Merge mainline llama.cpp (#3)

* Merging mainline - WIP

* Merging mainline - WIP

AVX2 and CUDA appear to work.
CUDA performance seems slightly (~1-2%) lower as it is so often
the case with llama.cpp/ggml after some "improvements" have been made.

* Merging mainline - fix Metal

* Remove check

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>

1 files changed, 15 insertions, 0 deletions

diff --git a/examples/llama-bench/llama-bench.cpp b/examples/llama-bench/llama-bench.cpp
index d641a9f1..a6497b6e 100644
--- a/examples/llama-bench/llama-bench.cpp
+++ b/examples/llama-bench/llama-bench.cpp
@@ -23,6 +23,10 @@
 #include "ggml-cuda.h"
 #include "ggml-sycl.h"
 
+#ifdef GGML_USE_CANN
+#include "ggml-cann.h"
+#endif
+
 // utils
 static uint64_t get_time_ns() {
     using clock = std::chrono::high_resolution_clock;
@@ -121,6 +125,17 @@ static std::string get_gpu_info() {
         }
     }
 #endif
+#ifdef GGML_USE_CANN
+    uint32_t count = ggml_backend_cann_get_device_count();
+    for (uint32_t i = 0; i < count; i++) {
+        char buf[128];
+        ggml_backend_cann_get_device_description(i, buf, sizeof(buf));
+        id += buf;
+        if (i < count - 1) {
+            id += "/";
+        }
+    }
+#endif
     // TODO: other backends
     return id;
 }