Skip to content

Allow disabling the linking of libdevice in CUDACompilerParams #2611

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

Draft
gbaraldi wants to merge 3 commits into
base: master
Choose a base branch
from
+11 −6
Draft

Allow disabling the linking of libdevice in CUDACompilerParams #2611

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
4b58767
Allow disabling the linking of libdevice in CUDACompilerParams
gbaraldi Jan 8, 2025
b3364d3
Address some comments
gbaraldi Feb 6, 2025
d199899
Merge branch 'master' into gb/reactant-fun
gbaraldi Feb 6, 2025
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
17 changes: 11 additions & 6 deletions src/compiler/compilation.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,12 +3,13 @@
Base.@kwdef struct CUDACompilerParams <: AbstractCompilerParams
cap::VersionNumber
ptx::VersionNumber
link_libdevice::Bool = true # Used by Reactant.jl
end

function Base.hash(params::CUDACompilerParams, h::UInt)
h = hash(params.cap, h)
h = hash(params.ptx, h)

h = hash(params.link_libdevice, h)
return h
end

Expand All @@ -18,15 +19,19 @@ const CUDACompilerJob = CompilerJob{PTXCompilerTarget,CUDACompilerParams}
GPUCompiler.runtime_module(@nospecialize(job::CUDACompilerJob)) = CUDA

# filter out functions from libdevice and cudadevrt
GPUCompiler.isintrinsic(@nospecialize(job::CUDACompilerJob), fn::String) =
invoke(GPUCompiler.isintrinsic,
Tuple{CompilerJob{PTXCompilerTarget}, typeof(fn)},
job, fn) ||
fn == "__nvvm_reflect" || startswith(fn, "cuda")
function GPUCompiler.isintrinsic(@nospecialize(job::CUDACompilerJob), fn::String)
is_intrinsic = invoke(GPUCompiler.isintrinsic,
Tuple{CompilerJob{PTXCompilerTarget}, typeof(fn)}, job, fn)
is_intrinsic |= fn == "__nvvm_reflect"
is_intrinsic |= startswith(fn, "cuda")
is_intrinsic |= !job.config.params.link_libdevice ? startswith(fn, "__nv_") : false # Reactant.jl wants to handle __nv_ functions
return is_intrinsic
end

# link libdevice
function GPUCompiler.link_libraries!(@nospecialize(job::CUDACompilerJob), mod::LLVM.Module,
undefined_fns::Vector{String})
job.config.params.link_libdevice || return
# only link if there's undefined __nv_ functions
if !any(fn->startswith(fn, "__nv_"), undefined_fns)
return
Expand Down