Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

metal: add col2im_1d op (f32/f16/bf16) (#25176)

Browse Source 
* metal: add col2im_1d op (f32/f16/bf16)

Gather kernel mirroring the CPU/CUDA path: each output (t_out, oc)
reads its ceil(K/s0) source columns with an F32 accumulator, a single
write and no atomics. One thread per output element, 256 per
threadgroup.

* metal: check dst contiguity and type match in supports_op for COL2IM_1D

Align the GGML_OP_COL2IM_1D predicate with the CPU, CUDA, and Vulkan
backends: the kernel writes dst with linear indexing and assumes the
same type as src0, so supports_op must also require a contiguous dst
and op->type == op->src[0]->type.

* Update ggml/src/ggml-metal/ggml-metal.metal

Co-authored-by: YiChen Lv <63285796+forforever73@users.noreply.github.com>

---------

Co-authored-by: YiChen Lv <63285796+forforever73@users.noreply.github.com>
This commit is contained in:
Pascal
2026-07-06 20:47:36 +02:00
committed by GitHub
co-authored by YiChen Lv
parent 74976e1aef
commit f36e5c348b
7 changed files with 125 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-metal/ggml-metal-ops.cpp
+45
View File
@@ -395,6 +395,10 @@ static int ggml_metal_op_encode_impl(ggml_metal_op_t ctx, int idx) {
{
n_fuse = ggml_metal_op_conv_transpose_2d(ctx, idx);
} break;
case GGML_OP_COL2IM_1D:
{
n_fuse = ggml_metal_op_col2im_1d(ctx, idx);
} break;
case GGML_OP_CONV_3D:
{
n_fuse = ggml_metal_op_conv_3d(ctx, idx);
@@ -3854,6 +3858,47 @@ int ggml_metal_op_conv_transpose_1d(ggml_metal_op_t ctx, int idx) {
return 1;
}
int ggml_metal_op_col2im_1d(ggml_metal_op_t ctx, int idx) {
ggml_tensor * op = ctx->node(idx);
ggml_metal_library_t lib = ctx->lib;
ggml_metal_encoder_t enc = ctx->enc;
const int32_t s0 = ((const int32_t *)(op->op_params))[0];
const int32_t OC = ((const int32_t *)(op->op_params))[1];
const int32_t p0 = ((const int32_t *)(op->op_params))[2];
const int32_t K_OC = (int32_t) op->src[0]->ne[0];
const int32_t T_in = (int32_t) op->src[0]->ne[1];
const int32_t K = K_OC / OC;
const int32_t T_out = (int32_t) op->ne[0];
ggml_metal_kargs_col2im_1d args = {
/*.T_in =*/ T_in,
/*.T_out =*/ T_out,
/*.OC =*/ OC,
/*.K =*/ K,
/*.K_OC =*/ K_OC,
/*.s0 =*/ s0,
/*.p0 =*/ p0,
};
auto pipeline = ggml_metal_library_get_pipeline_col2im_1d(lib, op);
const int total = T_out * OC;
const int nth = 256;
const int ntg = (total + nth - 1) / nth;
ggml_metal_encoder_set_pipeline(enc, pipeline);
ggml_metal_encoder_set_bytes (enc, &args, sizeof(args), 0);
ggml_metal_encoder_set_buffer (enc, ggml_metal_get_buffer_id(op->src[0]), 1);
ggml_metal_encoder_set_buffer (enc, ggml_metal_get_buffer_id(op), 2);
ggml_metal_encoder_dispatch_threadgroups(enc, ntg, 1, 1, nth, 1, 1);
return 1;
}
int ggml_metal_op_conv_transpose_2d(ggml_metal_op_t ctx, int idx) {
ggml_tensor * op = ctx->node(idx);