Merge branch 'main' into semaphore_negative_clCreateSemaphoreWithProp…

…ertiesKHR

.github/workflows/presubmit.yml

@@ -63,7 +63,7 @@ jobs:
  runs-on: ubuntu-22.04
  steps:
  - name: Install packages
- run: sudo apt install -y clang-format clang-format-11
+ run: sudo apt install -y clang-format clang-format-14
  - uses: actions/checkout@v4
  with:
  fetch-depth: 0

check-format.sh

@@ -2,7 +2,7 @@
 
 # Arg used to specify non-'origin/main' comparison branch
 ORIGIN_BRANCH=${1:-"origin/main"}
-CLANG_BINARY=${2:-"`which clang-format-11`"}
+CLANG_BINARY=${2:-"`which clang-format-14`"}
 
 # Run git-clang-format to check for violations
 CLANG_FORMAT_OUTPUT=$(git-clang-format --diff $ORIGIN_BRANCH --extensions c,cpp,h,hpp --binary $CLANG_BINARY)

test_common/gl/setup_x11.cpp

@@ -26,20 +26,26 @@ class X11GLEnvironment : public GLEnvironment
 private:
  cl_device_id m_devices[64];
  cl_uint m_device_count;
+ bool m_glut_init;
 
 public:
  X11GLEnvironment()
  {
  m_device_count = 0;
+ m_glut_init = false;
  }
  virtual int Init( int *argc, char **argv, int use_opencl_32 )
  {
  // Create a GLUT window to render into
- glutInit( argc, argv );
- glutInitWindowSize( 512, 512 );
- glutInitDisplayMode( GLUT_RGB | GLUT_DOUBLE );
- glutCreateWindow( "OpenCL <-> OpenGL Test" );
- glewInit();
+ if (!m_glut_init)
+ {
+ glutInit(argc, argv);
+ glutInitWindowSize(512, 512);
+ glutInitDisplayMode(GLUT_RGB | GLUT_DOUBLE);
+ glutCreateWindow("OpenCL <-> OpenGL Test");
+ glewInit();
+ m_glut_init = true;
+ }
  return 0;
  }
 

test_common/harness/errorHelpers.cpp

@@ -202,6 +202,8 @@ const char *GetChannelTypeName(cl_channel_type type)
  case CL_SFIXED14_APPLE: return "CL_SFIXED14_APPLE";
 #endif
  case CL_UNORM_INT24: return "CL_UNORM_INT24";
+ case CL_UNSIGNED_INT_RAW10_EXT: return "CL_UNSIGNED_INT_RAW10_EXT";
+ case CL_UNSIGNED_INT_RAW12_EXT: return "CL_UNSIGNED_INT_RAW12_EXT";
  default: return NULL;
  }
 }
@@ -369,7 +371,7 @@ static float Ulp_Error_Half_Float(float test, double reference)
  return (float)scalbn(testVal - reference, ulp_exp);
 }
 
-float Ulp_Error_Half(cl_half test, float reference)
+float Ulp_Error_Half(cl_half test, double reference)
 {
  return Ulp_Error_Half_Float(cl_half_to_float(test), reference);
 }

test_common/harness/errorHelpers.h

@@ -185,7 +185,7 @@ static int vlog_win32(const char *format, ...);
 
 extern const char *IGetErrorString(int clErrorCode);
 
-extern float Ulp_Error_Half(cl_half test, float reference);
+extern float Ulp_Error_Half(cl_half test, double reference);
 extern float Ulp_Error(float test, double reference);
 extern float Ulp_Error_Double(double test, long double reference);
 

test_common/harness/imageHelpers.cpp

@@ -288,6 +288,9 @@ uint32_t get_pixel_size(const cl_image_format *format)
  return get_format_channel_count(format) * sizeof(cl_float);
  case CL_UNORM_INT_101010_2: return 4;
 
+ case CL_UNSIGNED_INT_RAW10_EXT:
+ case CL_UNSIGNED_INT_RAW12_EXT: return 2;
+
  default: return 0;
  }
 }
@@ -1125,13 +1128,15 @@ cl_ulong get_image_size_mb(image_descriptor const *imageInfo)
 uint64_t gRoundingStartValue = 0;
 
 
-void escape_inf_nan_values(char *data, size_t allocSize)
+void escape_inf_nan_subnormal_values(char *data, size_t allocSize)
 {
  // filter values with 8 not-quite-highest bits
  unsigned int *intPtr = (unsigned int *)data;
  for (size_t i = 0; i<allocSize>> 2; i++)
  {
  if ((intPtr[i] & 0x7F800000) == 0x7F800000) intPtr[i] ^= 0x40000000;
+ else if ((intPtr[i] & 0x7F800000) == 0)
+ intPtr[i] ^= 0x40000000;
  }
 
  // Ditto with half floats (16-bit numbers with the 5 not-quite-highest bits
@@ -1140,6 +1145,8 @@ void escape_inf_nan_values(char *data, size_t allocSize)
  for (size_t i = 0; i<allocSize>> 1; i++)
  {
  if ((shortPtr[i] & 0x7C00) == 0x7C00) shortPtr[i] ^= 0x4000;
+ else if ((shortPtr[i] & 0x7C00) == 0)
+ shortPtr[i] ^= 0x4000;
  }
 }
 
@@ -1218,7 +1225,7 @@ char *generate_random_image_data(image_descriptor *imageInfo,
 
  // Note: inf or nan float values would cause problems, although we don't
  // know this will actually be a float, so we just know what to look for
- escape_inf_nan_values(data, allocSize);
+ escape_inf_nan_subnormal_values(data, allocSize);
  return data;
  }
 
@@ -1230,7 +1237,7 @@ char *generate_random_image_data(image_descriptor *imageInfo,
 
  // Note: inf or nan float values would cause problems, although we don't
  // know this will actually be a float, so we just know what to look for
- escape_inf_nan_values(data, allocSize);
+ escape_inf_nan_subnormal_values(data, allocSize);
 
  if (/*!gTestMipmaps*/ imageInfo->num_mip_levels < 2)
  {

test_common/harness/imageHelpers.h

@@ -191,6 +191,42 @@ cl_uint compute_max_mip_levels(size_t width, size_t height, size_t depth);
 cl_ulong compute_mipmapped_image_size(image_descriptor imageInfo);
 size_t compute_mip_level_offset(image_descriptor *imageInfo, size_t lod);
 
+constexpr size_t RAW10_EXT_CLUMP_SIZE = 5;
+constexpr size_t RAW10_EXT_CLUMP_NUM_PIXELS = 4;
+constexpr size_t RAW12_EXT_CLUMP_SIZE = 3;
+constexpr size_t RAW12_EXT_CLUMP_NUM_PIXELS = 2;
+
+inline bool is_width_compatible(image_descriptor imageInfo)
+{
+ if (imageInfo.format->image_channel_data_type == CL_UNSIGNED_INT_RAW10_EXT
+ && (imageInfo.width % RAW10_EXT_CLUMP_NUM_PIXELS) != 0)
+ {
+ return false;
+ }
+ if (imageInfo.format->image_channel_data_type == CL_UNSIGNED_INT_RAW12_EXT
+ && (imageInfo.width % RAW12_EXT_CLUMP_NUM_PIXELS) != 0)
+ {
+ return false;
+ }
+ return true;
+}
+
+inline size_t calculate_row_pitch(image_descriptor imageInfo, size_t pixelSize)
+{
+ if (imageInfo.format->image_channel_data_type == CL_UNSIGNED_INT_RAW10_EXT)
+ {
+ return (imageInfo.width * RAW10_EXT_CLUMP_SIZE)
+ / RAW10_EXT_CLUMP_NUM_PIXELS;
+ }
+ if (imageInfo.format->image_channel_data_type == CL_UNSIGNED_INT_RAW12_EXT)
+ {
+ return (imageInfo.width * RAW12_EXT_CLUMP_SIZE)
+ / RAW12_EXT_CLUMP_NUM_PIXELS;
+ }
+
+ return imageInfo.width * pixelSize;
+}
+
 template <class T>
 void read_image_pixel(void *imageData, image_descriptor *imageInfo, int x,
  int y, int z, T *outData, int lod)
@@ -255,10 +291,24 @@ void read_image_pixel(void *imageData, image_descriptor *imageInfo, int x,
 
  // Advance to the right spot
  char *ptr = (char *)imageData;
- size_t pixelSize = get_pixel_size(format);
-
- ptr += z * slice_pitch_lod + y * row_pitch_lod + x * pixelSize;
-
+ switch (format->image_channel_data_type)
+ {
+ case CL_UNSIGNED_INT_RAW10_EXT: {
+ ptr += z * slice_pitch_lod + y * row_pitch_lod
+ + (x / RAW10_EXT_CLUMP_NUM_PIXELS) * RAW10_EXT_CLUMP_SIZE;
+ break;
+ }
+ case CL_UNSIGNED_INT_RAW12_EXT: {
+ ptr += z * slice_pitch_lod + y * row_pitch_lod
+ + (x / RAW12_EXT_CLUMP_NUM_PIXELS) * RAW12_EXT_CLUMP_SIZE;
+ break;
+ }
+ default: {
+ size_t pixelSize = get_pixel_size(format);
+ ptr += z * slice_pitch_lod + y * row_pitch_lod + x * pixelSize;
+ break;
+ }
+ }
  // OpenCL only supports reading floats from certain formats
  switch (format->image_channel_data_type)
  {
@@ -377,6 +427,26 @@ void read_image_pixel(void *imageData, image_descriptor *imageInfo, int x,
  break;
  }
 #endif
+ case CL_UNSIGNED_INT_RAW10_EXT: {
+ cl_uchar *dPtr = (cl_uchar *)ptr;
+ i = x % RAW10_EXT_CLUMP_NUM_PIXELS;
+ uint8_t bit_index = i << 1;
+ uint16_t hi_val = dPtr[i] << 2;
+ uint16_t lo_val = (dPtr[4] & (0x3 << bit_index)) >> bit_index;
+
+ tempData[0] = (T)(hi_val | lo_val);
+ break;
+ }
+ case CL_UNSIGNED_INT_RAW12_EXT: {
+ cl_uchar *dPtr = (cl_uchar *)ptr;
+ i = x % RAW12_EXT_CLUMP_NUM_PIXELS;
+ uint8_t bit_index = i << 2;
+ uint16_t hi_val = dPtr[i] << 4;
+ uint16_t lo_val = (dPtr[2] & (0xF << bit_index)) >> bit_index;
+
+ tempData[0] = (T)(hi_val | lo_val);
+ break;
+ }
  }
 
 

test_common/harness/integer_ops_test_info.h

@@ -20,6 +20,8 @@
 #include "conversions.h"
 #include "testHarness.h"
 
+#include <vector>
+
 // TODO: expand usage to other tests.
 
 template <typename T> struct TestInfo

test_common/harness/os_helpers.cpp

@@ -209,7 +209,7 @@ static std::string _err_msg(int err, int level)
 
 #if (defined(__ANDROID__) && __ANDROID_API__ < 23) \
  || ((_POSIX_C_SOURCE >= 200112L || _XOPEN_SOURCE >= 600) && !_GNU_SOURCE) \
- || (defined(_GNU_SOURCE) && !defined(__GLIBC__))
+ || (defined(_GNU_SOURCE) && !defined(__GLIBC__) && !defined(__USE_GNU))
 
 // XSI version of strerror_r.
 #warning Not tested!

test_common/harness/rounding_mode.h

@@ -1,5 +1,5 @@
 //
-// Copyright (c) 2017 The Khronos Group Inc.
+// Copyright (c) 2017-2024 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.
@@ -42,10 +42,11 @@ typedef enum
  kshort = 3,
  kuint = 4,
  kint = 5,
- kfloat = 6,
- kdouble = 7,
- kulong = 8,
- klong = 9,
+ khalf = 6,
+ kfloat = 7,
+ kdouble = 8,
+ kulong = 9,
+ klong = 10,
 
  // This goes last
  kTypeCount

test_conformance/SVM/CMakeLists.txt

@@ -3,6 +3,7 @@ set(MODULE_NAME SVM)
 set(${MODULE_NAME}_SOURCES
  main.cpp
  test_allocate_shared_buffer.cpp
+ test_allocate_shared_buffer_negative.cpp
  test_byte_granularity.cpp
  test_cross_buffer_pointers.cpp
  test_enqueue_api.cpp

test_conformance/SVM/common.h

@@ -92,6 +92,10 @@ extern int test_svm_shared_address_space_fine_grain(cl_device_id deviceID, cl
 extern int test_svm_cross_buffer_pointers_coarse_grain(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
 extern int test_svm_pointer_passing(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
 extern int test_svm_allocate_shared_buffer(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
+extern int test_svm_allocate_shared_buffer_negative(cl_device_id deviceID,
+ cl_context context,
+ cl_command_queue queue,
+ int num_elements);
 extern int test_svm_shared_sub_buffers(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
 extern int test_svm_enqueue_api(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);
 extern int test_svm_migrate(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements);

test_conformance/SVM/main.cpp

@@ -261,20 +261,21 @@ cl_int create_cl_objects(cl_device_id device_from_harness, const char** ppCodeSt
 }
 
 test_definition test_list[] = {
- ADD_TEST( svm_byte_granularity),
- ADD_TEST( svm_set_kernel_exec_info_svm_ptrs ),
- ADD_TEST( svm_fine_grain_memory_consistency ),
- ADD_TEST( svm_fine_grain_sync_buffers ),
- ADD_TEST( svm_shared_address_space_fine_grain ),
- ADD_TEST( svm_shared_sub_buffers ),
- ADD_TEST( svm_shared_address_space_fine_grain_buffers ),
- ADD_TEST( svm_allocate_shared_buffer ),
- ADD_TEST( svm_shared_address_space_coarse_grain_old_api ),
- ADD_TEST( svm_shared_address_space_coarse_grain_new_api ),
- ADD_TEST( svm_cross_buffer_pointers_coarse_grain ),
- ADD_TEST( svm_pointer_passing ),
- ADD_TEST( svm_enqueue_api ),
- ADD_TEST_VERSION( svm_migrate, Version(2, 1)),
+ ADD_TEST(svm_byte_granularity),
+ ADD_TEST(svm_set_kernel_exec_info_svm_ptrs),
+ ADD_TEST(svm_fine_grain_memory_consistency),
+ ADD_TEST(svm_fine_grain_sync_buffers),
+ ADD_TEST(svm_shared_address_space_fine_grain),
+ ADD_TEST(svm_shared_sub_buffers),
+ ADD_TEST(svm_shared_address_space_fine_grain_buffers),
+ ADD_TEST(svm_allocate_shared_buffer),
+ ADD_TEST(svm_allocate_shared_buffer_negative),
+ ADD_TEST(svm_shared_address_space_coarse_grain_old_api),
+ ADD_TEST(svm_shared_address_space_coarse_grain_new_api),
+ ADD_TEST(svm_cross_buffer_pointers_coarse_grain),
+ ADD_TEST(svm_pointer_passing),
+ ADD_TEST(svm_enqueue_api),
+ ADD_TEST_VERSION(svm_migrate, Version(2, 1)),
 };
 
 const int test_num = ARRAY_SIZE( test_list );