_capabilities capabilities;
error = clGetDeviceInfo(device, CL_DEVICE_EXECUTION_CAPABILITIES, sizeof(capabilities), &capabilities, NULL);
if (!(capabilities & CL_EXEC_NATIVE_KERNEL)) {
log_info("Device does not support CL_EXEC_NATIVE_KERNEL.\n");
return 0;
}
clMemWrapper streams[ 2 ];
#if !(defined (_WIN32) && defined (_MSC_VER))
cl_int inBuffer[ n_elems ], outBuffer[ n_elems ];
#else
cl_int* inBuffer = (cl_int *)_malloca( n_elems * sizeof(cl_int) );
cl_int* outBuffer = (cl_int *)_malloca( n_elems * sizeof(cl_int) );
#endif
clEventWrapper finishEvent;
struct arg_struct
{
cl_mem inputStream;
cl_mem outputStream;
cl_int count;
} args;
// Create some input values
generate_random_data( kInt, n_elems, seed, inBuffer );
// Create I/O streams
streams[ 0 ] = clCreateBuffer( context, CL_MEM_COPY_HOST_PTR, n_elems * sizeof(cl_int), inBuffer, &error );
test_error( error, "Unable to create I/O stream" );
streams[ 1 ] = clCreateBuffer( context, 0, n_elems * sizeof(cl_int), NULL, &error );
test_error( error, "Unable to create I/O stream" );
// Set up the arrays to call with
args.inputStream = streams[ 0 ];
args.outputStream = streams[ 1 ];
args.count = n_elems;
void * memLocs[ 2 ] = { &args.inputStream, &args.outputStream };
// Run the kernel
error = clEnqueueNativeKernel( queue, test_native_kernel_fn,
&args, sizeof( args ),
2, &streams[ 0 ],
(const void **)memLocs,
0, NULL, &finishEvent );
test_error( error, "Unable to queue native kernel" );
// Finish and wait for the kernel to complete
error = clFinish( queue );
test_error(error, "clFinish failed");
error = clWaitForEvents( 1, &finishEvent );
test_error(error, "clWaitForEvents failed");
// Now read the results and verify
error = clEnqueueReadBuffer( queue, streams[ 1 ], CL_TRUE, 0, n_elems * sizeof(cl_int), outBuffer, 0, NULL, NULL );
test_error( error, "Unable to read results" );
for( int i = 0; i < n_elems; i++ )
{
if( inBuffer[ i ] != outBuffer[ i ] )
{
log_error( "ERROR: Data sample %d for native kernel did not validate (expected %d, got %d)\n",
i, (int)inBuffer[ i ], (int)outBuffer[ i ] );
return 1;
}
}
return 0;
}

324
test_conformance/api/test_null_buffer_arg.c
View File

@@ -1,162 +1,162 @@
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include <stdio.h>
#if defined(__APPLE__)
#include <OpenCL/opencl.h>
#include <OpenCL/cl_platform.h>
#else
#include <CL/opencl.h>
#include <CL/cl_platform.h>
#endif
#include "procs.h"
enum { SUCCESS, FAILURE };
typedef enum { NON_NULL_PATH, ADDROF_NULL_PATH, NULL_PATH } test_type;
#define NITEMS 4096
/* places the casted long value of the src ptr into each element of the output
* array, to allow testing that the kernel actually _gets_ the NULL value */
const char *kernel_string =
"kernel void test_kernel(global float *src, global long *dst)\n"
"{\n"
" uint tid = get_global_id(0);\n"
" dst[tid] = (long)src;\n"
"}\n";
/*
* The guts of the test:
* call setKernelArgs with a regular buffer, &NULL, or NULL depending on
* the value of 'test_type'
*/
static int test_setargs_and_execution(cl_command_queue queue, cl_kernel kernel,
cl_mem test_buf, cl_mem result_buf, test_type type)
{
unsigned int test_success = 0;
unsigned int i;
cl_int status;
char *typestr;
if (type == NON_NULL_PATH) {
status = clSetKernelArg(kernel, 0, sizeof(cl_mem), &test_buf);
typestr = "non-NULL";
} else if (type == ADDROF_NULL_PATH) {
test_buf = NULL;
status = clSetKernelArg(kernel, 0, sizeof(cl_mem), &test_buf);
typestr = "&NULL";
} else if (type == NULL_PATH) {
status = clSetKernelArg(kernel, 0, sizeof(cl_mem), NULL);
typestr = "NULL";
}
log_info("Testing setKernelArgs with %s buffer.\n", typestr);
if (status != CL_SUCCESS) {
log_error("clSetKernelArg failed with status: %d\n", status);
return FAILURE; // no point in continuing *this* test
}
size_t global = NITEMS;
status = clEnqueueNDRangeKernel(queue, kernel, 1, NULL, &global,
NULL, 0, NULL, NULL);
test_error(status, "NDRangeKernel failed.");
cl_long* host_result = (cl_long*)malloc(NITEMS*sizeof(cl_long));
status = clEnqueueReadBuffer(queue, result_buf, CL_TRUE, 0,
sizeof(cl_long)*NITEMS, host_result, 0, NULL, NULL);
test_error(status, "ReadBuffer failed.");
// in the non-null case, we expect NONZERO values:
if (type == NON_NULL_PATH) {
for (i=0; i<NITEMS; i++) {
if (host_result[i] == 0) {
log_error("failure: item %d in the result buffer was unexpectedly NULL.\n", i);
test_success = FAILURE; break;
}
}
} else if (type == ADDROF_NULL_PATH || type == NULL_PATH) {
for (i=0; i<NITEMS; i++) {
if (host_result[i] != 0) {
log_error("failure: item %d in the result buffer was unexpectedly non-NULL.\n", i);
test_success = FAILURE; break;
}
}
}
free(host_result);
if (test_success == SUCCESS) {
log_info("\t%s ok.\n", typestr);
}
return test_success;
}
int test_null_buffer_arg(cl_device_id device, cl_context context,
cl_command_queue queue, int num_elements)
{
unsigned int test_success = 0;
unsigned int i;
cl_int status;
cl_program program;
cl_kernel kernel;
// prep kernel:
program = clCreateProgramWithSource(context, 1, &kernel_string, NULL, &status);
test_error(status, "CreateProgramWithSource failed.");
status = clBuildProgram(program, 0, NULL, NULL, NULL, NULL);
test_error(status, "BuildProgram failed.");
kernel = clCreateKernel(program, "test_kernel", &status);
test_error(status, "CreateKernel failed.");
cl_mem dev_src = clCreateBuffer(context, CL_MEM_READ_ONLY, NITEMS*sizeof(cl_float),
NULL, NULL);
cl_mem dev_dst = clCreateBuffer(context, CL_MEM_WRITE_ONLY, NITEMS*sizeof(cl_long),
NULL, NULL);
// set the destination buffer normally:
status = clSetKernelArg(kernel, 1, sizeof(cl_mem), &dev_dst);
test_error(status, "SetKernelArg failed.");
//
// we test three cases:
//
// - typical case, used everyday: non-null buffer
// - the case of src as &NULL (the spec-compliance test)
// - the case of src as NULL (the backwards-compatibility test, Apple only)
//
test_success = test_setargs_and_execution(queue, kernel, dev_src, dev_dst, NON_NULL_PATH);
test_success |= test_setargs_and_execution(queue, kernel, dev_src, dev_dst, ADDROF_NULL_PATH);
#ifdef __APPLE__
test_success |= test_setargs_and_execution(queue, kernel, dev_src, dev_dst, NULL_PATH);
#endif
// clean up:
if (dev_src) clReleaseMemObject(dev_src);
clReleaseMemObject(dev_dst);
clReleaseKernel(kernel);
clReleaseProgram(program);
return test_success;
}
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include <stdio.h>
#if defined(__APPLE__)
#include <OpenCL/opencl.h>
#include <OpenCL/cl_platform.h>
#else
#include <CL/opencl.h>
#include <CL/cl_platform.h>
#endif
#include "procs.h"
enum { SUCCESS, FAILURE };
typedef enum { NON_NULL_PATH, ADDROF_NULL_PATH, NULL_PATH } test_type;
#define NITEMS 4096
/* places the casted long value of the src ptr into each element of the output
* array, to allow testing that the kernel actually _gets_ the NULL value */
const char *kernel_string =
"kernel void test_kernel(global float *src, global long *dst)\n"
"{\n"
" uint tid = get_global_id(0);\n"
" dst[tid] = (long)src;\n"
"}\n";
/*
* The guts of the test:
* call setKernelArgs with a regular buffer, &NULL, or NULL depending on
* the value of 'test_type'
*/
static int test_setargs_and_execution(cl_command_queue queue, cl_kernel kernel,
cl_mem test_buf, cl_mem result_buf, test_type type)
{
unsigned int test_success = 0;
unsigned int i;
cl_int status;
char *typestr;
if (type == NON_NULL_PATH) {
status = clSetKernelArg(kernel, 0, sizeof(cl_mem), &test_buf);
typestr = "non-NULL";
} else if (type == ADDROF_NULL_PATH) {
test_buf = NULL;
status = clSetKernelArg(kernel, 0, sizeof(cl_mem), &test_buf);
typestr = "&NULL";
} else if (type == NULL_PATH) {
status = clSetKernelArg(kernel, 0, sizeof(cl_mem), NULL);
typestr = "NULL";
}
log_info("Testing setKernelArgs with %s buffer.\n", typestr);
if (status != CL_SUCCESS) {
log_error("clSetKernelArg failed with status: %d\n", status);
return FAILURE; // no point in continuing *this* test
}
size_t global = NITEMS;
status = clEnqueueNDRangeKernel(queue, kernel, 1, NULL, &global,
NULL, 0, NULL, NULL);
test_error(status, "NDRangeKernel failed.");
cl_long* host_result = (cl_long*)malloc(NITEMS*sizeof(cl_long));
status = clEnqueueReadBuffer(queue, result_buf, CL_TRUE, 0,
sizeof(cl_long)*NITEMS, host_result, 0, NULL, NULL);
test_error(status, "ReadBuffer failed.");
// in the non-null case, we expect NONZERO values:
if (type == NON_NULL_PATH) {
for (i=0; i<NITEMS; i++) {
if (host_result[i] == 0) {
log_error("failure: item %d in the result buffer was unexpectedly NULL.\n", i);
test_success = FAILURE; break;
}
}
} else if (type == ADDROF_NULL_PATH || type == NULL_PATH) {
for (i=0; i<NITEMS; i++) {
if (host_result[i] != 0) {
log_error("failure: item %d in the result buffer was unexpectedly non-NULL.\n", i);
test_success = FAILURE; break;
}
}
}
free(host_result);
if (test_success == SUCCESS) {
log_info("\t%s ok.\n", typestr);
}
return test_success;
}
int test_null_buffer_arg(cl_device_id device, cl_context context,
cl_command_queue queue, int num_elements)
{
unsigned int test_success = 0;
unsigned int i;
cl_int status;
cl_program program;
cl_kernel kernel;
// prep kernel:
program = clCreateProgramWithSource(context, 1, &kernel_string, NULL, &status);
test_error(status, "CreateProgramWithSource failed.");
status = clBuildProgram(program, 0, NULL, NULL, NULL, NULL);
test_error(status, "BuildProgram failed.");
kernel = clCreateKernel(program, "test_kernel", &status);
test_error(status, "CreateKernel failed.");
cl_mem dev_src = clCreateBuffer(context, CL_MEM_READ_ONLY, NITEMS*sizeof(cl_float),
NULL, NULL);
cl_mem dev_dst = clCreateBuffer(context, CL_MEM_WRITE_ONLY, NITEMS*sizeof(cl_long),
NULL, NULL);
// set the destination buffer normally:
status = clSetKernelArg(kernel, 1, sizeof(cl_mem), &dev_dst);
test_error(status, "SetKernelArg failed.");
//
// we test three cases:
//
// - typical case, used everyday: non-null buffer
// - the case of src as &NULL (the spec-compliance test)
// - the case of src as NULL (the backwards-compatibility test, Apple only)
//
test_success = test_setargs_and_execution(queue, kernel, dev_src, dev_dst, NON_NULL_PATH);
test_success |= test_setargs_and_execution(queue, kernel, dev_src, dev_dst, ADDROF_NULL_PATH);
#ifdef __APPLE__
test_success |= test_setargs_and_execution(queue, kernel, dev_src, dev_dst, NULL_PATH);
#endif
// clean up:
if (dev_src) clReleaseMemObject(dev_src);
clReleaseMemObject(dev_dst);
clReleaseKernel(kernel);
clReleaseProgram(program);
return test_success;
}

578
test_conformance/api/test_platform.cpp
View File

@@ -1,289 +1,289 @@
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
#include <string.h>
#define EXTENSION_NAME_BUF_SIZE 4096
#define PRINT_EXTENSION_INFO 0
int test_platform_extensions(cl_device_id deviceID, cl_context context,
cl_command_queue queue, int num_elements)
{
const char * extensions[] = {
"cl_khr_byte_addressable_store",
// "cl_APPLE_SetMemObjectDestructor",
"cl_khr_global_int32_base_atomics",
"cl_khr_global_int32_extended_atomics",
"cl_khr_local_int32_base_atomics",
"cl_khr_local_int32_extended_atomics",
"cl_khr_int64_base_atomics",
"cl_khr_int64_extended_atomics",
// need to put in entires for various atomics
"cl_khr_3d_image_writes",
"cl_khr_fp16",
"cl_khr_fp64",
NULL
};
bool extensionsSupported[] = {
false, //"cl_khr_byte_addressable_store",
false, // need to put in entires for various atomics
false, // "cl_khr_global_int32_base_atomics",
false, // "cl_khr_global_int32_extended_atomics",
false, // "cl_khr_local_int32_base_atomics",
false, // "cl_khr_local_int32_extended_atomics",
false, // "cl_khr_int64_base_atomics",
false, // "cl_khr_int64_extended_atomics",
false, //"cl_khr_3d_image_writes",
false, //"cl_khr_fp16",
false, //"cl_khr_fp64",
false //NULL
};
int extensionIndex;
cl_platform_id platformID;
cl_int err;
char platform_extensions[EXTENSION_NAME_BUF_SIZE];
char device_extensions[EXTENSION_NAME_BUF_SIZE];
// Okay, so what we're going to do is just check the device indicated by
// deviceID against the platform that includes this device
// pass CL_DEVICE_PLATFORM to clGetDeviceInfo
// to get a result of type cl_platform_id
err = clGetDeviceInfo(deviceID,
CL_DEVICE_PLATFORM,
sizeof(cl_platform_id),
(void *)(&platformID),
NULL);
if(err != CL_SUCCESS)
{
vlog_error("test_platform_extensions : could not get platformID from device\n");
return -1;
}
// now we grab the set of extensions specified by the platform
err = clGetPlatformInfo(platformID,
CL_PLATFORM_EXTENSIONS,
sizeof(platform_extensions),
(void *)(&platform_extensions[0]),
NULL);
if(err != CL_SUCCESS)
{
vlog_error("test_platform_extensions : could not get extension string from platform\n");
return -1;
}
#if PRINT_EXTENSION_INFO
log_info("Platform extensions include \"%s\"\n\n", platform_extensions);
#endif
// here we parse the platform extensions, to look for the "important" ones
for(extensionIndex=0; extensions[extensionIndex] != NULL; ++extensionIndex)
{
if(strstr(platform_extensions, extensions[extensionIndex]) != NULL)
{
// we found it
#if PRINT_EXTENSION_INFO
log_info("Found \"%s\" in platform extensions\n",
extensions[extensionIndex]);
#endif
extensionsSupported[extensionIndex] = true;
}
}
// and then we grab the set of extensions specified by the device
// (this can be turned into a "loop over all devices in this platform")
err = clGetDeviceInfo(deviceID,
CL_DEVICE_EXTENSIONS,
sizeof(device_extensions),
(void *)(&device_extensions[0]),
NULL);
if(err != CL_SUCCESS)
{
vlog_error("test_platform_extensions : could not get extension string from device\n");
return -1;
}
#if PRINT_EXTENSION_INFO
log_info("Device extensions include \"%s\"\n\n", device_extensions);
#endif
for(extensionIndex=0; extensions[extensionIndex] != NULL; ++extensionIndex)
{
if(extensionsSupported[extensionIndex] == false)
{
continue; // skip this one
}
if(strstr(device_extensions, extensions[extensionIndex]) == NULL)
{
// device does not support it
vlog_error("Platform supports extension \"%s\" but device does not\n",
extensions[extensionIndex]);
return -1;
}
}
return 0;
}
int test_get_platform_ids(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements) {
cl_platform_id platforms[16];
cl_uint num_platforms;
char *string_returned;
string_returned = (char*)malloc(8192);
int total_errors = 0;
int err = CL_SUCCESS;
err = clGetPlatformIDs(16, platforms, &num_platforms);
test_error(err, "clGetPlatformIDs failed");
if (num_platforms <= 16) {
// Try with NULL
err = clGetPlatformIDs(num_platforms, platforms, NULL);
test_error(err, "clGetPlatformIDs failed with NULL for return size");
}
if (num_platforms < 1) {
log_error("Found 0 platforms.\n");
return -1;
}
log_info("Found %d platforms.\n", num_platforms);
for (int p=0; p<(int)num_platforms; p++) {
cl_device_id *devices;
cl_uint num_devices;
size_t size;
log_info("Platform %d (%p):\n", p, platforms[p]);
memset(string_returned, 0, 8192);
err = clGetPlatformInfo(platforms[p], CL_PLATFORM_PROFILE, 8192, string_returned, &size);
test_error(err, "clGetPlatformInfo for CL_PLATFORM_PROFILE failed");
log_info("\tCL_PLATFORM_PROFILE: %s\n", string_returned);
if (strlen(string_returned)+1 != size) {
log_error("Returned string length %ld does not equal reported one %ld.\n", strlen(string_returned)+1, size);
total_errors++;
}
memset(string_returned, 0, 8192);
err = clGetPlatformInfo(platforms[p], CL_PLATFORM_VERSION, 8192, string_returned, &size);
test_error(err, "clGetPlatformInfo for CL_PLATFORM_VERSION failed");
log_info("\tCL_PLATFORM_VERSION: %s\n", string_returned);
if (strlen(string_returned)+1 != size) {
log_error("Returned string length %ld does not equal reported one %ld.\n", strlen(string_returned)+1, size);
total_errors++;
}
memset(string_returned, 0, 8192);
err = clGetPlatformInfo(platforms[p], CL_PLATFORM_NAME, 8192, string_returned, &size);
test_error(err, "clGetPlatformInfo for CL_PLATFORM_NAME failed");
log_info("\tCL_PLATFORM_NAME: %s\n", string_returned);
if (strlen(string_returned)+1 != size) {
log_error("Returned string length %ld does not equal reported one %ld.\n", strlen(string_returned)+1, size);
total_errors++;
}
memset(string_returned, 0, 8192);
err = clGetPlatformInfo(platforms[p], CL_PLATFORM_VENDOR, 8192, string_returned, &size);
test_error(err, "clGetPlatformInfo for CL_PLATFORM_VENDOR failed");
log_info("\tCL_PLATFORM_VENDOR: %s\n", string_returned);
if (strlen(string_returned)+1 != size) {
log_error("Returned string length %ld does not equal reported one %ld.\n", strlen(string_returned)+1, size);
total_errors++;
}
memset(string_returned, 0, 8192);
err = clGetPlatformInfo(platforms[p], CL_PLATFORM_EXTENSIONS, 8192, string_returned, &size);
test_error(err, "clGetPlatformInfo for CL_PLATFORM_EXTENSIONS failed");
log_info("\tCL_PLATFORM_EXTENSIONS: %s\n", string_returned);
if (strlen(string_returned)+1 != size) {
log_error("Returned string length %ld does not equal reported one %ld.\n", strlen(string_returned)+1, size);
total_errors++;
}
err = clGetDeviceIDs(platforms[p], CL_DEVICE_TYPE_ALL, 0, NULL, &num_devices);
test_error(err, "clGetDeviceIDs size failed.\n");
devices = (cl_device_id *)malloc(num_devices*sizeof(cl_device_id));
memset(devices, 0, sizeof(cl_device_id)*num_devices);
err = clGetDeviceIDs(platforms[p], CL_DEVICE_TYPE_ALL, num_devices, devices, NULL);
test_error(err, "clGetDeviceIDs failed.\n");
log_info("\tPlatform has %d devices.\n", (int)num_devices);
for (int d=0; d<(int)num_devices; d++) {
size_t returned_size;
cl_platform_id returned_platform;
cl_context context;
cl_context_properties properties[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)platforms[p], 0 };
err = clGetDeviceInfo(devices[d], CL_DEVICE_PLATFORM, sizeof(cl_platform_id), &returned_platform, &returned_size);
test_error(err, "clGetDeviceInfo failed for CL_DEVICE_PLATFORM\n");
if (returned_size != sizeof(cl_platform_id)) {
log_error("Reported return size (%ld) does not match expected size (%ld).\n", returned_size, sizeof(cl_platform_id));
total_errors++;
}
memset(string_returned, 0, 8192);
err = clGetDeviceInfo(devices[d], CL_DEVICE_NAME, 8192, string_returned, NULL);
test_error(err, "clGetDeviceInfo failed for CL_DEVICE_NAME\n");
log_info("\t\tPlatform for device %d (%s) is %p.\n", d, string_returned, returned_platform);
log_info("\t\t\tTesting clCreateContext for the platform/device...\n");
// Try creating a context for the platform
context = clCreateContext(properties, 1, &devices[d], NULL, NULL, &err);
test_error(err, "\t\tclCreateContext failed for device with platform properties\n");
memset(properties, 0, sizeof(cl_context_properties)*3);
err = clGetContextInfo(context, CL_CONTEXT_PROPERTIES, sizeof(cl_context_properties)*3, properties, &returned_size);
test_error(err, "clGetContextInfo for CL_CONTEXT_PROPERTIES failed");
if (returned_size != sizeof(cl_context_properties)*3) {
log_error("Invalid size returned from clGetContextInfo for CL_CONTEXT_PROPERTIES. Got %ld, expected %ld.\n",
returned_size, sizeof(cl_context_properties)*3);
total_errors++;
}
if (properties[0] != (cl_context_properties)CL_CONTEXT_PLATFORM || properties[1] != (cl_context_properties)platforms[p]) {
log_error("Wrong properties returned. Expected: [%p %p], got [%p %p]\n",
(void*)CL_CONTEXT_PLATFORM, platforms[p], (void*)properties[0], (void*)properties[1]);
total_errors++;
}
err = clReleaseContext(context);
test_error(err, "clReleaseContext failed");
}
free(devices);
}
free(string_returned);
return total_errors;
}
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
#include <string.h>
#define EXTENSION_NAME_BUF_SIZE 4096
#define PRINT_EXTENSION_INFO 0
int test_platform_extensions(cl_device_id deviceID, cl_context context,
cl_command_queue queue, int num_elements)
{
const char * extensions[] = {
"cl_khr_byte_addressable_store",
// "cl_APPLE_SetMemObjectDestructor",
"cl_khr_global_int32_base_atomics",
"cl_khr_global_int32_extended_atomics",
"cl_khr_local_int32_base_atomics",
"cl_khr_local_int32_extended_atomics",
"cl_khr_int64_base_atomics",
"cl_khr_int64_extended_atomics",
// need to put in entires for various atomics
"cl_khr_3d_image_writes",
"cl_khr_fp16",
"cl_khr_fp64",
NULL
};
bool extensionsSupported[] = {
false, //"cl_khr_byte_addressable_store",
false, // need to put in entires for various atomics
false, // "cl_khr_global_int32_base_atomics",
false, // "cl_khr_global_int32_extended_atomics",
false, // "cl_khr_local_int32_base_atomics",
false, // "cl_khr_local_int32_extended_atomics",
false, // "cl_khr_int64_base_atomics",
false, // "cl_khr_int64_extended_atomics",
false, //"cl_khr_3d_image_writes",
false, //"cl_khr_fp16",
false, //"cl_khr_fp64",
false //NULL
};
int extensionIndex;
cl_platform_id platformID;
cl_int err;
char platform_extensions[EXTENSION_NAME_BUF_SIZE];
char device_extensions[EXTENSION_NAME_BUF_SIZE];
// Okay, so what we're going to do is just check the device indicated by
// deviceID against the platform that includes this device
// pass CL_DEVICE_PLATFORM to clGetDeviceInfo
// to get a result of type cl_platform_id
err = clGetDeviceInfo(deviceID,
CL_DEVICE_PLATFORM,
sizeof(cl_platform_id),
(void *)(&platformID),
NULL);
if(err != CL_SUCCESS)
{
vlog_error("test_platform_extensions : could not get platformID from device\n");
return -1;
}
// now we grab the set of extensions specified by the platform
err = clGetPlatformInfo(platformID,
CL_PLATFORM_EXTENSIONS,
sizeof(platform_extensions),
(void *)(&platform_extensions[0]),
NULL);
if(err != CL_SUCCESS)
{
vlog_error("test_platform_extensions : could not get extension string from platform\n");
return -1;
}
#if PRINT_EXTENSION_INFO
log_info("Platform extensions include \"%s\"\n\n", platform_extensions);
#endif
// here we parse the platform extensions, to look for the "important" ones
for(extensionIndex=0; extensions[extensionIndex] != NULL; ++extensionIndex)
{
if(strstr(platform_extensions, extensions[extensionIndex]) != NULL)
{
// we found it
#if PRINT_EXTENSION_INFO
log_info("Found \"%s\" in platform extensions\n",
extensions[extensionIndex]);
#endif
extensionsSupported[extensionIndex] = true;
}
}
// and then we grab the set of extensions specified by the device
// (this can be turned into a "loop over all devices in this platform")
err = clGetDeviceInfo(deviceID,
CL_DEVICE_EXTENSIONS,
sizeof(device_extensions),
(void *)(&device_extensions[0]),
NULL);
if(err != CL_SUCCESS)
{
vlog_error("test_platform_extensions : could not get extension string from device\n");
return -1;
}
#if PRINT_EXTENSION_INFO
log_info("Device extensions include \"%s\"\n\n", device_extensions);
#endif
for(extensionIndex=0; extensions[extensionIndex] != NULL; ++extensionIndex)
{
if(extensionsSupported[extensionIndex] == false)
{
continue; // skip this one
}
if(strstr(device_extensions, extensions[extensionIndex]) == NULL)
{
// device does not support it
vlog_error("Platform supports extension \"%s\" but device does not\n",
extensions[extensionIndex]);
return -1;
}
}
return 0;
}
int test_get_platform_ids(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements) {
cl_platform_id platforms[16];
cl_uint num_platforms;
char *string_returned;
string_returned = (char*)malloc(8192);
int total_errors = 0;
int err = CL_SUCCESS;
err = clGetPlatformIDs(16, platforms, &num_platforms);
test_error(err, "clGetPlatformIDs failed");
if (num_platforms <= 16) {
// Try with NULL
err = clGetPlatformIDs(num_platforms, platforms, NULL);
test_error(err, "clGetPlatformIDs failed with NULL for return size");
}
if (num_platforms < 1) {
log_error("Found 0 platforms.\n");
return -1;
}
log_info("Found %d platforms.\n", num_platforms);
for (int p=0; p<(int)num_platforms; p++) {
cl_device_id *devices;
cl_uint num_devices;
size_t size;
log_info("Platform %d (%p):\n", p, platforms[p]);
memset(string_returned, 0, 8192);
err = clGetPlatformInfo(platforms[p], CL_PLATFORM_PROFILE, 8192, string_returned, &size);
test_error(err, "clGetPlatformInfo for CL_PLATFORM_PROFILE failed");
log_info("\tCL_PLATFORM_PROFILE: %s\n", string_returned);
if (strlen(string_returned)+1 != size) {
log_error("Returned string length %ld does not equal reported one %ld.\n", strlen(string_returned)+1, size);
total_errors++;
}
memset(string_returned, 0, 8192);
err = clGetPlatformInfo(platforms[p], CL_PLATFORM_VERSION, 8192, string_returned, &size);
test_error(err, "clGetPlatformInfo for CL_PLATFORM_VERSION failed");
log_info("\tCL_PLATFORM_VERSION: %s\n", string_returned);
if (strlen(string_returned)+1 != size) {
log_error("Returned string length %ld does not equal reported one %ld.\n", strlen(string_returned)+1, size);
total_errors++;
}
memset(string_returned, 0, 8192);
err = clGetPlatformInfo(platforms[p], CL_PLATFORM_NAME, 8192, string_returned, &size);
test_error(err, "clGetPlatformInfo for CL_PLATFORM_NAME failed");
log_info("\tCL_PLATFORM_NAME: %s\n", string_returned);
if (strlen(string_returned)+1 != size) {
log_error("Returned string length %ld does not equal reported one %ld.\n", strlen(string_returned)+1, size);
total_errors++;
}
memset(string_returned, 0, 8192);
err = clGetPlatformInfo(platforms[p], CL_PLATFORM_VENDOR, 8192, string_returned, &size);
test_error(err, "clGetPlatformInfo for CL_PLATFORM_VENDOR failed");
log_info("\tCL_PLATFORM_VENDOR: %s\n", string_returned);
if (strlen(string_returned)+1 != size) {
log_error("Returned string length %ld does not equal reported one %ld.\n", strlen(string_returned)+1, size);
total_errors++;
}
memset(string_returned, 0, 8192);
err = clGetPlatformInfo(platforms[p], CL_PLATFORM_EXTENSIONS, 8192, string_returned, &size);
test_error(err, "clGetPlatformInfo for CL_PLATFORM_EXTENSIONS failed");
log_info("\tCL_PLATFORM_EXTENSIONS: %s\n", string_returned);
if (strlen(string_returned)+1 != size) {
log_error("Returned string length %ld does not equal reported one %ld.\n", strlen(string_returned)+1, size);
total_errors++;
}
err = clGetDeviceIDs(platforms[p], CL_DEVICE_TYPE_ALL, 0, NULL, &num_devices);
test_error(err, "clGetDeviceIDs size failed.\n");
devices = (cl_device_id *)malloc(num_devices*sizeof(cl_device_id));
memset(devices, 0, sizeof(cl_device_id)*num_devices);
err = clGetDeviceIDs(platforms[p], CL_DEVICE_TYPE_ALL, num_devices, devices, NULL);
test_error(err, "clGetDeviceIDs failed.\n");
log_info("\tPlatform has %d devices.\n", (int)num_devices);
for (int d=0; d<(int)num_devices; d++) {
size_t returned_size;
cl_platform_id returned_platform;
cl_context context;
cl_context_properties properties[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)platforms[p], 0 };
err = clGetDeviceInfo(devices[d], CL_DEVICE_PLATFORM, sizeof(cl_platform_id), &returned_platform, &returned_size);
test_error(err, "clGetDeviceInfo failed for CL_DEVICE_PLATFORM\n");
if (returned_size != sizeof(cl_platform_id)) {
log_error("Reported return size (%ld) does not match expected size (%ld).\n", returned_size, sizeof(cl_platform_id));
total_errors++;
}
memset(string_returned, 0, 8192);
err = clGetDeviceInfo(devices[d], CL_DEVICE_NAME, 8192, string_returned, NULL);
test_error(err, "clGetDeviceInfo failed for CL_DEVICE_NAME\n");
log_info("\t\tPlatform for device %d (%s) is %p.\n", d, string_returned, returned_platform);
log_info("\t\t\tTesting clCreateContext for the platform/device...\n");
// Try creating a context for the platform
context = clCreateContext(properties, 1, &devices[d], NULL, NULL, &err);
test_error(err, "\t\tclCreateContext failed for device with platform properties\n");
memset(properties, 0, sizeof(cl_context_properties)*3);
err = clGetContextInfo(context, CL_CONTEXT_PROPERTIES, sizeof(cl_context_properties)*3, properties, &returned_size);
test_error(err, "clGetContextInfo for CL_CONTEXT_PROPERTIES failed");
if (returned_size != sizeof(cl_context_properties)*3) {
log_error("Invalid size returned from clGetContextInfo for CL_CONTEXT_PROPERTIES. Got %ld, expected %ld.\n",
returned_size, sizeof(cl_context_properties)*3);
total_errors++;
}
if (properties[0] != (cl_context_properties)CL_CONTEXT_PLATFORM || properties[1] != (cl_context_properties)platforms[p]) {
log_error("Wrong properties returned. Expected: [%p %p], got [%p %p]\n",
(void*)CL_CONTEXT_PLATFORM, platforms[p], (void*)properties[0], (void*)properties[1]);
total_errors++;
}
err = clReleaseContext(context);
test_error(err, "clReleaseContext failed");
}
free(devices);
}
free(string_returned);
return total_errors;
}

1270
test_conformance/api/test_queries.cpp
View File

File diff suppressed because it is too large Load Diff

174
test_conformance/api/test_queue_properties.cpp Normal file
View File

@@ -0,0 +1,174 @@
//
// Copyright (c) 2018 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
#include "../../test_common/harness/typeWrappers.h"
#include "../../test_common/harness/conversions.h"
#include <sstream>
#include <string>
#include <vector>
using namespace std;
/*
The test against cl_khr_create_command_queue extension. It validates if devices with Opencl 1.X can use clCreateCommandQueueWithPropertiesKHR function.
Based on device capabilities test will create queue with NULL properties, CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE property and
CL_QUEUE_PROFILING_ENABLE property. Finally simple kernel will be executed on such queue.
*/
const char *queue_test_kernel[] = {
"__kernel void vec_cpy(__global int *src, __global int *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = src[tid];\n"
"\n"
"}\n" };
int enqueue_kernel(cl_context context, const cl_queue_properties_khr *queue_prop_def, cl_device_id deviceID, clKernelWrapper& kernel, size_t num_elements)
{
clMemWrapper streams[2];
int error;
std::vector<int> buf(num_elements);
clCreateCommandQueueWithPropertiesKHR_fn clCreateCommandQueueWithPropertiesKHR = NULL;
cl_platform_id platform;
clEventWrapper event;
error = clGetDeviceInfo(deviceID, CL_DEVICE_PLATFORM, sizeof(cl_platform_id), &platform, NULL);
test_error(error, "clGetDeviceInfo for CL_DEVICE_PLATFORM failed");
clCreateCommandQueueWithPropertiesKHR = (clCreateCommandQueueWithPropertiesKHR_fn) clGetExtensionFunctionAddressForPlatform(platform, "clCreateCommandQueueWithPropertiesKHR");
if (clCreateCommandQueueWithPropertiesKHR == NULL)
{
log_error("ERROR: clGetExtensionFunctionAddressForPlatform failed\n");
return -1;
}
clCommandQueueWrapper queue = clCreateCommandQueueWithPropertiesKHR(context, deviceID, queue_prop_def, &error);
test_error(error, "clCreateCommandQueueWithPropertiesKHR failed");
for (int i = 0; i < num_elements; ++i)
{
buf[i] = i;
}
streams[0] = clCreateBuffer(context, CL_MEM_READ_WRITE | CL_MEM_USE_HOST_PTR, num_elements * sizeof(int), buf.data(), &error);
test_error( error, "clCreateBuffer failed." );
streams[1] = clCreateBuffer(context, CL_MEM_READ_WRITE | CL_MEM_ALLOC_HOST_PTR, num_elements * sizeof(int), NULL, &error);
test_error( error, "clCreateBuffer failed." );
error = clSetKernelArg(kernel, 0, sizeof(streams[0]), &streams[0]);
test_error( error, "clSetKernelArg failed." );
error = clSetKernelArg(kernel, 1, sizeof(streams[1]), &streams[1]);
test_error( error, "clSetKernelArg failed." );
error = clEnqueueNDRangeKernel(queue, kernel, 1, NULL, &num_elements, NULL, 0, NULL, &event);
test_error( error, "clEnqueueNDRangeKernel failed." );
error = clWaitForEvents(1, &event);
test_error(error, "clWaitForEvents failed.");
error = clEnqueueReadBuffer(queue, streams[1], CL_TRUE, 0, num_elements, buf.data(), 0, NULL, NULL);
test_error( error, "clEnqueueReadBuffer failed." );
for (int i = 0; i < num_elements; ++i)
{
if (buf[i] != i)
{
log_error("ERROR: Incorrect vector copy result.");
return -1;
}
}
return 0;
}
int test_queue_properties(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
if (num_elements <= 0)
{
num_elements = 128;
}
int error = 0;
clProgramWrapper program;
clKernelWrapper kernel;
size_t strSize;
std::string strExt(0, '\0');
cl_queue_properties_khr device_props = NULL;
cl_queue_properties_khr queue_prop_def[] = { CL_QUEUE_PROPERTIES, 0, 0 };
// Query extension
error = clGetDeviceInfo(deviceID, CL_DEVICE_EXTENSIONS, 0, NULL, &strSize);
test_error(error, "clGetDeviceInfo for CL_DEVICE_EXTENSIONS failed");
strExt.resize(strSize);
error = clGetDeviceInfo(deviceID, CL_DEVICE_EXTENSIONS, strExt.size(), &strExt[0], NULL);
test_error(error, "clGetDeviceInfo for CL_DEVICE_EXTENSIONS failed");
log_info("CL_DEVICE_EXTENSIONS:\n%s\n\n", strExt.c_str());
if (strExt.find("cl_khr_create_command_queue") == string::npos)
{
log_info("extension cl_khr_create_command_queue is not supported.\n");
return 0;
}
error = create_single_kernel_helper(context, &program, &kernel, 1, queue_test_kernel, "vec_cpy");
test_error(error, "create_single_kernel_helper failed");
log_info("Queue property NULL. Testing ... \n");
error = enqueue_kernel(context, NULL,deviceID, kernel, (size_t)num_elements);
test_error(error, "enqueue_kernel failed");
error = clGetDeviceInfo(deviceID, CL_DEVICE_QUEUE_PROPERTIES, sizeof(device_props), &device_props, NULL);
test_error(error, "clGetDeviceInfo for CL_DEVICE_QUEUE_PROPERTIES failed");
if (device_props & CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE)
{
log_info("Queue property CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE supported. Testing ... \n");
queue_prop_def[1] = CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE;
error = enqueue_kernel(context, queue_prop_def, deviceID, kernel, (size_t)num_elements);
test_error(error, "enqueue_kernel failed");
} else
{
log_info("Queue property CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE not supported \n");
}
if (device_props & CL_QUEUE_PROFILING_ENABLE)
{
log_info("Queue property CL_QUEUE_PROFILING_ENABLE supported. Testing ... \n");
queue_prop_def[1] = CL_QUEUE_PROFILING_ENABLE;
error = enqueue_kernel(context, queue_prop_def, deviceID, kernel, (size_t)num_elements);
test_error(error, "enqueue_kernel failed");
} else
{
log_info("Queue property CL_QUEUE_PROFILING_ENABLE not supported \n");
}
if (device_props & CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE && device_props & CL_QUEUE_PROFILING_ENABLE)
{
log_info("Queue property CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE & CL_QUEUE_PROFILING_ENABLE supported. Testing ... \n");
queue_prop_def[1] = CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE|CL_QUEUE_PROFILING_ENABLE;
error = enqueue_kernel(context, queue_prop_def, deviceID, kernel, (size_t)num_elements);
test_error(error, "enqueue_kernel failed");
}
else
{
log_info("Queue property CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE or CL_QUEUE_PROFILING_ENABLE not supported \n");
}
return 0;
}

468
test_conformance/api/test_retain.cpp
View File

@@ -1,234 +1,234 @@
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
#if !defined(_WIN32)
#include <unistd.h>
#endif // !_WIN32
// Note: According to spec, the various functions to get instance counts should return an error when passed in an object
// that has already been released. However, the spec is out of date. If it gets re-updated to allow such action, re-enable
// this define.
//#define VERIFY_AFTER_RELEASE 1
#define GET_QUEUE_INSTANCE_COUNT(p) numInstances = ( (err = clGetCommandQueueInfo(p, CL_QUEUE_REFERENCE_COUNT, sizeof( numInstances ), &numInstances, NULL)) == CL_SUCCESS ? numInstances : 0 )
#define GET_MEM_INSTANCE_COUNT(p) numInstances = ( (err = clGetMemObjectInfo(p, CL_MEM_REFERENCE_COUNT, sizeof( numInstances ), &numInstances, NULL)) == CL_SUCCESS ? numInstances : 0 )
#define VERIFY_INSTANCE_COUNT(c,rightValue) if( c != rightValue ) { \
log_error( "ERROR: Instance count for test object is not valid! (should be %d, really is %d)\n", rightValue, c ); \
return -1; }
int test_retain_queue_single(cl_device_id deviceID, cl_context context, cl_command_queue queueNotUsed, int num_elements)
{
cl_command_queue queue;
cl_uint numInstances;
int err;
/* Create a test queue */
queue = clCreateCommandQueue( context, deviceID, 0, &err );
test_error( err, "Unable to create command queue to test with" );
/* Test the instance count */
GET_QUEUE_INSTANCE_COUNT( queue );
test_error( err, "Unable to get queue instance count" );
VERIFY_INSTANCE_COUNT( numInstances, 1 );
/* Now release the program */
clReleaseCommandQueue( queue );
#ifdef VERIFY_AFTER_RELEASE
/* We're not allowed to get the instance count after the object has been completely released. But that's
exactly how we can tell the release worked--by making sure getting the instance count fails! */
GET_QUEUE_INSTANCE_COUNT( queue );
if( err != CL_INVALID_COMMAND_QUEUE )
{
print_error( err, "Command queue was not properly released" );
return -1;
}
#endif
return 0;
}
int test_retain_queue_multiple(cl_device_id deviceID, cl_context context, cl_command_queue queueNotUsed, int num_elements)
{
cl_command_queue queue;
unsigned int numInstances, i;
int err;
/* Create a test program */
queue = clCreateCommandQueue( context, deviceID, 0, &err );
test_error( err, "Unable to create command queue to test with" );
/* Increment 9 times, which should bring the count to 10 */
for( i = 0; i < 9; i++ )
{
clRetainCommandQueue( queue );
}
/* Test the instance count */
GET_QUEUE_INSTANCE_COUNT( queue );
test_error( err, "Unable to get queue instance count" );
VERIFY_INSTANCE_COUNT( numInstances, 10 );
/* Now release 5 times, which should take us to 5 */
for( i = 0; i < 5; i++ )
{
clReleaseCommandQueue( queue );
}
GET_QUEUE_INSTANCE_COUNT( queue );
test_error( err, "Unable to get queue instance count" );
VERIFY_INSTANCE_COUNT( numInstances, 5 );
/* Retain again three times, which should take us to 8 */
for( i = 0; i < 3; i++ )
{
clRetainCommandQueue( queue );
}
GET_QUEUE_INSTANCE_COUNT( queue );
test_error( err, "Unable to get queue instance count" );
VERIFY_INSTANCE_COUNT( numInstances, 8 );
/* Release 7 times, which should take it to 1 */
for( i = 0; i < 7; i++ )
{
clReleaseCommandQueue( queue );
}
GET_QUEUE_INSTANCE_COUNT( queue );
test_error( err, "Unable to get queue instance count" );
VERIFY_INSTANCE_COUNT( numInstances, 1 );
/* And one last one */
clReleaseCommandQueue( queue );
#ifdef VERIFY_AFTER_RELEASE
/* We're not allowed to get the instance count after the object has been completely released. But that's
exactly how we can tell the release worked--by making sure getting the instance count fails! */
GET_QUEUE_INSTANCE_COUNT( queue );
if( err != CL_INVALID_COMMAND_QUEUE )
{
print_error( err, "Command queue was not properly released" );
return -1;
}
#endif
return 0;
}
int test_retain_mem_object_single(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
cl_mem object;
cl_uint numInstances;
int err;
/* Create a test object */
object = clCreateBuffer( context, CL_MEM_READ_ONLY, 32, NULL, &err );
test_error( err, "Unable to create buffer to test with" );
/* Test the instance count */
GET_MEM_INSTANCE_COUNT( object );
test_error( err, "Unable to get mem object count" );
VERIFY_INSTANCE_COUNT( numInstances, 1 );
/* Now release the program */
clReleaseMemObject( object );
#ifdef VERIFY_AFTER_RELEASE
/* We're not allowed to get the instance count after the object has been completely released. But that's
exactly how we can tell the release worked--by making sure getting the instance count fails! */
GET_MEM_INSTANCE_COUNT( object );
if( err != CL_INVALID_MEM_OBJECT )
{
print_error( err, "Mem object was not properly released" );
return -1;
}
#endif
return 0;
}
int test_retain_mem_object_multiple(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
cl_mem object;
unsigned int numInstances, i;
int err;
/* Create a test object */
object = clCreateBuffer( context, CL_MEM_READ_ONLY, 32, NULL, &err );
test_error( err, "Unable to create buffer to test with" );
/* Increment 9 times, which should bring the count to 10 */
for( i = 0; i < 9; i++ )
{
clRetainMemObject( object );
}
/* Test the instance count */
GET_MEM_INSTANCE_COUNT( object );
test_error( err, "Unable to get mem object count" );
VERIFY_INSTANCE_COUNT( numInstances, 10 );
/* Now release 5 times, which should take us to 5 */
for( i = 0; i < 5; i++ )
{
clReleaseMemObject( object );
}
GET_MEM_INSTANCE_COUNT( object );
test_error( err, "Unable to get mem object count" );
VERIFY_INSTANCE_COUNT( numInstances, 5 );
/* Retain again three times, which should take us to 8 */
for( i = 0; i < 3; i++ )
{
clRetainMemObject( object );
}
GET_MEM_INSTANCE_COUNT( object );
test_error( err, "Unable to get mem object count" );
VERIFY_INSTANCE_COUNT( numInstances, 8 );
/* Release 7 times, which should take it to 1 */
for( i = 0; i < 7; i++ )
{
clReleaseMemObject( object );
}
GET_MEM_INSTANCE_COUNT( object );
test_error( err, "Unable to get mem object count" );
VERIFY_INSTANCE_COUNT( numInstances, 1 );
/* And one last one */
clReleaseMemObject( object );
#ifdef VERIFY_AFTER_RELEASE
/* We're not allowed to get the instance count after the object has been completely released. But that's
exactly how we can tell the release worked--by making sure getting the instance count fails! */
GET_MEM_INSTANCE_COUNT( object );
if( err != CL_INVALID_MEM_OBJECT )
{
print_error( err, "Mem object was not properly released" );
return -1;
}
#endif
return 0;
}
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
#if !defined(_WIN32)
#include <unistd.h>
#endif // !_WIN32
// Note: According to spec, the various functions to get instance counts should return an error when passed in an object
// that has already been released. However, the spec is out of date. If it gets re-updated to allow such action, re-enable
// this define.
//#define VERIFY_AFTER_RELEASE 1
#define GET_QUEUE_INSTANCE_COUNT(p) numInstances = ( (err = clGetCommandQueueInfo(p, CL_QUEUE_REFERENCE_COUNT, sizeof( numInstances ), &numInstances, NULL)) == CL_SUCCESS ? numInstances : 0 )
#define GET_MEM_INSTANCE_COUNT(p) numInstances = ( (err = clGetMemObjectInfo(p, CL_MEM_REFERENCE_COUNT, sizeof( numInstances ), &numInstances, NULL)) == CL_SUCCESS ? numInstances : 0 )
#define VERIFY_INSTANCE_COUNT(c,rightValue) if( c != rightValue ) { \
log_error( "ERROR: Instance count for test object is not valid! (should be %d, really is %d)\n", rightValue, c ); \
return -1; }
int test_retain_queue_single(cl_device_id deviceID, cl_context context, cl_command_queue queueNotUsed, int num_elements)
{
cl_command_queue queue;
cl_uint numInstances;
int err;
/* Create a test queue */
queue = clCreateCommandQueue( context, deviceID, 0, &err );
test_error( err, "Unable to create command queue to test with" );
/* Test the instance count */
GET_QUEUE_INSTANCE_COUNT( queue );
test_error( err, "Unable to get queue instance count" );
VERIFY_INSTANCE_COUNT( numInstances, 1 );
/* Now release the program */
clReleaseCommandQueue( queue );
#ifdef VERIFY_AFTER_RELEASE
/* We're not allowed to get the instance count after the object has been completely released. But that's
exactly how we can tell the release worked--by making sure getting the instance count fails! */
GET_QUEUE_INSTANCE_COUNT( queue );
if( err != CL_INVALID_COMMAND_QUEUE )
{
print_error( err, "Command queue was not properly released" );
return -1;
}
#endif
return 0;
}
int test_retain_queue_multiple(cl_device_id deviceID, cl_context context, cl_command_queue queueNotUsed, int num_elements)
{
cl_command_queue queue;
unsigned int numInstances, i;
int err;
/* Create a test program */
queue = clCreateCommandQueue( context, deviceID, 0, &err );
test_error( err, "Unable to create command queue to test with" );
/* Increment 9 times, which should bring the count to 10 */
for( i = 0; i < 9; i++ )
{
clRetainCommandQueue( queue );
}
/* Test the instance count */
GET_QUEUE_INSTANCE_COUNT( queue );
test_error( err, "Unable to get queue instance count" );
VERIFY_INSTANCE_COUNT( numInstances, 10 );
/* Now release 5 times, which should take us to 5 */
for( i = 0; i < 5; i++ )
{
clReleaseCommandQueue( queue );
}
GET_QUEUE_INSTANCE_COUNT( queue );
test_error( err, "Unable to get queue instance count" );
VERIFY_INSTANCE_COUNT( numInstances, 5 );
/* Retain again three times, which should take us to 8 */
for( i = 0; i < 3; i++ )
{
clRetainCommandQueue( queue );
}
GET_QUEUE_INSTANCE_COUNT( queue );
test_error( err, "Unable to get queue instance count" );
VERIFY_INSTANCE_COUNT( numInstances, 8 );
/* Release 7 times, which should take it to 1 */
for( i = 0; i < 7; i++ )
{
clReleaseCommandQueue( queue );
}
GET_QUEUE_INSTANCE_COUNT( queue );
test_error( err, "Unable to get queue instance count" );
VERIFY_INSTANCE_COUNT( numInstances, 1 );
/* And one last one */
clReleaseCommandQueue( queue );
#ifdef VERIFY_AFTER_RELEASE
/* We're not allowed to get the instance count after the object has been completely released. But that's
exactly how we can tell the release worked--by making sure getting the instance count fails! */
GET_QUEUE_INSTANCE_COUNT( queue );
if( err != CL_INVALID_COMMAND_QUEUE )
{
print_error( err, "Command queue was not properly released" );
return -1;
}
#endif
return 0;
}
int test_retain_mem_object_single(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
cl_mem object;
cl_uint numInstances;
int err;
/* Create a test object */
object = clCreateBuffer( context, CL_MEM_READ_ONLY, 32, NULL, &err );
test_error( err, "Unable to create buffer to test with" );
/* Test the instance count */
GET_MEM_INSTANCE_COUNT( object );
test_error( err, "Unable to get mem object count" );
VERIFY_INSTANCE_COUNT( numInstances, 1 );
/* Now release the program */
clReleaseMemObject( object );
#ifdef VERIFY_AFTER_RELEASE
/* We're not allowed to get the instance count after the object has been completely released. But that's
exactly how we can tell the release worked--by making sure getting the instance count fails! */
GET_MEM_INSTANCE_COUNT( object );
if( err != CL_INVALID_MEM_OBJECT )
{
print_error( err, "Mem object was not properly released" );
return -1;
}
#endif
return 0;
}
int test_retain_mem_object_multiple(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
cl_mem object;
unsigned int numInstances, i;
int err;
/* Create a test object */
object = clCreateBuffer( context, CL_MEM_READ_ONLY, 32, NULL, &err );
test_error( err, "Unable to create buffer to test with" );
/* Increment 9 times, which should bring the count to 10 */
for( i = 0; i < 9; i++ )
{
clRetainMemObject( object );
}
/* Test the instance count */
GET_MEM_INSTANCE_COUNT( object );
test_error( err, "Unable to get mem object count" );
VERIFY_INSTANCE_COUNT( numInstances, 10 );
/* Now release 5 times, which should take us to 5 */
for( i = 0; i < 5; i++ )
{
clReleaseMemObject( object );
}
GET_MEM_INSTANCE_COUNT( object );
test_error( err, "Unable to get mem object count" );
VERIFY_INSTANCE_COUNT( numInstances, 5 );
/* Retain again three times, which should take us to 8 */
for( i = 0; i < 3; i++ )
{
clRetainMemObject( object );
}
GET_MEM_INSTANCE_COUNT( object );
test_error( err, "Unable to get mem object count" );
VERIFY_INSTANCE_COUNT( numInstances, 8 );
/* Release 7 times, which should take it to 1 */
for( i = 0; i < 7; i++ )
{
clReleaseMemObject( object );
}
GET_MEM_INSTANCE_COUNT( object );
test_error( err, "Unable to get mem object count" );
VERIFY_INSTANCE_COUNT( numInstances, 1 );
/* And one last one */
clReleaseMemObject( object );
#ifdef VERIFY_AFTER_RELEASE
/* We're not allowed to get the instance count after the object has been completely released. But that's
exactly how we can tell the release worked--by making sure getting the instance count fails! */
GET_MEM_INSTANCE_COUNT( object );
if( err != CL_INVALID_MEM_OBJECT )
{
print_error( err, "Mem object was not properly released" );
return -1;
}
#endif
return 0;
}

218
test_conformance/api/test_retain_program.c
View File

@@ -1,109 +1,109 @@
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
#if !defined(_WIN32)
#include <unistd.h>
#endif
#include "../../test_common/harness/compat.h"
int test_release_kernel_order(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
cl_program program;
cl_kernel kernel;
int error;
const char *testProgram[] = { "__kernel void sample_test(__global int *data){}" };
/* Create a test program */
program = clCreateProgramWithSource( context, 1, testProgram, NULL, &error);
test_error( error, "Unable to create program to test with" );
/* Compile the program */
error = clBuildProgram( program, 1, &deviceID, NULL, NULL, NULL );
test_error( error, "Unable to build sample program to test with" );
/* And create a kernel from it */
kernel = clCreateKernel( program, "sample_test", &error );
test_error( error, "Unable to create kernel" );
/* Now try freeing the program first, then the kernel. If refcounts are right, this should work just fine */
clReleaseProgram( program );
clReleaseKernel( kernel );
/* If we got here fine, we succeeded. If not, well, we won't be able to return an error :) */
return 0;
}
const char *sample_delay_kernel[] = {
"__kernel void sample_test(__global float *src, __global int *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
" for( int i = 0; i < 1000000; i++ ); \n"
" dst[tid] = (int)src[tid];\n"
"\n"
"}\n" };
int test_release_during_execute( cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
int error;
cl_program program;
cl_kernel kernel;
cl_mem streams[2];
size_t threads[1] = { 10 }, localThreadSize;
/* We now need an event to test. So we'll execute a kernel to get one */
if( create_single_kernel_helper( context, &program, &kernel, 1, sample_delay_kernel, "sample_test" ) )
{
return -1;
}
streams[0] = clCreateBuffer(context, (cl_mem_flags)(CL_MEM_READ_WRITE), sizeof(cl_float) * 10, NULL, &error);
test_error( error, "Creating test array failed" );
streams[1] = clCreateBuffer(context, (cl_mem_flags)(CL_MEM_READ_WRITE), sizeof(cl_int) * 10, NULL, &error);
test_error( error, "Creating test array failed" );
/* Set the arguments */
error = clSetKernelArg(kernel, 0, sizeof( streams[0] ), &streams[ 0 ]);
test_error( error, "Unable to set indexed kernel arguments" );
error = clSetKernelArg(kernel, 1, sizeof( streams[1] ), &streams[ 1 ]);
test_error( error, "Unable to set indexed kernel arguments" );
error = get_max_common_work_group_size( context, kernel, threads[0], &localThreadSize );
test_error( error, "Unable to calc local thread size" );
/* Execute the kernel */
error = clEnqueueNDRangeKernel( queue, kernel, 1, NULL, threads, &localThreadSize, 0, NULL, NULL );
test_error( error, "Unable to execute test kernel" );
/* The kernel should still be executing, but we should still be able to release it. It's not terribly
useful, but we should be able to do it, if the internal refcounting is indeed correct. */
clReleaseMemObject( streams[ 1 ] );
clReleaseMemObject( streams[ 0 ] );
clReleaseKernel( kernel );
clReleaseProgram( program );
/* Now make sure we're really finished before we go on. */
error = clFinish(queue);
test_error( error, "Unable to finish context.");
return 0;
}
//
// Copyright (c) 2017 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#include "testBase.h"
#if !defined(_WIN32)
#include <unistd.h>
#endif
#include "../../test_common/harness/compat.h"
int test_release_kernel_order(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
cl_program program;
cl_kernel kernel;
int error;
const char *testProgram[] = { "__kernel void sample_test(__global int *data){}" };
/* Create a test program */
program = clCreateProgramWithSource( context, 1, testProgram, NULL, &error);
test_error( error, "Unable to create program to test with" );
/* Compile the program */
error = clBuildProgram( program, 1, &deviceID, NULL, NULL, NULL );
test_error( error, "Unable to build sample program to test with" );
/* And create a kernel from it */
kernel = clCreateKernel( program, "sample_test", &error );
test_error( error, "Unable to create kernel" );
/* Now try freeing the program first, then the kernel. If refcounts are right, this should work just fine */
clReleaseProgram( program );
clReleaseKernel( kernel );
/* If we got here fine, we succeeded. If not, well, we won't be able to return an error :) */
return 0;
}
const char *sample_delay_kernel[] = {
"__kernel void sample_test(__global float *src, __global int *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
" for( int i = 0; i < 1000000; i++ ); \n"
" dst[tid] = (int)src[tid];\n"
"\n"
"}\n" };
int test_release_during_execute( cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
int error;
cl_program program;
cl_kernel kernel;
cl_mem streams[2];
size_t threads[1] = { 10 }, localThreadSize;
/* We now need an event to test. So we'll execute a kernel to get one */
if( create_single_kernel_helper( context, &program, &kernel, 1, sample_delay_kernel, "sample_test" ) )
{
return -1;
}
streams[0] = clCreateBuffer(context, (cl_mem_flags)(CL_MEM_READ_WRITE), sizeof(cl_float) * 10, NULL, &error);
test_error( error, "Creating test array failed" );
streams[1] = clCreateBuffer(context, (cl_mem_flags)(CL_MEM_READ_WRITE), sizeof(cl_int) * 10, NULL, &error);
test_error( error, "Creating test array failed" );
/* Set the arguments */
error = clSetKernelArg(kernel, 0, sizeof( streams[0] ), &streams[ 0 ]);
test_error( error, "Unable to set indexed kernel arguments" );
error = clSetKernelArg(kernel, 1, sizeof( streams[1] ), &streams[ 1 ]);
test_error( error, "Unable to set indexed kernel arguments" );
error = get_max_common_work_group_size( context, kernel, threads[0], &localThreadSize );
test_error( error, "Unable to calc local thread size" );
/* Execute the kernel */
error = clEnqueueNDRangeKernel( queue, kernel, 1, NULL, threads, &localThreadSize, 0, NULL, NULL );
test_error( error, "Unable to execute test kernel" );
/* The kernel should still be executing, but we should still be able to release it. It's not terribly
useful, but we should be able to do it, if the internal refcounting is indeed correct. */
clReleaseMemObject( streams[ 1 ] );
clReleaseMemObject( streams[ 0 ] );
clReleaseKernel( kernel );
clReleaseProgram( program );
/* Now make sure we're really finished before we go on. */
error = clFinish(queue);
test_error( error, "Unable to finish context.");
return 0;
}