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

metal : add solve_tri (#19302)

Browse Source
This commit is contained in:
Georgi Gerganov
2026-02-03 23:43:14 +02:00
committed by GitHub
parent 6a9bf2f788
commit 44008ce8f9
7 changed files with 200 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-metal/ggml-metal-ops.cpp
+61
View File
@@ -341,6 +341,10 @@ static int ggml_metal_op_encode_impl(ggml_metal_op_t ctx, int idx) {
{
n_fuse = ggml_metal_op_rwkv(ctx, idx);
} break;
case GGML_OP_SOLVE_TRI:
{
n_fuse = ggml_metal_op_solve_tri(ctx, idx);
} break;
case GGML_OP_MUL_MAT:
{
n_fuse = ggml_metal_op_mul_mat(ctx, idx);
@@ -1557,6 +1561,63 @@ int ggml_metal_op_rwkv(ggml_metal_op_t ctx, int idx) {
return 1;
}
int ggml_metal_op_solve_tri(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;
GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
GGML_TENSOR_LOCALS( int32_t, ne1, op->src[1], ne);
GGML_TENSOR_LOCALS(uint64_t, nb1, op->src[1], nb);
GGML_TENSOR_LOCALS( int32_t, ne, op, ne);
GGML_TENSOR_LOCALS(uint64_t, nb, op, nb);
ggml_metal_kargs_solve_tri args = {
/*.ne00 =*/ ne00,
/*.ne01 =*/ ne01,
/*.ne02 =*/ ne02,
/*.ne03 =*/ ne03,
/*.nb00 =*/ nb00,
/*.nb01 =*/ nb01,
/*.nb02 =*/ nb02,
/*.nb03 =*/ nb03,
/*.ne10 =*/ ne10,
/*.ne11 =*/ ne11,
/*.ne12 =*/ ne12,
/*.ne13 =*/ ne13,
/*.nb10 =*/ nb10,
/*.nb11 =*/ nb11,
/*.nb12 =*/ nb12,
/*.nb13 =*/ nb13,
/*.ne0 =*/ ne0,
/*.ne1 =*/ ne1,
/*.ne2 =*/ ne2,
/*.ne3 =*/ ne3,
/*.nb0 =*/ nb0,
/*.nb1 =*/ nb1,
/*.nb2 =*/ nb2,
/*.nb3 =*/ nb3,
};
auto pipeline = ggml_metal_library_get_pipeline_solve_tri(lib, op);
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->src[1]), 2);
ggml_metal_encoder_set_buffer (enc, ggml_metal_get_buffer_id(op), 3);
const int nsg = pipeline.nsg;
ggml_metal_encoder_set_threadgroup_memory_size(enc, pipeline.smem, 0);
ggml_metal_encoder_dispatch_threadgroups(enc, (ne10 + nsg - 1)/nsg, ne02, ne03, 32, nsg, 1);
return 1;
}
int ggml_metal_op_cpy(ggml_metal_op_t ctx, int idx) {
ggml_tensor * op = ctx->node(idx);