blob: a3d025441da749bb8fbe2e8ad7a34ac592263548 [file] [log] [blame]
//
// 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 "harness/compat.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <time.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <CL/cl_half.h>
#include "procs.h"
//#define HK_DO_NOT_RUN_SHORT_ASYNC 1
//#define HK_DO_NOT_RUN_USHORT_ASYNC 1
//#define HK_DO_NOT_RUN_CHAR_ASYNC 1
//#define HK_DO_NOT_RUN_UCHAR_ASYNC 1
#define TEST_PRIME_INT ((1<<16)+1)
#define TEST_PRIME_UINT ((1U<<16)+1U)
#define TEST_PRIME_LONG ((1LL<<32)+1LL)
#define TEST_PRIME_ULONG ((1ULL<<32)+1ULL)
#define TEST_PRIME_SHORT ((1S<<8)+1S)
#define TEST_PRIME_FLOAT (float)3.40282346638528860e+38
#define TEST_PRIME_HALF 119.f
#define TEST_BOOL true
#define TEST_PRIME_CHAR 0x77
#ifndef ulong
typedef unsigned long ulong;
#endif
#ifndef uchar
typedef unsigned char uchar;
#endif
#ifndef TestStruct
typedef struct{
int a;
float b;
} TestStruct;
#endif
//--- the code for the kernel executables
static const char *buffer_read_int_kernel_code[] = {
"__kernel void test_buffer_read_int(__global int *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1<<16)+1);\n"
"}\n",
"__kernel void test_buffer_read_int2(__global int2 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1<<16)+1);\n"
"}\n",
"__kernel void test_buffer_read_int4(__global int4 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1<<16)+1);\n"
"}\n",
"__kernel void test_buffer_read_int8(__global int8 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1<<16)+1);\n"
"}\n",
"__kernel void test_buffer_read_int16(__global int16 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1<<16)+1);\n"
"}\n" };
static const char *int_kernel_name[] = { "test_buffer_read_int", "test_buffer_read_int2", "test_buffer_read_int4", "test_buffer_read_int8", "test_buffer_read_int16" };
static const char *buffer_read_uint_kernel_code[] = {
"__kernel void test_buffer_read_uint(__global uint *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1U<<16)+1U);\n"
"}\n",
"__kernel void test_buffer_read_uint2(__global uint2 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1U<<16)+1U);\n"
"}\n",
"__kernel void test_buffer_read_uint4(__global uint4 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1U<<16)+1U);\n"
"}\n",
"__kernel void test_buffer_read_uint8(__global uint8 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1U<<16)+1U);\n"
"}\n",
"__kernel void test_buffer_read_uint16(__global uint16 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1U<<16)+1U);\n"
"}\n" };
static const char *uint_kernel_name[] = { "test_buffer_read_uint", "test_buffer_read_uint2", "test_buffer_read_uint4", "test_buffer_read_uint8", "test_buffer_read_uint16" };
static const char *buffer_read_long_kernel_code[] = {
"__kernel void test_buffer_read_long(__global long *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1L<<32)+1L);\n"
"}\n",
"__kernel void test_buffer_read_long2(__global long2 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1L<<32)+1L);\n"
"}\n",
"__kernel void test_buffer_read_long4(__global long4 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1L<<32)+1L);\n"
"}\n",
"__kernel void test_buffer_read_long8(__global long8 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1L<<32)+1L);\n"
"}\n",
"__kernel void test_buffer_read_long16(__global long16 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1L<<32)+1L);\n"
"}\n" };
static const char *long_kernel_name[] = { "test_buffer_read_long", "test_buffer_read_long2", "test_buffer_read_long4", "test_buffer_read_long8", "test_buffer_read_long16" };
static const char *buffer_read_ulong_kernel_code[] = {
"__kernel void test_buffer_read_ulong(__global ulong *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1UL<<32)+1UL);\n"
"}\n",
"__kernel void test_buffer_read_ulong2(__global ulong2 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1UL<<32)+1UL);\n"
"}\n",
"__kernel void test_buffer_read_ulong4(__global ulong4 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1UL<<32)+1UL);\n"
"}\n",
"__kernel void test_buffer_read_ulong8(__global ulong8 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1UL<<32)+1UL);\n"
"}\n",
"__kernel void test_buffer_read_ulong16(__global ulong16 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = ((1UL<<32)+1UL);\n"
"}\n" };
static const char *ulong_kernel_name[] = { "test_buffer_read_ulong", "test_buffer_read_ulong2", "test_buffer_read_ulong4", "test_buffer_read_ulong8", "test_buffer_read_ulong16" };
static const char *buffer_read_short_kernel_code[] = {
"__kernel void test_buffer_read_short(__global short *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (short)((1<<8)+1);\n"
"}\n",
"__kernel void test_buffer_read_short2(__global short2 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (short)((1<<8)+1);\n"
"}\n",
"__kernel void test_buffer_read_short4(__global short4 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (short)((1<<8)+1);\n"
"}\n",
"__kernel void test_buffer_read_short8(__global short8 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (short)((1<<8)+1);\n"
"}\n",
"__kernel void test_buffer_read_short16(__global short16 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (short)((1<<8)+1);\n"
"}\n" };
static const char *short_kernel_name[] = { "test_buffer_read_short", "test_buffer_read_short2", "test_buffer_read_short4", "test_buffer_read_short8", "test_buffer_read_short16" };
static const char *buffer_read_ushort_kernel_code[] = {
"__kernel void test_buffer_read_ushort(__global ushort *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (ushort)((1<<8)+1);\n"
"}\n",
"__kernel void test_buffer_read_ushort2(__global ushort2 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (ushort)((1<<8)+1);\n"
"}\n",
"__kernel void test_buffer_read_ushort4(__global ushort4 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (ushort)((1<<8)+1);\n"
"}\n",
"__kernel void test_buffer_read_ushort8(__global ushort8 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (ushort)((1<<8)+1);\n"
"}\n",
"__kernel void test_buffer_read_ushort16(__global ushort16 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (ushort)((1<<8)+1);\n"
"}\n" };
static const char *ushort_kernel_name[] = { "test_buffer_read_ushort", "test_buffer_read_ushort2", "test_buffer_read_ushort4", "test_buffer_read_ushort8", "test_buffer_read_ushort16" };
static const char *buffer_read_float_kernel_code[] = {
"__kernel void test_buffer_read_float(__global float *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (float)3.40282346638528860e+38;\n"
"}\n",
"__kernel void test_buffer_read_float2(__global float2 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (float)3.40282346638528860e+38;\n"
"}\n",
"__kernel void test_buffer_read_float4(__global float4 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (float)3.40282346638528860e+38;\n"
"}\n",
"__kernel void test_buffer_read_float8(__global float8 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (float)3.40282346638528860e+38;\n"
"}\n",
"__kernel void test_buffer_read_float16(__global float16 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (float)3.40282346638528860e+38;\n"
"}\n" };
static const char *float_kernel_name[] = { "test_buffer_read_float", "test_buffer_read_float2", "test_buffer_read_float4", "test_buffer_read_float8", "test_buffer_read_float16" };
static const char *buffer_read_half_kernel_code[] = {
"#pragma OPENCL EXTENSION cl_khr_fp16 : enable\n"
"__kernel void test_buffer_read_half(__global half *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (half)119;\n"
"}\n",
"#pragma OPENCL EXTENSION cl_khr_fp16 : enable\n"
"__kernel void test_buffer_read_half2(__global half2 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (half)119;\n"
"}\n",
"#pragma OPENCL EXTENSION cl_khr_fp16 : enable\n"
"__kernel void test_buffer_read_half4(__global half4 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (half)119;\n"
"}\n",
"#pragma OPENCL EXTENSION cl_khr_fp16 : enable\n"
"__kernel void test_buffer_read_half8(__global half8 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (half)119;\n"
"}\n",
"#pragma OPENCL EXTENSION cl_khr_fp16 : enable\n"
"__kernel void test_buffer_read_half16(__global half16 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (half)119;\n"
"}\n"
};
static const char *half_kernel_name[] = { "test_buffer_read_half", "test_buffer_read_half2", "test_buffer_read_half4", "test_buffer_read_half8", "test_buffer_read_half16" };
static const char *buffer_read_char_kernel_code[] = {
"__kernel void test_buffer_read_char(__global char *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (char)'w';\n"
"}\n",
"__kernel void test_buffer_read_char2(__global char2 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (char)'w';\n"
"}\n",
"__kernel void test_buffer_read_char4(__global char4 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (char)'w';\n"
"}\n",
"__kernel void test_buffer_read_char8(__global char8 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (char)'w';\n"
"}\n",
"__kernel void test_buffer_read_char16(__global char16 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (char)'w';\n"
"}\n" };
static const char *char_kernel_name[] = { "test_buffer_read_char", "test_buffer_read_char2", "test_buffer_read_char4", "test_buffer_read_char8", "test_buffer_read_char16" };
static const char *buffer_read_uchar_kernel_code[] = {
"__kernel void test_buffer_read_uchar(__global uchar *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = 'w';\n"
"}\n",
"__kernel void test_buffer_read_uchar2(__global uchar2 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (uchar)'w';\n"
"}\n",
"__kernel void test_buffer_read_uchar4(__global uchar4 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (uchar)'w';\n"
"}\n",
"__kernel void test_buffer_read_uchar8(__global uchar8 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (uchar)'w';\n"
"}\n",
"__kernel void test_buffer_read_uchar16(__global uchar16 *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid] = (uchar)'w';\n"
"}\n" };
static const char *uchar_kernel_name[] = { "test_buffer_read_uchar", "test_buffer_read_uchar2", "test_buffer_read_uchar4", "test_buffer_read_uchar8", "test_buffer_read_uchar16" };
static const char *buffer_read_struct_kernel_code =
"typedef struct{\n"
"int a;\n"
"float b;\n"
"} TestStruct;\n"
"__kernel void test_buffer_read_struct(__global TestStruct *dst)\n"
"{\n"
" int tid = get_global_id(0);\n"
"\n"
" dst[tid].a = ((1<<16)+1);\n"
" dst[tid].b = (float)3.40282346638528860e+38;\n"
"}\n";
//--- the verify functions
static int verify_read_int(void *ptr, int n)
{
int i;
cl_int *outptr = (cl_int *)ptr;
for (i=0; i<n; i++){
if ( outptr[i] != TEST_PRIME_INT )
return -1;
}
return 0;
}
static int verify_read_uint(void *ptr, int n)
{
int i;
cl_uint *outptr = (cl_uint *)ptr;
for (i=0; i<n; i++){
if ( outptr[i] != TEST_PRIME_UINT )
return -1;
}
return 0;
}
static int verify_read_long(void *ptr, int n)
{
int i;
cl_long *outptr = (cl_long *)ptr;
for (i=0; i<n; i++){
if ( outptr[i] != TEST_PRIME_LONG )
return -1;
}
return 0;
}
static int verify_read_ulong(void *ptr, int n)
{
int i;
cl_ulong *outptr = (cl_ulong *)ptr;
for (i=0; i<n; i++){
if ( outptr[i] != TEST_PRIME_ULONG )
return -1;
}
return 0;
}
static int verify_read_short(void *ptr, int n)
{
int i;
cl_short *outptr = (cl_short *)ptr;
for (i=0; i<n; i++){
if ( outptr[i] != (cl_short)((1<<8)+1) )
return -1;
}
return 0;
}
static int verify_read_ushort(void *ptr, int n)
{
int i;
cl_ushort *outptr = (cl_ushort *)ptr;
for (i=0; i<n; i++){
if ( outptr[i] != (cl_ushort)((1<<8)+1) )
return -1;
}
return 0;
}
static int verify_read_float( void *ptr, int n )
{
int i;
cl_float *outptr = (cl_float *)ptr;
for (i=0; i<n; i++){
if ( outptr[i] != TEST_PRIME_FLOAT )
return -1;
}
return 0;
}
static int verify_read_half( void *ptr, int n )
{
int i;
cl_half *outptr = (cl_half *)ptr;
for (i = 0; i < n; i++)
{
if (cl_half_to_float(outptr[i]) != TEST_PRIME_HALF) return -1;
}
return 0;
}
static int verify_read_char(void *ptr, int n)
{
int i;
cl_char *outptr = (cl_char *)ptr;
for (i=0; i<n; i++){
if ( outptr[i] != TEST_PRIME_CHAR )
return -1;
}
return 0;
}
static int verify_read_uchar(void *ptr, int n)
{
int i;
cl_uchar *outptr = (cl_uchar *)ptr;
for (i=0; i<n; i++){
if ( outptr[i] != TEST_PRIME_CHAR )
return -1;
}
return 0;
}
static int verify_read_struct(TestStruct *outptr, int n)
{
int i;
for (i=0; i<n; i++)
{
if ( ( outptr[i].a != TEST_PRIME_INT ) ||
( outptr[i].b != TEST_PRIME_FLOAT ) )
return -1;
}
return 0;
}
//----- the test functions
int test_buffer_read( cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements, size_t size, char *type, int loops,
const char *kernelCode[], const char *kernelName[], int (*fn)(void *,int) )
{
cl_mem buffers[5];
void *outptr[5];
void *inptr[5];
cl_program program[5];
cl_kernel kernel[5];
size_t global_work_size[3];
cl_int err;
int i;
size_t ptrSizes[5];
int src_flag_id;
int total_errors = 0;
size_t min_alignment = get_min_alignment(context);
global_work_size[0] = (cl_uint)num_elements;
ptrSizes[0] = size;
ptrSizes[1] = ptrSizes[0] << 1;
ptrSizes[2] = ptrSizes[1] << 1;
ptrSizes[3] = ptrSizes[2] << 1;
ptrSizes[4] = ptrSizes[3] << 1;
//skip devices that don't support long
if (! gHasLong && strstr(type,"long") )
{
log_info( "Device does not support 64-bit integers. Skipping test.\n" );
return CL_SUCCESS;
}
for (src_flag_id=0; src_flag_id < NUM_FLAGS; src_flag_id++) {
log_info("Testing with cl_mem_flags src: %s\n", flag_set_names[src_flag_id]);
for ( i = 0; i < loops; i++ ){
outptr[i] = align_malloc( ptrSizes[i] * num_elements, min_alignment);
if ( ! outptr[i] ){
log_error( " unable to allocate %d bytes for outptr\n", (int)( ptrSizes[i] * num_elements ) );
return -1;
}
inptr[i] = align_malloc( ptrSizes[i] * num_elements, min_alignment);
if ( ! inptr[i] ){
log_error( " unable to allocate %d bytes for inptr\n", (int)( ptrSizes[i] * num_elements ) );
return -1;
}
if ((flag_set[src_flag_id] & CL_MEM_USE_HOST_PTR) || (flag_set[src_flag_id] & CL_MEM_COPY_HOST_PTR))
buffers[i] = clCreateBuffer(context, flag_set[src_flag_id], ptrSizes[i] * num_elements, inptr[i], &err);
else
buffers[i] = clCreateBuffer(context, flag_set[src_flag_id], ptrSizes[i] * num_elements, NULL, &err);
if ( err != CL_SUCCESS ){
print_error(err, " clCreateBuffer failed\n" );
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
err = create_single_kernel_helper( context, &program[i], &kernel[i], 1, &kernelCode[i], kernelName[i] );
if ( err ){
log_error("Creating program for %s\n", type);
print_error(err, " Error creating program " );
clReleaseMemObject(buffers[i]);
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
err = clSetKernelArg( kernel[i], 0, sizeof( cl_mem ), (void *)&buffers[i] );
if ( err != CL_SUCCESS ){
print_error( err, "clSetKernelArg failed" );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
err = clEnqueueNDRangeKernel( queue, kernel[i], 1, NULL, global_work_size, NULL, 0, NULL, NULL );
if ( err != CL_SUCCESS ){
print_error( err, "clEnqueueNDRangeKernel failed" );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
err = clEnqueueReadBuffer( queue, buffers[i], CL_TRUE, 0, ptrSizes[i]*num_elements, outptr[i], 0, NULL, NULL );
if ( err != CL_SUCCESS ){
print_error( err, "clEnqueueReadBuffer failed" );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
if (fn(outptr[i], num_elements*(1<<i))){
log_error( " %s%d test failed\n", type, 1<<i );
total_errors++;
}
else{
log_info( " %s%d test passed\n", type, 1<<i );
}
err = clEnqueueReadBuffer( queue, buffers[i], CL_TRUE, 0, ptrSizes[i]*num_elements, inptr[i], 0, NULL, NULL );
if ( err != CL_SUCCESS ){
print_error( err, "clEnqueueReadBuffer failed" );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
if (fn(inptr[i], num_elements*(1<<i))){
log_error( " %s%d test failed in-place readback\n", type, 1<<i );
total_errors++;
}
else{
log_info( " %s%d test passed in-place readback\n", type, 1<<i );
}
// cleanup
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
align_free( inptr[i] );
}
} // mem flag
return total_errors;
} // end test_buffer_read()
int test_buffer_read_async( cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements, size_t size, char *type, int loops,
const char *kernelCode[], const char *kernelName[], int (*fn)(void *,int) )
{
cl_mem buffers[5];
cl_program program[5];
cl_kernel kernel[5];
cl_event event;
void *outptr[5];
void *inptr[5];
size_t global_work_size[3];
cl_int err;
int i;
size_t lastIndex;
size_t ptrSizes[5];
int src_flag_id;
int total_errors = 0;
size_t min_alignment = get_min_alignment(context);
global_work_size[0] = (cl_uint)num_elements;
ptrSizes[0] = size;
ptrSizes[1] = ptrSizes[0] << 1;
ptrSizes[2] = ptrSizes[1] << 1;
ptrSizes[3] = ptrSizes[2] << 1;
ptrSizes[4] = ptrSizes[3] << 1;
//skip devices that don't support long
if (! gHasLong && strstr(type,"long") )
{
log_info( "Device does not support 64-bit integers. Skipping test.\n" );
return CL_SUCCESS;
}
for (src_flag_id=0; src_flag_id < NUM_FLAGS; src_flag_id++) {
log_info("Testing with cl_mem_flags src: %s\n", flag_set_names[src_flag_id]);
for ( i = 0; i < loops; i++ ){
outptr[i] = align_malloc(ptrSizes[i] * num_elements, min_alignment);
if ( ! outptr[i] ){
log_error( " unable to allocate %d bytes for outptr\n", (int)(ptrSizes[i] * num_elements) );
return -1;
}
memset( outptr[i], 0, ptrSizes[i] * num_elements ); // initialize to zero to tell difference
inptr[i] = align_malloc(ptrSizes[i] * num_elements, min_alignment);
if ( ! inptr[i] ){
log_error( " unable to allocate %d bytes for inptr\n", (int)(ptrSizes[i] * num_elements) );
return -1;
}
memset( inptr[i], 0, ptrSizes[i] * num_elements ); // initialize to zero to tell difference
if ((flag_set[src_flag_id] & CL_MEM_USE_HOST_PTR) || (flag_set[src_flag_id] & CL_MEM_COPY_HOST_PTR))
buffers[i] = clCreateBuffer(context, flag_set[src_flag_id], ptrSizes[i] * num_elements, inptr[i], &err);
else
buffers[i] = clCreateBuffer(context, flag_set[src_flag_id], ptrSizes[i] * num_elements, NULL, &err);
if ( err != CL_SUCCESS ){
print_error(err, " clCreateBuffer failed\n" );
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
err = create_single_kernel_helper( context, &program[i], &kernel[i], 1, &kernelCode[i], kernelName[i]);
if ( err ){
log_error( " Error creating program for %s\n", type );
clReleaseMemObject( buffers[i] );
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
err = clSetKernelArg( kernel[i], 0, sizeof( cl_mem ), (void *)&buffers[i] );
if ( err != CL_SUCCESS ){
print_error( err, "clSetKernelArg failed" );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
err = clEnqueueNDRangeKernel( queue, kernel[i], 1, NULL, global_work_size, NULL, 0, NULL, NULL );
if ( err != CL_SUCCESS ){
print_error( err, "clEnqueueNDRangeKernel failed" );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
lastIndex = ( num_elements * ( 1 << i ) - 1 ) * ptrSizes[0];
err = clEnqueueReadBuffer( queue, buffers[i], false, 0, ptrSizes[i]*num_elements, outptr[i], 0, NULL, &event );
#ifdef CHECK_FOR_NON_WAIT
if ( ((uchar *)outptr[i])[lastIndex] ){
log_error( " clEnqueueReadBuffer() possibly returned only after inappropriately waiting for execution to be finished\n" );
log_error( " Function was run asynchornously, but last value in array was set in code line following clEnqueueReadBuffer()\n" );
}
#endif
if ( err != CL_SUCCESS ){
print_error( err, "clEnqueueReadBuffer failed" );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
err = clWaitForEvents(1, &event );
if ( err != CL_SUCCESS ){
print_error( err, "clWaitForEvents() failed" );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
if ( fn(outptr[i], num_elements*(1<<i)) ){
log_error( " %s%d test failed\n", type, 1<<i );
total_errors++;
}
else{
log_info( " %s%d test passed\n", type, 1<<i );
}
// cleanup
clReleaseEvent( event );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
align_free( inptr[i] );
}
} // mem flags
return total_errors;
} // end test_buffer_read_array_async()
int test_buffer_read_array_barrier( cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements, size_t size, char *type, int loops,
const char *kernelCode[], const char *kernelName[], int (*fn)(void *,int) )
{
cl_mem buffers[5];
cl_program program[5];
cl_kernel kernel[5];
cl_event event;
void *outptr[5], *inptr[5];
size_t global_work_size[3];
cl_int err;
int i;
size_t lastIndex;
size_t ptrSizes[5];
int src_flag_id;
int total_errors = 0;
size_t min_alignment = get_min_alignment(context);
global_work_size[0] = (cl_uint)num_elements;
ptrSizes[0] = size;
ptrSizes[1] = ptrSizes[0] << 1;
ptrSizes[2] = ptrSizes[1] << 1;
ptrSizes[3] = ptrSizes[2] << 1;
ptrSizes[4] = ptrSizes[3] << 1;
//skip devices that don't support long
if (! gHasLong && strstr(type,"long") )
{
log_info( "Device does not support 64-bit integers. Skipping test.\n" );
return CL_SUCCESS;
}
for (src_flag_id=0; src_flag_id < NUM_FLAGS; src_flag_id++) {
log_info("Testing with cl_mem_flags src: %s\n", flag_set_names[src_flag_id]);
for ( i = 0; i < loops; i++ ){
outptr[i] = align_malloc(ptrSizes[i] * num_elements, min_alignment);
if ( ! outptr[i] ){
log_error( " unable to allocate %d bytes for outptr\n", (int)(ptrSizes[i] * num_elements) );
return -1;
}
memset( outptr[i], 0, ptrSizes[i] * num_elements ); // initialize to zero to tell difference
inptr[i] = align_malloc(ptrSizes[i] * num_elements, min_alignment);
if ( ! inptr[i] ){
log_error( " unable to allocate %d bytes for inptr\n", (int)(ptrSizes[i] * num_elements) );
return -1;
}
memset( inptr[i], 0, ptrSizes[i] * num_elements ); // initialize to zero to tell difference
if ((flag_set[src_flag_id] & CL_MEM_USE_HOST_PTR) || (flag_set[src_flag_id] & CL_MEM_COPY_HOST_PTR))
buffers[i] = clCreateBuffer(context, flag_set[src_flag_id], ptrSizes[i] * num_elements, inptr[i], &err);
else
buffers[i] = clCreateBuffer(context, flag_set[src_flag_id], ptrSizes[i] * num_elements, NULL, &err);
if ( err != CL_SUCCESS ){
print_error(err, " clCreateBuffer failed\n" );
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
err = create_single_kernel_helper( context, &program[i], &kernel[i], 1, &kernelCode[i], kernelName[i] );
if ( err ){
log_error( " Error creating program for %s\n", type );
clReleaseMemObject( buffers[i] );
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
err = clSetKernelArg( kernel[i], 0, sizeof( cl_mem ), (void *)&buffers[i] );
if ( err != CL_SUCCESS ){
print_error( err, "clSetKernelArgs failed" );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
err = clEnqueueNDRangeKernel( queue, kernel[i], 1, NULL, global_work_size, NULL, 0, NULL, NULL );
if ( err != CL_SUCCESS ){
print_error( err, "clEnqueueNDRangeKernel failed" );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
lastIndex = ( num_elements * ( 1 << i ) - 1 ) * ptrSizes[0];
err = clEnqueueReadBuffer( queue, buffers[i], false, 0, ptrSizes[i]*num_elements, (void *)(outptr[i]), 0, NULL, &event );
#ifdef CHECK_FOR_NON_WAIT
if ( ((uchar *)outptr[i])[lastIndex] ){
log_error( " clEnqueueReadBuffer() possibly returned only after inappropriately waiting for execution to be finished\n" );
log_error( " Function was run asynchornously, but last value in array was set in code line following clEnqueueReadBuffer()\n" );
}
#endif
if ( err != CL_SUCCESS ){
print_error( err, "clEnqueueReadBuffer failed" );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
err = clEnqueueBarrierWithWaitList(queue, 0, NULL, NULL);
if ( err != CL_SUCCESS ){
print_error( err, "clEnqueueBarrierWithWaitList() failed" );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
return -1;
}
err = clWaitForEvents(1, &event);
if ( err != CL_SUCCESS ){
print_error( err, "clWaitForEvents() failed" );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
align_free( inptr[i] );
return -1;
}
if ( fn(outptr[i], num_elements*(1<<i)) ){
log_error(" %s%d test failed\n", type, 1<<i);
total_errors++;
}
else{
log_info(" %s%d test passed\n", type, 1<<i);
}
// cleanup
clReleaseEvent( event );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
align_free( inptr[i] );
}
} // cl_mem flags
return total_errors;
} // end test_buffer_read_array_barrier()
#define DECLARE_READ_TEST(type, realType) \
int test_buffer_read_##type( cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements ) \
{ \
return test_buffer_read( deviceID, context, queue, num_elements, sizeof( realType ), (char*)#type, 5, \
buffer_read_##type##_kernel_code, type##_kernel_name, verify_read_##type ); \
}
DECLARE_READ_TEST(int, cl_int)
DECLARE_READ_TEST(uint, cl_uint)
DECLARE_READ_TEST(long, cl_long)
DECLARE_READ_TEST(ulong, cl_ulong)
DECLARE_READ_TEST(short, cl_short)
DECLARE_READ_TEST(ushort, cl_ushort)
DECLARE_READ_TEST(float, cl_float)
DECLARE_READ_TEST(char, cl_char)
DECLARE_READ_TEST(uchar, cl_uchar)
int test_buffer_read_half(cl_device_id deviceID, cl_context context,
cl_command_queue queue, int num_elements)
{
PASSIVE_REQUIRE_FP16_SUPPORT(deviceID)
return test_buffer_read( deviceID, context, queue, num_elements, sizeof( cl_float ) / 2, (char*)"half", 5,
buffer_read_half_kernel_code, half_kernel_name, verify_read_half );
}
#define DECLARE_ASYNC_TEST(type, realType) \
int test_buffer_read_async_##type( cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements ) \
{ \
return test_buffer_read_async( deviceID, context, queue, num_elements, sizeof( realType ), (char*)#type, 5, \
buffer_read_##type##_kernel_code, type##_kernel_name, verify_read_##type ); \
}
DECLARE_ASYNC_TEST(char, cl_char)
DECLARE_ASYNC_TEST(uchar, cl_uchar)
DECLARE_ASYNC_TEST(short, cl_short)
DECLARE_ASYNC_TEST(ushort, cl_ushort)
DECLARE_ASYNC_TEST(int, cl_int)
DECLARE_ASYNC_TEST(uint, cl_uint)
DECLARE_ASYNC_TEST(long, cl_long)
DECLARE_ASYNC_TEST(ulong, cl_ulong)
DECLARE_ASYNC_TEST(float, cl_float)
#define DECLARE_BARRIER_TEST(type, realType) \
int test_buffer_read_array_barrier_##type( cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements ) \
{ \
return test_buffer_read_array_barrier( deviceID, context, queue, num_elements, sizeof( realType ), (char*)#type, 5, \
buffer_read_##type##_kernel_code, type##_kernel_name, verify_read_##type ); \
}
DECLARE_BARRIER_TEST(int, cl_int)
DECLARE_BARRIER_TEST(uint, cl_uint)
DECLARE_BARRIER_TEST(long, cl_long)
DECLARE_BARRIER_TEST(ulong, cl_ulong)
DECLARE_BARRIER_TEST(short, cl_short)
DECLARE_BARRIER_TEST(ushort, cl_ushort)
DECLARE_BARRIER_TEST(char, cl_char)
DECLARE_BARRIER_TEST(uchar, cl_uchar)
DECLARE_BARRIER_TEST(float, cl_float)
int test_buffer_read_struct(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
cl_mem buffers[1];
TestStruct *output_ptr;
cl_program program[1];
cl_kernel kernel[1];
size_t global_work_size[3];
cl_int err;
size_t objSize = sizeof(TestStruct);
size_t min_alignment = get_min_alignment(context);
global_work_size[0] = (cl_uint)num_elements;
output_ptr = (TestStruct*)align_malloc(objSize * num_elements, min_alignment);
if ( ! output_ptr ){
log_error( " unable to allocate %d bytes for output_ptr\n", (int)(objSize * num_elements) );
return -1;
}
buffers[0] = clCreateBuffer(context, CL_MEM_READ_WRITE,
objSize * num_elements, NULL, &err);
if ( err != CL_SUCCESS ){
print_error( err, " clCreateBuffer failed\n" );
align_free( output_ptr );
return -1;
}
err = create_single_kernel_helper( context, &program[0], &kernel[0], 1, &buffer_read_struct_kernel_code, "test_buffer_read_struct" );
if ( err ){
clReleaseProgram( program[0] );
align_free( output_ptr );
return -1;
}
err = clSetKernelArg( kernel[0], 0, sizeof( cl_mem ), (void *)&buffers[0] );
if ( err != CL_SUCCESS){
print_error( err, "clSetKernelArg failed" );
clReleaseMemObject( buffers[0] );
clReleaseKernel( kernel[0] );
clReleaseProgram( program[0] );
align_free( output_ptr );
return -1;
}
err = clEnqueueNDRangeKernel( queue, kernel[0], 1, NULL, global_work_size, NULL, 0, NULL, NULL );
if ( err != CL_SUCCESS ){
print_error( err, "clEnqueueNDRangeKernel failed" );
clReleaseMemObject( buffers[0] );
clReleaseKernel( kernel[0] );
clReleaseProgram( program[0] );
align_free( output_ptr );
return -1;
}
err = clEnqueueReadBuffer( queue, buffers[0], true, 0, objSize*num_elements, (void *)output_ptr, 0, NULL, NULL );
if ( err != CL_SUCCESS){
print_error( err, "clEnqueueReadBuffer failed" );
clReleaseMemObject( buffers[0] );
clReleaseKernel( kernel[0] );
clReleaseProgram( program[0] );
align_free( output_ptr );
return -1;
}
if (verify_read_struct(output_ptr, num_elements)){
log_error(" struct test failed\n");
err = -1;
}
else{
log_info(" struct test passed\n");
err = 0;
}
// cleanup
clReleaseMemObject( buffers[0] );
clReleaseKernel( kernel[0] );
clReleaseProgram( program[0] );
align_free( output_ptr );
return err;
}
static int testRandomReadSize( cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements, cl_uint startOfRead, size_t sizeOfRead )
{
cl_mem buffers[3];
int *outptr[3];
cl_program program[3];
cl_kernel kernel[3];
size_t global_work_size[3];
cl_int err;
int i, j;
size_t ptrSizes[3]; // sizeof(int), sizeof(int2), sizeof(int4)
int total_errors = 0;
size_t min_alignment = get_min_alignment(context);
global_work_size[0] = (cl_uint)num_elements;
ptrSizes[0] = sizeof(cl_int);
ptrSizes[1] = ptrSizes[0] << 1;
ptrSizes[2] = ptrSizes[1] << 1;
for ( i = 0; i < 3; i++ ){
outptr[i] = (int *)align_malloc( ptrSizes[i] * num_elements, min_alignment);
if ( ! outptr[i] ){
log_error( " Unable to allocate %d bytes for outptr[%d]\n", (int)(ptrSizes[i] * num_elements), i );
for ( j = 0; j < i; j++ ){
clReleaseMemObject( buffers[j] );
align_free( outptr[j] );
}
return -1;
}
buffers[i] = clCreateBuffer(context, CL_MEM_READ_WRITE,
ptrSizes[i] * num_elements, NULL, &err);
if ( err != CL_SUCCESS ){
print_error(err, " clCreateBuffer failed\n" );
for ( j = 0; j < i; j++ ){
clReleaseMemObject( buffers[j] );
align_free( outptr[j] );
}
align_free( outptr[i] );
return -1;
}
}
err = create_single_kernel_helper( context, &program[0], &kernel[0], 1, &buffer_read_int_kernel_code[0], "test_buffer_read_int" );
if ( err ){
log_error( " Error creating program for int\n" );
for ( i = 0; i < 3; i++ ){
clReleaseMemObject( buffers[i] );
align_free( outptr[i] );
}
return -1;
}
err = create_single_kernel_helper( context, &program[1], &kernel[1], 1, &buffer_read_int_kernel_code[1], "test_buffer_read_int2" );
if ( err ){
log_error( " Error creating program for int2\n" );
clReleaseKernel( kernel[0] );
clReleaseProgram( program[0] );
for ( i = 0; i < 3; i++ ){
clReleaseMemObject( buffers[i] );
align_free( outptr[i] );
}
return -1;
}
err = create_single_kernel_helper( context, &program[2], &kernel[2], 1, &buffer_read_int_kernel_code[2], "test_buffer_read_int4" );
if ( err ){
log_error( " Error creating program for int4\n" );
clReleaseKernel( kernel[0] );
clReleaseProgram( program[0] );
clReleaseKernel( kernel[1] );
clReleaseProgram( program[1] );
for ( i = 0; i < 3; i++ ){
clReleaseMemObject( buffers[i] );
align_free( outptr[i] );
}
return -1;
}
for (i=0; i<3; i++){
err = clSetKernelArg( kernel[i], 0, sizeof( cl_mem ), (void *)&buffers[i] );
if ( err != CL_SUCCESS ){
print_error( err, "clSetKernelArgs failed" );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
return -1;
}
err = clEnqueueNDRangeKernel( queue, kernel[i], 1, NULL, global_work_size, NULL, 0, NULL, NULL );
if ( err != CL_SUCCESS ){
print_error( err, "clEnqueueNDRangeKernel failed" );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
return -1;
}
err = clEnqueueReadBuffer( queue, buffers[i], true, startOfRead*ptrSizes[i], ptrSizes[i]*sizeOfRead, (void *)(outptr[i]), 0, NULL, NULL );
if ( err != CL_SUCCESS ){
print_error( err, "clEnqueueReadBuffer failed" );
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
return -1;
}
if ( verify_read_int( outptr[i], (int)sizeOfRead*(1<<i) ) ){
log_error(" random size from %d, size: %d test failed on i%d\n", (int)startOfRead, (int)sizeOfRead, 1<<i);
total_errors++;
}
else{
log_info(" random size from %d, size: %d test passed on i%d\n", (int)startOfRead, (int)sizeOfRead, 1<<i);
}
// cleanup
clReleaseMemObject( buffers[i] );
clReleaseKernel( kernel[i] );
clReleaseProgram( program[i] );
align_free( outptr[i] );
}
return total_errors;
} // end testRandomReadSize()
int test_buffer_read_random_size(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
int err = 0;
int i;
cl_uint start;
size_t size;
MTdata d = init_genrand( gRandomSeed );
// now test for random sizes of array being read
for ( i = 0; i < 8; i++ ){
start = (cl_uint)get_random_float( 0.f, (float)(num_elements - 8), d );
size = (size_t)get_random_float( 8.f, (float)(num_elements - start), d );
if (testRandomReadSize( deviceID, context, queue, num_elements, start, size ))
err++;
}
free_mtdata(d);
return err;
}