vx_spawn_warps redesign using opencl's style scheduler
This commit is contained in:
@@ -4,7 +4,8 @@
|
||||
#define KERNEL_ARG_DEV_MEM_ADDR 0x7ffff000
|
||||
|
||||
struct kernel_arg_t {
|
||||
uint32_t count;
|
||||
uint32_t num_tasks;
|
||||
uint32_t task_size;
|
||||
uint32_t src0_ptr;
|
||||
uint32_t src1_ptr;
|
||||
uint32_t dst_ptr;
|
||||
|
||||
@@ -117,7 +117,8 @@ int main(int argc, char *argv[]) {
|
||||
RT_CHECK(vx_dev_caps(device, VX_CAPS_MAX_WARPS, &max_warps));
|
||||
RT_CHECK(vx_dev_caps(device, VX_CAPS_MAX_THREADS, &max_threads));
|
||||
|
||||
uint32_t num_points = count * max_cores * max_warps * max_threads;
|
||||
uint32_t num_tasks = max_cores * max_warps * max_threads;
|
||||
uint32_t num_points = count * num_tasks;
|
||||
uint32_t buf_size = num_points * sizeof(uint32_t);
|
||||
|
||||
std::cout << "number of points: " << num_points << std::endl;
|
||||
@@ -137,7 +138,8 @@ int main(int argc, char *argv[]) {
|
||||
RT_CHECK(vx_alloc_dev_mem(device, buf_size, &value));
|
||||
kernel_arg.dst_ptr = value;
|
||||
|
||||
kernel_arg.count = count;
|
||||
kernel_arg.num_tasks = num_tasks;
|
||||
kernel_arg.task_size = count;
|
||||
|
||||
std::cout << "dev_src0=" << std::hex << kernel_arg.src0_ptr << std::endl;
|
||||
std::cout << "dev_src1=" << std::hex << kernel_arg.src1_ptr << std::endl;
|
||||
|
||||
Binary file not shown.
@@ -3,14 +3,14 @@
|
||||
#include <vx_spawn.h>
|
||||
#include "common.h"
|
||||
|
||||
void kernel_body(void* arg) {
|
||||
void kernel_body(int task_id, void* arg) {
|
||||
struct kernel_arg_t* _arg = (struct kernel_arg_t*)(arg);
|
||||
uint32_t count = _arg->count;
|
||||
uint32_t count = _arg->task_size;
|
||||
int32_t* src0_ptr = (int32_t*)_arg->src0_ptr;
|
||||
int32_t* src1_ptr = (int32_t*)_arg->src1_ptr;
|
||||
int32_t* dst_ptr = (int32_t*)_arg->dst_ptr;
|
||||
|
||||
uint32_t offset = vx_thread_gid() * count;
|
||||
uint32_t offset = task_id * count;
|
||||
|
||||
for (uint32_t i = 0; i < count; ++i) {
|
||||
dst_ptr[offset+i] = src0_ptr[offset+i] + src1_ptr[offset+i];
|
||||
@@ -19,7 +19,5 @@ void kernel_body(void* arg) {
|
||||
|
||||
void main() {
|
||||
struct kernel_arg_t* arg = (struct kernel_arg_t*)KERNEL_ARG_DEV_MEM_ADDR;
|
||||
int num_warps = vx_num_warps();
|
||||
int num_threads = vx_num_threads();
|
||||
vx_spawn_warps(num_warps, num_threads, kernel_body, arg);
|
||||
vx_spawn_tasks(arg->num_tasks, kernel_body, arg);
|
||||
}
|
||||
File diff suppressed because it is too large
Load Diff
Binary file not shown.
Reference in New Issue
Block a user