vulkan: support fattn sinks (#15126)

This commit is contained in:
Jeff Bolz
2025-08-07 15:44:20 -05:00
committed by GitHub
parent a0552c8bee
commit c4f53563df
6 changed files with 148 additions and 20 deletions

View File

@@ -50,10 +50,13 @@ layout (push_constant) uniform parameter {
uint32_t k_num;
} p;
#define SINK_ENABLE_BIT (1<<24)
#define MASK_ENABLE_BIT (1<<16)
#define N_LOG2_MASK 0xFFFF
layout (binding = 4) writeonly buffer O {D_TYPE data_o[];};
layout (binding = 4) readonly buffer S {float data_s[];};
layout (binding = 5) writeonly buffer O {D_TYPE data_o[];};
#if defined(A_TYPE_PACKED16)
#define BINDING_IDX_K 0
@@ -111,6 +114,14 @@ ACC_TYPE perElemOpComputeSlope(const in uint32_t r, const in uint32_t c, const i
return ACC_TYPE(pow(base, ACC_TYPE(exph)));
}
// Load the sink value, indexed by Q's dimension 2.
ACC_TYPE perElemOpGetSink(const in uint32_t r, const in uint32_t c, const in ACC_TYPE elem, const in uint32_t iq2)
{
const uint32_t h = iq2 + (r % p.gqa_ratio);
return ACC_TYPE(data_s[h]);
}
uint32_t i, N, KV, split_k_index, Tr, start_j, end_j,
iq2, iq3, rk2, rk3, rv2, rv3, ik2, ik3, iv2, iv3,
q_stride, k_stride, v_stride, m_stride;