1#include "backend-dispatched.h"
2#include "backend-virgl-apir.h"
3#include "ggml-backend-impl.h"
4#include "ggml-backend.h"
5#include "ggml-impl.h"
6#include "shared/apir_backend.h"
7
8#include <cstdint>
9
10uint32_t backend_backend_graph_compute(apir_encoder * enc, apir_decoder * dec, virgl_apir_context * ctx) {
11 GGML_UNUSED(ctx);
12 GGML_UNUSED(enc);
13
14 static bool async_backend_initialized = false;
15 static bool async_backend;
16
17 if (!async_backend_initialized) {
18 ggml_backend_dev_props props;
19
20 dev->iface.get_props(dev, &props);
21 async_backend = props.caps.async;
22 async_backend_initialized = true;
23 }
24
25 uint32_t shmem_res_id;
26 apir_decode_virtgpu_shmem_res_id(dec, &shmem_res_id);
27
28 const void * shmem_data = ctx->iface->get_shmem_ptr(ctx->ctx_id, shmem_res_id);
29 if (!shmem_data) {
30 GGML_LOG_ERROR(GGML_VIRTGPU_BCK "%s: Couldn't get the shmem addr from virgl\n", __func__);
31 apir_decoder_set_fatal(dec);
32 return 1;
33 }
34 size_t cgraph_size;
35 apir_decode_size_t(dec, &cgraph_size);
36
37 apir_decoder secondary_dec = apir_new_decoder((const char *) shmem_data, cgraph_size);
38
39 ggml_cgraph * cgraph = apir_decode_ggml_cgraph(&secondary_dec, cgraph_size);
40
41 ggml_status status;
42#if APIR_BACKEND_CHECK_SUPPORTS_OP == 1
43 for (int idx = 0; idx < cgraph->n_nodes; idx++) {
44 ggml_tensor * op = ggml_graph_node(cgraph, idx);
45 if (dev->iface.supports_op(dev, op)) {
46 continue;
47 }
48 GGML_LOG_ERROR(GGML_VIRTGPU_BCK "%s: Graph node %d (%s) not supported by the backend\n", idx, ggml_op_desc(op));
49
50 status = GGML_STATUS_ABORTED;
51 apir_encode_ggml_status(enc, &status);
52
53 return 0;
54 }
55#endif
56 status = bck->iface.graph_compute(bck, cgraph);
57
58 if (async_backend) {
59 bck->iface.synchronize(bck);
60 }
61
62 apir_encode_ggml_status(enc, &status);
63
64 return 0;
65}