aboutsummaryrefslogtreecommitdiff
path: root/embed_mlx.c
diff options
context:
space:
mode:
authorOnur Gündüz <ogunduz@gmail.com>2026-06-15 09:57:00 +0200
committerOnur Gündüz <ogunduz@gmail.com>2026-06-15 09:57:00 +0200
commit07d3b036de430fb86935e029ca03bcf6011a02e1 (patch)
tree525309d20ef18ac1f49c135dda968d779f1eae2f /embed_mlx.c
parent43fdb308879fb0c61cb71553d9ac885c011d3a66 (diff)
downloadffwd-07d3b036de430fb86935e029ca03bcf6011a02e1.tar.gz
ffwd-07d3b036de430fb86935e029ca03bcf6011a02e1.zip
late: restore MLX device select/concat used by interaction check
Diffstat (limited to 'embed_mlx.c')
-rw-r--r--embed_mlx.c101
1 files changed, 101 insertions, 0 deletions
diff --git a/embed_mlx.c b/embed_mlx.c
index 1058d07..e18a0a4 100644
--- a/embed_mlx.c
+++ b/embed_mlx.c
@@ -1344,6 +1344,107 @@ cleanup:
return result;
}
+/* Device-resident select/concat primitives. The server rerank path batches
+ * candidates through embed_mlx_late_encode_docs_device, but the late-interaction
+ * verification harness (scripts/check_late_interaction.py) drives these directly
+ * to validate per-document gather and concat against the CPU MaxSim reference. */
+embed_mlx_late_vectors_t *embed_mlx_late_vectors_concat(
+ embed_mlx_late_ctx_t *ctx, const embed_mlx_late_vectors_t *const *items, int count) {
+ if (!ctx || !ctx->base || !items || count <= 0)
+ return NULL;
+ int dim = 0;
+ int total = 0;
+ for (int i = 0; i < count; i++) {
+ const embed_mlx_late_vectors_t *v = items[i];
+ if (!v || v->owner != ctx || !arr_ok(v->vectors) || v->tokens <= 0 || v->dim <= 0)
+ return NULL;
+ if (i == 0)
+ dim = v->dim;
+ if (v->dim != dim || total > INT_MAX - v->tokens)
+ return NULL;
+ total += v->tokens;
+ }
+
+ mlx_array *parts = (mlx_array *)malloc((size_t)count * sizeof(*parts));
+ if (!parts)
+ return NULL;
+ for (int i = 0; i < count; i++)
+ parts[i] = items[i]->vectors;
+
+ mlx_stream S = ctx->base->stream;
+ mlx_vector_array vec = mlx_vector_array_new_data(parts, (size_t)count);
+ free(parts);
+ if (!vec.ctx)
+ return NULL;
+
+ mlx_array joined = mlx_array_new();
+ mlx_concatenate_axis(&joined, vec, 0, S);
+ mlx_vector_array_free(vec);
+ if (!arr_ok(joined)) {
+ mlx_array_free(joined);
+ return NULL;
+ }
+
+ embed_mlx_late_vectors_t *out = (embed_mlx_late_vectors_t *)calloc(1, sizeof(*out));
+ if (!out) {
+ mlx_array_free(joined);
+ return NULL;
+ }
+ out->owner = ctx;
+ out->vectors = joined;
+ out->tokens = total;
+ out->dim = dim;
+ return out;
+}
+
+embed_mlx_late_vectors_t *embed_mlx_late_vectors_select(embed_mlx_late_ctx_t *ctx,
+ const embed_mlx_late_vectors_t *vecs,
+ const int *token_indices,
+ int count) {
+ if (!ctx || !ctx->base || !vecs || vecs->owner != ctx || !arr_ok(vecs->vectors) ||
+ !token_indices || count <= 0 || vecs->tokens <= 0 || vecs->dim <= 0)
+ return NULL;
+
+ for (int i = 0; i < count; i++) {
+ if (token_indices[i] < 0 || token_indices[i] >= vecs->tokens)
+ return NULL;
+ }
+
+ size_t idx_bytes;
+ if (mlx_mul_size((size_t)count, sizeof(int), &idx_bytes) != 0)
+ return NULL;
+ int *idx_data = (int *)malloc(idx_bytes);
+ if (!idx_data)
+ return NULL;
+ memcpy(idx_data, token_indices, idx_bytes);
+
+ int idx_shape[] = {count};
+ mlx_array idx = mlx_array_new_data(idx_data, idx_shape, 1, MLX_INT32);
+ free(idx_data);
+ if (!arr_ok(idx))
+ return NULL;
+
+ mlx_stream S = ctx->base->stream;
+ mlx_array selected = mlx_array_new();
+ mlx_take_axis(&selected, vecs->vectors, idx, 0, S);
+ mlx_array_free(idx);
+ if (!arr_ok(selected)) {
+ mlx_array_free(selected);
+ return NULL;
+ }
+
+ embed_mlx_late_vectors_t *out = (embed_mlx_late_vectors_t *)calloc(1, sizeof(*out));
+ if (!out) {
+ mlx_array_free(selected);
+ return NULL;
+ }
+ out->owner = ctx;
+ out->vectors = selected;
+ out->tokens = count;
+ out->dim = vecs->dim;
+ return out;
+}
+
int embed_mlx_late_maxsim_batch_device(embed_mlx_late_ctx_t *ctx,
const embed_mlx_late_vectors_t *query,
const embed_mlx_late_vectors_t *docs,