Skip to content

Precompile the compiler for devices without Float64 support #746

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

Merged
maleadt merged 1 commit into from
Oct 27, 2026
Merged

Precompile the compiler for devices without Float64 support #746

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
38 changes: 22 additions & 16 deletions src/compiler/precompile.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,24 +23,30 @@ if _precompile_lts ? SPIRV_LLVM_Translator_jll.is_available() :
# Build a device-independent compiler config. `_compiler_config` normally derives
# these knobs from the device; here we use conservative, portable defaults (the
# workload only exercises the pipeline, it does not target a specific device).
target = SPIRVCompilerTarget(;
backend = (_precompile_lts ? :khronos : :llvm),
extensions = "", supports_fp16 = true,
supports_fp64 = true, supports_bfloat16 = false, driver = :intel,
)
params = oneAPICompilerParams()
config = CompilerConfig(target, params; kernel=true, name=nothing,
always_inline=false)
# Cover both Float64 settings: each gives inference a different method table
# type (see `method_tables`).
for supports_fp64 in (true, false)
target = SPIRVCompilerTarget(;
backend = (_precompile_lts ? :khronos : :llvm),
extensions = "", supports_fp16 = true,
supports_fp64, supports_bfloat16 = false, driver = :intel,
)
params = oneAPICompilerParams()
config = CompilerConfig(target, params; kernel=true, name=nothing,
always_inline=false)

tt = Tuple{oneDeviceArray{Float32,1,AS.CrossWorkgroup}}
source = methodinstance(typeof(_precompile_kernel), tt)
job = CompilerJob(source, config)
tt = Tuple{oneDeviceArray{Float32,1,AS.CrossWorkgroup}}
source = methodinstance(typeof(_precompile_kernel), tt)
job = CompilerJob(source, config)

# On Julia < 1.12, driving GPU compilation during precompilation can leak foreign
# MethodInstances into host native compilation; only run the full compile on 1.12+.
@static if VERSION >= v"1.12-"
# Exercise the launch-side cache path and serialize its portable SPIR-V image.
compile_or_lookup(job)
# On Julia < 1.12, driving GPU compilation during precompilation can leak
# foreign MethodInstances into host native compilation; only run the full
# compile on 1.12+.
@static if VERSION >= v"1.12-"
# Exercise the launch-side cache path and serialize its portable SPIR-V
# image.
compile_or_lookup(job)
end
end
end
end
Expand Down
Loading