Implement max_size for backend buffer types to limit the size of a si…

…ngle allocation

ggml-alloc.c

@@ -780,6 +780,7 @@ ggml_backend_buffer_t ggml_backend_alloc_ctx_tensors_from_buft(struct ggml_conte
     GGML_ASSERT(ggml_get_no_alloc(ctx) == true);
 
     size_t alignment = ggml_backend_buft_get_alignment(buft);
+    size_t max_size = ggml_backend_buft_get_max_size(buft);
 
     size_t nbytes = 0;
     for (struct ggml_tensor * t = ggml_get_first_tensor(ctx); t != NULL; t = ggml_get_next_tensor(ctx, t)) {
@@ -796,35 +797,115 @@ ggml_backend_buffer_t ggml_backend_alloc_ctx_tensors_from_buft(struct ggml_conte
         return NULL;
     }
 
-    ggml_backend_buffer_t buffer = ggml_backend_buft_alloc_buffer(buft, nbytes);
-    if (buffer == NULL) {
-        // failed to allocate buffer
+    // single buffer allocation
+    if (nbytes <= max_size) {
+        ggml_backend_buffer_t buffer = ggml_backend_buft_alloc_buffer(buft, nbytes);
+        if (buffer == NULL) {
+            // failed to allocate buffer
 #ifndef NDEBUG
-        fprintf(stderr, "%s: failed to allocate buffer\n", __func__);
+            fprintf(stderr, "%s: failed to allocate buffer\n", __func__);
 #endif
-        return NULL;
+            return NULL;
+        }
+
+        ggml_tallocr_t tallocr = ggml_tallocr_new_from_buffer(buffer);
+
+        for (struct ggml_tensor * t = ggml_get_first_tensor(ctx); t != NULL; t = ggml_get_next_tensor(ctx, t)) {
+            if (t->data == NULL) {
+                if (t->view_src == NULL) {
+                    ggml_tallocr_alloc(tallocr, t);
+                } else {
+                    ggml_backend_view_init(buffer, t);
+                }
+            } else {
+                if (t->view_src != NULL) {
+                    // view of a pre-allocated tensor
+                    ggml_backend_view_init(buffer, t);
+                }
+            }
+        }
+
+        ggml_tallocr_free(tallocr);
+
+        return buffer;
     }
 
-    ggml_tallocr_t tallocr = ggml_tallocr_new_from_buffer(buffer);
+    // multi-buffer
+    size_t n_allocs = (nbytes - 1 + max_size) / max_size;
+    size_t * nbytes_per_alloc = (size_t *) malloc(n_allocs * sizeof(size_t));
+    memset(nbytes_per_alloc, 0, n_allocs * sizeof(size_t));
 
+    // Calculate nbytes per alloc
+    size_t alloc_idx = 0;
     for (struct ggml_tensor * t = ggml_get_first_tensor(ctx); t != NULL; t = ggml_get_next_tensor(ctx, t)) {
-        if (t->data == NULL) {
-            if (t->view_src == NULL) {
-                ggml_tallocr_alloc(tallocr, t);
-            } else {
-                ggml_backend_view_init(buffer, t);
+        if (t->data == NULL && t->view_src == NULL) {
+            size_t tensor_size = GGML_PAD(ggml_backend_buft_get_alloc_size(buft, t), alignment);
+            if (nbytes_per_alloc[alloc_idx] + tensor_size > max_size) {
+                // Move to next allocation
+                alloc_idx += 1;
             }
-        } else {
-            if (t->view_src != NULL) {
-                // view of a pre-allocated tensor
-                ggml_backend_view_init(buffer, t);
+            nbytes_per_alloc[alloc_idx] += tensor_size;
+        }
+    }
+
+    ggml_backend_buffer_t multi_buffer = ggml_backend_multi_buffer_alloc_buffer(n_allocs, buft, nbytes);
+    ggml_backend_multi_buffer_context_t multi_ctx = (ggml_backend_multi_buffer_context_t) multi_buffer->context;
+
+    size_t bytes_counter = 0;
+    struct ggml_tensor * current_tensor = ggml_get_first_tensor(ctx);
+
+    for (alloc_idx = 0; alloc_idx < n_allocs; alloc_idx++) {
+        ggml_backend_buffer_t buffer = ggml_backend_buft_alloc_buffer(buft, nbytes_per_alloc[alloc_idx]);
+        if (buffer == NULL) {
+            // failed to allocate buffer
+#ifndef NDEBUG
+            fprintf(stderr, "%s: failed to allocate buffer\n", __func__);
+#endif
+
+            // free previously allocated buffers
+            for (size_t dealloc_idx = 0; dealloc_idx < alloc_idx; dealloc_idx++) {
+                ggml_backend_buffer_free(multi_ctx->buffers[dealloc_idx]);
+            }
+
+            free(nbytes_per_alloc);
+
+            return NULL;
+        }
+
+        multi_ctx->buffers[alloc_idx] = buffer;
+
+        ggml_tallocr_t tallocr = ggml_tallocr_new_from_buffer(buffer);
+
+        for (; current_tensor != NULL; current_tensor = ggml_get_next_tensor(ctx, current_tensor)) {
+            size_t tensor_size = GGML_PAD(ggml_backend_buft_get_alloc_size(buft, current_tensor), alignment);
+
+            if (bytes_counter + tensor_size > max_size) {
+                // tensor uses next buffer
+                bytes_counter = 0;
+                break;
+            }
+
+            bytes_counter += tensor_size;
+            if (current_tensor->data == NULL) {
+                if (current_tensor->view_src == NULL) {
+                    ggml_tallocr_alloc(tallocr, current_tensor);
+                } else {
+                    ggml_backend_view_init(buffer, current_tensor);
+                }
+            } else {
+                if (current_tensor->view_src != NULL) {
+                    // view of a pre-allocated tensor
+                    ggml_backend_view_init(buffer, current_tensor);
+                }
             }
         }
+
+        ggml_tallocr_free(tallocr);
     }
 
-    ggml_tallocr_free(tallocr);
+    free(nbytes_per_alloc);
 
-    return buffer;
+    return multi_buffer;
 }
 
 ggml_backend_buffer_t ggml_backend_alloc_ctx_tensors(struct ggml_context * ctx, ggml_backend_t backend) {

ggml-backend-impl.h

@@ -19,6 +19,7 @@ extern "C" {
         const char *          (*GGML_CALL get_name)        (ggml_backend_buffer_type_t buft);
         ggml_backend_buffer_t (*GGML_CALL alloc_buffer)    (ggml_backend_buffer_type_t buft, size_t size);
         size_t                (*GGML_CALL get_alignment)   (ggml_backend_buffer_type_t buft); // tensor alignment
+        size_t                (*GGML_CALL get_max_size)    (ggml_backend_buffer_type_t buft); // allocation max size
         size_t                (*GGML_CALL get_alloc_size)  (ggml_backend_buffer_type_t buft, const struct ggml_tensor * tensor); // data size needed to allocate the tensor, including padding
         bool                  (*GGML_CALL supports_backend)(ggml_backend_buffer_type_t buft, ggml_backend_t backend); // check if the buffer type is usable by the backend
         // check if tensor data is in host memory
@@ -63,6 +64,20 @@ extern "C" {
     // do not use directly, use ggml_backend_tensor_copy instead
     bool ggml_backend_buffer_copy_tensor(const struct ggml_tensor * src, struct ggml_tensor * dst);
 
+    // multi-buffer
+    struct ggml_backend_multi_buffer_context {
+        ggml_backend_buffer_t * buffers;
+        size_t n_buffers;
+    };
+
+    typedef struct ggml_backend_multi_buffer_context * ggml_backend_multi_buffer_context_t;
+
+    GGML_CALL const char* ggml_backend_multi_buffer_get_name(ggml_backend_buffer_t buffer);
+    GGML_CALL ggml_backend_buffer_t ggml_backend_multi_buffer_alloc_buffer(size_t n_buffers, ggml_backend_buffer_type_t buft, size_t nbytes);
+    GGML_CALL void ggml_backend_multi_buffer_free_buffer(ggml_backend_buffer_t buffer);
+    GGML_CALL void ggml_backend_multi_buffer_clear(ggml_backend_buffer_t buffer, uint8_t value);
+    struct ggml_backend_buffer_i ggml_backend_multi_buffer_context_interface(void);
+
     //
     // Backend
     //

ggml-backend.c

@@ -27,6 +27,14 @@ size_t ggml_backend_buft_get_alignment(ggml_backend_buffer_type_t buft) {
     return buft->iface.get_alignment(buft);
 }
 
+size_t ggml_backend_buft_get_max_size(ggml_backend_buffer_type_t buft) {
+    // get_max_size is optional, defaults to UINT64_MAX
+    if (buft->iface.get_max_size) {
+        return buft->iface.get_max_size(buft);
+    }
+    return UINT64_MAX;
+}
+
 GGML_CALL size_t ggml_backend_buft_get_alloc_size(ggml_backend_buffer_type_t buft, struct ggml_tensor * tensor) {
     // get_alloc_size is optional, defaults to ggml_nbytes
     if (buft->iface.get_alloc_size) {
@@ -55,8 +63,6 @@ GGML_CALL ggml_backend_buffer_t ggml_backend_buffer_init(
                size_t                          size) {
     ggml_backend_buffer_t buffer = malloc(sizeof(struct ggml_backend_buffer));
 
-    GGML_ASSERT(iface.get_base != NULL);
-
     (*buffer) = (struct ggml_backend_buffer) {
         /* .interface = */ iface,
         /* .buft      = */ buft,
@@ -106,6 +112,10 @@ size_t ggml_backend_buffer_get_alignment (ggml_backend_buffer_t buffer) {
     return ggml_backend_buft_get_alignment(ggml_backend_buffer_get_type(buffer));
 }
 
+size_t ggml_backend_buffer_get_max_size(ggml_backend_buffer_t buffer) {
+    return ggml_backend_buft_get_max_size(ggml_backend_buffer_get_type(buffer));
+}
+
 size_t ggml_backend_buffer_get_alloc_size(ggml_backend_buffer_t buffer, struct ggml_tensor * tensor) {
     return ggml_backend_buft_get_alloc_size(ggml_backend_buffer_get_type(buffer), tensor);
 }
@@ -169,6 +179,10 @@ size_t ggml_backend_get_alignment(ggml_backend_t backend) {
     return ggml_backend_buft_get_alignment(ggml_backend_get_default_buffer_type(backend));
 }
 
+size_t ggml_backend_get_max_size(ggml_backend_t backend) {
+    return ggml_backend_buft_get_max_size(ggml_backend_get_default_buffer_type(backend));
+}
+
 void ggml_backend_tensor_set_async(ggml_backend_t backend, struct ggml_tensor * tensor, const void * data, size_t offset, size_t size) {
     GGML_ASSERT(tensor->data != NULL && "tensor not allocated");
     GGML_ASSERT(offset + size <= ggml_nbytes(tensor) && "tensor write out of bounds");
@@ -342,6 +356,11 @@ GGML_CALL static void ggml_backend_registry_init(void) {
     extern GGML_CALL ggml_backend_buffer_type_t ggml_backend_metal_buffer_type(void);
     ggml_backend_register("Metal", ggml_backend_reg_metal_init, ggml_backend_metal_buffer_type(), NULL);
 #endif
+
+#ifdef GGML_USE_VULKAN
+    extern GGML_CALL int ggml_backend_vk_reg_devices(void);
+    ggml_backend_vk_reg_devices();
+#endif
 }
 
 GGML_CALL void ggml_backend_register(const char * name, ggml_backend_init_fn init_fn, ggml_backend_buffer_type_t default_buffer_type, void * user_data) {
@@ -545,6 +564,7 @@ GGML_CALL ggml_backend_buffer_type_t ggml_backend_cpu_buffer_type(void) {
             /* .get_name         = */ ggml_backend_cpu_buffer_type_get_name,
             /* .alloc_buffer     = */ ggml_backend_cpu_buffer_type_alloc_buffer,
             /* .get_alignment    = */ ggml_backend_cpu_buffer_type_get_alignment,
+            /* .get_max_size     = */ NULL, // defaults to UINT64_MAX
             /* .get_alloc_size   = */ NULL, // defaults to ggml_nbytes
             /* .supports_backend = */ ggml_backend_cpu_buffer_type_supports_backend,
             /* .is_host          = */ ggml_backend_cpu_buffer_type_is_host,
@@ -600,6 +620,7 @@ ggml_backend_buffer_type_t ggml_backend_cpu_hbm_buffer_type(void) {
             /* .get_name         = */ ggml_backend_cpu_hbm_buffer_type_get_name,
             /* .alloc_buffer     = */ ggml_backend_cpu_hbm_buffer_type_alloc_buffer,
             /* .get_alignment    = */ ggml_backend_cpu_buffer_type_get_alignment,
+            /* .get_max_size     = */ NULL, // defaults to UINT64_MAX
             /* .get_alloc_size   = */ NULL, // defaults to ggml_nbytes
             /* .supports_backend = */ ggml_backend_cpu_buffer_type_supports_backend,
             /* .is_host          = */ ggml_backend_cpu_buffer_type_is_host,
@@ -755,6 +776,56 @@ GGML_CALL static ggml_backend_t ggml_backend_reg_cpu_init(const char * params, v
 }
 
 
+// multi-buffer buffer
+
+GGML_CALL const char * ggml_backend_multi_buffer_get_name(ggml_backend_buffer_t buffer) {
+    ggml_backend_multi_buffer_context_t ctx = (ggml_backend_multi_buffer_context_t) buffer->context;
+
+    return ctx->buffers[0]->iface.get_name(ctx->buffers[0]);
+}
+
+GGML_CALL ggml_backend_buffer_t ggml_backend_multi_buffer_alloc_buffer(size_t n_buffers, ggml_backend_buffer_type_t buft, size_t nbytes) {
+    ggml_backend_multi_buffer_context_t ctx = (ggml_backend_multi_buffer_context_t) malloc(sizeof(struct ggml_backend_multi_buffer_context));
+    ctx->n_buffers = n_buffers;
+    ctx->buffers = (ggml_backend_buffer_t *) malloc(n_buffers * sizeof(ggml_backend_buffer_t));
+
+    return ggml_backend_buffer_init(buft, ggml_backend_multi_buffer_context_interface(), ctx, nbytes);
+}
+
+GGML_CALL void ggml_backend_multi_buffer_free_buffer(ggml_backend_buffer_t buffer) {
+    ggml_backend_multi_buffer_context_t ctx = (ggml_backend_multi_buffer_context_t) buffer->context;
+    for (size_t i = 0; i < ctx->n_buffers; i++) {
+        ggml_backend_buffer_free(ctx->buffers[i]);
+    }
+
+    free(ctx->buffers);
+    free(ctx);
+}
+
+GGML_CALL void ggml_backend_multi_buffer_clear(ggml_backend_buffer_t buffer, uint8_t value) {
+    ggml_backend_multi_buffer_context_t ctx = (ggml_backend_multi_buffer_context_t) buffer->context;
+    for (size_t i = 0; i < ctx->n_buffers; i++) {
+        ggml_backend_buffer_clear(ctx->buffers[i], value);
+    }
+}
+
+struct ggml_backend_buffer_i ggml_backend_multi_buffer_context_interface(void) {
+    static struct ggml_backend_buffer_i multi_backend_buffer_i = {
+        /* .get_name        = */ ggml_backend_multi_buffer_get_name,
+        /* .free_buffer     = */ ggml_backend_multi_buffer_free_buffer,
+        /* .get_base        = */ NULL,
+        /* .init_tensor     = */ NULL,
+        /* .set_tensor      = */ NULL,
+        /* .get_tensor      = */ NULL,
+        /* .cpy_tensor      = */ NULL,
+        /* .clear           = */ ggml_backend_multi_buffer_clear,
+        /* .reset           = */ NULL,
+    };
+
+    return multi_backend_buffer_i;
+}
+
+
 // scheduler
 
 #define GGML_MAX_BACKENDS 16

ggml-backend.h

@@ -20,6 +20,7 @@ extern "C" {
     GGML_API           const char *          ggml_backend_buft_name            (ggml_backend_buffer_type_t buft);
     GGML_API GGML_CALL ggml_backend_buffer_t ggml_backend_buft_alloc_buffer    (ggml_backend_buffer_type_t buft, size_t size);
     GGML_API           size_t                ggml_backend_buft_get_alignment   (ggml_backend_buffer_type_t buft);
+    GGML_API           size_t                ggml_backend_buft_get_max_size    (ggml_backend_buffer_type_t buft);
     GGML_API GGML_CALL size_t                ggml_backend_buft_get_alloc_size  (ggml_backend_buffer_type_t buft, struct ggml_tensor * tensor);
     GGML_API           bool                  ggml_backend_buft_supports_backend(ggml_backend_buffer_type_t buft, ggml_backend_t backend);
     GGML_API           bool                  ggml_backend_buft_is_host         (ggml_backend_buffer_type_t buft);
@@ -36,6 +37,7 @@ extern "C" {
     GGML_API           size_t                     ggml_backend_buffer_get_size      (ggml_backend_buffer_t buffer);
     GGML_API GGML_CALL void                       ggml_backend_buffer_init_tensor   (ggml_backend_buffer_t buffer, struct ggml_tensor * tensor);
     GGML_API           size_t                     ggml_backend_buffer_get_alignment (ggml_backend_buffer_t buffer);
+    GGML_API           size_t                     ggml_backend_buffer_get_max_size  (ggml_backend_buffer_t buffer);
     GGML_API           size_t                     ggml_backend_buffer_get_alloc_size(ggml_backend_buffer_t buffer, struct ggml_tensor * tensor);
     GGML_API           void                       ggml_backend_buffer_clear         (ggml_backend_buffer_t buffer, uint8_t value);
     GGML_API           bool                       ggml_backend_buffer_is_host       (ggml_backend_buffer_t buffer);
@@ -54,6 +56,7 @@ extern "C" {
     GGML_API ggml_backend_buffer_type_t ggml_backend_get_default_buffer_type(ggml_backend_t backend);
     GGML_API ggml_backend_buffer_t      ggml_backend_alloc_buffer(ggml_backend_t backend, size_t size);
     GGML_API size_t                     ggml_backend_get_alignment(ggml_backend_t backend);
+    GGML_API size_t                     ggml_backend_get_max_size(ggml_backend_t backend);
 
     GGML_API void ggml_backend_tensor_set_async(ggml_backend_t backend,       struct ggml_tensor * tensor, const void * data, size_t offset, size_t size);
     GGML_API void ggml_backend_tensor_get_async(ggml_backend_t backend, const struct ggml_tensor * tensor,       void * data, size_t offset, size_t size);

ggml-cuda.cu

@@ -10428,6 +10428,7 @@ static ggml_backend_buffer_type_i ggml_backend_cuda_buffer_type_interface = {
     /* .get_name         = */ ggml_backend_cuda_buffer_type_name,
     /* .alloc_buffer     = */ ggml_backend_cuda_buffer_type_alloc_buffer,
     /* .get_alignment    = */ ggml_backend_cuda_buffer_type_get_alignment,
+    /* .get_max_size     = */ NULL, // defaults to UINT64_MAX
     /* .get_alloc_size   = */ ggml_backend_cuda_buffer_type_get_alloc_size,
     /* .supports_backend = */ ggml_backend_cuda_buffer_type_supports_backend,
     /* .is_host          = */ NULL,
@@ -10703,6 +10704,7 @@ static ggml_backend_buffer_type_i ggml_backend_cuda_split_buffer_type_interface
     /* .get_name         = */ ggml_backend_cuda_split_buffer_type_name,
     /* .alloc_buffer     = */ ggml_backend_cuda_split_buffer_type_alloc_buffer,
     /* .get_alignment    = */ ggml_backend_cuda_split_buffer_type_get_alignment,
+    /* .get_max_size     = */ NULL, // defaults to UINT64_MAX
     /* .get_alloc_size   = */ ggml_backend_cuda_split_buffer_type_get_alloc_size,
     /* .supports_backend = */ ggml_backend_cuda_split_buffer_type_supports_backend,
     /* .is_host          = */ ggml_backend_cuda_split_buffer_type_is_host,
@@ -10782,6 +10784,7 @@ GGML_CALL ggml_backend_buffer_type_t ggml_backend_cuda_host_buffer_type() {
             /* .get_name         = */ ggml_backend_cuda_host_buffer_type_name,
             /* .alloc_buffer     = */ ggml_backend_cuda_host_buffer_type_alloc_buffer,
             /* .get_alignment    = */ ggml_backend_cpu_buffer_type()->iface.get_alignment,
+            /* .get_max_size     = */ NULL, // defaults to UINT64_MAX
             /* .get_alloc_size   = */ ggml_backend_cpu_buffer_type()->iface.get_alloc_size,
             /* .supports_backend = */ ggml_backend_cpu_buffer_type()->iface.supports_backend,
             /* .is_host          = */ ggml_backend_cpu_buffer_type()->iface.is_host,

ggml-metal.m

@@ -2445,6 +2445,7 @@ GGML_CALL ggml_backend_buffer_type_t ggml_backend_metal_buffer_type(void) {
             /* .get_name         = */ ggml_backend_metal_buffer_type_get_name,
             /* .alloc_buffer     = */ ggml_backend_metal_buffer_type_alloc_buffer,
             /* .get_alignment    = */ ggml_backend_metal_buffer_type_get_alignment,
+            /* .get_max_size     = */ NULL, // defaults to UINT64_MAX
             /* .get_alloc_size   = */ NULL, // defaults to ggml_nbytes
             /* .supports_backend = */ ggml_backend_metal_buffer_type_supports_backend,
             /* .is_host          = */ ggml_backend_metal_buffer_type_is_host,

ggml-opencl.cpp

@@ -2055,6 +2055,7 @@ static ggml_backend_buffer_type_i ggml_backend_opencl_buffer_type_interface = {
     /* .get_name         = */ ggml_backend_opencl_buffer_type_name,
     /* .alloc_buffer     = */ ggml_backend_opencl_buffer_type_alloc_buffer,
     /* .get_alignment    = */ ggml_backend_opencl_buffer_type_get_alignment,
+    /* .get_max_size     = */ NULL, // defaults to UINT64_MAX
     /* .get_alloc_size   = */ NULL,
     /* .supports_backend = */ ggml_backend_opencl_buffer_type_supports_backend,
     /* .is_host          = */ NULL,
@@ -2111,6 +2112,7 @@ ggml_backend_buffer_type_t ggml_backend_opencl_host_buffer_type() {
             /* .get_name         = */ ggml_backend_opencl_host_buffer_type_name,
             /* .alloc_buffer     = */ ggml_backend_opencl_host_buffer_type_alloc_buffer,
             /* .get_alignment    = */ ggml_backend_cpu_buffer_type()->iface.get_alignment,
+            /* .get_max_size     = */ NULL, // defaults to UINT64_MAX
             /* .get_alloc_size   = */ ggml_backend_cpu_buffer_type()->iface.get_alloc_size,
             /* .supports_backend = */ ggml_backend_cpu_buffer_type()->iface.supports_backend,
             /* .is_host          = */ ggml_backend_cpu_buffer_type()->iface.is_host,