ggml : introduce GGML_CALL function annotation (#4850)

This change makes it possible to build ggml-cuda.cu and ggml-metal.m as
independent dynamic shared objects, that may be conditionally linked at
runtime in a multiplatform binary. It introduces a GGML_CALL annotation
that documents which functions have a cyclic call relationship, between
the application code and GPU modules.

This change does nothing, unless the build defines -DGGML_MULTIPLATFORM
which causes back-references and function pointers to conform to MS ABI
which is supported by NVCC, ROCm, XCode, GCC and Clang across platforms

1 files changed, 34 insertions, 24 deletions

diff --git a/ggml.h b/ggml.h
index 1187074f..837c52e6 100644
--- a/ggml.h
+++ b/ggml.h
@@ -187,6 +187,16 @@
 #    define GGML_API
 #endif
 
+#ifdef GGML_MULTIPLATFORM
+#    if defined(_WIN32)
+#        define GGML_CALL
+#    else
+#        define GGML_CALL __attribute__((__ms_abi__))
+#    endif
+#else
+#    define GGML_CALL
+#endif
+
 // TODO: support for clang
 #ifdef __GNUC__
 #    define GGML_DEPRECATED(func, hint) func __attribute__((deprecated(hint)))
@@ -649,41 +659,41 @@ extern "C" {
     GGML_API void    ggml_print_object (const struct ggml_object * obj);
     GGML_API void    ggml_print_objects(const struct ggml_context * ctx);
 
-    GGML_API int64_t ggml_nelements   (const struct ggml_tensor * tensor);
-    GGML_API int64_t ggml_nrows       (const struct ggml_tensor * tensor);
-    GGML_API size_t  ggml_nbytes      (const struct ggml_tensor * tensor);
-    GGML_API size_t  ggml_nbytes_pad  (const struct ggml_tensor * tensor); // same as ggml_nbytes() but padded to GGML_MEM_ALIGN
+    GGML_API GGML_CALL int64_t ggml_nelements   (const struct ggml_tensor * tensor);
+    GGML_API GGML_CALL int64_t ggml_nrows       (const struct ggml_tensor * tensor);
+    GGML_API GGML_CALL size_t  ggml_nbytes      (const struct ggml_tensor * tensor);
+    GGML_API           size_t  ggml_nbytes_pad  (const struct ggml_tensor * tensor); // same as ggml_nbytes() but padded to GGML_MEM_ALIGN
 
-    GGML_API int    ggml_blck_size(enum ggml_type type);
-    GGML_API size_t ggml_type_size(enum ggml_type type);             // size in bytes for all elements in a block
-    GGML_API size_t ggml_row_size (enum ggml_type type, int64_t ne); // size in bytes for all elements in a row
+    GGML_API GGML_CALL int    ggml_blck_size(enum ggml_type type);
+    GGML_API GGML_CALL size_t ggml_type_size(enum ggml_type type);             // size in bytes for all elements in a block
+    GGML_API GGML_CALL size_t ggml_row_size (enum ggml_type type, int64_t ne); // size in bytes for all elements in a row
 
     GGML_DEPRECATED(
     GGML_API double ggml_type_sizef(enum ggml_type type), // ggml_type_size()/ggml_blck_size() as float
     "use ggml_row_size() instead");
 
-    GGML_API const char * ggml_type_name(enum ggml_type type);
-    GGML_API const char * ggml_op_name  (enum ggml_op   op);
-    GGML_API const char * ggml_op_symbol(enum ggml_op   op);
+    GGML_API GGML_CALL const char * ggml_type_name(enum ggml_type type);
+    GGML_API GGML_CALL const char * ggml_op_name  (enum ggml_op   op);
+    GGML_API           const char * ggml_op_symbol(enum ggml_op   op);
 
-    GGML_API const char * ggml_unary_op_name(enum ggml_unary_op op);
-    GGML_API const char * ggml_op_desc(const struct ggml_tensor * t); // unary or op name
+    GGML_API           const char * ggml_unary_op_name(enum ggml_unary_op op);
+    GGML_API GGML_CALL const char * ggml_op_desc(const struct ggml_tensor * t); // unary or op name
 
-    GGML_API size_t  ggml_element_size(const struct ggml_tensor * tensor);
+    GGML_API GGML_CALL size_t  ggml_element_size(const struct ggml_tensor * tensor);
 
-    GGML_API bool    ggml_is_quantized(enum ggml_type type);
+    GGML_API GGML_CALL bool    ggml_is_quantized(enum ggml_type type);
 
     // TODO: temporary until model loading of ggml examples is refactored
     GGML_API enum ggml_type ggml_ftype_to_ggml_type(enum ggml_ftype ftype);
 
-    GGML_API bool ggml_is_transposed(const struct ggml_tensor * tensor);
-    GGML_API bool ggml_is_contiguous(const struct ggml_tensor * tensor);
-    GGML_API bool ggml_is_permuted  (const struct ggml_tensor * tensor);
-    GGML_API bool ggml_is_scalar    (const struct ggml_tensor * tensor);
-    GGML_API bool ggml_is_vector    (const struct ggml_tensor * tensor);
-    GGML_API bool ggml_is_matrix    (const struct ggml_tensor * tensor);
-    GGML_API bool ggml_is_3d        (const struct ggml_tensor * tensor);
-    GGML_API int  ggml_n_dims       (const struct ggml_tensor * tensor); // returns 1 for scalars
+    GGML_API GGML_CALL bool ggml_is_transposed(const struct ggml_tensor * tensor);
+    GGML_API GGML_CALL bool ggml_is_contiguous(const struct ggml_tensor * tensor);
+    GGML_API GGML_CALL bool ggml_is_permuted  (const struct ggml_tensor * tensor);
+    GGML_API           bool ggml_is_scalar    (const struct ggml_tensor * tensor);
+    GGML_API           bool ggml_is_vector    (const struct ggml_tensor * tensor);
+    GGML_API           bool ggml_is_matrix    (const struct ggml_tensor * tensor);
+    GGML_API           bool ggml_is_3d        (const struct ggml_tensor * tensor);
+    GGML_API           int  ggml_n_dims       (const struct ggml_tensor * tensor); // returns 1 for scalars
 
     GGML_API bool ggml_are_same_shape(const struct ggml_tensor * t0, const struct ggml_tensor * t1);
 
@@ -770,7 +780,7 @@ extern "C" {
     GGML_API void *  ggml_get_data    (const struct ggml_tensor * tensor);
     GGML_API float * ggml_get_data_f32(const struct ggml_tensor * tensor);
 
-    GGML_API enum ggml_unary_op ggml_get_unary_op(const struct ggml_tensor * tensor);
+    GGML_API GGML_CALL enum ggml_unary_op ggml_get_unary_op(const struct ggml_tensor * tensor);
 
     GGML_API const char *         ggml_get_name   (const struct ggml_tensor * tensor);
     GGML_API struct ggml_tensor * ggml_set_name   (      struct ggml_tensor * tensor, const char * name);
@@ -1413,7 +1423,7 @@ extern "C" {
             float                 beta_slow);
 
     // compute correction dims for YaRN RoPE scaling
-    void ggml_rope_yarn_corr_dims(
+    GGML_CALL void ggml_rope_yarn_corr_dims(
         int n_dims, int n_orig_ctx, float freq_base, float beta_fast, float beta_slow, float dims[2]);
 
     // xPos RoPE, in-place, returns view(a)