add utest for load spir binary.

To generate SPIR binary, please refer to the page
https://github.com/KhronosGroup/SPIR.
For llvm3.2, the command is "clang -cc1 -emit-llvm-bc -triple
spir-unknown-unknown -cl-std=CL1.2 -include opencl_spir.h
compiler_ceil.cl -o compiler_ceil32.spir"
For llvm3.5, the option -cl-kernel-arg-info is required,
and option -fno-builtin is required to avoid warning.

v2: add missing load_program_from_spir.cpp file.

Signed-off-by: Luo Xionghu <xionghu.luo@intel.com>
Reviewed-by: Zhigang Gong <zhigang.gong@linux.intel.com>

1 files changed, 90 insertions, 0 deletions

diff --git a/utests/load_program_from_spir.cpp b/utests/load_program_from_spir.cpp
new file mode 100644
index 00000000..3e4534c7
--- /dev/null
+++ b/utests/load_program_from_spir.cpp
@@ -0,0 +1,90 @@
+#include "utest_helper.hpp"
+#include "utest_file_map.hpp"
+#include <cstring>
+#include <cmath>
+#include <algorithm>
+
+using namespace std;
+
+static void cpu(int global_id, float *src, float *dst) {
+    dst[global_id] = ceilf(src[global_id]);
+}
+
+static void test_load_program_from_spir(void)
+{
+    size_t param_value_size;
+    std::string extensionStr;
+    OCL_CALL (clGetPlatformInfo, platform, CL_PLATFORM_EXTENSIONS, 0, 0, &param_value_size);
+    std::vector<char> param_value(param_value_size);
+    OCL_CALL (clGetDeviceInfo, device, CL_DEVICE_EXTENSIONS, param_value_size, param_value.empty() ? NULL : &param_value.front(), &param_value_size);
+    if (!param_value.empty())
+      extensionStr = std::string(&param_value.front(), param_value_size-1);
+
+    if (!std::strstr(extensionStr.c_str(), "cl_khr_spir")) {
+      OCL_ASSERT(0);
+    }
+
+    const size_t n = 16;
+    float cpu_dst[16], cpu_src[16];
+    cl_int status;
+    cl_int binary_status;
+    char *ker_path = NULL;
+
+    cl_file_map_t *fm = cl_file_map_new();
+    ker_path = cl_do_kiss_path("compiler_ceil32.spir", device);
+    OCL_ASSERT (cl_file_map_open(fm, ker_path) == CL_FILE_MAP_SUCCESS);
+
+    const unsigned char *src = (const unsigned char *)cl_file_map_begin(fm);
+    const size_t sz = cl_file_map_size(fm);
+
+    program = clCreateProgramWithBinary(ctx, 1,
+              &device, &sz, &src, &binary_status, &status);
+
+    OCL_ASSERT(program && status == CL_SUCCESS);
+
+    /* OCL requires to build the program even if it is created from a binary */
+    OCL_ASSERT(clBuildProgram(program, 1, &device, "-x spir", NULL, NULL) == CL_SUCCESS);
+
+    kernel = clCreateKernel(program, "compiler_ceil", &status);
+    OCL_ASSERT(status == CL_SUCCESS);
+
+    OCL_CREATE_BUFFER(buf[0], 0, n * sizeof(float), NULL);
+    OCL_CREATE_BUFFER(buf[1], 0, n * sizeof(float), NULL);
+    OCL_SET_ARG(0, sizeof(cl_mem), &buf[0]);
+    OCL_SET_ARG(1, sizeof(cl_mem), &buf[1]);
+    globals[0] = 16;
+    locals[0] = 16;
+
+    // Run random tests
+    for (uint32_t pass = 0; pass < 8; ++pass) {
+        OCL_MAP_BUFFER(0);
+        for (int32_t i = 0; i < (int32_t) n; ++i)
+            cpu_src[i] = ((float*)buf_data[0])[i] = .1f * (rand() & 15) - .75f;
+        OCL_UNMAP_BUFFER(0);
+
+        // Run the kernel on GPU
+        OCL_NDRANGE(1);
+
+        // Run on CPU
+        for (int32_t i = 0; i < (int32_t) n; ++i) cpu(i, cpu_src, cpu_dst);
+
+        // Compare
+        OCL_MAP_BUFFER(1);
+
+#if 0
+        printf("#### GPU:\n");
+        for (int32_t i = 0; i < (int32_t) n; ++i)
+            printf(" %f", ((float *)buf_data[1])[i]);
+        printf("\n#### CPU:\n");
+        for (int32_t i = 0; i < (int32_t) n; ++i)
+            printf(" %f", cpu_dst[i]);
+        printf("\n");
+#endif
+
+        for (int32_t i = 0; i < (int32_t) n; ++i)
+            OCL_ASSERT(((float *)buf_data[1])[i] == cpu_dst[i]);
+        OCL_UNMAP_BUFFER(1);
+    }
+}
+
+MAKE_UTEST_FROM_FUNCTION(test_load_program_from_spir);