refactor: tidy attention params and launcher interfaces

- rename output pointer field o to o_ptr for consistency with q_ptr/k_ptr/v_ptr
- regroup AttentionParams fields by responsibility and fix misleading comments
- drop unused max_seq_len/total_q fields and paged decode max_seq_len arg
- drop redundant group_size param from decode launchers (computed from p)
This commit is contained in:
2026-08-09 20:52:06 +08:00
parent a5a3cc1fc2
commit cd31f1f62f
14 changed files with 69 additions and 84 deletions
+2 -2
View File
@@ -143,13 +143,13 @@ __global__ void attn_prefill_split_q_mma_kernel(AttentionParams<bf16> p) {
__nv_bfloat162 v = __floats2bfloat162_rn(Oacc[dn8][0] * rl0,
Oacc[dn8][1] * rl0);
*reinterpret_cast<__nv_bfloat162*>(
&p.o[o_base + qr0 * p.q_l_stride + d * p.q_d_stride]) = v;
&p.o_ptr[o_base + qr0 * p.q_l_stride + d * p.q_d_stride]) = v;
}
if (qr1 < q_len) {
__nv_bfloat162 v = __floats2bfloat162_rn(Oacc[dn8][2] * rl1,
Oacc[dn8][3] * rl1);
*reinterpret_cast<__nv_bfloat162*>(
&p.o[o_base + qr1 * p.q_l_stride + d * p.q_d_stride]) = v;
&p.o_ptr[o_base + qr1 * p.q_l_stride + d * p.q_d_stride]) = v;
}
}
}