Compile bug: can't build with MSVC + Vulkan

[foldl]

Git commit

git rev-parse HEAD
d7b31a9d84297b493a61c9a8ee3a458e7ccc64a7

Operating systems

Windows

GGML backends

Vulkan

Problem description & steps to reproduce

I'm trying to build with MSVC + Vulkan:

cmake -B build-vulkan -DGGML_VULKAN=1
cmake --build build-vulkan --config Release

CMake log about Vulkan:

-- Vulkan found
-- GL_KHR_cooperative_matrix supported by glslc
-- GL_NV_cooperative_matrix2 not supported by glslc
-- Including Vulkan backend

Lots of undefined identifiers when building:

llama.cpp\ggml-vulkan\ggml-vulkan.cpp(1732,13): error C2065: “matmul_iq2_xxs_f32_f16acc_coopmat_len”: ...
llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(1732,13): error C2065: “matmul_iq2_xxs_f32_f16acc_coopmat_data”: ...
llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(1732,13): error C2065: “matmul_iq2_xxs_f32_aligned_f16acc_coopmat_len”: ...

First Bad Commit

b685daf

Compile command

cmake -B build-vulkan -DGGML_VULKAN=1
cmake --build build-vulkan --config Release

Relevant log output

Lots of undefined identifiers when building commit b685daf3867c54e42a9db484d7b92619021d4510:


llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(1923,90): error C2065: “rope_norm_f16_rte_len”: ...

[jeffbolznv]

Please try installing the latest vulkan sdk, and if it still fails attach a complete log for a clean build.

[foldl]

Updated to VulkanSDK-1.4.304.1, there are still lots of undefined identifiers:

ggml\src\ggml-vulkan\ggml-vulkan.cpp(1613,9): error C2065: “flash_attn_f32_f16_f16_f16acc_cm2_len”: ...
...

[jeffbolznv]

Please share a complete log for a clean build (like delete the whole build folder, to avoid any cmake caching).

[foldl]

@jeffbolznv full log (commit 89daa25).

cmake -B build-vulkan -DGGML_VULKAN=1

-- Building for: Visual Studio 17 2022
-- Selecting Windows SDK version 10.0.22621.0 to target Windows 10.0.26100.
-- The C compiler identification is MSVC 19.41.34120.0
-- The CXX compiler identification is MSVC 19.41.34120.0
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: C:/Program Files/Microsoft Visual Studio/2022/Community/VC/Tools/MSVC/14.41.34120/bin/Hostx64/x64/cl.exe - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: C:/Program Files/Microsoft Visual Studio/2022/Community/VC/Tools/MSVC/14.41.34120/bin/Hostx64/x64/cl.exe - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Found Git: C:/Programs/Git/cmd/git.exe (found version "2.40.0.windows.1") 
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Failed
-- Looking for pthread_create in pthreads
-- Looking for pthread_create in pthreads - not found
-- Looking for pthread_create in pthread
-- Looking for pthread_create in pthread - not found
-- Found Threads: TRUE  
-- ccache found, compilation results will be cached. Disable with GGML_CCACHE=OFF.
-- CMAKE_SYSTEM_PROCESSOR: AMD64
-- CMAKE_GENERATOR_PLATFORM: 
-- Including CPU backend
-- Found OpenMP_C: -openmp (found version "2.0") 
-- Found OpenMP_CXX: -openmp (found version "2.0") 
-- Found OpenMP: TRUE (found version "2.0")  
-- x86 detected
-- Performing Test HAS_AVX_1
-- Performing Test HAS_AVX_1 - Success
-- Performing Test HAS_AVX2_1
-- Performing Test HAS_AVX2_1 - Success
-- Performing Test HAS_FMA_1
-- Performing Test HAS_FMA_1 - Success
-- Performing Test HAS_AVX512_1
-- Performing Test HAS_AVX512_1 - Success
-- Adding CPU backend variant ggml-cpu: /arch:AVX512 GGML_AVX512
-- Found Vulkan: C:/VulkanSDK/1.4.304.1/Lib/vulkan-1.lib (found version "1.4.304") found components: glslc glslangValidator 
-- Vulkan found
-- GL_KHR_cooperative_matrix supported by glslc
-- GL_NV_cooperative_matrix2 supported by glslc
-- Including Vulkan backend
-- Configuring done (22.0s)
-- Generating done (0.6s)
-- Build files have been written to: C:/projects/llama.cpp/build-vulkan

cmake --build build-vulkan --config Release -j8

.NET Framework MSBuild Version 17.11.2+c078802d4

  Checking Build System
  Building Custom Rule C:/projects/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/CMakeLists.txt
  Building Custom Rule C:/projects/llama.cpp/ggml/src/CMakeLists.txt
  Building Custom Rule C:/projects/llama.cpp/examples/gguf-hash/CMakeLists.txt
  Generating build details from Git
  Building Custom Rule C:/projects/llama.cpp/examples/gguf-hash/CMakeLists.txt
  Building Custom Rule C:/projects/llama.cpp/examples/gguf-hash/CMakeLists.txt
  vulkan-shaders-gen.cpp
  -- Found Git: C:/Programs/Git/cmd/git.exe (found version "2.40.0.windows.1") 
  xxhash.c
  sha1.c
  ggml.c
  sha256.c
  Building Custom Rule C:/projects/llama.cpp/common/CMakeLists.txt
  ggml-alloc.c
  build-info.cpp
  ggml-quants.c
  sha1.vcxproj -> C:\projects\llama.cpp\build-vulkan\examples\gguf-hash\sha1.dir\Release\sha1.lib
  sha256.vcxproj -> C:\projects\llama.cpp\build-vulkan\examples\gguf-hash\sha256.dir\Release\sha256.lib
  Generating Code...
C:\projects\llama.cpp\ggml\src\ggml.c(1955,26): warning C4267: 'initializing': conversion from 'size_t' to 'int32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(1955,31): warning C4267: 'initializing': conversion from 'size_t' to 'int32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(1955,36): warning C4267: 'initializing': conversion from 'size_t' to 'int32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(1955,41): warning C4267: 'initializing': conversion from 'size_t' to 'int32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(2839,26): warning C4267: 'initializing': conversion from 'size_t' to 'int32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(2839,31): warning C4267: 'initializing': conversion from 'size_t' to 'int32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(2839,36): warning C4267: 'initializing': conversion from 'size_t' to 'int32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(2839,41): warning C4267: 'initializing': conversion from 'size_t' to 'int32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(3227,22): warning C4244: '=': conversion from 'int64_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(3228,22): warning C4244: '=': conversion from 'int64_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(3229,22): warning C4244: '=': conversion from 'int64_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(3230,22): warning C4244: '=': conversion from 'int64_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(3232,22): warning C4267: '=': conversion from 'size_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(3233,22): warning C4267: '=': conversion from 'size_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(3234,22): warning C4267: '=': conversion from 'size_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(3235,22): warning C4267: '=': conversion from 'size_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(3874,48): warning C4244: 'function': conversion from 'int64_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(3906,52): warning C4244: 'function': conversion from 'int64_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(3980,41): warning C4244: 'function': conversion from 'int64_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(3980,51): warning C4244: 'function': conversion from 'int64_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(3989,51): warning C4244: 'function': conversion from 'int64_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(3989,65): warning C4244: 'function': conversion from 'int64_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(4050,35): warning C4244: 'return': conversion from 'float' to 'int64_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(4063,54): warning C4244: 'function': conversion from 'int' to 'float', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(4100,46): warning C4244: 'initializing': conversion from 'float' to 'int32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]       
C:\projects\llama.cpp\ggml\src\ggml.c(4100,50): warning C4244: 'initializing': conversion from 'float' to 'int32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]       
C:\projects\llama.cpp\ggml\src\ggml.c(4123,46): warning C4244: 'initializing': conversion from 'float' to 'int32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]       
C:\projects\llama.cpp\ggml\src\ggml.c(4123,50): warning C4244: 'initializing': conversion from 'float' to 'int32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]       
C:\projects\llama.cpp\ggml\src\ggml.c(4159,47): warning C4244: 'function': conversion from 'int64_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(4159,72): warning C4244: 'function': conversion from 'int64_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(4159,93): warning C4244: 'function': conversion from 'int64_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(4159,103): warning C4244: 'function': conversion from 'int64_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml.c(4530,36): warning C4244: 'initializing': conversion from 'int64_t' to 'const int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]   
C:\projects\llama.cpp\ggml\src\ggml.c(4531,36): warning C4244: 'initializing': conversion from 'int64_t' to 'const int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]   
C:\projects\llama.cpp\ggml\src\ggml.c(5667,108): warning C4244: 'function': conversion from 'const int32_t' to 'float', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]    
C:\projects\llama.cpp\ggml\src\ggml.c(5667,112): warning C4244: 'function': conversion from 'const int32_t' to 'float', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]    
C:\projects\llama.cpp\ggml\src\ggml.c(5945,29): warning C4267: 'initializing': conversion from 'size_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\Program Files (x86)\Windows Kits\10\Include\10.0.22621.0\ucrt\assert.h(21,9): warning C4005: 'static_assert': macro redefinition [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
  (compiling source file '../../../ggml/src/ggml-quants.c')
      C:\projects\llama.cpp\ggml\src\ggml-common.h(79,9):
      see previous definition of 'static_assert'

  build_info.vcxproj -> C:\projects\llama.cpp\build-vulkan\common\build_info.dir\Release\build_info.lib
  xxhash.vcxproj -> C:\projects\llama.cpp\build-vulkan\examples\gguf-hash\xxhash.dir\Release\xxhash.lib
  ggml-backend.cpp
  ggml-opt.cpp
C:\projects\llama.cpp\ggml\src\ggml-opt.cpp(599,77): warning C4244: 'argument': conversion from 'int64_t' to 'float', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml-opt.cpp(600,77): warning C4244: 'argument': conversion from 'int64_t' to 'float', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml-opt.cpp(810,30): warning C4244: '=': conversion from 'const int64_t' to 'int32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]     
  ggml-threading.cpp
  gguf.cpp
C:\projects\llama.cpp\ggml\src\gguf.cpp(457,48): warning C4244: 'initializing': conversion from 'int64_t' to 'int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\gguf.cpp(457,33): warning C4244: 'initializing': conversion from 'int64_t' to 'const int', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
C:\projects\llama.cpp\ggml\src\gguf.cpp(592,21): warning C4267: 'argument': conversion from 'size_t' to 'long', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-base.vcxproj]
  vulkan-shaders-gen.vcxproj -> C:\projects\llama.cpp\build-vulkan\bin\vulkan-shaders-gen.exe
  Generating Code...
     Creating library C:/projects/llama.cpp/build-vulkan/ggml/src/Release/ggml-base.lib and object C:/projects/llama.cpp/build-vulkan/ggml/src/Release/ggml-base.exp
  ggml-base.vcxproj -> C:\projects\llama.cpp\build-vulkan\bin\Release\ggml-base.dll
  Building Custom Rule C:/projects/llama.cpp/ggml/src/CMakeLists.txt
  Generate vulkan shaders
  ggml_vulkan: Generating and compiling shaders to SPIR-V
  ggml-cpu.c
  ggml-cpu.cpp
  ggml-cpu-aarch64.cpp
  ggml-cpu-hbm.cpp
  ggml-cpu-traits.cpp
  amx.cpp
  mmq.cpp
  sgemm.cpp
  Generating Code...
  ggml-cpu-quants.c
C:\Program Files (x86)\Windows Kits\10\Include\10.0.22621.0\ucrt\assert.h(21,9): warning C4005: 'static_assert': macro redefinition [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-cpu.vcxproj]
  (compiling source file '../../../ggml/src/ggml-cpu/ggml-cpu-quants.c')
      C:\projects\llama.cpp\ggml\src\ggml-common.h(79,9):
      see previous definition of 'static_assert'

     Creating library C:/projects/llama.cpp/build-vulkan/ggml/src/Release/ggml-cpu.lib and object C:/projects/llama.cpp/build-vulkan/ggml/src/Release/ggml-cpu.exp
  ggml-cpu.vcxproj -> C:\projects\llama.cpp\build-vulkan\bin\Release\ggml-cpu.dll
  Building Custom Rule C:/projects/llama.cpp/ggml/src/ggml-vulkan/CMakeLists.txt
  ggml-vulkan.cpp
  ggml-vulkan-shaders.cpp
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(824,48): warning C4267: '=': conversion from 'size_t' to 'uint32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-vulkan\g
gml-vulkan.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(825,46): warning C4267: '=': conversion from 'size_t' to 'uint32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-vulkan\g
gml-vulkan.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(830,36): warning C4267: 'argument': conversion from 'size_t' to 'uint32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-v 
ulkan\ggml-vulkan.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(932,27): warning C4267: 'initializing': conversion from 'size_t' to 'uint32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\gg 
ml-vulkan\ggml-vulkan.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(934,27): warning C4267: 'initializing': conversion from 'size_t' to 'uint32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\gg 
ml-vulkan\ggml-vulkan.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(1075,52): warning C4267: 'initializing': conversion from 'size_t' to 'uint32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\g 
gml-vulkan\ggml-vulkan.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(1075,27): warning C4267: 'initializing': conversion from 'size_t' to 'const uint32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml 
\src\ggml-vulkan\ggml-vulkan.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(1087,20): warning C4267: 'return': conversion from 'size_t' to 'uint32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-vu 
lkan\ggml-vulkan.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(1094,20): warning C4267: 'return': conversion from 'size_t' to 'uint32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-vu 
lkan\ggml-vulkan.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(1101,20): warning C4267: 'return': conversion from 'size_t' to 'uint32_t', possible loss of data [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-vu 
lkan\ggml-vulkan.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(1614,9): error C2065: 'flash_attn_f32_f16_f16_f16acc_cm2_len': undeclared identifier [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-vulkan\ggml-vu 
lkan.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(1614,9): error C2065: 'flash_attn_f32_f16_f16_f16acc_cm2_data': undeclared identifier [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-vulkan\ggml-v 
ulkan.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(1614,9): error C2065: 'flash_attn_f32_f16_f16_cm2_len': undeclared identifier [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-vulkan\ggml-vulkan.vc 
xproj]
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(1614,9): error C2065: 'flash_attn_f32_f16_f16_cm2_data': undeclared identifier [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-vulkan\ggml-vulkan.v 
cxproj]
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(1615,9): error C2065: 'flash_attn_f32_f16_q4_0_f16acc_cm2_len': undeclared identifier [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-vulkan\ggml-v 
ulkan.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(1615,9): error C2065: 'flash_attn_f32_f16_q4_0_f16acc_cm2_data': undeclared identifier [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-vulkan\ggml- 
vulkan.vcxproj]
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(1615,9): error C2065: 'flash_attn_f32_f16_q4_0_cm2_len': undeclared identifier [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-vulkan\ggml-vulkan.v 
cxproj]
C:\projects\llama.cpp\ggml\src\ggml-vulkan\ggml-vulkan.cpp(1615,9): error C1003: error count exceeds 100; stopping compilation [C:\projects\llama.cpp\build-vulkan\ggml\src\ggml-vulkan\ggml-vulkan.vcxproj]     

[jeffbolznv]

I tried with a clean tree and same commit and build steps, and haven't been able to reproduce it. It seems like GGML_VULKAN_COOPMAT2_GLSLC_SUPPORT is not getting set during the shader generation, but there's nothing in the log to indicate why that might happen.

[jeffbolznv]

It's a longshot, but please check if https://github.com/jeffbolznv/llama.cpp/tree/glslc_mismatch helps.

[foldl]

This is because the generated ggml-vulkan-shaders.cpp/hpp is moved from ggml/src/ggml-vulkan to CMAKE_DIR/ggml/src/ggml-vulkan. Unfortunately, in my case, after updating the repo, the elder version of ggml-vulkan-shaders.cpp/hpp are still there in ggml/src/ggml-vulkan, so, the newly generated .hpp is not use. A clean build would not help.

Just delete ggml-vulkan-shaders.cpp/hpp in ggml/src/ggml-vulkan. Case resolved.