blob: 2f392f5499f4203f3281ab7a108bb1ccbab064ea [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 "testBase.h"
#include "harness/conversions.h"
#include "harness/typeWrappers.h"
#include "harness/testHarness.h"
#include "structs.h"
#include "defines.h"
#include "type_replacer.h"
size_t get_align(size_t vecSize)
{
if (vecSize == 3)
{
return 4;
}
return vecSize;
}
/* // Lots of conditionals means this is not gonna be an optimal min on intel.
*/
/* // That's okay, make sure we only call a few times per test, not for every */
/* // element */
/* size_t min_of_nonzero(size_t a, size_t b) */
/* { */
/* if(a != 0 && (a<=b || b==0)) */
/* { */
/* return a; */
/* } */
/* if(b != 0 && (b<a || a==0)) */
/* { */
/* return b; */
/* } */
/* return 0; */
/* } */
/* size_t get_min_packed_alignment(size_t preSize, size_t typeMultiplePreSize,
*/
/* size_t postSize, size_t typeMultiplePostSize, */
/* ExplicitType kType, size_t vecSize) */
/* { */
/* size_t pre_min = min_of_nonzero(preSize, */
/* typeMultiplePreSize* */
/* get_explicit_type_size(kType)); */
/* size_t post_min = min_of_nonzero(postSize, */
/* typeMultiplePostSize* */
/* get_explicit_type_size(kType)); */
/* size_t struct_min = min_of_nonzero(pre_min, post_min); */
/* size_t result = min_of_nonzero(struct_min, get_align(vecSize) */
/* *get_explicit_type_size(kType)); */
/* return result; */
/* } */
int test_vec_internal(cl_device_id deviceID, cl_context context,
cl_command_queue queue, const char* pattern,
const char* testName, size_t bufSize, size_t preSize,
size_t typeMultiplePreSize, size_t postSize,
size_t typeMultiplePostSize)
{
int err;
int typeIdx, vecSizeIdx;
char tmpBuffer[2048];
char srcBuffer[2048];
size_t preSizeBytes, postSizeBytes, typeSize, totSize;
clState* pClState = newClState(deviceID, context, queue);
bufferStruct* pBuffers = newBufferStruct(
bufSize, bufSize * sizeof(cl_uint) / sizeof(cl_char), pClState);
if (pBuffers == NULL)
{
destroyClState(pClState);
vlog_error("%s : Could not create buffer\n", testName);
return -1;
}
for (typeIdx = 0; types[typeIdx] != kNumExplicitTypes; ++typeIdx)
{
// Skip doubles if it is not supported otherwise enable pragma
if (types[typeIdx] == kDouble)
{
if (!is_extension_available(deviceID, "cl_khr_fp64"))
{
continue;
}
else
{
doReplace(tmpBuffer, 2048, pattern, ".PRAGMA.",
"#pragma OPENCL EXTENSION cl_khr_fp64: ", ".STATE.",
"enable");
}
}
else
{
if (types[typeIdx] == kLong || types[typeIdx] == kULong)
{
if (gIsEmbedded) continue;
}
doReplace(tmpBuffer, 2048, pattern, ".PRAGMA.", " ", ".STATE.",
" ");
}
typeSize = get_explicit_type_size(types[typeIdx]);
preSizeBytes = preSize + typeSize * typeMultiplePreSize;
postSizeBytes = postSize + typeSize * typeMultiplePostSize;
for (vecSizeIdx = 1; vecSizeIdx < NUM_VECTOR_SIZES; ++vecSizeIdx)
{
totSize = preSizeBytes + postSizeBytes
+ typeSize * get_align(g_arrVecSizes[vecSizeIdx]);
doReplace(srcBuffer, 2048, tmpBuffer, ".TYPE.",
g_arrTypeNames[typeIdx], ".NUM.",
g_arrVecSizeNames[vecSizeIdx]);
if (srcBuffer[0] == '\0')
{
vlog_error("%s: failed to fill source buf for type %s%s\n",
testName, g_arrTypeNames[typeIdx],
g_arrVecSizeNames[vecSizeIdx]);
destroyBufferStruct(pBuffers, pClState);
destroyClState(pClState);
return -1;
}
// log_info("Buffer is \"\n%s\n\"\n", srcBuffer);
// fflush(stdout);
err = clStateMakeProgram(pClState, srcBuffer, testName);
if (err)
{
vlog_error("%s: Error compiling \"\n%s\n\"", testName,
srcBuffer);
destroyBufferStruct(pBuffers, pClState);
destroyClState(pClState);
return -1;
}
err = pushArgs(pBuffers, pClState);
if (err != 0)
{
vlog_error("%s: failed to push args %s%s\n", testName,
g_arrTypeNames[typeIdx],
g_arrVecSizeNames[vecSizeIdx]);
destroyBufferStruct(pBuffers, pClState);
destroyClState(pClState);
return -1;
}
// log_info("About to Run kernel\n"); fflush(stdout);
// now we run the kernel
err = runKernel(
pClState,
bufSize
/ (g_arrVecSizes[vecSizeIdx] * g_arrTypeSizes[typeIdx]));
if (err != 0)
{
vlog_error("%s: runKernel fail (%ld threads) %s%s\n", testName,
pClState->m_numThreads, g_arrTypeNames[typeIdx],
g_arrVecSizeNames[vecSizeIdx]);
destroyBufferStruct(pBuffers, pClState);
destroyClState(pClState);
return -1;
}
// log_info("About to retrieve results\n"); fflush(stdout);
err = retrieveResults(pBuffers, pClState);
if (err != 0)
{
vlog_error("%s: failed to retrieve results %s%s\n", testName,
g_arrTypeNames[typeIdx],
g_arrVecSizeNames[vecSizeIdx]);
destroyBufferStruct(pBuffers, pClState);
destroyClState(pClState);
return -1;
}
if (preSizeBytes + postSizeBytes == 0)
{
// log_info("About to Check Correctness\n"); fflush(stdout);
err = checkCorrectnessAlign(pBuffers, pClState,
get_align(g_arrVecSizes[vecSizeIdx])
* typeSize);
}
else
{
// we're checking for an aligned struct
err = checkPackedCorrectness(pBuffers, pClState, totSize,
preSizeBytes);
}
if (err != 0)
{
vlog_error("%s: incorrect results %s%s\n", testName,
g_arrTypeNames[typeIdx],
g_arrVecSizeNames[vecSizeIdx]);
vlog_error("%s: Source was \"\n%s\n\"", testName, srcBuffer);
destroyBufferStruct(pBuffers, pClState);
destroyClState(pClState);
return -1;
}
clStateDestroyProgramAndKernel(pClState);
}
}
destroyBufferStruct(pBuffers, pClState);
destroyClState(pClState);
// vlog_error("%s : implementation incomplete : FAIL\n", testName);
return 0; // -1; // fails on account of not being written.
}
static const char* patterns[] = {
".PRAGMA..STATE.\n"
"__kernel void test_vec_align_array(.SRC_SCOPE. .TYPE..NUM. *source, "
".DST_SCOPE. uint *dest)\n"
"{\n"
" int tid = get_global_id(0);\n"
" dest[tid] = (uint)((.SRC_SCOPE. uchar *)(source+tid));\n"
"}\n",
".PRAGMA..STATE.\n"
"typedef struct myUnpackedStruct { \n"
".PRE."
" .TYPE..NUM. vec;\n"
".POST."
"} testStruct;\n"
"__kernel void test_vec_align_struct(__constant .TYPE..NUM. *source, "
".DST_SCOPE. uint *dest)\n"
"{\n"
" .SRC_SCOPE. testStruct test;\n"
" int tid = get_global_id(0);\n"
" dest[tid] = (uint)((.SRC_SCOPE. uchar *)&(test.vec));\n"
"}\n",
".PRAGMA..STATE.\n"
"typedef struct __attribute__ ((packed)) myPackedStruct { \n"
".PRE."
" .TYPE..NUM. vec;\n"
".POST."
"} testStruct;\n"
"__kernel void test_vec_align_packed_struct(__constant .TYPE..NUM. "
"*source, .DST_SCOPE. uint *dest)\n"
"{\n"
" .SRC_SCOPE. testStruct test;\n"
" int tid = get_global_id(0);\n"
" dest[tid] = (uint)((.SRC_SCOPE. uchar *)&(test.vec) - (.SRC_SCOPE. "
"uchar *)&test);\n"
"}\n",
".PRAGMA..STATE.\n"
"typedef struct myStruct { \n"
".PRE."
" .TYPE..NUM. vec;\n"
".POST."
"} testStruct;\n"
"__kernel void test_vec_align_struct_arr(.SRC_SCOPE. testStruct *source, "
".DST_SCOPE. uint *dest)\n"
"{\n"
" int tid = get_global_id(0);\n"
" dest[tid] = (uint)((.SRC_SCOPE. uchar *)&(source[tid].vec));\n"
"}\n",
".PRAGMA..STATE.\n"
"typedef struct __attribute__ ((packed)) myPackedStruct { \n"
".PRE."
" .TYPE..NUM. vec;\n"
".POST."
"} testStruct;\n"
"__kernel void test_vec_align_packed_struct_arr(.SRC_SCOPE. testStruct "
"*source, .DST_SCOPE. uint *dest)\n"
"{\n"
" int tid = get_global_id(0);\n"
" dest[tid] = (uint)((.SRC_SCOPE. uchar *)&(source[tid].vec) - "
"(.SRC_SCOPE. uchar *)&(source[0]));\n"
"}\n",
// __attribute__ ((packed))
};
const char* pre_substitution_arr[] = { "",
"char c;\n",
"short3 s;",
".TYPE.3 tPre;\n",
".TYPE. arrPre[5];\n",
".TYPE. arrPre[12];\n",
NULL };
// alignments of everything in pre_substitution_arr as raw alignments
// 0 if such a thing is meaningless
size_t pre_align_arr[] = { 0,
sizeof(cl_char),
4 * sizeof(cl_short),
0, // taken care of in type_multiple_pre_align_arr
0,
0 };
// alignments of everything in pre_substitution_arr as multiples of
// sizeof(.TYPE.)
// 0 if such a thing is meaningless
size_t type_multiple_pre_align_arr[] = { 0, 0, 0, 4, 5, 12 };
const char* post_substitution_arr[] = { "",
"char cPost;\n",
".TYPE. arrPost[3];\n",
".TYPE. arrPost[5];\n",
".TYPE.3 arrPost;\n",
".TYPE. arrPost[12];\n",
NULL };
// alignments of everything in post_substitution_arr as raw alignments
// 0 if such a thing is meaningless
size_t post_align_arr[] = { 0, sizeof(cl_char),
0, // taken care of in type_multiple_post_align_arr
0, 0,
0 };
// alignments of everything in post_substitution_arr as multiples of
// sizeof(.TYPE.)
// 0 if such a thing is meaningless
size_t type_multiple_post_align_arr[] = { 0, 0, 3, 5, 4, 12 };
// there hsould be a packed version of this?
int test_vec_align_array(cl_device_id deviceID, cl_context context,
cl_command_queue queue, int num_elements)
{
char tmp[2048];
int result;
log_info("Testing global\n");
doReplace(tmp, (size_t)2048, patterns[0], ".SRC_SCOPE.", "__global",
".DST_SCOPE.", "__global"); //
result = test_vec_internal(deviceID, context, queue, tmp,
"test_vec_align_array", BUFFER_SIZE, 0, 0, 0, 0);
return result;
}
int test_vec_align_struct(cl_device_id deviceID, cl_context context,
cl_command_queue queue, int num_elements)
{
char tmp1[2048], tmp2[2048];
int result = 0;
int preIdx, postIdx;
log_info("testing __private\n");
doReplace(tmp2, (size_t)2048, patterns[1], ".SRC_SCOPE.", "__private",
".DST_SCOPE.", "__global"); //
for (preIdx = 0; pre_substitution_arr[preIdx] != NULL; ++preIdx)
{
for (postIdx = 0; post_substitution_arr[postIdx] != NULL; ++postIdx)
{
doReplace(tmp1, (size_t)2048, tmp2, ".PRE.",
pre_substitution_arr[preIdx], ".POST.",
post_substitution_arr[postIdx]);
result =
test_vec_internal(deviceID, context, queue, tmp1,
"test_vec_align_struct", 512, 0, 0, 0, 0);
if (result != 0)
{
return result;
}
}
}
log_info("testing __local\n");
doReplace(tmp2, (size_t)2048, patterns[1], ".SRC_SCOPE.", "__local",
".DST_SCOPE.", "__global"); //
for (preIdx = 0; pre_substitution_arr[preIdx] != NULL; ++preIdx)
{
for (postIdx = 0; post_substitution_arr[postIdx] != NULL; ++postIdx)
{
doReplace(tmp1, (size_t)2048, tmp2, ".PRE.",
pre_substitution_arr[preIdx], ".POST.",
post_substitution_arr[postIdx]);
result =
test_vec_internal(deviceID, context, queue, tmp1,
"test_vec_align_struct", 512, 0, 0, 0, 0);
if (result != 0)
{
return result;
}
}
}
return 0;
}
int test_vec_align_packed_struct(cl_device_id deviceID, cl_context context,
cl_command_queue queue, int num_elements)
{
char tmp1[2048], tmp2[2048];
int result = 0;
int preIdx, postIdx;
log_info("Testing __private\n");
doReplace(tmp2, (size_t)2048, patterns[2], ".SRC_SCOPE.", "__private",
".DST_SCOPE.", "__global"); //
for (preIdx = 0; pre_substitution_arr[preIdx] != NULL; ++preIdx)
{
for (postIdx = 0; post_substitution_arr[postIdx] != NULL; ++postIdx)
{
doReplace(tmp1, (size_t)2048, tmp2, ".PRE.",
pre_substitution_arr[preIdx], ".POST.",
post_substitution_arr[postIdx]);
result = test_vec_internal(
deviceID, context, queue, tmp1, "test_vec_align_packed_struct",
512, pre_align_arr[preIdx], type_multiple_pre_align_arr[preIdx],
post_align_arr[postIdx], type_multiple_post_align_arr[postIdx]);
if (result != 0)
{
return result;
}
}
}
log_info("testing __local\n");
doReplace(tmp2, (size_t)2048, patterns[2], ".SRC_SCOPE.", "__local",
".DST_SCOPE.", "__global"); //
for (preIdx = 0; pre_substitution_arr[preIdx] != NULL; ++preIdx)
{
for (postIdx = 0; post_substitution_arr[postIdx] != NULL; ++postIdx)
{
doReplace(tmp1, (size_t)2048, tmp2, ".PRE.",
pre_substitution_arr[preIdx], ".POST.",
post_substitution_arr[postIdx]);
result = test_vec_internal(
deviceID, context, queue, tmp1, "test_vec_align_packed_struct",
512, pre_align_arr[preIdx], type_multiple_pre_align_arr[preIdx],
post_align_arr[postIdx], type_multiple_post_align_arr[postIdx]);
if (result != 0)
{
return result;
}
}
}
return 0;
}
int test_vec_align_struct_arr(cl_device_id deviceID, cl_context context,
cl_command_queue queue, int num_elements)
{
char tmp1[2048], tmp2[2048];
int result = 0;
int preIdx, postIdx;
log_info("testing __global\n");
doReplace(tmp2, (size_t)2048, patterns[3], ".SRC_SCOPE.", "__global",
".DST_SCOPE.", "__global"); //
for (preIdx = 0; pre_substitution_arr[preIdx] != NULL; ++preIdx)
{
for (postIdx = 0; post_substitution_arr[postIdx] != NULL; ++postIdx)
{
doReplace(tmp1, (size_t)2048, tmp2, ".PRE.",
pre_substitution_arr[preIdx], ".POST.",
post_substitution_arr[postIdx]);
result = test_vec_internal(deviceID, context, queue, tmp1,
"test_vec_align_struct_arr", BUFFER_SIZE,
0, 0, 0, 0);
if (result != 0)
{
return result;
}
}
}
return 0;
}
int test_vec_align_packed_struct_arr(cl_device_id deviceID, cl_context context,
cl_command_queue queue, int num_elements)
{
char tmp1[2048], tmp2[2048];
int result = 0;
int preIdx, postIdx;
log_info("Testing __global\n");
doReplace(tmp2, (size_t)2048, patterns[4], ".SRC_SCOPE.", "__global",
".DST_SCOPE.", "__global"); //
for (preIdx = 0; pre_substitution_arr[preIdx] != NULL; ++preIdx)
{
for (postIdx = 0; post_substitution_arr[postIdx] != NULL; ++postIdx)
{
doReplace(tmp1, (size_t)2048, tmp2, ".PRE.",
pre_substitution_arr[preIdx], ".POST.",
post_substitution_arr[postIdx]);
result = test_vec_internal(
deviceID, context, queue, tmp1,
"test_vec_align_packed_struct_arr", BUFFER_SIZE,
pre_align_arr[preIdx], type_multiple_pre_align_arr[preIdx],
post_align_arr[postIdx], type_multiple_post_align_arr[postIdx]);
if (result != 0) return result;
}
}
return 0;
}