Skip to content

fix(meson): propagate HAVE_CUDA/HAVE_SYCL to 3 more static libs (ODR class follow-up to #840) #842

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
lusoris wants to merge 1 commit into from
Closed

fix(meson): propagate HAVE_CUDA/HAVE_SYCL to 3 more static libs (ODR class follow-up to #840) #842

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
40 changes: 30 additions & 10 deletions core/src/meson.build
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -899,6 +899,27 @@ if is_nvtx_enabled
include_directories: include_directories('/usr/local/cuda/include'))
endif

# ODR fix (follow-up to PR #840): propagate HAVE_CUDA / HAVE_NVTX / HAVE_SYCL
# into vmaf_cflags_common HERE — before any static_library target captures it —
# so that cuda_static_lib (line ~1178), libvmaf_feature_static_lib (line ~1700),
# and every other lib that uses `c_args: vmaf_cflags_common` sees the defines.
# Previously these three lines appeared after the libvmaf sources list (line ~1856),
# which meant every static-lib target built before that point compiled without the
# defines, producing an ODR violation: VmafPicturePrivate had different sizes in
# the allocator TU vs the consumer TU. Moving the block here (before
# `if is_cuda_enabled`) fixes cuda_static_lib and libvmaf_feature_static_lib.
# gpu_picture_pool_cpp23_lib carries an explicit cpp_args below (it never uses
# vmaf_cflags_common directly).
if is_cuda_enabled
vmaf_cflags_common += '-DHAVE_CUDA'
endif
if is_nvtx_enabled
vmaf_cflags_common += '-DHAVE_NVTX'
endif
if is_sycl_enabled
vmaf_cflags_common += '-DHAVE_SYCL'
endif

if is_cuda_enabled
# Check for required CUDA headers
if not cc.has_header('ffnvcodec/dynlink_cuda.h')
Expand Down Expand Up @@ -1796,6 +1817,15 @@ gpu_picture_pool_cpp23_lib = static_library(
include_directories : [vmaf_base_include, libvmaf_include],
# libvmaf_cpu_cpp_std: compiler-aware token (ADR-0860 follow-up, PR #692 fix).
override_options : ['cpp_std=' + libvmaf_cpu_cpp_std],
# ODR fix (follow-up to PR #840): gpu_picture_pool.cpp includes picture.h
# (line 30) and gpu_picture_pool.h (line 31, which re-includes picture.h).
# VmafPicturePrivate's cuda sub-struct changes size when HAVE_CUDA is absent,
# shifting buf_type and other fields. This lib does not inherit vmaf_cflags_common
# (it carries no c_args at all), so the defines must be passed explicitly —
# mirroring the same fix applied to picture_pool_cpp23_lib in PR #840.
cpp_args : (is_cuda_enabled ? ['-DHAVE_CUDA'] : []) +
(is_nvtx_enabled ? ['-DHAVE_NVTX'] : []) +
(is_sycl_enabled ? ['-DHAVE_SYCL'] : []),
dependencies : [pthread_dependency],
pic : true,
install : false,
Expand Down Expand Up @@ -1853,16 +1883,6 @@ libvmaf_sources = [
# binaries that link against libvmaf.a without rust_tad_dep).
] + rust_tad_direct_sources

if is_cuda_enabled
vmaf_cflags_common += '-DHAVE_CUDA'
endif
if is_nvtx_enabled
vmaf_cflags_common += '-DHAVE_NVTX'
endif
if is_sycl_enabled
vmaf_cflags_common += '-DHAVE_SYCL'
endif

# `-fsycl` is embedded in sycl_dependency.link_args (see the
# declare_dependency block above) so every consumer — libvmaf.so and all
# SYCL test executables — gets it automatically without per-target
Expand Down
Loading