Support Julia 1.13

[eschnett]

Closes #3019.

[github-actions]

CUDA.jl Benchmarks

Details

Benchmark suite	Current: cb74c7a	Previous: 7a27d77	Ratio
latency/precompile	44381498055.5 ns	44455759835 ns	1.00
latency/ttfp	13144733912 ns	13140153243 ns	1.00
latency/import	3768995999 ns	3755312424 ns	1.00
integration/volumerhs	9445468 ns	9442840 ns	1.00
integration/byval/slices=1	145662 ns	145598 ns	1.00
integration/byval/slices=3	422646.5 ns	422554 ns	1.00
integration/byval/reference	143785 ns	143811 ns	1.00
integration/byval/slices=2	284112 ns	284011 ns	1.00
integration/cudadevrt	102551 ns	102397 ns	1.00
kernel/indexing	13367.5 ns	13434 ns	1.00
kernel/indexing_checked	14039 ns	13908 ns	1.01
kernel/occupancy	643.6987951807229 ns	644.5636363636364 ns	1.00
kernel/launch	2085.6 ns	2090.3 ns	1.00
kernel/rand	14817.5 ns	14479 ns	1.02
array/reverse/1d	18616 ns	18661 ns	1.00
array/reverse/2dL_inplace	66221 ns	66252 ns	1.00
array/reverse/1dL	68793 ns	68893 ns	1.00
array/reverse/2d	20790 ns	21087 ns	0.99
array/reverse/1d_inplace	10501 ns	10503.833333333332 ns	1.00
array/reverse/2d_inplace	10453 ns	11399.5 ns	0.92
array/reverse/2dL	72780 ns	73163 ns	0.99
array/reverse/1dL_inplace	66066 ns	66146 ns	1.00
array/copy	18142 ns	18502.5 ns	0.98
array/iteration/findall/int	145556.5 ns	146476.5 ns	0.99
array/iteration/findall/bool	130349 ns	130795 ns	1.00
array/iteration/findfirst/int	84137 ns	84133 ns	1.00
array/iteration/findfirst/bool	81322 ns	81624.5 ns	1.00
array/iteration/scalar	66718 ns	65804 ns	1.01
array/iteration/logical	195274.5 ns	198187.5 ns	0.99
array/iteration/findmin/1d	84764.5 ns	86504 ns	0.98
array/iteration/findmin/2d	116397 ns	117154 ns	0.99
array/reductions/reduce/Int64/1d	39356 ns	41088.5 ns	0.96
array/reductions/reduce/Int64/dims=1	51634 ns	52190.5 ns	0.99
array/reductions/reduce/Int64/dims=2	58848 ns	59179 ns	0.99
array/reductions/reduce/Int64/dims=1L	87073.5 ns	87126 ns	1.00
array/reductions/reduce/Int64/dims=2L	84606 ns	84418.5 ns	1.00
array/reductions/reduce/Float32/1d	33956 ns	34001 ns	1.00
array/reductions/reduce/Float32/dims=1	39729.5 ns	39890 ns	1.00
array/reductions/reduce/Float32/dims=2	56655.5 ns	55899 ns	1.01
array/reductions/reduce/Float32/dims=1L	51474 ns	51535 ns	1.00
array/reductions/reduce/Float32/dims=2L	69730 ns	69798 ns	1.00
array/reductions/mapreduce/Int64/1d	39196 ns	40980.5 ns	0.96
array/reductions/mapreduce/Int64/dims=1	41696.5 ns	41741 ns	1.00
array/reductions/mapreduce/Int64/dims=2	58701 ns	59036 ns	0.99
array/reductions/mapreduce/Int64/dims=1L	87052 ns	87134 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	84512 ns	84427 ns	1.00
array/reductions/mapreduce/Float32/1d	33698 ns	33457 ns	1.01
array/reductions/mapreduce/Float32/dims=1	48904 ns	48711 ns	1.00
array/reductions/mapreduce/Float32/dims=2	55684 ns	55941 ns	1.00
array/reductions/mapreduce/Float32/dims=1L	51470 ns	51352 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	68897.5 ns	68956 ns	1.00
array/broadcast	20174 ns	20251 ns	1.00
array/copyto!/gpu_to_gpu	10408.166666666668 ns	10684.333333333334 ns	0.97
array/copyto!/cpu_to_gpu	214860 ns	214898 ns	1.00
array/copyto!/gpu_to_cpu	285587 ns	281876 ns	1.01
array/accumulate/Int64/1d	118031 ns	118336 ns	1.00
array/accumulate/Int64/dims=1	79306.5 ns	79780 ns	0.99
array/accumulate/Int64/dims=2	155444.5 ns	155968.5 ns	1.00
array/accumulate/Int64/dims=1L	1694458 ns	1694089 ns	1.00
array/accumulate/Int64/dims=2L	960288 ns	960949 ns	1.00
array/accumulate/Float32/1d	100200 ns	100823 ns	0.99
array/accumulate/Float32/dims=1	75596 ns	76350 ns	0.99
array/accumulate/Float32/dims=2	143914 ns	144365 ns	1.00
array/accumulate/Float32/dims=1L	1584412.5 ns	1584729 ns	1.00
array/accumulate/Float32/dims=2L	656263 ns	656302 ns	1.00
array/construct	1274.4 ns	1283.1 ns	0.99
array/random/randn/Float32	36142.5 ns	36610 ns	0.99
array/random/randn!/Float32	26781.5 ns	30335 ns	0.88
array/random/rand!/Int64	34387 ns	26934 ns	1.28
array/random/rand!/Float32	8197.25 ns	8186.666666666667 ns	1.00
array/random/rand/Int64	29742 ns	30201.5 ns	0.98
array/random/rand/Float32	12222 ns	12396 ns	0.99
array/permutedims/4d	51717.5 ns	52729 ns	0.98
array/permutedims/2d	52258 ns	52645 ns	0.99
array/permutedims/3d	52636 ns	53080 ns	0.99
array/sorting/1d	2735795 ns	2736443 ns	1.00
array/sorting/by	3305363 ns	3305811 ns	1.00
array/sorting/2d	1066907 ns	1071655.5 ns	1.00
cuda/synchronization/stream/auto	974.15 ns	1034.5263157894738 ns	0.94
cuda/synchronization/stream/nonblocking	7304.9 ns	7705.9 ns	0.95
cuda/synchronization/stream/blocking	803.0625 ns	784.4516129032259 ns	1.02
cuda/synchronization/context/auto	1169.5 ns	1133.5 ns	1.03
cuda/synchronization/context/nonblocking	7610.5 ns	7594.6 ns	1.00
cuda/synchronization/context/blocking	930.6842105263158 ns	885.6792452830189 ns	1.05

This comment was automatically generated by workflow using github-action-benchmark.

[eschnett]

The self-tests fail because the linear algebra functions (e.g. matrix exponential) as implemented in LinearAlgebra use scalar iteration. See e.g. exp! in https://github.com/JuliaLang/LinearAlgebra.jl/blob/f55e4736fb6dce08fee8a7ac7f0aba1f2b54838e/src/dense.jl#L784.

How should this be handled? Rewrite exp!? Find a respective CUDA library function to call and add a new method to exp? Fall back to the Julia 1.12 implementation? How does this work in Julia 1.12?

[eschnett]

I think it's JuliaGPU/GPUArrays.jl#679.

[eschnett]

The buildkite error is

  ptxas /tmp/jl_PALmvKnqta.ptx, line 226; error   : Modifier '.NaN' requires .target sm_80 or higher
  ptxas /tmp/jl_PALmvKnqta.ptx, line 226; error   : Feature 'min.f16 or min.f16x2' requires .target sm_80 or higher

This seems unrelated to my changes, except that I am now running CI tests on Julia 1.12 and Julia 1.13...

[maleadt]

I guess #3025 needs to be active for all LLVM versions.

[eschnett]

Good news: CUDA.jl now works for Julia 1.12.
Bad news: There's an LLVM segfault for Julia 1.13.

�_bk;t=1770145810814�      From worker 5:	[271397] signal 11 (1): Segmentation fault
�_bk;t=1770145810814�      From worker 5:	in expression starting at /var/lib/buildkite-agent/builds/gpuci-9/julialang/cuda-dot-jl/test/base/texture.jl:41
�_bk;t=1770145810920�      From worker 5:	_ZN12_GLOBAL__N_124NVPTXReplaceImageHandles18findIndexForHandleERN4llvm14MachineOperandERNS1_15MachineFunctionERj.isra.0 at /root/.cache/julia-buildkite-plugin/julia_installs/bin/linux/x64/1.13/julia-1.13-latest-linux-x86_64/bin/../lib/julia/libLLVM.so.20.1jl (unknown line)
�_bk;t=1770145810920�      From worker 5:	_ZN12_GLOBAL__N_124NVPTXReplaceImageHandles18findIndexForHandleERN4llvm14MachineOperandERNS1_15MachineFunctionERj.isra.0 at /root/.cache/julia-buildkite-plugin/julia_installs/bin/linux/x64/1.13/julia-1.13-latest-linux-x86_64/bin/../lib/julia/libLLVM.so.20.1jl (unknown line)
�_bk;t=1770145810920�      From worker 5:	_ZN12_GLOBAL__N_124NVPTXReplaceImageHandles20runOnMachineFunctionERN4llvm15MachineFunctionE at /root/.cache/julia-buildkite-plugin/julia_installs/bin/linux/x64/1.13/julia-1.13-latest-linux-x86_64/bin/../lib/julia/libLLVM.so.20.1jl (unknown line)
�_bk;t=1770145810921�      From worker 5:	_ZN4llvm19MachineFunctionPass13runOnFunctionERNS_8FunctionE.part.0 at /root/.cache/julia-buildkite-plugin/julia_installs/bin/linux/x64/1.13/julia-1.13-latest-linux-x86_64/bin/../lib/julia/libLLVM.so.20.1jl (unknown line)
�_bk;t=1770145810921�      From worker 5:	_ZN4llvm13FPPassManager13runOnFunctionERNS_8FunctionE at /root/.cache/julia-buildkite-plugin/julia_installs/bin/linux/x64/1.13/julia-1.13-latest-linux-x86_64/bin/../lib/julia/libLLVM.so.20.1jl (unknown line)
�_bk;t=1770145810921�      From worker 5:	_ZN4llvm13FPPassManager11runOnModuleERNS_6ModuleE at /root/.cache/julia-buildkite-plugin/julia_installs/bin/linux/x64/1.13/julia-1.13-latest-linux-x86_64/bin/../lib/julia/libLLVM.so.20.1jl (unknown line)
�_bk;t=1770145810922�      From worker 5:	_ZN4llvm6legacy15PassManagerImpl3runERNS_6ModuleE at /root/.cache/julia-buildkite-plugin/julia_installs/bin/linux/x64/1.13/julia-1.13-latest-linux-x86_64/bin/../lib/julia/libLLVM.so.20.1jl (unknown line)
�_bk;t=1770145810922�      From worker 5:	_ZL21LLVMTargetMachineEmitP23LLVMOpaqueTargetMachineP16LLVMOpaqueModuleRN4llvm17raw_pwrite_streamE19LLVMCodeGenFileTypePPc at /root/.cache/julia-buildkite-plugin/julia_installs/bin/linux/x64/1.13/julia-1.13-latest-linux-x86_64/bin/../lib/julia/libLLVM.so.20.1jl (unknown line)

[eschnett]

I think it's texture interpolation that is broken on 1.13. This line segfaults LLVM:

dst[i] = texture[u]

in test/base/texture.jl (function kernel_texture_warp_native).

[eschnett]

We will need to update KernelAbstractions.jl as well JuliaGPU/KernelAbstractions.jl#679.

[codecov]

Codecov Report

❌ Patch coverage is 50.00000% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 89.48%. Comparing base (7a27d77) to head (cb74c7a).

Files with missing lines	Patch %	Lines
lib/nvml/NVML.jl	50.00%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #3020      +/-   ##
==========================================
+ Coverage   89.46%   89.48%   +0.01%     
==========================================
  Files         148      148              
  Lines       13047    13044       -3     
==========================================
- Hits        11673    11672       -1     
+ Misses       1374     1372       -2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[eschnett]

All green!