mirrors/ollama-for-amd
1
0
Fork 0
mirror of https://github.com/likelovewant/ollama-for-amd.git synced 2025-12-22 23:03:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update GGML to b6646 (#12245)

Browse Source 
Notable EOLs with this change:
- MacOS v12 and v13 are no longer supported (v14+ required)
- AMD gfx900 and gfx906 are no longer supported
This commit is contained in:
Daniel Hiltgen
2025-10-02 14:47:10 -07:00
committed by GitHub
parent fdb109469f
commit c68f367ef6
326 changed files with 30615 additions and 20624 deletions
Download Patch File Download Diff File Expand all files Collapse all files

32
llama/patches/0024-ggml-Enable-resetting-backend-devices.patch
View File

@@ -15,10 +15,10 @@ unused then it can be reset to free these data structures.
5 files changed, 29 insertions(+), 2 deletions(-)
diff --git a/ggml/include/ggml-backend.h b/ggml/include/ggml-backend.h
index b602a7c78..fda5ceb24 100644
index d4352663..0a2dae26 100644
--- a/ggml/include/ggml-backend.h
+++ b/ggml/include/ggml-backend.h
@@ -167,6 +167,7 @@ extern "C" {
@@ -178,6 +178,7 @@ extern "C" {
GGML_API void ggml_backend_dev_get_props(ggml_backend_dev_t device, struct ggml_backend_dev_props * props);
GGML_API ggml_backend_reg_t ggml_backend_dev_backend_reg(ggml_backend_dev_t device);
GGML_API ggml_backend_t ggml_backend_dev_init(ggml_backend_dev_t device, const char * params);
@@ -27,10 +27,10 @@ index b602a7c78..fda5ceb24 100644
GGML_API ggml_backend_buffer_type_t ggml_backend_dev_host_buffer_type(ggml_backend_dev_t device);
GGML_API ggml_backend_buffer_t ggml_backend_dev_buffer_from_host_ptr(ggml_backend_dev_t device, void * ptr, size_t size, size_t max_tensor_size);
diff --git a/ggml/src/ggml-backend-impl.h b/ggml/src/ggml-backend-impl.h
index 81749a5a3..6f10c353b 100644
index 869dc07d..4889df79 100644
--- a/ggml/src/ggml-backend-impl.h
+++ b/ggml/src/ggml-backend-impl.h
@@ -178,6 +178,10 @@ extern "C" {
@@ -195,6 +195,10 @@ extern "C" {
ggml_backend_event_t (*event_new) (ggml_backend_dev_t dev);
void (*event_free) (ggml_backend_dev_t dev, ggml_backend_event_t event);
void (*event_synchronize) (ggml_backend_dev_t dev, ggml_backend_event_t event);
@@ -42,10 +42,10 @@ index 81749a5a3..6f10c353b 100644
struct ggml_backend_device {
diff --git a/ggml/src/ggml-backend.cpp b/ggml/src/ggml-backend.cpp
index 05a842ed5..6556943b0 100644
index 6ef5eeaf..0b757af5 100644
--- a/ggml/src/ggml-backend.cpp
+++ b/ggml/src/ggml-backend.cpp
@@ -477,6 +477,14 @@ ggml_backend_t ggml_backend_dev_init(ggml_backend_dev_t device, const char * par
@@ -526,6 +526,14 @@ ggml_backend_t ggml_backend_dev_init(ggml_backend_dev_t device, const char * par
return device->iface.init_backend(device, params);
}
@@ -58,13 +58,13 @@ index 05a842ed5..6556943b0 100644
+}
+
ggml_backend_buffer_type_t ggml_backend_dev_buffer_type(ggml_backend_dev_t device) {
GGML_ASSERT(device);
return device->iface.get_buffer_type(device);
}
diff --git a/ggml/src/ggml-cuda/ggml-cuda.cu b/ggml/src/ggml-cuda/ggml-cuda.cu
index c7f9dc3a5..e43fde523 100644
index d324bc68..531d6e27 100644
--- a/ggml/src/ggml-cuda/ggml-cuda.cu
+++ b/ggml/src/ggml-cuda/ggml-cuda.cu
@@ -103,6 +103,11 @@ int ggml_cuda_get_device() {
@@ -107,6 +107,11 @@ int ggml_cuda_get_device() {
return id;
}
@@ -76,10 +76,10 @@ index c7f9dc3a5..e43fde523 100644
static cudaError_t ggml_cuda_device_malloc(void ** ptr, size_t size, int device) {
ggml_cuda_set_device(device);
cudaError_t err;
@@ -3243,7 +3248,10 @@ static void ggml_backend_cuda_device_get_props(ggml_backend_dev_t dev, ggml_back
props->description = ggml_backend_cuda_device_get_description(dev);
@@ -3512,7 +3517,10 @@ static void ggml_backend_cuda_device_get_props(ggml_backend_dev_t dev, ggml_back
props->id = ggml_backend_cuda_device_get_id(dev);
props->type = ggml_backend_cuda_device_get_type(dev);
props->device_id = ctx->pci_bus_id.empty() ? nullptr : ctx->pci_bus_id.c_str();
- ggml_backend_cuda_device_get_memory(dev, &props->memory_free, &props->memory_total);
+
+ // Memory reporting is disabled to avoid allocation of a CUDA primary context (~300 MB per device).
@@ -88,7 +88,7 @@ index c7f9dc3a5..e43fde523 100644
bool host_buffer = getenv("GGML_CUDA_NO_PINNED") == nullptr;
#ifdef GGML_CUDA_NO_PEER_COPY
@@ -3700,6 +3708,11 @@ static void ggml_backend_cuda_device_event_synchronize(ggml_backend_dev_t dev, g
@@ -3945,6 +3953,11 @@ static void ggml_backend_cuda_device_event_synchronize(ggml_backend_dev_t dev, g
CUDA_CHECK(cudaEventSynchronize((cudaEvent_t)event->context));
}
@@ -100,7 +100,7 @@ index c7f9dc3a5..e43fde523 100644
static const ggml_backend_device_i ggml_backend_cuda_device_interface = {
/* .get_name = */ ggml_backend_cuda_device_get_name,
/* .get_description = */ ggml_backend_cuda_device_get_description,
@@ -3716,6 +3729,7 @@ static const ggml_backend_device_i ggml_backend_cuda_device_interface = {
@@ -3961,6 +3974,7 @@ static const ggml_backend_device_i ggml_backend_cuda_device_interface = {
/* .event_new = */ ggml_backend_cuda_device_event_new,
/* .event_free = */ ggml_backend_cuda_device_event_free,
/* .event_synchronize = */ ggml_backend_cuda_device_event_synchronize,
@@ -108,7 +108,7 @@ index c7f9dc3a5..e43fde523 100644
};
// backend reg
@@ -3835,7 +3849,6 @@ ggml_backend_reg_t ggml_backend_cuda_reg() {
@@ -4076,7 +4090,6 @@ ggml_backend_reg_t ggml_backend_cuda_reg() {
dev_ctx->device = i;
dev_ctx->name = GGML_CUDA_NAME + std::to_string(i);
@@ -117,10 +117,10 @@ index c7f9dc3a5..e43fde523 100644
CUDA_CHECK(cudaGetDeviceProperties(&prop, i));
dev_ctx->description = prop.name;
diff --git a/ggml/src/ggml-cuda/vendors/hip.h b/ggml/src/ggml-cuda/vendors/hip.h
index c31f31923..cf22e60d2 100644
index 37386afc..06f9e7c1 100644
--- a/ggml/src/ggml-cuda/vendors/hip.h
+++ b/ggml/src/ggml-cuda/vendors/hip.h
@@ -40,6 +40,7 @@
@@ -41,6 +41,7 @@
#define cudaDeviceDisablePeerAccess hipDeviceDisablePeerAccess
#define cudaDeviceEnablePeerAccess hipDeviceEnablePeerAccess
#define cudaDeviceProp hipDeviceProp_t