fix: synchronize final decode async copy
- wait for the final split-KV tile before reading shared memory - cover long decode with production context capacity
This commit is contained in:
@@ -121,7 +121,10 @@ __global__ void attn_decode_split_kv_mma_kernel(AttentionParams<bf16> p) {
|
||||
load_tile(ti_begin + i, i);
|
||||
|
||||
for (int it = 0; it < ntiles; it++) {
|
||||
cp_async_wait_group<STAGES - 1>();
|
||||
if (it + 1 == ntiles)
|
||||
cp_async_wait_group<0>();
|
||||
else
|
||||
cp_async_wait_group<STAGES - 1>();
|
||||
__syncwarp();
|
||||
process_tile(it, it & (STAGES - 1));
|
||||
__syncwarp();
|
||||
|
||||
Reference in New Issue
Block a user