Skip to content

ggml: reserve in gguf_writer and added const pointers as params #11297

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 1 commit into
base: master
Choose a base branch
from
+80 −79
Open

ggml: reserve in gguf_writer and added const pointers as params #11297

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
6 changes: 3 additions & 3 deletions ggml/include/ggml-alloc.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -46,17 +46,17 @@ GGML_API void ggml_tallocr_alloc(struct ggml_tallocr * talloc, st
typedef struct ggml_gallocr * ggml_gallocr_t;

GGML_API ggml_gallocr_t ggml_gallocr_new(ggml_backend_buffer_type_t buft);
GGML_API ggml_gallocr_t ggml_gallocr_new_n(ggml_backend_buffer_type_t * bufts, int n_bufs);
GGML_API ggml_gallocr_t ggml_gallocr_new_n(const ggml_backend_buffer_type_t * bufts, int n_bufs);
GGML_API void ggml_gallocr_free(ggml_gallocr_t galloc);

// pre-allocate buffers from a measure graph - does not allocate or modify the graph
// call with a worst-case graph to avoid buffer reallocations
// not strictly required for single buffer usage: ggml_gallocr_alloc_graph will reallocate the buffers automatically if needed
// returns false if the buffer allocation failed
GGML_API bool ggml_gallocr_reserve(ggml_gallocr_t galloc, struct ggml_cgraph * graph);
GGML_API bool ggml_gallocr_reserve(ggml_gallocr_t galloc, const struct ggml_cgraph * graph);
GGML_API bool ggml_gallocr_reserve_n(
ggml_gallocr_t galloc,
struct ggml_cgraph * graph,
const struct ggml_cgraph * graph,
const int * node_buffer_ids,
const int * leaf_buffer_ids);

Expand Down
31 changes: 16 additions & 15 deletions ggml/include/ggml.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -698,7 +698,7 @@ extern "C" {

GGML_API size_t ggml_used_mem(const struct ggml_context * ctx);

GGML_API bool ggml_get_no_alloc(struct ggml_context * ctx);
GGML_API bool ggml_get_no_alloc(const struct ggml_context * ctx);
GGML_API void ggml_set_no_alloc(struct ggml_context * ctx, bool no_alloc);

GGML_API void * ggml_get_mem_buffer (const struct ggml_context * ctx);
Expand Down Expand Up @@ -745,7 +745,7 @@ extern "C" {
// Context tensor enumeration and lookup
GGML_API struct ggml_tensor * ggml_get_first_tensor(const struct ggml_context * ctx);
GGML_API struct ggml_tensor * ggml_get_next_tensor (const struct ggml_context * ctx, struct ggml_tensor * tensor);
GGML_API struct ggml_tensor * ggml_get_tensor(struct ggml_context * ctx, const char * name);
GGML_API struct ggml_tensor * ggml_get_tensor(const struct ggml_context * ctx, const char * name);

// Converts a flat index into coordinates
GGML_API void ggml_unravel_index(const struct ggml_tensor * tensor, int64_t i, int64_t * i0, int64_t * i1, int64_t * i2, int64_t * i3);
Expand All @@ -763,7 +763,7 @@ extern "C" {
// Tensor flags
GGML_API void ggml_set_input(struct ggml_tensor * tensor);
GGML_API void ggml_set_output(struct ggml_tensor * tensor);
GGML_API void ggml_set_param(struct ggml_context * ctx, struct ggml_tensor * tensor);
GGML_API void ggml_set_param(const struct ggml_context * ctx, struct ggml_tensor * tensor);
GGML_API void ggml_set_loss(struct ggml_tensor * tensor);

//
Expand Down Expand Up @@ -927,13 +927,13 @@ extern "C" {
GGML_API struct ggml_tensor * ggml_repeat(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b);
const struct ggml_tensor * b);

// sums repetitions in a into shape of b
GGML_API struct ggml_tensor * ggml_repeat_back(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b);
const struct ggml_tensor * b);

// concat a and b along dim
// used in stable-diffusion
Expand Down Expand Up @@ -1243,7 +1243,7 @@ extern "C" {
GGML_API struct ggml_tensor * ggml_reshape(
struct ggml_context * ctx,
struct ggml_tensor * a,
struct ggml_tensor * b);
const struct ggml_tensor * b);

// return view(a)
// TODO: when we start computing gradient, make a copy instead of view
Expand Down Expand Up @@ -1335,7 +1335,7 @@ extern "C" {
struct ggml_context * ctx,
struct ggml_tensor * a, // gradients of ggml_get_rows result
struct ggml_tensor * b, // row indices
struct ggml_tensor * c); // data for ggml_get_rows, only used for its shape
const struct ggml_tensor * c); // data for ggml_get_rows, only used for its shape

GGML_API struct ggml_tensor * ggml_diag(
struct ggml_context * ctx,
Expand Down Expand Up @@ -1563,7 +1563,7 @@ extern "C" {
struct ggml_context * ctx,
struct ggml_tensor * a, // convolution kernel
struct ggml_tensor * b, // gradient of im2col output
int64_t * ne, // shape of im2col input
const int64_t * ne, // shape of im2col input
int s0, // stride dimension 0
int s1, // stride dimension 1
int p0, // padding dimension 0
Expand Down Expand Up @@ -2062,15 +2062,16 @@ extern "C" {
// graph allocation in a context
GGML_API struct ggml_cgraph * ggml_new_graph (struct ggml_context * ctx); // size = GGML_DEFAULT_GRAPH_SIZE, grads = false
GGML_API struct ggml_cgraph * ggml_new_graph_custom(struct ggml_context * ctx, size_t size, bool grads);
GGML_API struct ggml_cgraph * ggml_graph_dup (struct ggml_context * ctx, struct ggml_cgraph * cgraph);
GGML_API void ggml_graph_cpy (struct ggml_cgraph * src, struct ggml_cgraph * dst);
GGML_API void ggml_graph_reset (struct ggml_cgraph * cgraph); // set regular grads + optimizer momenta to 0, set loss grad to 1
GGML_API struct ggml_cgraph * ggml_graph_dup (struct ggml_context * ctx, const struct ggml_cgraph * cgraph);
GGML_API void ggml_graph_cpy (const struct ggml_cgraph * src, struct ggml_cgraph * dst);
GGML_API void ggml_graph_reset (
const struct ggml_cgraph * cgraph); // set regular grads + optimizer momenta to 0, set loss grad to 1
GGML_API void ggml_graph_clear (struct ggml_cgraph * cgraph);

GGML_API int ggml_graph_size (struct ggml_cgraph * cgraph);
GGML_API struct ggml_tensor * ggml_graph_node (struct ggml_cgraph * cgraph, int i); // if i < 0, returns nodes[n_nodes + i]
GGML_API struct ggml_tensor ** ggml_graph_nodes (struct ggml_cgraph * cgraph);
GGML_API int ggml_graph_n_nodes(struct ggml_cgraph * cgraph);
GGML_API int ggml_graph_size (const struct ggml_cgraph * cgraph);
GGML_API struct ggml_tensor * ggml_graph_node (const struct ggml_cgraph * cgraph, int i); // if i < 0, returns nodes[n_nodes + i]
GGML_API struct ggml_tensor ** ggml_graph_nodes (const struct ggml_cgraph * cgraph);
GGML_API int ggml_graph_n_nodes(const struct ggml_cgraph * cgraph);

GGML_API void ggml_graph_add_node(struct ggml_cgraph * cgraph, struct ggml_tensor * tensor);

Expand Down
18 changes: 9 additions & 9 deletions ggml/src/ggml-alloc.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -377,7 +377,7 @@ struct ggml_gallocr {
int n_leafs;
};

ggml_gallocr_t ggml_gallocr_new_n(ggml_backend_buffer_type_t * bufts, int n_bufs) {
ggml_gallocr_t ggml_gallocr_new_n(const ggml_backend_buffer_type_t * bufts, int n_bufs) {
ggml_gallocr_t galloc = (ggml_gallocr_t)calloc(1, sizeof(struct ggml_gallocr));
GGML_ASSERT(galloc != NULL);

Expand Down Expand Up @@ -563,7 +563,7 @@ static int get_node_buffer_id(const int * node_buffer_ids, int i) {
return node_buffer_ids ? node_buffer_ids[i] : 0;
}

static void ggml_gallocr_alloc_graph_impl(ggml_gallocr_t galloc, struct ggml_cgraph * graph, const int * node_buffer_ids, const int * leaf_buffer_ids) {
static void ggml_gallocr_alloc_graph_impl(ggml_gallocr_t galloc, const struct ggml_cgraph * graph, const int * node_buffer_ids, const int * leaf_buffer_ids) {
// clear hash tables
ggml_hash_set_reset(&galloc->hash_set);
memset(galloc->hash_values, 0, sizeof(struct hash_node) * galloc->hash_set.size);
Expand Down Expand Up @@ -670,7 +670,7 @@ static void ggml_gallocr_alloc_graph_impl(ggml_gallocr_t galloc, struct ggml_cgr
}
}

bool ggml_gallocr_reserve_n(ggml_gallocr_t galloc, struct ggml_cgraph * graph, const int * node_buffer_ids, const int * leaf_buffer_ids) {
bool ggml_gallocr_reserve_n(ggml_gallocr_t galloc, const struct ggml_cgraph * graph, const int * node_buffer_ids, const int * leaf_buffer_ids) {
size_t min_hash_size = graph->n_nodes + graph->n_leafs;
// add 25% margin to avoid hash collisions
min_hash_size += min_hash_size / 4;
Expand Down Expand Up @@ -780,11 +780,11 @@ bool ggml_gallocr_reserve_n(ggml_gallocr_t galloc, struct ggml_cgraph * graph, c
return true;
}

bool ggml_gallocr_reserve(ggml_gallocr_t galloc, struct ggml_cgraph *graph) {
bool ggml_gallocr_reserve(ggml_gallocr_t galloc, const struct ggml_cgraph *graph) {
return ggml_gallocr_reserve_n(galloc, graph, NULL, NULL);
}

static void ggml_gallocr_init_tensor(ggml_gallocr_t galloc, struct ggml_tensor * tensor, struct tensor_alloc * tensor_alloc) {
static void ggml_gallocr_init_tensor(ggml_gallocr_t galloc, struct ggml_tensor * tensor, const struct tensor_alloc * tensor_alloc) {
int buffer_id = tensor_alloc->buffer_id;
assert(tensor->data || tensor->view_src || ggml_backend_buffer_get_alloc_size(galloc->buffers[buffer_id], tensor) <= tensor_alloc->size_max);

Expand Down Expand Up @@ -813,7 +813,7 @@ static void ggml_gallocr_init_tensor(ggml_gallocr_t galloc, struct ggml_tensor *
}
}

static bool ggml_gallocr_node_needs_realloc(ggml_gallocr_t galloc, struct ggml_tensor * node, struct tensor_alloc * talloc) {
static bool ggml_gallocr_node_needs_realloc(ggml_gallocr_t galloc, struct ggml_tensor * node, const struct tensor_alloc * talloc) {
size_t node_size = 0;
if (!node->data && !node->view_src) {
GGML_ASSERT(talloc->buffer_id >= 0); // prevent segfault when misusing the API
Expand All @@ -822,7 +822,7 @@ static bool ggml_gallocr_node_needs_realloc(ggml_gallocr_t galloc, struct ggml_t
return talloc->size_max >= node_size;
}

static bool ggml_gallocr_needs_realloc(ggml_gallocr_t galloc, struct ggml_cgraph * graph) {
static bool ggml_gallocr_needs_realloc(ggml_gallocr_t galloc, const struct ggml_cgraph * graph) {
if (galloc->n_nodes != graph->n_nodes) {
#ifndef NDEBUG
GGML_LOG_DEBUG("%s: graph has different number of nodes\n", __func__);
Expand Down Expand Up @@ -933,8 +933,8 @@ size_t ggml_gallocr_get_buffer_size(ggml_gallocr_t galloc, int buffer_id) {

// utils

static bool alloc_tensor_range(struct ggml_context * ctx,
struct ggml_tensor * first, struct ggml_tensor * last,
static bool alloc_tensor_range(const struct ggml_context * ctx,
struct ggml_tensor * first, const struct ggml_tensor * last,
ggml_backend_buffer_type_t buft, size_t size,
ggml_backend_buffer_t ** buffers, size_t * n_buffers) {
ggml_backend_buffer_t buffer = ggml_backend_buft_alloc_buffer(buft, size);
Expand Down
2 changes: 1 addition & 1 deletion ggml/src/ggml-backend-reg.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -124,7 +124,7 @@ static void * dl_get_sym(dl_handle * handle, const char * name) {
using dl_handle = void;

struct dl_handle_deleter {
void operator()(void * handle) {
void operator()(void * handle) const {
dlclose(handle);
}
};
Expand Down
6 changes: 3 additions & 3 deletions ggml/src/ggml-impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -188,13 +188,13 @@ struct ggml_hash_set {
};

struct ggml_hash_set ggml_hash_set_new(size_t size);
void ggml_hash_set_free(struct ggml_hash_set * hash_set);
void ggml_hash_set_free(const struct ggml_hash_set * hash_set);

// returns the minimum size for a hash set that can hold min_sz elements
size_t ggml_hash_size(size_t min_sz);

// remove all elements from the hash set
void ggml_hash_set_reset(struct ggml_hash_set * hash_set);
void ggml_hash_set_reset(const struct ggml_hash_set * hash_set);

// returns true if key is in the hash set
static bool ggml_hash_contains(const struct ggml_hash_set * hash_set, struct ggml_tensor * key);
Expand Down Expand Up @@ -302,7 +302,7 @@ struct ggml_cgraph {
// returns a slice of cgraph with nodes [i0, i1)
// the slice does not have leafs or gradients
// if you need the gradients, get them from the original graph
struct ggml_cgraph ggml_graph_view(struct ggml_cgraph * cgraph, int i0, int i1);
struct ggml_cgraph ggml_graph_view(const struct ggml_cgraph * cgraph, int i0, int i1);

// Memory allocation

Expand Down
Loading
Loading