Merge branch 'main' into negative_command_copy_image

test_common/harness/imageHelpers.cpp

@@ -769,10 +769,14 @@ void get_max_sizes(
 
  (*numberOfSizes) = 0;
 
- if (image_type == CL_MEM_OBJECT_IMAGE1D)
+ if (image_type == CL_MEM_OBJECT_IMAGE1D
+ || image_type == CL_MEM_OBJECT_IMAGE1D_BUFFER)
  {
 
  size_t M = maximum_sizes[0];
+ size_t A = max_pixels;
+
+ M = static_cast<size_t>(fmax(1, fmin(A / M, M)));
 
  // Store the size
  sizes[(*numberOfSizes)][0] = M;
@@ -860,6 +864,7 @@ void get_max_sizes(
  {
  switch (image_type)
  {
+ case CL_MEM_OBJECT_IMAGE1D_BUFFER:
  case CL_MEM_OBJECT_IMAGE1D:
  log_info(" size[%d] = [%zu] (%g MB image)\n", j, sizes[j][0],
  raw_pixel_size * sizes[j][0] * sizes[j][1]
@@ -1080,6 +1085,7 @@ cl_ulong get_image_size(image_descriptor const *imageInfo)
  {
  switch (imageInfo->type)
  {
+ case CL_MEM_OBJECT_IMAGE1D_BUFFER:
  case CL_MEM_OBJECT_IMAGE1D: imageSize = imageInfo->rowPitch; break;
  case CL_MEM_OBJECT_IMAGE2D:
  imageSize = imageInfo->height * imageInfo->rowPitch;
@@ -2317,6 +2323,7 @@ int debug_find_vector_in_image(void *imagePtr, image_descriptor *imageInfo,
 
  switch (imageInfo->type)
  {
+ case CL_MEM_OBJECT_IMAGE1D_BUFFER:
  case CL_MEM_OBJECT_IMAGE1D:
  width = (imageInfo->width >> lod) ? (imageInfo->width >> lod) : 1;
  height = 1;
@@ -3513,6 +3520,7 @@ void copy_image_data(image_descriptor *srcImageInfo,
 
  switch (srcImageInfo->type)
  {
+ case CL_MEM_OBJECT_IMAGE1D_BUFFER:
  case CL_MEM_OBJECT_IMAGE1D:
  src_lod = sourcePos[1];
  sourcePos_lod[1] = sourcePos_lod[2] = 0;
@@ -3558,6 +3566,7 @@ void copy_image_data(image_descriptor *srcImageInfo,
  size_t dst_height_lod = 1 /*dstImageInfo->height*/;
  switch (dstImageInfo->type)
  {
+ case CL_MEM_OBJECT_IMAGE1D_BUFFER:
  case CL_MEM_OBJECT_IMAGE1D:
  dst_lod = destPos[1];
  destPos_lod[1] = destPos_lod[2] = 0;
@@ -4022,6 +4031,7 @@ cl_ulong compute_mipmapped_image_size(image_descriptor imageInfo)
  retSize += (cl_ulong)curr_width * curr_height
  * get_pixel_size(imageInfo.format);
  break;
+ case CL_MEM_OBJECT_IMAGE1D_BUFFER:
  case CL_MEM_OBJECT_IMAGE1D:
  retSize +=
  (cl_ulong)curr_width * get_pixel_size(imageInfo.format);
@@ -4043,6 +4053,7 @@ cl_ulong compute_mipmapped_image_size(image_descriptor imageInfo)
  case CL_MEM_OBJECT_IMAGE2D:
  case CL_MEM_OBJECT_IMAGE2D_ARRAY:
  curr_height = curr_height >> 1 ? curr_height >> 1 : 1;
+ case CL_MEM_OBJECT_IMAGE1D_BUFFER:
  case CL_MEM_OBJECT_IMAGE1D:
  case CL_MEM_OBJECT_IMAGE1D_ARRAY:
  curr_width = curr_width >> 1 ? curr_width >> 1 : 1;
@@ -4080,6 +4091,7 @@ size_t compute_mip_level_offset(image_descriptor *imageInfo, size_t lod)
  retOffset +=
  (size_t)width * height * get_pixel_size(imageInfo->format);
  break;
+ case CL_MEM_OBJECT_IMAGE1D_BUFFER:
  case CL_MEM_OBJECT_IMAGE1D:
  retOffset += (size_t)width * get_pixel_size(imageInfo->format);
  break;
@@ -4092,6 +4104,7 @@ size_t compute_mip_level_offset(image_descriptor *imageInfo, size_t lod)
  case CL_MEM_OBJECT_IMAGE2D:
  case CL_MEM_OBJECT_IMAGE2D_ARRAY:
  height = (height >> 1) ? (height >> 1) : 1;
+ case CL_MEM_OBJECT_IMAGE1D_BUFFER:
  case CL_MEM_OBJECT_IMAGE1D_ARRAY:
  case CL_MEM_OBJECT_IMAGE1D: width = (width >> 1) ? (width >> 1) : 1;
  }

test_conformance/api/CMakeLists.txt

@@ -35,6 +35,7 @@ set(${MODULE_NAME}_SOURCES
  test_mem_object_properties_queries.cpp
  test_queue_properties_queries.cpp
  test_pipe_properties_queries.cpp
+ test_wg_suggested_local_work_size.cpp
 )
 
 include(../CMakeCommon.txt)

test_conformance/api/main.cpp

@@ -148,6 +148,10 @@ test_definition test_list[] = {
  ADD_TEST(min_image_formats),
  ADD_TEST(negative_get_platform_info),
  ADD_TEST(negative_get_platform_ids),
+
+ ADD_TEST(work_group_suggested_local_size_1D),
+ ADD_TEST(work_group_suggested_local_size_2D),
+ ADD_TEST(work_group_suggested_local_size_3D),
 };
 
 const int test_num = ARRAY_SIZE(test_list);

test_conformance/api/procs.h

@@ -1,6 +1,6 @@
 //
 // 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
@@ -204,3 +204,16 @@ extern int test_negative_get_platform_ids(cl_device_id deviceID,
  int num_elements);
 extern int test_kernel_attributes(cl_device_id deviceID, cl_context context,
  cl_command_queue queue, int num_elements);
+
+extern int test_work_group_suggested_local_size_1D(cl_device_id device,
+ cl_context context,
+ cl_command_queue queue,
+ int n_elems);
+extern int test_work_group_suggested_local_size_2D(cl_device_id device,
+ cl_context context,
+ cl_command_queue queue,
+ int n_elems);
+extern int test_work_group_suggested_local_size_3D(cl_device_id device,
+ cl_context context,
+ cl_command_queue queue,
+ int n_elems);

test_conformance/workgroups/test_wg_suggested_local_work_size.cpp → test_conformance/api/test_wg_suggested_local_work_size.cpp

@@ -72,7 +72,7 @@ const char* wg_scan_local_work_group_size = R"(
  global uint *output)
  {
  __local char c[LOCAL_MEM_SIZE];
- 
+
  if(!is_zero_linear_id()) return;
  for (uint i = 0; i < 3; i++)
  {

test_conformance/common/vulkan_wrapper/opencl_vulkan_wrapper.cpp

@@ -853,7 +853,7 @@ clExternalImportableSemaphore::~clExternalImportableSemaphore()
  cl_int err = clReleaseSemaphoreKHRptr(m_externalSemaphore);
  if (err != CL_SUCCESS)
  {
- throw std::runtime_error("clReleaseSemaphoreKHR failed!");
+ log_error("clReleaseSemaphoreKHR failed with %d\n", err);
  }
 }
 
@@ -935,7 +935,7 @@ clExternalExportableSemaphore::~clExternalExportableSemaphore()
  cl_int err = clReleaseSemaphoreKHRptr(m_externalSemaphore);
  if (err != CL_SUCCESS)
  {
- throw std::runtime_error("clReleaseSemaphoreKHR failed!");
+ log_error("clReleaseSemaphoreKHR failed with %d\n", err);
  }
 }
 
@@ -1052,4 +1052,4 @@ VulkanImageTiling vkClExternalMemoryHandleTilingAssumption(
  }
 
  return mode;
-}
+}

test_conformance/conversions/CMakeLists.txt

@@ -8,14 +8,6 @@ if("${CLConform_TARGET_ARCH}" STREQUAL "ARM" OR "${CLConform_TARGET_ARCH}" STREQ
  list(APPEND ${MODULE_NAME}_SOURCES fplib.cpp)
 endif()
 
-if(NOT CMAKE_CL_64 AND NOT MSVC AND NOT ANDROID)
-# -march is needed for CPU atomics, default arch on gcc is i386
-# that does not support atomics.
-set_source_files_properties(
- ${MODULE_NAME}_SOURCES
- COMPILE_FLAGS -march=i686)
-endif(NOT CMAKE_CL_64 AND NOT MSVC AND NOT ANDROID)
-
 set_gnulike_module_compile_flags("-Wno-sign-compare")
 
 include(../CMakeCommon.txt)

test_conformance/device_execution/enqueue_block.cpp

@@ -641,7 +641,9 @@ int test_enqueue_block(cl_device_id device, cl_context context, cl_command_queue
  if (!gKernelName.empty() && gKernelName != sources_enqueue_block[i].kernel_name)
  continue;
 
- log_info("Running '%s' kernel (%d of %d) ...\n", sources_enqueue_block[i].kernel_name, i + 1, num_kernels_enqueue_block);
+ log_info("Running '%s' kernel (%d of %zu) ...\n",
+ sources_enqueue_block[i].kernel_name, i + 1,
+ num_kernels_enqueue_block);
  err_ret = run_n_kernel_args(context, queue, sources_enqueue_block[i].lines, sources_enqueue_block[i].num_lines, sources_enqueue_block[i].kernel_name, local_size, global_size, kernel_results, sizeof(kernel_results), 0, NULL);
  if(check_error(err_ret, "'%s' kernel execution failed", sources_enqueue_block[i].kernel_name)) { ++failCnt; res = -1; }
  else if((n = check_kernel_results(kernel_results, arr_size(kernel_results))) >= 0 && check_error(-1, "'%s' kernel results validation failed: [%d] returned %d expected 0", sources_enqueue_block[i].kernel_name, n, kernel_results[n])) res = -1;
@@ -650,7 +652,8 @@ int test_enqueue_block(cl_device_id device, cl_context context, cl_command_queue
 
  if (failCnt > 0)
  {
- log_error("ERROR: %d of %d kernels failed.\n", failCnt, num_kernels_enqueue_block);
+ log_error("ERROR: %zu of %zu kernels failed.\n", failCnt,
+ num_kernels_enqueue_block);
  }
 
  return res;

test_conformance/device_execution/enqueue_flags.cpp

@@ -714,7 +714,9 @@ int test_enqueue_flags(cl_device_id device, cl_context context, cl_command_queue
  if (!gKernelName.empty() && gKernelName != sources_enqueue_block_flags[i].kernel_name)
  continue;
 
- log_info("Running '%s' kernel (%d of %d) ...\n", sources_enqueue_block_flags[i].kernel_name, i + 1, num_enqueue_block_flags);
+ log_info("Running '%s' kernel (%d of %zu) ...\n",
+ sources_enqueue_block_flags[i].kernel_name, i + 1,
+ num_enqueue_block_flags);
 
  clMemWrapper mem = clCreateBuffer(context, CL_MEM_READ_WRITE|CL_MEM_COPY_HOST_PTR, global_size * BITS_DEPTH * sizeof(cl_int), buff, &err_ret);
  test_error(err_ret, "clCreateBuffer() failed");
@@ -749,7 +751,8 @@ int test_enqueue_flags(cl_device_id device, cl_context context, cl_command_queue
 
  if (failCnt > 0)
  {
- log_error("ERROR: %d of %d kernels failed.\n", failCnt, num_enqueue_block_flags);
+ log_error("ERROR: %zu of %zu kernels failed.\n", failCnt,
+ num_enqueue_block_flags);
  }
 
  return res;

test_conformance/device_execution/enqueue_ndrange.cpp

@@ -700,7 +700,9 @@ int test_enqueue_ndrange(cl_device_id device, cl_context context, cl_command_que
  { sizeof(cl_mem), &mem4 },
  };
 
- log_info("Running '%s' kernel (%d of %d) ...\n", sources_ndrange_Xd[i].src.kernel_name, i + 1, num_kernels_ndrange_Xd);
+ log_info("Running '%s' kernel (%d of %zu) ...\n",
+ sources_ndrange_Xd[i].src.kernel_name, i + 1,
+ num_kernels_ndrange_Xd);
  err_ret = run_single_kernel_args(context, queue, sources_ndrange_Xd[i].src.lines, sources_ndrange_Xd[i].src.num_lines, sources_ndrange_Xd[i].src.kernel_name, kernel_results, sizeof(kernel_results), arr_size(args), args);
 
  cl_int *ptr = (cl_int *)clEnqueueMapBuffer(queue, mem3, CL_TRUE, CL_MAP_READ, 0, glob_results.size() * sizeof(cl_int), 0, 0, 0, &err_ret);
@@ -718,7 +720,8 @@ int test_enqueue_ndrange(cl_device_id device, cl_context context, cl_command_que
 
  if (failCnt > 0)
  {
- log_error("ERROR: %d of %d kernels failed.\n", failCnt, num_kernels_ndrange_Xd);
+ log_error("ERROR: %zu of %zu kernels failed.\n", failCnt,
+ num_kernels_ndrange_Xd);
  }
 
  return res;