Merge remote-tracking branch 'github/main' into fp16_conversions

.github/workflows/presubmit.yml

@@ -13,16 +13,16 @@ jobs:
       fail-fast: false
       matrix:
         mainmatrix: [true]
-        os: [ubuntu-20.04, macos-latest, windows-latest]
+        os: [ubuntu-22.04, macos-latest, windows-latest]
         include:
-          - os: ubuntu-20.04
+          - os: ubuntu-22.04
             mainmatrix: true
             gl: 1
             extra: " gl"
-          - os: ubuntu-20.04
+          - os: ubuntu-22.04
             mainmatrix: false
             arch: arm
-          - os: ubuntu-20.04
+          - os: ubuntu-22.04
             mainmatrix: false
             arch: aarch64
             debug: 1
@@ -55,10 +55,10 @@ jobs:
         run: ./presubmit.sh
   formatcheck:
     name: Check code format
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-22.04
     steps:
       - name: Install packages
-        run: sudo apt install -y clang-format clang-format-9
+        run: sudo apt install -y clang-format clang-format-11
       - uses: actions/checkout@v3
         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-9`"}
+CLANG_BINARY=${2:-"`which clang-format-11`"}
 
 # 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/helpers.cpp

@@ -1715,7 +1715,7 @@ void * CreateGLRenderbuffer( GLsizei width, GLsizei height,
         // Reverse and reorder to validate since in the
         // kernel the read_imagef() call always returns RGBA
         cl_uchar *p = (cl_uchar *)buffer;
-        for( size_t i = 0; i < (size_t)width * height; i++ )
+        for (GLsizei i = 0; i < width * height; i++)
         {
             cl_uchar uc0 = p[i * 4 + 0];
             cl_uchar uc1 = p[i * 4 + 1];
@@ -1733,7 +1733,7 @@ void * CreateGLRenderbuffer( GLsizei width, GLsizei height,
       // Reverse and reorder to validate since in the
       // kernel the read_imagef() call always returns RGBA
       cl_uchar *p = (cl_uchar *)buffer;
-      for( size_t i = 0; i < width * height; i++ )
+      for (GLsizei i = 0; i < width * height; i++)
       {
         cl_uchar uc0 = p[i * 4 + 0];
         cl_uchar uc1 = p[i * 4 + 1];

test_common/harness/mt19937.cpp

@@ -51,6 +51,7 @@
 #include "harness/alloc.h"
 
 #ifdef __SSE2__
+#include <mutex>
 #include <emmintrin.h>
 #endif
 
@@ -107,7 +108,7 @@ cl_uint genrand_int32(MTdata d)
     /* mag01[x] = x * MATRIX_A  for x=0,1 */
     static const cl_uint mag01[2] = { 0x0UL, MATRIX_A };
 #ifdef __SSE2__
-    static volatile int init = 0;
+    static std::once_flag init_flag;
     static union {
         __m128i v;
         cl_uint s[4];
@@ -123,8 +124,7 @@ cl_uint genrand_int32(MTdata d)
         int kk;
 
 #ifdef __SSE2__
-        if (0 == init)
-        {
+        auto init_fn = []() {
             upper_mask.s[0] = upper_mask.s[1] = upper_mask.s[2] =
                 upper_mask.s[3] = UPPER_MASK;
             lower_mask.s[0] = lower_mask.s[1] = lower_mask.s[2] =
@@ -134,8 +134,8 @@ cl_uint genrand_int32(MTdata d)
                 MATRIX_A;
             c0.s[0] = c0.s[1] = c0.s[2] = c0.s[3] = (cl_uint)0x9d2c5680UL;
             c1.s[0] = c1.s[1] = c1.s[2] = c1.s[3] = (cl_uint)0xefc60000UL;
-            init = 1;
-        }
+        };
+        std::call_once(init_flag, init_fn);
 #endif
 
         kk = 0;

test_common/harness/stringHelpers.h

@@ -14,10 +14,11 @@
 // limitations under the License.
 //
 
-#ifndef BASIC_UTILS_H
-#define BASIC_UTILS_H
+#ifndef STRING_HELPERS_H
+#define STRING_HELPERS_H
 
 #include <memory>
+#include <stdexcept>
 #include <string>
 
 inline std::string concat_kernel(const char *sstr[], int num)
@@ -38,4 +39,4 @@ inline std::string str_sprintf(const std::string &str, Args... args)
     return std::string(buffer.get(), buffer.get() + s - 1);
 }
 
-#endif // BASIC_UTIL_H
+#endif // STRING_HELPERS_H

test_common/harness/testHarness.cpp

@@ -835,9 +835,9 @@ void callTestFunctions(test_definition testList[],
         std::vector<std::thread *> threads;
         test_harness_state state = { testList, resultTestList, deviceToUse,
                                      config };
-        for (int i = 0; i < config.numWorkerThreads; i++)
+        for (unsigned i = 0; i < config.numWorkerThreads; i++)
         {
-            log_info("Spawning worker thread %i\n", i);
+            log_info("Spawning worker thread %u\n", i);
             threads.push_back(new std::thread(test_function_runner, &state));
         }
 

test_conformance/allocations/main.cpp

@@ -326,6 +326,7 @@ int main(int argc, const char *argv[])
         else if ( strcmp( argv[i], "--help" ) == 0 || strcmp( argv[i], "-h" ) == 0 )
         {
             printUsage( argv[0] );
+            free(argList);
             return -1;
         }
 

test_conformance/api/test_queries.cpp

@@ -799,8 +799,8 @@ int test_kernel_required_group_size(cl_device_id deviceID, cl_context context, c
         test_error(error, "clFinish failed");
 
         if (max_dimensions == 2) {
-            return 0;
             free(source);
+            return 0;
         }
 
         local[1]--; local[2]++;

test_conformance/atomics/test_indexed_cases.cpp

@@ -13,6 +13,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 //
+
+#include <memory>
+
 #include "testBase.h"
 #include "harness/conversions.h"
 
@@ -226,13 +229,13 @@ int add_index_bin_test(size_t *global_threads, cl_command_queue queue,
              (int)global_threads[0], (int)local_threads[0]);
 
     // Allocate our storage
-    cl_mem bin_counters =
+    clMemWrapper bin_counters =
         clCreateBuffer(context, CL_MEM_READ_WRITE,
                        sizeof(cl_int) * number_of_bins, NULL, NULL);
-    cl_mem bins = clCreateBuffer(
+    clMemWrapper bins = clCreateBuffer(
         context, CL_MEM_READ_WRITE,
         sizeof(cl_int) * number_of_bins * max_counts_per_bin, NULL, NULL);
-    cl_mem bin_assignments =
+    clMemWrapper bin_assignments =
         clCreateBuffer(context, CL_MEM_READ_ONLY,
                        sizeof(cl_int) * number_of_items, NULL, NULL);
 
@@ -253,7 +256,7 @@ int add_index_bin_test(size_t *global_threads, cl_command_queue queue,
     }
 
     // Initialize our storage
-    cl_int *l_bin_counts = (cl_int *)malloc(sizeof(cl_int) * number_of_bins);
+    std::unique_ptr<cl_int[]> l_bin_counts(new cl_int[number_of_bins]);
     if (!l_bin_counts)
     {
         log_error("add_index_bin_test FAILED to allocate initial values for "
@@ -263,8 +266,8 @@ int add_index_bin_test(size_t *global_threads, cl_command_queue queue,
     int i;
     for (i = 0; i < number_of_bins; i++) l_bin_counts[i] = 0;
     err = clEnqueueWriteBuffer(queue, bin_counters, true, 0,
-                               sizeof(cl_int) * number_of_bins, l_bin_counts, 0,
-                               NULL, NULL);
+                               sizeof(cl_int) * number_of_bins,
+                               l_bin_counts.get(), 0, NULL, NULL);
     if (err)
     {
         log_error("add_index_bin_test FAILED to set initial values for "
@@ -273,8 +276,8 @@ int add_index_bin_test(size_t *global_threads, cl_command_queue queue,
         return -1;
     }
 
-    cl_int *values =
-        (cl_int *)malloc(sizeof(cl_int) * number_of_bins * max_counts_per_bin);
+    std::unique_ptr<cl_int[]> values(
+        new cl_int[number_of_bins * max_counts_per_bin]);
     if (!values)
     {
         log_error(
@@ -285,18 +288,16 @@ int add_index_bin_test(size_t *global_threads, cl_command_queue queue,
     err = clEnqueueWriteBuffer(queue, bins, true, 0,
                                sizeof(cl_int) * number_of_bins
                                    * max_counts_per_bin,
-                               values, 0, NULL, NULL);
+                               values.get(), 0, NULL, NULL);
     if (err)
     {
         log_error(
             "add_index_bin_test FAILED to set initial values for bins: %d\n",
             err);
         return -1;
     }
-    free(values);
 
-    cl_int *l_bin_assignments =
-        (cl_int *)malloc(sizeof(cl_int) * number_of_items);
+    std::unique_ptr<cl_int[]> l_bin_assignments(new cl_int[number_of_items]);
     if (!l_bin_assignments)
     {
         log_error("add_index_bin_test FAILED to allocate initial values for "
@@ -326,7 +327,7 @@ int add_index_bin_test(size_t *global_threads, cl_command_queue queue,
     }
     err = clEnqueueWriteBuffer(queue, bin_assignments, true, 0,
                                sizeof(cl_int) * number_of_items,
-                               l_bin_assignments, 0, NULL, NULL);
+                               l_bin_assignments.get(), 0, NULL, NULL);
     if (err)
     {
         log_error("add_index_bin_test FAILED to set initial values for "
@@ -355,8 +356,8 @@ int add_index_bin_test(size_t *global_threads, cl_command_queue queue,
         return -1;
     }
 
-    cl_int *final_bin_assignments =
-        (cl_int *)malloc(sizeof(cl_int) * number_of_bins * max_counts_per_bin);
+    std::unique_ptr<cl_int[]> final_bin_assignments(
+        new cl_int[number_of_bins * max_counts_per_bin]);
     if (!final_bin_assignments)
     {
         log_error("add_index_bin_test FAILED to allocate initial values for "
@@ -366,24 +367,23 @@ int add_index_bin_test(size_t *global_threads, cl_command_queue queue,
     err = clEnqueueReadBuffer(queue, bins, true, 0,
                               sizeof(cl_int) * number_of_bins
                                   * max_counts_per_bin,
-                              final_bin_assignments, 0, NULL, NULL);
+                              final_bin_assignments.get(), 0, NULL, NULL);
     if (err)
     {
         log_error("add_index_bin_test FAILED to read back bins: %d\n", err);
         return -1;
     }
 
-    cl_int *final_bin_counts =
-        (cl_int *)malloc(sizeof(cl_int) * number_of_bins);
+    std::unique_ptr<cl_int[]> final_bin_counts(new cl_int[number_of_bins]);
     if (!final_bin_counts)
     {
         log_error("add_index_bin_test FAILED to allocate initial values for "
                   "final_bin_counts.\n");
         return -1;
     }
     err = clEnqueueReadBuffer(queue, bin_counters, true, 0,
-                              sizeof(cl_int) * number_of_bins, final_bin_counts,
-                              0, NULL, NULL);
+                              sizeof(cl_int) * number_of_bins,
+                              final_bin_counts.get(), 0, NULL, NULL);
     if (err)
     {
         log_error("add_index_bin_test FAILED to read back bin_counters: %d\n",
@@ -460,13 +460,7 @@ int add_index_bin_test(size_t *global_threads, cl_command_queue queue,
             errors++;
         }
     }
-    free(l_bin_counts);
-    free(l_bin_assignments);
-    free(final_bin_assignments);
-    free(final_bin_counts);
-    clReleaseMemObject(bin_counters);
-    clReleaseMemObject(bins);
-    clReleaseMemObject(bin_assignments);
+
     if (errors == 0)
     {
         log_info("add_index_bin_test passed. Each item was put in the correct "

test_conformance/basic/CMakeLists.txt

@@ -52,14 +52,12 @@ set(${MODULE_NAME}_SOURCES
     test_kernel_call_kernel_function.cpp
     test_local_kernel_scope.cpp
     test_progvar.cpp
-    test_wg_barrier.cpp
     test_global_linear_id.cpp
     test_local_linear_id.cpp
     test_enqueued_local_size.cpp
     test_simple_image_pitch.cpp
     test_get_linear_ids.cpp
     test_rw_image_access_qualifier.cpp
-    test_wg_barrier.cpp
     test_enqueued_local_size.cpp
     test_global_linear_id.cpp
     test_local_linear_id.cpp