Build on FreeBSD

[abelbabel]

Hi,

I could compile this on FreeBSD 13.1-RELEASE-p2 amd64, having devel/gmake installed (using gmake then instead of make) and using the following modifications:

--- Makefile_ori        2022-10-16 21:19:22.498824000 +0200
+++ Makefile    2022-10-16 22:40:53.787014000 +0200
@@ -22,10 +22,17 @@
        CFLAGS   += -pthread
        CXXFLAGS += -pthread
 endif
+ifeq ($(UNAME_S),FreeBSD)
+       CFLAGS   += -pthread
+       CXXFLAGS += -pthread
+endif
 
 # Architecture specific
 # TODO: probably these flags need to be tweaked on some architectures
 ifeq ($(UNAME_M),x86_64)
+       CFLAGS += -mavx -mavx2 -mfma -mf16c
+endif
+ifeq ($(UNAME_M),amd64)
        CFLAGS += -mavx -mavx2 -mfma -mf16c
 endif
 ifneq ($(filter arm%,$(UNAME_M)),)

(don't know gmake-Makefiles too much, could be prettier with logical or here ...)

--- ggml.c_ori  2022-10-16 21:19:22.502786000 +0200
+++ ggml.c      2022-10-16 21:28:00.140594000 +0200
@@ -2,7 +2,7 @@

 #if defined(_MSC_VER) || defined(__MINGW32__)
 #include <malloc.h> // using malloc.h with MSC/MINGW
-#else
+#elif !defined(__FreeBSD__)
 #include <alloca.h>
 #endif

Seems not so hard to merge changes into upstream ...

For downloading models ftp/wget is needed.

Kind regards,
abelbabel

[abelbabel]

Found recently that shells/bash also has to be installed in order to be able to download models.

[abelbabel]

For downloading samples (as defined by Makefile) multimedia/ffmpeg is also needed ...

[abelbabel]

And an additional note, since I saw whisper.cpp being compiled with GCC: on the above FreeBSD it uses clang by default:

$ cc --version
FreeBSD clang version 13.0.0 (git@github.com:llvm/llvm-project.git llvmorg-13.0.0-0-gd7b669b3a303)
...
$ cpp --version
FreeBSD clang version 13.0.0 (git@github.com:llvm/llvm-project.git llvmorg-13.0.0-0-gd7b669b3a303)
...

[Baenwort]

I would be interested in getting this to work for me @abelbabel as I'm starting to plan how to use Whisper to create srt subtitles for broadcasts I've recorded that either didn't have functioning CC or only had them in English for other language audio.

Any chance you might make a port package or some instructions that are step by step from a fresh jail?

[abelbabel]

In fact everything should already be here.

To sum it up:

1. Install pre-requisites

# pkg install devel/gmake shells/bash ftp/wget git-lite

2. Clone the git repo ...
3. in root of git repo do gmake
4. use the program like $ ./main -m ...

I recently got the impression that if you want/need to compile it for other architectures than amd64, you might need to improve Makefile ...

Please have a try and if you run in some trouble, share them.

[abelbabel]

You could also use an "ad-hoc"-jail using the main systems bins and libs (so you won't need an extra jails) - to be 1000%-sure the whisper-implementation does not call home, you can even disallow all ip-access ...

example:

# jail -c -u USER path=/ ip4=disable ip6=disable command=WHISPER_ROOT/main -m WHISPER_ROOT/models/ggml-small.bin WAVE-FILE-WITH-FULL-PATH

[abelbabel]

While being here: depending on what "supported" means, FreeBSD could also be included in the "supported platforms" section in readme.

[leonardorame]

Hi, I cannot compile the examples/stream program on FreeBSD 13.1 amd64. Just went to examples/stream and tried make and/or gmake, but nothing was built.

Any pointers on how to build it?

[leonardorame]

I found a way!:

sudo pkg install sdl2
cmake -D WHISPER_SUPPORT_SDL2=ON .
make