whisper.rn 0.4.0-rc.6 → 0.4.0-rc.8

package/cpp/ggml-metal.h

@@ -36,70 +36,22 @@ struct wsp_ggml_cgraph;
 extern "C" {
 #endif
 
-//
-// internal API
-// temporary exposed to user-code
-//
-
-struct wsp_ggml_metal_context;
-
-void wsp_ggml_metal_log_set_callback(wsp_ggml_log_callback log_callback, void * user_data);
-
-// number of command buffers to use
-struct wsp_ggml_metal_context * wsp_ggml_metal_init(int n_cb);
-void wsp_ggml_metal_free(struct wsp_ggml_metal_context * ctx);
-
-void * wsp_ggml_metal_host_malloc(size_t n);
-void   wsp_ggml_metal_host_free  (void * data);
-
-// set the number of command buffers to use
-void wsp_ggml_metal_set_n_cb(struct wsp_ggml_metal_context * ctx, int n_cb);
-
-// creates a mapping between a host memory buffer and a device memory buffer
-// - make sure to map all buffers used in the graph before calling wsp_ggml_metal_graph_compute
-// - the mapping is used during computation to determine the arguments of the compute kernels
-// - you don't need to keep the host memory buffer allocated as it is never accessed by Metal
-// - max_size specifies the maximum size of a tensor and is used to create shared views such
-//   that it is guaranteed that the tensor will fit in at least one of the views
-//
-bool wsp_ggml_metal_add_buffer(
-        struct wsp_ggml_metal_context * ctx,
-                       const char * name,
-                             void * data,
-                           size_t   size,
-                           size_t   max_size);
-
-// set data from host memory into the device
-void wsp_ggml_metal_set_tensor(struct wsp_ggml_metal_context * ctx, struct wsp_ggml_tensor * t);
-
-// get data from the device into host memory
-void wsp_ggml_metal_get_tensor(struct wsp_ggml_metal_context * ctx, struct wsp_ggml_tensor * t);
-
-// try to find operations that can be run concurrently in the graph
-// you should run it again if the topology of your graph changes
-void wsp_ggml_metal_graph_find_concurrency(struct wsp_ggml_metal_context * ctx, struct wsp_ggml_cgraph * gf, bool check_mem);
-
-// if the graph has been optimized for concurrently dispatch, return length of the concur_list if optimized
-int wsp_ggml_metal_if_optimized(struct wsp_ggml_metal_context * ctx);
-
-// output the concur_list for wsp_ggml_alloc
-int * wsp_ggml_metal_get_concur_list(struct wsp_ggml_metal_context * ctx);
-
-// same as wsp_ggml_graph_compute but uses Metal
-// creates gf->n_threads command buffers in parallel
-void wsp_ggml_metal_graph_compute(struct wsp_ggml_metal_context * ctx, struct wsp_ggml_cgraph * gf);
-
 //
 // backend API
 // user-code should use only these functions
 //
 
+WSP_GGML_API void wsp_ggml_backend_metal_log_set_callback(wsp_ggml_log_callback log_callback, void * user_data);
+
 WSP_GGML_API wsp_ggml_backend_t wsp_ggml_backend_metal_init(void);
 
 WSP_GGML_API bool wsp_ggml_backend_is_metal(wsp_ggml_backend_t backend);
 
+WSP_GGML_API WSP_GGML_CALL wsp_ggml_backend_buffer_t wsp_ggml_backend_metal_buffer_from_ptr(void * data, size_t size, size_t max_size);
+
 WSP_GGML_API void wsp_ggml_backend_metal_set_n_cb(wsp_ggml_backend_t backend, int n_cb);
-WSP_GGML_API wsp_ggml_backend_buffer_type_t wsp_ggml_backend_metal_buffer_type(void);
+
+WSP_GGML_API WSP_GGML_CALL wsp_ggml_backend_buffer_type_t wsp_ggml_backend_metal_buffer_type(void);
 
 // helper to check if the device supports a specific family
 // ideally, the user code should be doing these checks