[maleadt]

Every launch resolves the method instance, builds the compiler configuration and job, takes a global lock and queries GPUCompiler's cache: about 5500 instructions, a third of an asynchronous @cuda. This wraps that lookup in GPUCompiler.cached_lookup (JuliaGPU/GPUCompiler.jl#1061), keyed on the context and the compiler options.

RTX 5080, Julia 1.12:

	before	after
cufunction (hit)	0.20 µs, 5476 insns	0.011 µs, 273 insns
@cuda (async)	1.63–1.71 µs, 16965 insns	1.35–1.42 µs, 11810 insns

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 85.63%. Comparing base (a3dfbfb) to head (35a8e24).
⚠️ Report is 6 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3482      +/-   ##
==========================================
- Coverage   87.75%   85.63%   -2.12%     
==========================================
  Files         193      192       -1     
  Lines       20275    19992     -283     
==========================================
- Hits        17792    17121     -671     
- Misses       2483     2871     +388     

☔ 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.

[github-actions]

CUDA.jl Benchmarks

Details

Benchmark suite	Current: 35a8e24	Previous: a3dfbfb	Ratio
array/accumulate/Float32/1d	42886 ns	43170 ns	0.99
array/accumulate/Float32/dims=1	27701 ns	28020 ns	0.99
array/accumulate/Float32/dims=1L	82480 ns	82829 ns	1.00
array/accumulate/Float32/dims=2	50679 ns	50386 ns	1.01
array/accumulate/Float32/dims=2L	374251 ns	373912 ns	1.00
array/accumulate/Int64/1d	72746 ns	72779 ns	1.00
array/accumulate/Int64/dims=1	45162 ns	45411 ns	0.99
array/accumulate/Int64/dims=1L	140435 ns	140892 ns	1.00
array/accumulate/Int64/dims=2	69799 ns	69499 ns	1.00
array/accumulate/Int64/dims=2L	768023 ns	765169 ns	1.00
array/broadcast	10311 ns	10581 ns	0.97
array/broadcast launch	6514.6 ns	7015.5 ns	0.93
array/construct	923.9459459459459 ns	908.8285714285714 ns	1.02
array/copy	17396 ns	17431 ns	1.00
array/copyto!/cpu_to_gpu	209862 ns	208849 ns	1.00
array/copyto!/gpu_to_cpu	242766 ns	241782 ns	1.00
array/copyto!/gpu_to_gpu	9060 ns	9026 ns	1.00
array/iteration/findall/bool	47765 ns	48091 ns	0.99
array/iteration/findall/int	62597 ns	62463 ns	1.00
array/iteration/findfirst/bool	32929 ns	36113 ns	0.91
array/iteration/findfirst/int	33660 ns	36206 ns	0.93
array/iteration/findmin/1d	50407 ns	51439 ns	0.98
array/iteration/findmin/2d	51315 ns	53794 ns	0.95
array/iteration/logical	48119 ns	48853 ns	0.98
array/iteration/scalar	61325 ns	61985 ns	0.99
array/permutedims/2d	47237 ns	47251 ns	1.00
array/permutedims/3d	44711 ns	44716 ns	1.00
array/permutedims/4d	37356 ns	48107 ns	0.78
array/random/rand/Float32	11545 ns	11443 ns	1.01
array/random/rand/Int64	17580 ns	17963 ns	0.98
array/random/rand!/Float32	8297.333333333334 ns	8423.666666666666 ns	0.99
array/random/rand!/Int64	15450 ns	15714 ns	0.98
array/random/randn/Float32	34484 ns	34085 ns	1.01
array/random/randn!/Float32	25364 ns	25383 ns	1.00
array/reductions/mapreduce/Float32/1d	34192 ns	36650 ns	0.93
array/reductions/mapreduce/Float32/dims=1	19634 ns	20606 ns	0.95
array/reductions/mapreduce/Float32/dims=1L	57137 ns	58262 ns	0.98
array/reductions/mapreduce/Float32/dims=2	26621 ns	28542 ns	0.93
array/reductions/mapreduce/Float32/dims=2L	57519 ns	58546 ns	0.98
array/reductions/mapreduce/Int64/1d	34279 ns	36832 ns	0.93
array/reductions/mapreduce/Int64/dims=1	21624 ns	23076 ns	0.94
array/reductions/mapreduce/Int64/dims=1L	96345 ns	97631 ns	0.99
array/reductions/mapreduce/Int64/dims=2	27486 ns	29611 ns	0.93
array/reductions/mapreduce/Int64/dims=2L	83320 ns	83824 ns	0.99
array/reductions/reduce/Float32/1d	34181 ns	36237 ns	0.94
array/reductions/reduce/Float32/dims=1	19332 ns	20754 ns	0.93
array/reductions/reduce/Float32/dims=1L	57113 ns	58298 ns	0.98
array/reductions/reduce/Float32/dims=2	26635 ns	28206 ns	0.94
array/reductions/reduce/Float32/dims=2L	57466 ns	58233 ns	0.99
array/reductions/reduce/Int64/1d	34569 ns	36907 ns	0.94
array/reductions/reduce/Int64/dims=1	22365 ns	22929 ns	0.98
array/reductions/reduce/Int64/dims=1L	97165 ns	97585 ns	1.00
array/reductions/reduce/Int64/dims=2	27688 ns	29990 ns	0.92
array/reductions/reduce/Int64/dims=2L	83594 ns	83516 ns	1.00
array/reverse/1d	19778 ns	19819 ns	1.00
array/reverse/1dL	73678 ns	73701 ns	1.00
array/reverse/1dL_inplace	69862 ns	70634 ns	0.99
array/reverse/1d_inplace	11779 ns	12449 ns	0.95
array/reverse/2d	22290 ns	22700 ns	0.98
array/reverse/2dL	80879 ns	81153 ns	1.00
array/reverse/2dL_inplace	70648 ns	71145 ns	0.99
array/reverse/2d_inplace	14380 ns	15174 ns	0.95
array/sorting/1d	748790 ns	750060 ns	1.00
array/sorting/2d	193447 ns	194579 ns	0.99
array/sorting/by	926358 ns	927982 ns	1.00
cuda/graph/capture	33777 ns	36378 ns	0.93
cuda/graph/eager	31905 ns	35636 ns	0.90
cuda/graph/launch	14026 ns	13878 ns	1.01
cuda/synchronization/context/auto	2355 ns	2377.777777777778 ns	0.99
cuda/synchronization/context/blocking	900.8205128205128 ns	908.2162162162163 ns	0.99
cuda/synchronization/context/nonblocking	10992 ns	10870 ns	1.01
cuda/synchronization/stream/auto	901.7096774193549 ns	866.7241379310345 ns	1.04
cuda/synchronization/stream/blocking	794.9387755102041 ns	790.26 ns	1.01
cuda/synchronization/stream/nonblocking	8640.333333333334 ns	8471 ns	1.02
integration/byval/reference	146702 ns	147937 ns	0.99
integration/byval/slices=1	147412 ns	148832 ns	0.99
integration/byval/slices=2	288987 ns	290301 ns	1.00
integration/byval/slices=3	430798 ns	432515 ns	1.00
integration/cudadevrt	104013 ns	105148 ns	0.99
integration/volumerhs	9610232 ns	9607952 ns	1.00
kernel/indexing	10384 ns	11654 ns	0.89
kernel/indexing_checked	10990 ns	12225 ns	0.90
kernel/launch	1347.6 ns	2436.3333333333335 ns	0.55
kernel/occupancy	34.18932527693857 ns	683.6513157894736 ns	0.05000988733190264
kernel/rand	13853 ns	15333 ns	0.90
latency/import	4058786002 ns	4047989887 ns	1.00
latency/precompile	4996941479 ns	5006008511 ns	1.00
latency/ttfp	5029410425 ns	4974010072 ns	1.01

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