[maleadt]

The launch configuration tests asserted which kind of launch KernelAbstractions picks for a few shapes. That choice is KernelAbstractions' policy, not something the back end controls, and the assertions break whenever it changes: JuliaGPU/KernelAbstractions.jl#919 launches 4-d ranges on the N-d grid, which fails them on every back end.

This keeps the checks of what the back end does affect: that a dynamic 3-d range is indexed without divisions, and that iteration spaces too large for 32-bit indices are indexed correctly. The latter now covers a 4-d range that is launched on the N-d grid with JuliaGPU/KernelAbstractions.jl#919, and one that is only large along its last dimension.

The tests pass with KernelAbstractions main and with JuliaGPU/KernelAbstractions.jl#919, so this can be merged first.

[github-actions]

CUDA.jl Benchmarks

Details

Benchmark suite	Current: 12e67b0	Previous: 4c4b3e5	Ratio
array/accumulate/Float32/1d	53513 ns	55804 ns	0.96
array/accumulate/Float32/dims=1	39257 ns	40028 ns	0.98
array/accumulate/Float32/dims=1L	82385 ns	82750 ns	1.00
array/accumulate/Float32/dims=2	49192 ns	49956 ns	0.98
array/accumulate/Float32/dims=2L	399160 ns	398401 ns	1.00
array/accumulate/Int64/1d	87912 ns	88814 ns	0.99
array/accumulate/Int64/dims=1	59858 ns	60000 ns	1.00
array/accumulate/Int64/dims=1L	139986 ns	140124 ns	1.00
array/accumulate/Int64/dims=2	68283 ns	69307 ns	0.99
array/accumulate/Int64/dims=2L	795586 ns	796279 ns	1.00
array/broadcast	10489 ns	10347 ns	1.01
array/broadcast launch	6656.8 ns	6790.2 ns	0.98
array/construct	933.3684210526316 ns	920.25 ns	1.01
array/copy	16883 ns	16941 ns	1.00
array/copyto!/cpu_to_gpu	207710 ns	208550 ns	1.00
array/copyto!/gpu_to_cpu	238972 ns	241531 ns	0.99
array/copyto!/gpu_to_gpu	8816 ns	9009 ns	0.98
array/iteration/findall/bool	49438 ns	50331 ns	0.98
array/iteration/findall/int	64239 ns	65676 ns	0.98
array/iteration/findfirst/bool	34409 ns	35523 ns	0.97
array/iteration/findfirst/int	34838 ns	35669 ns	0.98
array/iteration/findmin/1d	50110 ns	51799 ns	0.97
array/iteration/findmin/2d	51255 ns	53096 ns	0.97
array/iteration/logical	49705 ns	49885 ns	1.00
array/iteration/scalar	60995 ns	62101 ns	0.98
array/permutedims/2d	47105 ns	47294 ns	1.00
array/permutedims/3d	45069 ns	44494 ns	1.01
array/permutedims/4d	47794 ns	47646 ns	1.00
array/random/rand/Float32	11126 ns	11219 ns	0.99
array/random/rand/Int64	17683 ns	17674 ns	1.00
array/random/rand!/Float32	8263 ns	8372.333333333334 ns	0.99
array/random/rand!/Int64	15576 ns	15576 ns	1
array/random/randn/Float32	34148 ns	34174 ns	1.00
array/random/randn!/Float32	25029 ns	25289 ns	0.99
array/reductions/mapreduce/Float32/1d	34229 ns	35857 ns	0.95
array/reductions/mapreduce/Float32/dims=1	19671 ns	20204 ns	0.97
array/reductions/mapreduce/Float32/dims=1L	57672 ns	57641 ns	1.00
array/reductions/mapreduce/Float32/dims=2	27309 ns	28520 ns	0.96
array/reductions/mapreduce/Float32/dims=2L	57948 ns	58314 ns	0.99
array/reductions/mapreduce/Int64/1d	35517 ns	36343 ns	0.98
array/reductions/mapreduce/Int64/dims=1	21975 ns	22390 ns	0.98
array/reductions/mapreduce/Int64/dims=1L	96909 ns	97258 ns	1.00
array/reductions/mapreduce/Int64/dims=2	29223 ns	29457 ns	0.99
array/reductions/mapreduce/Int64/dims=2L	82897 ns	84461 ns	0.98
array/reductions/reduce/Float32/1d	34796 ns	35872 ns	0.97
array/reductions/reduce/Float32/dims=1	20050 ns	20932 ns	0.96
array/reductions/reduce/Float32/dims=1L	57591 ns	58293 ns	0.99
array/reductions/reduce/Float32/dims=2	27764 ns	28664 ns	0.97
array/reductions/reduce/Float32/dims=2L	57756 ns	58605 ns	0.99
array/reductions/reduce/Int64/1d	35553 ns	37120 ns	0.96
array/reductions/reduce/Int64/dims=1	22307 ns	22603 ns	0.99
array/reductions/reduce/Int64/dims=1L	97197 ns	97266 ns	1.00
array/reductions/reduce/Int64/dims=2	29032 ns	29768 ns	0.98
array/reductions/reduce/Int64/dims=2L	83404 ns	84012 ns	0.99
array/reverse/1d	19358 ns	19586 ns	0.99
array/reverse/1dL	73191 ns	73513 ns	1.00
array/reverse/1dL_inplace	70186 ns	70240 ns	1.00
array/reverse/1d_inplace	12030 ns	11903 ns	1.01
array/reverse/2d	22357 ns	22635 ns	0.99
array/reverse/2dL	80735 ns	81092 ns	1.00
array/reverse/2dL_inplace	70866 ns	70857 ns	1.00
array/reverse/2d_inplace	14948 ns	14785 ns	1.01
array/sorting/1d	750014 ns	751370 ns	1.00
array/sorting/2d	194457 ns	194940 ns	1.00
array/sorting/by	926389 ns	928189 ns	1.00
cuda/graph/capture	36518 ns	36432 ns	1.00
cuda/graph/eager	33816 ns	33669 ns	1.00
cuda/graph/launch	13925 ns	14067 ns	0.99
cuda/synchronization/context/auto	2379.8888888888887 ns	2293.222222222222 ns	1.04
cuda/synchronization/context/blocking	897.1818181818181 ns	900.9772727272727 ns	1.00
cuda/synchronization/context/nonblocking	10913 ns	10465 ns	1.04
cuda/synchronization/stream/auto	880.1521739130435 ns	891.8181818181819 ns	0.99
cuda/synchronization/stream/blocking	774.6923076923077 ns	778.6057692307693 ns	0.99
cuda/synchronization/stream/nonblocking	8447.666666666666 ns	8676.333333333334 ns	0.97
integration/byval/reference	147639 ns	147744 ns	1.00
integration/byval/slices=1	148871 ns	148699 ns	1.00
integration/byval/slices=2	290288 ns	290312 ns	1.00
integration/byval/slices=3	432215 ns	432327 ns	1.00
integration/cudadevrt	105276 ns	105111 ns	1.00
integration/volumerhs	9610078 ns	9611992 ns	1.00
kernel/indexing	11387 ns	11493 ns	0.99
kernel/indexing_checked	12064 ns	12050 ns	1.00
kernel/launch	2481.222222222222 ns	2329.4444444444443 ns	1.07
kernel/occupancy	715.8235294117648 ns	674.1812080536913 ns	1.06
kernel/rand	14894 ns	15046 ns	0.99
latency/import	4043455833 ns	4052266943 ns	1.00
latency/precompile	5012987199 ns	5021620476 ns	1.00
latency/ttfp	4964884821 ns	4946268237 ns	1.00

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

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 85.72%. Comparing base (4c4b3e5) to head (12e67b0).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3458      +/-   ##
==========================================
- Coverage   87.69%   85.72%   -1.97%     
==========================================
  Files         193      192       -1     
  Lines       20110    19823     -287     
==========================================
- Hits        17635    16994     -641     
- Misses       2475     2829     +354     

☔ View full report in Codecov by Harness.
📢 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.