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

Use GGML_LOG instead of GGML_PRINT

Browse source
This commit is contained in:
Mason M 2024-10-01 16:42:18 -03:00
parent 31c3b89c88
commit a731280e61
1 changed files with 23 additions and 25 deletions
Download patch file Download diff file Expand all files Collapse all files

48
ggml/src/ggml.c
View file

@ -357,25 +357,23 @@ void ggml_log_callback_default(enum ggml_log_level level, const char * text, voi
} }
#if (GGML_DEBUG >= 1) #if (GGML_DEBUG >= 1)
#define GGML_PRINT_DEBUG(...) printf(__VA_ARGS__) #define GGML_PRINT_DEBUG(...) GGML_LOG_DEBUG(__VA_ARGS__)
#else #else
#define GGML_PRINT_DEBUG(...) #define GGML_PRINT_DEBUG(...)
#endif #endif
#if (GGML_DEBUG >= 5) #if (GGML_DEBUG >= 5)
#define GGML_PRINT_DEBUG_5(...) printf(__VA_ARGS__) #define GGML_PRINT_DEBUG_5(...) GGML_LOG_DEBUG(__VA_ARGS__)
#else #else
#define GGML_PRINT_DEBUG_5(...) #define GGML_PRINT_DEBUG_5(...)
#endif #endif
#if (GGML_DEBUG >= 10) #if (GGML_DEBUG >= 10)
#define GGML_PRINT_DEBUG_10(...) printf(__VA_ARGS__) #define GGML_PRINT_DEBUG_10(...) GGML_LOG_DEBUG(__VA_ARGS__)
#else #else
#define GGML_PRINT_DEBUG_10(...) #define GGML_PRINT_DEBUG_10(...)
#endif #endif
#define GGML_PRINT(...) printf(__VA_ARGS__)
// //
// end of logging block // end of logging block
// //
@ -392,7 +390,7 @@ void ggml_log_callback_default(enum ggml_log_level level, const char * text, voi
#else #else
inline static void * ggml_aligned_malloc(size_t size) { inline static void * ggml_aligned_malloc(size_t size) {
if (size == 0) { if (size == 0) {
GGML_PRINT("WARNING: Behavior may be unexpected when allocating 0 bytes for ggml_aligned_malloc!\n"); GGML_LOG_WARN("Behavior may be unexpected when allocating 0 bytes for ggml_aligned_malloc!\n");
return NULL; return NULL;
} }
void * aligned_memory = NULL; void * aligned_memory = NULL;
@ -414,7 +412,7 @@ inline static void * ggml_aligned_malloc(size_t size) {
error_desc = "insufficient memory"; error_desc = "insufficient memory";
break; break;
} }
GGML_PRINT("%s: %s (attempted to allocate %6.2f MB)\n", __func__, error_desc, size/(1024.0*1024.0)); GGML_LOG_ERROR("%s: %s (attempted to allocate %6.2f MB)\n", __func__, error_desc, size/(1024.0*1024.0));
GGML_ABORT("fatal error"); GGML_ABORT("fatal error");
return NULL; return NULL;
} }
@ -430,12 +428,12 @@ inline static void * ggml_aligned_malloc(size_t size) {
inline static void * ggml_malloc(size_t size) { inline static void * ggml_malloc(size_t size) {
if (size == 0) { if (size == 0) {
GGML_PRINT("WARNING: Behavior may be unexpected when allocating 0 bytes for ggml_malloc!\n"); GGML_LOG_WARN("Behavior may be unexpected when allocating 0 bytes for ggml_malloc!\n");
return NULL; return NULL;
} }
void * result = malloc(size); void * result = malloc(size);
if (result == NULL) { if (result == NULL) {
GGML_PRINT("%s: failed to allocate %6.2f MB\n", __func__, size/(1024.0*1024.0)); GGML_LOG_ERROR("%s: failed to allocate %6.2f MB\n", __func__, size/(1024.0*1024.0));
GGML_ABORT("fatal error"); GGML_ABORT("fatal error");
} }
return result; return result;
@ -444,12 +442,12 @@ inline static void * ggml_malloc(size_t size) {
// calloc // calloc
inline static void * ggml_calloc(size_t num, size_t size) { inline static void * ggml_calloc(size_t num, size_t size) {
if (num == 0 || size == 0) { if (num == 0 || size == 0) {
GGML_PRINT("WARNING: Behavior may be unexpected when allocating 0 bytes for ggml_calloc!\n"); GGML_LOG_WARN("Behavior may be unexpected when allocating 0 bytes for ggml_calloc!\n");
return NULL; return NULL;
} }
void * result = calloc(num, size); void * result = calloc(num, size);
if (result == NULL) { if (result == NULL) {
GGML_PRINT("%s: failed to allocate %6.2f MB\n", __func__, size/(1024.0*1024.0)); GGML_LOG_ERROR("%s: failed to allocate %6.2f MB\n", __func__, size/(1024.0*1024.0));
GGML_ABORT("fatal error"); GGML_ABORT("fatal error");
} }
return result; return result;
@ -3384,7 +3382,7 @@ void ggml_numa_init(enum ggml_numa_strategy numa_flag) {
if (fptr != NULL) { if (fptr != NULL) {
char buf[42]; char buf[42];
if (fgets(buf, sizeof(buf), fptr) && strncmp(buf, "0\n", sizeof(buf)) != 0) { if (fgets(buf, sizeof(buf), fptr) && strncmp(buf, "0\n", sizeof(buf)) != 0) {
GGML_PRINT("WARNING: /proc/sys/kernel/numa_balancing is enabled, this has been observed to impair performance\n"); GGML_LOG_WARN("/proc/sys/kernel/numa_balancing is enabled, this has been observed to impair performance\n");
} }
fclose(fptr); fclose(fptr);
} }
@ -3402,21 +3400,21 @@ bool ggml_is_numa(void) {
//////////////////////////////////////////////////////////////////////////////// ////////////////////////////////////////////////////////////////////////////////
void ggml_print_object(const struct ggml_object * obj) { void ggml_print_object(const struct ggml_object * obj) {
GGML_PRINT(" - ggml_object: type = %d, offset = %zu, size = %zu, next = %p\n", GGML_LOG_INFO(" - ggml_object: type = %d, offset = %zu, size = %zu, next = %p\n",
obj->type, obj->offs, obj->size, (const void *) obj->next); obj->type, obj->offs, obj->size, (const void *) obj->next);
} }
void ggml_print_objects(const struct ggml_context * ctx) { void ggml_print_objects(const struct ggml_context * ctx) {
struct ggml_object * obj = ctx->objects_begin; struct ggml_object * obj = ctx->objects_begin;
GGML_PRINT("%s: objects in context %p:\n", __func__, (const void *) ctx); GGML_LOG_INFO("%s: objects in context %p:\n", __func__, (const void *) ctx);
while (obj != NULL) { while (obj != NULL) {
ggml_print_object(obj); ggml_print_object(obj);
obj = obj->next; obj = obj->next;
} }
GGML_PRINT("%s: --- end ---\n", __func__); GGML_LOG_INFO("%s: --- end ---\n", __func__);
} }
GGML_CALL int64_t ggml_nelements(const struct ggml_tensor * tensor) { GGML_CALL int64_t ggml_nelements(const struct ggml_tensor * tensor) {
@ -3999,7 +3997,7 @@ static struct ggml_object * ggml_new_object(struct ggml_context * ctx, enum ggml
struct ggml_object * const obj_new = (struct ggml_object *)(mem_buffer + cur_end); struct ggml_object * const obj_new = (struct ggml_object *)(mem_buffer + cur_end);
if (cur_end + size_needed + GGML_OBJECT_SIZE > ctx->mem_size) { if (cur_end + size_needed + GGML_OBJECT_SIZE > ctx->mem_size) {
GGML_PRINT("%s: not enough space in the context's memory pool (needed %zu, available %zu)\n", GGML_LOG_WARN("%s: not enough space in the context's memory pool (needed %zu, available %zu)\n",
__func__, cur_end + size_needed + GGML_OBJECT_SIZE, ctx->mem_size); __func__, cur_end + size_needed + GGML_OBJECT_SIZE, ctx->mem_size);
assert(false); assert(false);
return NULL; return NULL;
@ -4063,7 +4061,7 @@ static struct ggml_tensor * ggml_new_tensor_impl(
if (ctx->scratch.data != NULL) { if (ctx->scratch.data != NULL) {
// allocate tensor data in the scratch buffer // allocate tensor data in the scratch buffer
if (ctx->scratch.offs + data_size > ctx->scratch.size) { if (ctx->scratch.offs + data_size > ctx->scratch.size) {
GGML_PRINT("%s: not enough space in the scratch memory pool (needed %zu, available %zu)\n", GGML_LOG_WARN("%s: not enough space in the scratch memory pool (needed %zu, available %zu)\n",
__func__, ctx->scratch.offs + data_size, ctx->scratch.size); __func__, ctx->scratch.offs + data_size, ctx->scratch.size);
assert(false); assert(false);
return NULL; return NULL;
@ -20046,7 +20044,7 @@ enum ggml_status ggml_graph_compute(struct ggml_cgraph * cgraph, struct ggml_cpl
} }
#else #else
if (n_threads > threadpool->n_threads_max) { if (n_threads > threadpool->n_threads_max) {
GGML_PRINT("WARNING: cplan requested more threads (%d) than available (%d)\n", n_threads, threadpool->n_threads_max); GGML_LOG_WARN("cplan requested more threads (%d) than available (%d)\n", n_threads, threadpool->n_threads_max);
n_threads = threadpool->n_threads_max; n_threads = threadpool->n_threads_max;
} }
@ -20585,30 +20583,30 @@ struct ggml_cgraph * ggml_graph_import(const char * fname, struct ggml_context *
} }
void ggml_graph_print(const struct ggml_cgraph * cgraph) { void ggml_graph_print(const struct ggml_cgraph * cgraph) {
GGML_PRINT("=== GRAPH ===\n"); GGML_LOG_INFO("=== GRAPH ===\n");
GGML_PRINT("n_nodes = %d\n", cgraph->n_nodes); GGML_LOG_INFO("n_nodes = %d\n", cgraph->n_nodes);
for (int i = 0; i < cgraph->n_nodes; i++) { for (int i = 0; i < cgraph->n_nodes; i++) {
struct ggml_tensor * node = cgraph->nodes[i]; struct ggml_tensor * node = cgraph->nodes[i];
GGML_PRINT(" - %3d: [ %5" PRId64 ", %5" PRId64 ", %5" PRId64 "] %16s %s\n", GGML_LOG_INFO(" - %3d: [ %5" PRId64 ", %5" PRId64 ", %5" PRId64 "] %16s %s\n",
i, i,
node->ne[0], node->ne[1], node->ne[2], node->ne[0], node->ne[1], node->ne[2],
ggml_op_name(node->op), (node->flags & GGML_TENSOR_FLAG_PARAM) ? "x" : node->grad ? "g" : " "); ggml_op_name(node->op), (node->flags & GGML_TENSOR_FLAG_PARAM) ? "x" : node->grad ? "g" : " ");
} }
GGML_PRINT("n_leafs = %d\n", cgraph->n_leafs); GGML_LOG_INFO("n_leafs = %d\n", cgraph->n_leafs);
for (int i = 0; i < cgraph->n_leafs; i++) { for (int i = 0; i < cgraph->n_leafs; i++) {
struct ggml_tensor * node = cgraph->leafs[i]; struct ggml_tensor * node = cgraph->leafs[i];
GGML_PRINT(" - %3d: [ %5" PRId64 ", %5" PRId64 "] %8s %16s\n", GGML_LOG_INFO(" - %3d: [ %5" PRId64 ", %5" PRId64 "] %8s %16s\n",
i, i,
node->ne[0], node->ne[1], node->ne[0], node->ne[1],
ggml_op_name(node->op), ggml_op_name(node->op),
ggml_get_name(node)); ggml_get_name(node));
} }
GGML_PRINT("========================================\n"); GGML_LOG_INFO("========================================\n");
} }
// check if node is part of the graph // check if node is part of the graph
@ -20779,7 +20777,7 @@ void ggml_graph_dump_dot(const struct ggml_cgraph * gb, const struct ggml_cgraph
fclose(fp); fclose(fp);
GGML_PRINT("%s: dot -Tpng %s -o %s.png && open %s.png\n", __func__, filename, filename, filename); GGML_LOG_INFO("%s: dot -Tpng %s -o %s.png && open %s.png\n", __func__, filename, filename, filename);
} }
//////////////////////////////////////////////////////////////////////////////// ////////////////////////////////////////////////////////////////////////////////