ggerganov
fix building & UTF8 output with MSVC + SDL2
This PR could also solve the issue mentioned in #2804, in which, TDM64 (gcc compiler) is used because MSVC failed to build.
Note: example lsp talks to vim script through std IO, so not updated.
#2804 is 99.99% just SDL2.dll not being with the binaries. So I'm a little confused by this PR's naming.
#2804 is 99.99% just SDL2.dll not being with the binaries. So I'm a little confused by this PR's naming.
With this PR, MSVC can be used to build with SDL2. This is an issue mentioned in #2804, in which, since MSVC failed to build, TDM64 (gcc compiler) is used.
Thanks, this fixes the whisper-stream build for me on Win10 with MSVC, but the app is exiting immediately after launch with no indication of why:
C:\Users\ryan\Speech\whisper.cpp>.\build\bin\Release\whisper-stream.exe -m .\ggml-large-v3.bin
init: found 2 capture devices:
init: - Capture device #0: 'Headset Microphone (Logitech Mono H650e)'
init: - Capture device #1: 'Headset (Ryan’s AirPods Pro - Find My Hands-Free AG Audio)'
init: attempt to open default capture device ...
init: obtained spec for input device (SDL Id = 2):
init: - sample rate: 16000
init: - format: 33056 (required: 33056)
init: - channels: 1 (required: 1)
init: - samples per frame: 1024
whisper_init_from_file_with_params ...
whisper_init_from_file_with_params_no_state: loading model from '.\ggml-large-v3.bin'
whisper_init_with_params_no_state: use gpu = 1
whisper_init_with_params_no_state: flash attn = 0
whisper_init_with_params_no_state: gpu_device = 0
whisper_init_with_params_no_state: dtw = 0
ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = AMD Radeon RX 7900 XT (AMD proprietary driver) | uma: 0 | fp16: 1 | warp size: 64 | matrix cores: KHR_coopmat
whisper_init_with_params_no_state: devices = 2
whisper_init_with_params_no_state: backends = 2
whisper_model_load: loading model
whisper_model_load: n_vocab = 51866
whisper_model_load: n_audio_ctx = 1500
whisper_model_load: n_audio_state = 1280
whisper_model_load: n_audio_head = 20
whisper_model_load: n_audio_layer = 32
whisper_model_load: n_text_ctx = 448
whisper_model_load: n_text_state = 1280
whisper_model_load: n_text_head = 20
whisper_model_load: n_text_layer = 32
whisper_model_load: n_mels = 128
whisper_model_load: ftype = 1
whisper_model_load: qntvr = 0
whisper_model_load: type = 5 (large v3)
whisper_model_load: adding 1609 extra tokens
whisper_model_load: n_langs = 100
C:\Users\ryan\Speech\whisper.cpp>
@tokyovigilante You need to copy SDL2.dll to C:\Users\ryan\Speech\whisper.cpp\build\bin\Release, or add it to PATH.
Thanks for looking, I think this should be ok?
Directory of C:\Users\ryan\Speech\whisper.cpp\build\bin\Release
20/02/2025 10:05 pm <DIR> .
20/02/2025 10:05 pm <DIR> ..
20/02/2025 09:40 pm 27,136 bench.exe
20/02/2025 09:40 pm 27,136 command.exe
20/02/2025 09:40 pm 487,936 ggml-base.dll
20/02/2025 09:40 pm 322,048 ggml-cpu.dll
20/02/2025 09:40 pm 20,034,048 ggml-vulkan.dll
20/02/2025 09:40 pm 75,264 ggml.dll
20/02/2025 10:05 pm 184,832 lsp.exe
20/02/2025 09:40 pm 27,136 main.exe
20/02/2025 09:40 pm 107,520 quantize.exe
20/02/2025 09:11 pm 1,664,000 SDL2.dll
20/02/2025 09:40 pm 27,136 stream.exe
20/02/2025 10:05 pm 165,888 wchess.exe
20/02/2025 09:40 pm 20,480 whisper-bench.exe
20/02/2025 09:40 pm 183,808 whisper-cli.exe
20/02/2025 10:05 pm 167,936 whisper-command.exe
20/02/2025 10:05 pm 407,040 whisper-server.exe
20/02/2025 10:05 pm 81,920 whisper-stream.exe
20/02/2025 10:05 pm 1,200,128 whisper-talk-llama.exe
20/02/2025 09:40 pm 438,272 whisper.dll
19 File(s) 25,649,664 bytes
2 Dir(s) 33,750,409,216 bytes free
[EDIT]
Hmm, seems to run on Windows for me with the -ng flag, despite having the Vulkan SDK installed (7900 XT). Have just tried on Linux with the mainline code whiich works fine with my GPU. I really need to use this with a Windows app unfortunately, but this is clearly not the issue. Sorry for the noise.
@foldl Sorry about the delay on this. Would you be able rebase this and resolve conflicts here and we should be able to merge this.
