Rename kernels to kmain since main forbidden...

Author: cirosantilli

opencl/common.h

@@ -71,7 +71,7 @@ void common_create_kernel(
 ) {
     if (NULL != source) {
         common_create_program(common, source, options, &common->program);
-        common->kernel = clCreateKernel(common->program, "main", NULL);
+        common->kernel = clCreateKernel(common->program, "mymain", NULL);
     } else {
         common->kernel = NULL;
         common->program = NULL;

opencl/inc.c

@@ -19,7 +19,10 @@ int main(void) {
     const char *source =
         /* kernel pointer arguments must be __global, __constant, or __local. */
         /* https://www.khronos.org/registry/cl/sdk/2.1/docs/man/xhtml/restrictions.html */
-        "__kernel void main(__global int *out) {\n"
+        /**/
+        /* Kernel functions cannot be called main... NVIDIA's compiler may allow that
+         * but others don't, so don't do it. */
+        "__kernel void kmain(__global int *out) {\n"
         "    out[0]++;\n"
         "}\n";
     cl_command_queue command_queue;
@@ -37,7 +40,7 @@ int main(void) {
     context = clCreateContext(NULL, 1, &device, NULL, NULL, NULL);
     program = clCreateProgramWithSource(context, 1, &source, NULL, NULL);
 	clBuildProgram(program, 1, &device, "", NULL, NULL);
-    kernel = clCreateKernel(program, "main", NULL);
+    kernel = clCreateKernel(program, "kmain", NULL);
     buffer = clCreateBuffer(context, CL_MEM_READ_WRITE | CL_MEM_COPY_HOST_PTR, sizeof(cl_int), &input, NULL);
     clSetKernelArg(kernel, 0, sizeof(buffer), &buffer);
     command_queue = clCreateCommandQueue(context, device, 0, NULL);

opencl/inc_vector.c

@@ -13,7 +13,7 @@ it is just a clEnqueueNDRangeKernel + get_global_id hello world.
 
 int main(int argc, char **argv) {
     const char *source =
-        "__kernel void main(__global int *io) {\n"
+        "__kernel void kmain(__global int *io) {\n"
         "    io[get_global_id(0)]++;\n"
         "}\n";
     cl_int *io, *expected_output;

opencl/inc_vector_globals.c

@@ -3,7 +3,7 @@ API exercise, increment a vector with less global work groups than integers,
 which forces us to put a for loop in the kernel.
 
 I don't think we can get the size of each global work group from the kernel,
-so we just calculate it on CPU ans pass a sa parameter.
+so we just calculate it on CPU and pass as a parameter.
 
 This is how the work will be split:
 
@@ -17,7 +17,7 @@ This is how the work will be split:
 
 int main(void) {
     const char *source =
-        "__kernel void main(uint group_nlems, __global int *out) {\n"
+        "__kernel void kmain(uint group_nlems, __global int *out) {\n"
         "    uint i_min = get_global_id(0) * group_nlems;\n"
         "    uint i_max = i_min + group_nlems;\n"
         "    for (uint i = i_min; i < i_max; ++i) {\n"

opencl/matmul.cl

@@ -1,4 +1,4 @@
-__kernel void main(
+__kernel void kmain(
     __global float *A,
     __global float *B,
     __global float *C,

opencl/matmul_block.cl

@@ -1,4 +1,4 @@
-__kernel void main(
+__kernel void kmain(
     __global const float* restrict A,
     __global const float* restrict B,
     __global float* restrict C,

opencl/matmul_row_local.cl

@@ -1,4 +1,4 @@
-__kernel void main(
+__kernel void kmain(
     __global float* A,
     __global float* B,
     __global float* C,

opencl/matmul_row_priv.cl

@@ -1,4 +1,4 @@
-__kernel void main(
+__kernel void kmain(
     __global float* A,
     __global float* B,
     __global float* C,

opencl/matmul_row_priv_col_local.cl

@@ -1,4 +1,4 @@
-__kernel void main(
+__kernel void kmain(
     __global float* A,
     __global float* B,
     __global float* C,

opencl/matmul_row_priv_cols_local.cl

@@ -1,4 +1,4 @@
-__kernel void main(
+__kernel void kmain(
     __global float* A,
     __global float* B,
     __global float* C,