vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

ggml : introduce ggml_status (ggml/750)

Browse source 
* using enum as an exit code instead of macros

* update return type from enum to unsigned int

* indentation fix

* compound update
ggml_compute_exit_code -> ggml_status
changed ggml_status from a bit-field type to simple codes
ggml_status to string cast

* ggml_status to string cast

* GGML_CALL was removed

Co-authored-by: slaren <slarengh@gmail.com>

---------

Co-authored-by: slaren <slarengh@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Michael Podvitskiy 2024-03-04 10:05:42 +01:00 • committed by Georgi Gerganov
parent fe52be11e3
commit 9fa2627347
11 changed files with 88 additions and 63 deletions
Download patch file Download diff file Expand all files Collapse all files

39
ggml-backend.c
View file

@ -262,11 +262,11 @@ void ggml_backend_graph_plan_free(ggml_backend_t backend, ggml_backend_graph_pla
backend->iface.graph_plan_free(backend, plan);
}
void ggml_backend_graph_plan_compute(ggml_backend_t backend, ggml_backend_graph_plan_t plan) {
backend->iface.graph_plan_compute(backend, plan);
enum ggml_status ggml_backend_graph_plan_compute(ggml_backend_t backend, ggml_backend_graph_plan_t plan) {
return backend->iface.graph_plan_compute(backend, plan);
}
bool ggml_backend_graph_compute(ggml_backend_t backend, struct ggml_cgraph * cgraph) {
enum ggml_status ggml_backend_graph_compute(ggml_backend_t backend, struct ggml_cgraph * cgraph) {
return backend->iface.graph_compute(backend, cgraph);
}
@ -732,15 +732,15 @@ GGML_CALL static void ggml_backend_cpu_graph_plan_free(ggml_backend_t backend, g
GGML_UNUSED(backend);
}
GGML_CALL static void ggml_backend_cpu_graph_plan_compute(ggml_backend_t backend, ggml_backend_graph_plan_t plan) {
GGML_CALL static enum ggml_status ggml_backend_cpu_graph_plan_compute(ggml_backend_t backend, ggml_backend_graph_plan_t plan) {
struct ggml_backend_plan_cpu * cpu_plan = (struct ggml_backend_plan_cpu *)plan;
ggml_graph_compute(&cpu_plan->cgraph, &cpu_plan->cplan);
return ggml_graph_compute(&cpu_plan->cgraph, &cpu_plan->cplan);
GGML_UNUSED(backend);
}
GGML_CALL static bool ggml_backend_cpu_graph_compute(ggml_backend_t backend, struct ggml_cgraph * cgraph) {
GGML_CALL static enum ggml_status ggml_backend_cpu_graph_compute(ggml_backend_t backend, struct ggml_cgraph * cgraph) {
struct ggml_backend_cpu_context * cpu_ctx = (struct ggml_backend_cpu_context *)backend->context;
struct ggml_cplan cplan = ggml_graph_plan(cgraph, cpu_ctx->n_threads);
@ -755,8 +755,7 @@ GGML_CALL static bool ggml_backend_cpu_graph_compute(ggml_backend_t backend, str
cplan.abort_callback = cpu_ctx->abort_callback;
cplan.abort_callback_data = cpu_ctx->abort_callback_data;
ggml_graph_compute(cgraph, &cplan);
return true;
return ggml_graph_compute(cgraph, &cplan);
}
GGML_CALL static bool ggml_backend_cpu_supports_op(ggml_backend_t backend, const struct ggml_tensor * op) {
@ -1437,7 +1436,7 @@ static bool ggml_backend_sched_alloc_splits(ggml_backend_sched_t sched) {
return true;
}
static bool ggml_backend_sched_compute_splits(ggml_backend_sched_t sched) {
static enum ggml_status ggml_backend_sched_compute_splits(ggml_backend_sched_t sched) {
uint64_t copy_us[GGML_MAX_BACKENDS] = {0};
uint64_t compute_us[GGML_MAX_BACKENDS] = {0};
@ -1472,8 +1471,9 @@ static bool ggml_backend_sched_compute_splits(ggml_backend_sched_t sched) {
uint64_t compute_start_us = ggml_time_us();
if (!sched->callback_eval) {
if (!ggml_backend_graph_compute(split_backend, &split->graph)) {
return false;
enum ggml_status ec = ggml_backend_graph_compute(split_backend, &split->graph);
if (ec != GGML_STATUS_SUCCESS) {
return ec;
}
//ggml_backend_synchronize(split_backend); // necessary to measure compute time
} else {
@ -1494,8 +1494,9 @@ static bool ggml_backend_sched_compute_splits(ggml_backend_sched_t sched) {
struct ggml_cgraph gv = ggml_graph_view(&split->graph, j0, j1 + 1);
if (!ggml_backend_graph_compute(split_backend, &gv)) {
return false;
enum ggml_status ec = ggml_backend_graph_compute(split_backend, &gv);
if (ec != GGML_STATUS_SUCCESS) {
return ec;
}
if (need && !sched->callback_eval(t, false, sched->callback_eval_user_data)) {
@ -1519,7 +1520,7 @@ static bool ggml_backend_sched_compute_splits(ggml_backend_sched_t sched) {
}
#endif
return true;
return GGML_STATUS_SUCCESS;
}
ggml_backend_sched_t ggml_backend_sched_new(ggml_backend_t * backends, ggml_backend_buffer_type_t * bufts, int n_backends, size_t graph_size) {
@ -1581,7 +1582,7 @@ bool ggml_backend_sched_reserve(ggml_backend_sched_t sched, struct ggml_cgraph *
return true;
}
bool ggml_backend_sched_graph_compute(ggml_backend_sched_t sched, struct ggml_cgraph * graph) {
enum ggml_status ggml_backend_sched_graph_compute(ggml_backend_sched_t sched, struct ggml_cgraph * graph) {
GGML_ASSERT((int)sched->hash_set.size >= graph->n_nodes + GGML_MAX_SPLITS*GGML_MAX_SPLIT_INPUTS);
if (!sched->is_reset) {
@ -1590,14 +1591,10 @@ bool ggml_backend_sched_graph_compute(ggml_backend_sched_t sched, struct ggml_cg
ggml_backend_sched_split_graph(sched, graph);
if (!ggml_backend_sched_alloc_splits(sched)) {
return false;
return GGML_STATUS_ALLOC_FAILED;
}
if (!ggml_backend_sched_compute_splits(sched)) {
return false;
}
return true;
return ggml_backend_sched_compute_splits(sched);
}
void ggml_backend_sched_set_eval_callback(ggml_backend_sched_t sched, ggml_backend_sched_eval_callback callback, void * user_data) {