wu-arch/kernels
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

new unaligned access kernel, update idle kernel

Browse Source
This commit is contained in:
Richard Yan
2024-10-24 17:13:13 -07:00
parent 8cc0c3bae4
commit fb928b5cda
9 changed files with 328 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
tests/regression/idle/.gitignore vendored Normal file
View File

@@ -0,0 +1,5 @@
*.bin
*.dump
*.elf
idle
.depend

2
tests/regression/idle/Makefile
View File

@@ -1,4 +1,4 @@
PROJECT = sgemm_gemmini_dma
PROJECT = idle
SRCS = main.cpp common.h

46
tests/regression/idle/kernel.cpp
View File

@@ -7,7 +7,7 @@
#include "gemmini_mmio.h"
#define NUM_CLUSTERS 1
#define NUM_THREADS_IN_CLUSTER 256
#define NUM_THREADS_IN_CLUSTER 512
#define HW_TID() ({uint32_t gtid; asm volatile ("csrr %0, mhartid" : "=r" (gtid)); gtid;})
@@ -22,9 +22,45 @@ void kernel_body(int task_id, kernel_arg_t *__UNIFORM__ arg) {
// reinterpret_cast<uint32_t *>(arg->addr_c)[0] = counter;
// call barrier in a divergent branch, which will hang the core
if ((vx_thread_id() % NUM_THREADS) == 0) {
vx_barrier(0, NUM_WARPS);
}
asm volatile("li x1, 0xa0a0a0a0");
asm volatile("li x2, 0xa0a0a0a0");
asm volatile("li x3, 0xa0a0a0a0");
asm volatile("li x4, 0xa0a0a0a0");
asm volatile("li x5, 0xa0a0a0a0");
asm volatile("li x6, 0xa0a0a0a0");
asm volatile("li x7, 0xa0a0a0a0");
asm volatile("li x8, 0xa0a0a0a0");
asm volatile("li x9, 0xa0a0a0a0");
asm volatile("li x10, 0xa0a0a0a0");
asm volatile("li x11, 0xa0a0a0a0");
asm volatile("li x12, 0xa0a0a0a0");
asm volatile("li x13, 0xa0a0a0a0");
asm volatile("li x14, 0xa0a0a0a0");
asm volatile("li x15, 0xa0a0a0a0");
asm volatile("li x16, 0xa0a0a0a0");
asm volatile("li x17, 0xa0a0a0a0");
asm volatile("li x18, 0xa0a0a0a0");
asm volatile("li x19, 0xa0a0a0a0");
asm volatile("li x20, 0xa0a0a0a0");
asm volatile("li x21, 0xa0a0a0a0");
asm volatile("li x22, 0xa0a0a0a0");
asm volatile("li x23, 0xa0a0a0a0");
asm volatile("li x24, 0xa0a0a0a0");
asm volatile("li x25, 0xa0a0a0a0");
asm volatile("li x26, 0xa0a0a0a0");
asm volatile("li x27, 0xa0a0a0a0");
asm volatile("li x28, 0xa0a0a0a0");
asm volatile("li x29, 0xa0a0a0a0");
asm volatile("li x30, 0xa0a0a0a0");
asm volatile("li x31, 0xa0a0a0a0");
asm volatile("csrr a0, 0xcc1");
asm volatile("beqz a0, bar");
asm volatile("vx_tmc zero");
asm volatile("bar:");
asm volatile("vx_bar zero, a0");
// if ((vx_thread_id() % NUM_THREADS) == 0) {
// vx_barrier(0, NUM_WARPS);
// }
vx_tmc(0);
}
@@ -35,7 +71,7 @@ int main() {
// spawn a single warp in every core
const uint32_t grid_size = NUM_THREADS * NUM_CORES;
#ifdef RADIANCE
vx_spawn_tasks_cluster(grid_size, (vx_spawn_tasks_cb)kernel_body, arg);
vx_spawn_tasks_cluster(NUM_THREADS_IN_CLUSTER, (vx_spawn_tasks_cb)kernel_body, arg);
#else
vx_spawn_tasks_contiguous(grid_size, (vx_spawn_tasks_cb)kernel_body, arg);
#endif