-
Notifications
You must be signed in to change notification settings - Fork 13.7k
Closed
Labels
bug-unconfirmedcritical severityUsed to report critical severity bugs in llama.cpp (e.g. Crashing, Corrupted, Dataloss)Used to report critical severity bugs in llama.cpp (e.g. Crashing, Corrupted, Dataloss)
Description
What happened?
When trying to build the lastest version of the Vulkan backend, the shader compilation fails.
I suspect commit 17eb6aa to have introduced the issue, but more testing is required to know for sure.
Name and Version
commit: 3807c3d
cmake flags: -DBUILD_SHARED_LIBS=OFF -DGGML_VULKAN=1 -G "Visual Studio 17 2022" -A x64
MSBuild version: 17.9.8+b34f75857
Vulkan Instance Version: 1.3.261
What operating system are you seeing the problem on?
Windows
Relevant log output
Version MSBuild 17.9.8+b34f75857 pour .NET Framework
[...]
Generate vulkan shaders
ggml_vulkan: Generating and compiling shaders to SPIR-V
cannot compile mul_mat_vec_id_f32_f32
C:/VulkanSDK/1.3.275.0/Bin/glslc.exe -fshader-stage=compute --target-env=vulkan1.2 -O C:/llama cpp/ggml/src/vulkan-shaders\mul_mat_vec.comp -o C:/llama cpp/build/ggml/src/vulkan-shaders.spv\mul_mat_vec_id
_f32_f32.spv -DB_TYPE=float -DDATA_A_F32=1 -DD_TYPE=float -DFLOAT_TYPE=float -DMUL_MAT_ID=1
glslc : error : linking multiple files is not supported yet. Use -c to compile files individually. [C:\llama cpp\build\ggml\src\ggml.vcxproj]
cannot compile cannot compile mul_mat_vec_q4_0_f32_f32mul_mat_vec_q4_0_f16_f32
C:/VulkanSDK/1.3.275.0/Bin/glslc.exe -fshader-stage=compute --target-env=vulkan1.2 -O C:/llama cpp/ggml/src/vulkan-shaders\mul_mat_vec.comp -o C:/llama cpp/build/ggml/src/vulkan-shaders.spv\mul_mat_vec_q4
_0_f32_f32.spv -DB_TYPE=float -DDATA_A_Q4_0=1 -DD_TYPE=float -DFLOAT_TYPE=float
glslc : error : linking multiple files is not supported yet. Use -c to compile files individually. [C:\llama cpp\build\ggml\src\ggml.vcxproj]
(this goes on for every op, and build crashes with code -1)Metadata
Metadata
Assignees
Labels
bug-unconfirmedcritical severityUsed to report critical severity bugs in llama.cpp (e.g. Crashing, Corrupted, Dataloss)Used to report critical severity bugs in llama.cpp (e.g. Crashing, Corrupted, Dataloss)