[SYCL] supprt Flash Attention for fp32/fp16/Q4/Q5/Q8 (#20190)

* support flash-attention for fp32/fp16/Q4/Q5/Q8 * rm warining * update for JIT
2026-03-08 12:00:07 +08:00
parent c5a778891b
commit 213c4a0b81
65 changed files with 20091 additions and 8593 deletions
@@ -18,7 +18,7 @@ static void count_equal(const T *__restrict__ x, const T *__restrict__ y,
        nequal += xi == yi;
    }

-    nequal = warp_reduce_sum(nequal);
+    nequal = warp_reduce_sum<WARP_SIZE>(nequal);

    if (item_ct1.get_local_id(2) != 0) {
        return;