Skip to content

SYCL: Add GGML_OP_MEAN operator support #16009

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
+34 −0
Open

SYCL: Add GGML_OP_MEAN operator support #16009

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ff3becb
SYCL: Add GGML_OP_MEAN operator support
yael-works Sep 15, 2025
8393ddc
SYCL: Fix indentation for ggml_sycl_op_mean implementation
yael-works Sep 16, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
34 changes: 34 additions & 0 deletions ggml/src/ggml-sycl/ggml-sycl.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2127,6 +2127,30 @@ inline void ggml_sycl_op_sum_rows(ggml_backend_sycl_context & ctx, ggml_tensor *
sum_rows_f32_sycl(src0_dd, dst_dd, ncols, nrows, main_stream);
}

inline void ggml_sycl_op_mean(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
GGML_ASSERT(dst->src[0]->type == GGML_TYPE_F32);
GGML_ASSERT(dst->type == GGML_TYPE_F32);

dpct::queue_ptr main_stream = ctx.stream();
SYCL_CHECK(ggml_sycl_set_device(ctx.device));

const float * src0_dd = static_cast<const float *>(dst->src[0]->data);
float * dst_dd = static_cast<float *>(dst->data);

const int64_t ncols = dst->src[0]->ne[0];
const int64_t nrows = ggml_nrows(dst->src[0]);

sum_rows_f32_sycl(src0_dd, dst_dd, ncols, nrows, main_stream);

main_stream->parallel_for(
sycl::range<1>(nrows),
[=](sycl::id<1> row) {
dst_dd[row] /= ncols;
}
);
}


inline void ggml_sycl_op_argsort(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
GGML_ASSERT(dst->src[0]->type == GGML_TYPE_F32);
GGML_ASSERT(dst->type == GGML_TYPE_I32);
Expand Down Expand Up @@ -3510,6 +3534,12 @@ static void ggml_sycl_sum_rows(ggml_backend_sycl_context & ctx, ggml_tensor * ds
ggml_sycl_op_sum_rows(ctx, dst);
}

static void ggml_sycl_mean(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
scope_op_debug_print scope_dbg_print(__func__, dst, /*num_src=*/1);
GGML_ASSERT(ggml_is_contiguous(dst->src[0]));
ggml_sycl_op_mean(ctx, dst);
}

static void ggml_sycl_argsort(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
scope_op_debug_print scope_dbg_print(__func__, dst, /*num_src=*/1);
GGML_ASSERT(ggml_is_contiguous(dst->src[0]));
Expand Down Expand Up @@ -3756,6 +3786,9 @@ static bool ggml_sycl_compute_forward(ggml_backend_sycl_context & ctx, struct gg
case GGML_OP_SUM_ROWS:
ggml_sycl_sum_rows(ctx, dst);
break;
case GGML_OP_MEAN:
ggml_sycl_mean(ctx, dst);
break;
case GGML_OP_ARGSORT:
ggml_sycl_argsort(ctx, dst);
break;
Expand Down Expand Up @@ -4406,6 +4439,7 @@ static bool ggml_backend_sycl_device_supports_op(ggml_backend_dev_t dev, const g
return op->src[0]->type == GGML_TYPE_F32 && op->op_params[0] == GGML_SCALE_MODE_NEAREST;
case GGML_OP_SUM:
case GGML_OP_SUM_ROWS:
case GGML_OP_MEAN:
case GGML_OP_ARGSORT:
return ggml_is_contiguous(op->src[0]);
case GGML_OP_POOL_2D:
Expand Down
Loading