flash: Disable rescale flag check
GEMM-II finishes much earlier than softmax for this to be a problem.
This commit is contained in:
@@ -602,17 +602,17 @@ void kernel_body(int task_id, kernel_arg_t *__UNIFORM__ arg) {
|
|||||||
}
|
}
|
||||||
|
|
||||||
#ifdef FENCE_GEMM_II
|
#ifdef FENCE_GEMM_II
|
||||||
asm volatile("rescale_fence_read_start_%=:" ::);
|
// asm volatile("rescale_fence_read_start_%=:" ::);
|
||||||
// check flag to make sure GEMM II finished and read-after-write
|
// // check flag to make sure GEMM II finished and read-after-write
|
||||||
// dependency on O tile is settled for rescale
|
// // dependency on O tile is settled for rescale
|
||||||
if (tid_in_warpgroup == 0) {
|
// if (tid_in_warpgroup == 0) {
|
||||||
while ((*smem_O_flag) != 1)
|
// while ((*smem_O_flag) != 1)
|
||||||
;
|
// ;
|
||||||
// set it back to 0 for the next tile iteration
|
// // set it back to 0 for the next tile iteration
|
||||||
*smem_O_flag = 0;
|
// *smem_O_flag = 0;
|
||||||
vx_fence();
|
// vx_fence();
|
||||||
}
|
// }
|
||||||
asm volatile("rescale_fence_read_end_%=:" ::);
|
// asm volatile("rescale_fence_read_end_%=:" ::);
|
||||||
#endif
|
#endif
|
||||||
|
|
||||||
#if 0
|
#if 0
|
||||||
|
|||||||
Reference in New Issue
Block a user