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

ggml/kompute: Rename ggml_kompute_context to ggml_backend_kompute_context

Browse source 
Signed-off-by: Ming Xie <xieming@kylinos.cn>
This commit is contained in:
Ming Xie 2024-08-21 10:54:09 +08:00 • committed by Feng Jiang
parent fc54ef0d1c
commit 3666c861d4
1 changed files with 11 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

22
ggml/src/ggml-kompute.cpp
View file

@ -66,12 +66,12 @@ static std::string ggml_kompute_format_name(int device) {
return "Kompute" + std::to_string(device); return "Kompute" + std::to_string(device);
} }
struct ggml_kompute_context { struct ggml_backend_kompute_context {
int device; int device;
std::string name; std::string name;
std::shared_ptr<vk::DescriptorPool> pool; std::shared_ptr<vk::DescriptorPool> pool;
ggml_kompute_context(int device) ggml_backend_kompute_context(int device)
: device(device), name(ggml_kompute_format_name(device)) {} : device(device), name(ggml_kompute_format_name(device)) {}
}; };
@ -79,7 +79,7 @@ struct ggml_kompute_context {
// and consolidate the init functions and simplify object lifetime management. As it currently stands, // and consolidate the init functions and simplify object lifetime management. As it currently stands,
// we *have* to have the kompute manager no matter what for device discovery, but the kompute context // we *have* to have the kompute manager no matter what for device discovery, but the kompute context
// is only created when a device is set and vulkan is explicitly turned on. // is only created when a device is set and vulkan is explicitly turned on.
static ggml_kompute_context *s_kompute_context = nullptr; static ggml_backend_kompute_context *s_kompute_context = nullptr;
class kompute_manager { class kompute_manager {
kp::Manager *s_mgr = nullptr; kp::Manager *s_mgr = nullptr;
@ -348,7 +348,7 @@ ggml_vk_device ggml_vk_current_device() {
} }
static static
void ggml_vk_allocate_descriptor_pool(struct ggml_kompute_context * ctx, size_t size) { void ggml_vk_allocate_descriptor_pool(struct ggml_backend_kompute_context * ctx, size_t size) {
std::vector<vk::DescriptorPoolSize> descriptorPoolSizes = { std::vector<vk::DescriptorPoolSize> descriptorPoolSizes = {
vk::DescriptorPoolSize( vk::DescriptorPoolSize(
vk::DescriptorType::eStorageBuffer, vk::DescriptorType::eStorageBuffer,
@ -370,7 +370,7 @@ void ggml_vk_allocate_descriptor_pool(struct ggml_kompute_context * ctx, size_t
} }
static static
void ggml_vk_free_descriptor_pool(struct ggml_kompute_context * ctx) { void ggml_vk_free_descriptor_pool(struct ggml_backend_kompute_context * ctx) {
if (ctx->pool) { if (ctx->pool) {
komputeManager()->device()->destroy( komputeManager()->device()->destroy(
*ctx->pool, *ctx->pool,
@ -1412,7 +1412,7 @@ static bool ggml_vk_supports_op(const struct ggml_tensor * op) {
return false; return false;
} }
static void ggml_vk_graph_compute(struct ggml_kompute_context * ctx, struct ggml_cgraph * gf) { static void ggml_vk_graph_compute(struct ggml_backend_kompute_context * ctx, struct ggml_cgraph * gf) {
const int n_seq = 8; const int n_seq = 8;
// FIXME: Figure out if we can somehow optimize the size of the pool... right now we're setting // FIXME: Figure out if we can somehow optimize the size of the pool... right now we're setting
@ -1935,12 +1935,12 @@ ggml_backend_buffer_type_t ggml_backend_kompute_buffer_type(int device) {
// backend // backend
static const char * ggml_backend_kompute_name(ggml_backend_t backend) { static const char * ggml_backend_kompute_name(ggml_backend_t backend) {
auto * ctx = static_cast<ggml_kompute_context *>(backend->context); auto * ctx = static_cast<ggml_backend_kompute_context *>(backend->context);
return ctx->name.c_str(); return ctx->name.c_str();
} }
static void ggml_backend_kompute_free(ggml_backend_t backend) { static void ggml_backend_kompute_free(ggml_backend_t backend) {
auto * ctx = static_cast<ggml_kompute_context *>(backend->context); auto * ctx = static_cast<ggml_backend_kompute_context *>(backend->context);
assert(ctx == s_kompute_context); assert(ctx == s_kompute_context);
s_kompute_context = nullptr; s_kompute_context = nullptr;
@ -1952,12 +1952,12 @@ static void ggml_backend_kompute_free(ggml_backend_t backend) {
} }
static ggml_backend_buffer_type_t ggml_backend_kompute_get_default_buffer_type(ggml_backend_t backend) { static ggml_backend_buffer_type_t ggml_backend_kompute_get_default_buffer_type(ggml_backend_t backend) {
auto * ctx = static_cast<ggml_kompute_context *>(backend->context); auto * ctx = static_cast<ggml_backend_kompute_context *>(backend->context);
return ggml_backend_kompute_buffer_type(ctx->device); return ggml_backend_kompute_buffer_type(ctx->device);
} }
static ggml_status ggml_backend_kompute_graph_compute(ggml_backend_t backend, struct ggml_cgraph * cgraph) { static ggml_status ggml_backend_kompute_graph_compute(ggml_backend_t backend, struct ggml_cgraph * cgraph) {
auto * ctx = static_cast<ggml_kompute_context *>(backend->context); auto * ctx = static_cast<ggml_backend_kompute_context *>(backend->context);
ggml_vk_graph_compute(ctx, cgraph); ggml_vk_graph_compute(ctx, cgraph);
return GGML_STATUS_SUCCESS; return GGML_STATUS_SUCCESS;
} }
@ -2002,7 +2002,7 @@ static ggml_guid_t ggml_backend_kompute_guid() {
ggml_backend_t ggml_backend_kompute_init(int device) { ggml_backend_t ggml_backend_kompute_init(int device) {
GGML_ASSERT(s_kompute_context == nullptr); GGML_ASSERT(s_kompute_context == nullptr);
s_kompute_context = new ggml_kompute_context(device); s_kompute_context = new ggml_backend_kompute_context(device);
ggml_backend_t kompute_backend = new ggml_backend { ggml_backend_t kompute_backend = new ggml_backend {
/* .guid = */ ggml_backend_kompute_guid(), /* .guid = */ ggml_backend_kompute_guid(),