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, 5 insertions, 5 deletions

diff --git a/scripts/build-info.sh b/scripts/build-info.sh
index 32682afb..fa9e7bac 100755
--- a/scripts/build-info.sh
+++ b/scripts/build-info.sh
@@ -8,20 +8,20 @@ build_compiler="unknown"
 build_target="unknown"
 
 if out=$(git rev-list --count HEAD); then
-  # git is broken on WSL so we need to strip extra newlines
-  build_number=$(printf '%s' "$out" | tr -d '\n')
+    # git is broken on WSL so we need to strip extra newlines
+    build_number=$(printf '%s' "$out" | tr -d '\n')
 fi
 
 if out=$(git rev-parse --short HEAD); then
-  build_commit=$(printf '%s' "$out" | tr -d '\n')
+    build_commit=$(printf '%s' "$out" | tr -d '\n')
 fi
 
 if out=$($CC --version | head -1); then
-  build_compiler=$out
+    build_compiler=$out
 fi
 
 if out=$($CC -dumpmachine); then
-  build_target=$out
+    build_target=$out
 fi
 
 echo "int LLAMA_BUILD_NUMBER = ${build_number};"