ggml: Remove allocation status reporting

For each memory allocation we report the size of the (attempted)
allocation and whether it succeeded or failed. The latter status
reporting proved to be not that useful in practice as systems
such as Windows can automatically overflow from VRAM into RAM,
resultings in successful allocations even when there isn't
enough memory where we wanted.

As a result, this information is only used for debug logging,
which isn't worthwhile enough for the amount of code. It
also isn't fully accurate, as multiple allocations may result
in partial failures.

llama/patches/0014-graph-memory-reporting-on-failure.patch

@@ -4,48 +4,38 @@ Date: Fri, 18 Apr 2025 15:58:19 -0700
 Subject: [PATCH] graph memory reporting on failure
 
 ---
- ggml/include/ggml-alloc.h   |  6 ++++++
- ggml/include/ggml-backend.h |  6 ++++++
- ggml/src/ggml-alloc.c       | 38 +++++++++++++++++++++++++++++++++----
- ggml/src/ggml-backend.cpp   | 10 ++++++++++
- 4 files changed, 56 insertions(+), 4 deletions(-)
+ ggml/include/ggml-alloc.h   |  1 +
+ ggml/include/ggml-backend.h |  1 +
+ ggml/src/ggml-alloc.c       | 36 ++++++++++++++++++++++++++++++++----
+ ggml/src/ggml-backend.cpp   |  7 +++++++
+ 4 files changed, 41 insertions(+), 4 deletions(-)
 
 diff --git a/ggml/include/ggml-alloc.h b/ggml/include/ggml-alloc.h
-index 2cb150fd..781b1e10 100644
+index 2cb150fd2..7ab3f0192 100644
 --- a/ggml/include/ggml-alloc.h
 +++ b/ggml/include/ggml-alloc.h
-@@ -66,6 +66,12 @@ GGML_API bool ggml_gallocr_alloc_graph(ggml_gallocr_t galloc, struct ggml_cgraph
+@@ -65,6 +65,7 @@ GGML_API bool ggml_gallocr_reserve_n(
+ GGML_API bool ggml_gallocr_alloc_graph(ggml_gallocr_t galloc, struct ggml_cgraph * graph);
  
  GGML_API size_t ggml_gallocr_get_buffer_size(ggml_gallocr_t galloc, int buffer_id);
++GGML_API size_t ggml_gallocr_get_attempted_buffer_size(ggml_gallocr_t galloc, int buffer_id);
  
-+struct ggml_allocr_buffer_status {
-+    size_t size;
-+    bool allocated;
-+};
-+GGML_API struct ggml_allocr_buffer_status ggml_gallocr_get_attempted_buffer_size(ggml_gallocr_t galloc, int buffer_id);
-+
  // Utils
  // Create a buffer and allocate all the tensors in a ggml_context
- GGML_API struct ggml_backend_buffer * ggml_backend_alloc_ctx_tensors_from_buft(struct ggml_context * ctx, ggml_backend_buffer_type_t buft);
 diff --git a/ggml/include/ggml-backend.h b/ggml/include/ggml-backend.h
-index a2977ea2..8a91b381 100644
+index a2977ea2e..e8cf30841 100644
 --- a/ggml/include/ggml-backend.h
 +++ b/ggml/include/ggml-backend.h
-@@ -304,6 +304,12 @@ extern "C" {
+@@ -303,6 +303,7 @@ extern "C" {
+     GGML_API int                  ggml_backend_sched_get_n_copies(ggml_backend_sched_t sched);
  
      GGML_API size_t               ggml_backend_sched_get_buffer_size(ggml_backend_sched_t sched, ggml_backend_t backend);
++    GGML_API size_t               ggml_backend_sched_get_attempted_buffer_size(ggml_backend_sched_t sched, ggml_backend_t backend);
  
-+    struct ggml_backend_buffer_status {
-+        size_t size;
-+        bool allocated;
-+    };
-+    GGML_API struct ggml_backend_buffer_status ggml_backend_sched_get_attempted_buffer_size(ggml_backend_sched_t sched, ggml_backend_t backend);
-+
      GGML_API void                 ggml_backend_sched_set_tensor_backend(ggml_backend_sched_t sched, struct ggml_tensor * node, ggml_backend_t backend);
      GGML_API ggml_backend_t       ggml_backend_sched_get_tensor_backend(ggml_backend_sched_t sched, struct ggml_tensor * node);
- 
 diff --git a/ggml/src/ggml-alloc.c b/ggml/src/ggml-alloc.c
-index 8b6e6028..41c8c4a2 100644
+index 8b6e60283..b58bd671d 100644
 --- a/ggml/src/ggml-alloc.c
 +++ b/ggml/src/ggml-alloc.c
 @@ -350,6 +350,7 @@ struct node_alloc {
@@ -108,11 +98,11 @@ index 8b6e6028..41c8c4a2 100644
  }
  
  bool ggml_gallocr_reserve(ggml_gallocr_t galloc, struct ggml_cgraph *graph) {
-@@ -920,6 +932,24 @@ size_t ggml_gallocr_get_buffer_size(ggml_gallocr_t galloc, int buffer_id) {
+@@ -920,6 +932,22 @@ size_t ggml_gallocr_get_buffer_size(ggml_gallocr_t galloc, int buffer_id) {
      return ggml_backend_buffer_get_size(galloc->buffers[buffer_id]);
  }
  
-+struct ggml_allocr_buffer_status ggml_gallocr_get_attempted_buffer_size(ggml_gallocr_t galloc, int buffer_id) {
++size_t ggml_gallocr_get_attempted_buffer_size(ggml_gallocr_t galloc, int buffer_id) {
 +    GGML_ASSERT(buffer_id >= 0 && buffer_id < galloc->n_buffers);
 +
 +    for (int i = 0; i < buffer_id; i++) {
@@ -121,34 +111,29 @@ index 8b6e6028..41c8c4a2 100644
 +            // (See above.) However, we need a different check because multiple buffers might be NULL in our
 +            // case and we still want to know the attempted size.
 +
-+            struct ggml_allocr_buffer_status status = {0, true};
-+            return status;
++            return 0;
 +        }
 +    }
 +
-+    struct ggml_allocr_buffer_status status = {galloc->buffer_sizes[buffer_id], galloc->buffers[buffer_id] != NULL};
-+    return status;
++    return galloc->buffer_sizes[buffer_id];
 +}
 +
  // utils
  
  static void free_buffers(ggml_backend_buffer_t ** buffers, const size_t * n_buffers) {
 diff --git a/ggml/src/ggml-backend.cpp b/ggml/src/ggml-backend.cpp
-index 97f47abd..eded0291 100644
+index 97f47abd2..d02a40e60 100644
 --- a/ggml/src/ggml-backend.cpp
 +++ b/ggml/src/ggml-backend.cpp
-@@ -1631,6 +1631,16 @@ size_t ggml_backend_sched_get_buffer_size(ggml_backend_sched_t sched, ggml_backe
+@@ -1631,6 +1631,13 @@ size_t ggml_backend_sched_get_buffer_size(ggml_backend_sched_t sched, ggml_backe
      return ggml_gallocr_get_buffer_size(sched->galloc, backend_index);
  }
  
-+struct ggml_backend_buffer_status ggml_backend_sched_get_attempted_buffer_size(ggml_backend_sched_t sched, ggml_backend_t backend) {
++size_t ggml_backend_sched_get_attempted_buffer_size(ggml_backend_sched_t sched, ggml_backend_t backend) {
 +    int backend_index = ggml_backend_sched_backend_id(sched, backend);
 +    GGML_ASSERT(backend_index >= 0 && backend_index < sched->n_backends);
 +
-+    struct ggml_allocr_buffer_status allocr_status = ggml_gallocr_get_attempted_buffer_size(sched->galloc, backend_index);
-+    struct ggml_backend_buffer_status status = {allocr_status.size, allocr_status.allocated};
-+
-+    return status;
++    return ggml_gallocr_get_attempted_buffer_size(sched->galloc, backend_index);
 +}
 +
  void ggml_backend_sched_set_tensor_backend(ggml_backend_sched_t sched, struct ggml_tensor * node, ggml_backend_t backend) {