Use relocatable IR linked at load time

[maleadt]

Adapts to JuliaGPU/GPUCompiler.jl#878

[github-actions]

CUDA.jl Benchmarks

Details

Benchmark suite	Current: 7d26d61	Previous: 069cdff	Ratio
array/accumulate/Float32/1d	97592 ns	98074 ns	1.00
array/accumulate/Float32/dims=1	71237 ns	74919 ns	0.95
array/accumulate/Float32/dims=1L	1599227 ns	1600002 ns	1.00
array/accumulate/Float32/dims=2	136868 ns	140367 ns	0.98
array/accumulate/Float32/dims=2L	659994 ns	659745 ns	1.00
array/accumulate/Int64/1d	117120 ns	117818 ns	0.99
array/accumulate/Int64/dims=1	75019 ns	79251 ns	0.95
array/accumulate/Int64/dims=1L	1715771 ns	1717606 ns	1.00
array/accumulate/Int64/dims=2	147864 ns	153507 ns	0.96
array/accumulate/Int64/dims=2L	986380 ns	986800 ns	1.00
array/broadcast	17598 ns	18060 ns	0.97
array/broadcast launch	8519 ns
array/construct	836.2714285714286 ns	869.1052631578947 ns	0.96
array/copy	16095 ns	15967 ns	1.01
array/copyto!/cpu_to_gpu	209326 ns	208599 ns	1.00
array/copyto!/gpu_to_cpu	240849 ns	241280 ns	1.00
array/copyto!/gpu_to_gpu	8881.333333333334 ns	9209.333333333334 ns	0.96
array/iteration/findall/bool	131187 ns	132021 ns	0.99
array/iteration/findall/int	144614 ns	145286 ns	1.00
array/iteration/findfirst/bool	67629 ns	67274 ns	1.01
array/iteration/findfirst/int	69032 ns	68529 ns	1.01
array/iteration/findmin/1d	61297 ns	64309 ns	0.95
array/iteration/findmin/2d	98475 ns	99940 ns	0.99
array/iteration/logical	182849 ns	185815 ns	0.98
array/iteration/scalar	60408 ns	63136 ns	0.96
array/permutedims/2d	47743 ns	48431 ns	0.99
array/permutedims/3d	48903 ns	50026 ns	0.98
array/permutedims/4d	49377 ns	49849 ns	0.99
array/random/rand/Float32	10879 ns	11669 ns	0.93
array/random/rand/Int64	20061 ns	22883 ns	0.88
array/random/rand!/Float32	7772.25 ns	7731.75 ns	1.01
array/random/rand!/Int64	19199 ns	20210 ns	0.95
array/random/randn/Float32	32449 ns	32574 ns	1.00
array/random/randn!/Float32	23327 ns	23189 ns	1.01
array/reductions/mapreduce/Float32/1d	31326 ns	31371 ns	1.00
array/reductions/mapreduce/Float32/dims=1	36979 ns	37491 ns	0.99
array/reductions/mapreduce/Float32/dims=1L	49742 ns	49977 ns	1.00
array/reductions/mapreduce/Float32/dims=2	54349 ns	54508 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	66057 ns	66329 ns	1.00
array/reductions/mapreduce/Int64/1d	38488 ns	37643 ns	1.02
array/reductions/mapreduce/Int64/dims=1	40321 ns	39961 ns	1.01
array/reductions/mapreduce/Int64/dims=1L	87930 ns	87917 ns	1.00
array/reductions/mapreduce/Int64/dims=2	56932 ns	57047 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	82277 ns	82751 ns	0.99
array/reductions/reduce/Float32/1d	31846 ns	31482 ns	1.01
array/reductions/reduce/Float32/dims=1	37376 ns	37630 ns	0.99
array/reductions/reduce/Float32/dims=1L	50055 ns	50073 ns	1.00
array/reductions/reduce/Float32/dims=2	54492 ns	54518 ns	1.00
array/reductions/reduce/Float32/dims=2L	67439 ns	67963 ns	0.99
array/reductions/reduce/Int64/1d	38401 ns	38054 ns	1.01
array/reductions/reduce/Int64/dims=1	39983 ns	39983 ns	1
array/reductions/reduce/Int64/dims=1L	87863 ns	87988 ns	1.00
array/reductions/reduce/Int64/dims=2	56796 ns	56860 ns	1.00
array/reductions/reduce/Int64/dims=2L	82392 ns	82191 ns	1.00
array/reverse/1d	16230 ns	16098 ns	1.01
array/reverse/1dL	68827 ns	68844 ns	1.00
array/reverse/1dL_inplace	67029 ns	66918 ns	1.00
array/reverse/1d_inplace	8374 ns	9757 ns	0.86
array/reverse/2d	19561 ns	19108 ns	1.02
array/reverse/2dL	72876 ns	72561 ns	1.00
array/reverse/2dL_inplace	66835 ns	66785 ns	1.00
array/reverse/2d_inplace	9480 ns	9397 ns	1.01
array/sorting/1d	2639323 ns	2656939 ns	0.99
array/sorting/2d	1027902 ns	1038211 ns	0.99
array/sorting/by	3182088 ns	3192363 ns	1.00
cuda/synchronization/context/auto	1041.8 ns	1015.3 ns	1.03
cuda/synchronization/context/blocking	801.010752688172 ns	784.6960784313726 ns	1.02
cuda/synchronization/context/nonblocking	5713 ns	5655.166666666667 ns	1.01
cuda/synchronization/stream/auto	877.8775510204082 ns	857.3174603174604 ns	1.02
cuda/synchronization/stream/blocking	679.5490196078431 ns	654.7730061349694 ns	1.04
cuda/synchronization/stream/nonblocking	5532.285714285715 ns	5491 ns	1.01
integration/byval/reference	147329 ns	147245 ns	1.00
integration/byval/slices=1	151637 ns	149267 ns	1.02
integration/byval/slices=2	294507 ns	292191 ns	1.01
integration/byval/slices=3	437547 ns	434868 ns	1.01
integration/cudadevrt	104428 ns	104357 ns	1.00
integration/volumerhs	9141010 ns	9315717 ns	0.98
kernel/indexing	12751 ns	12183 ns	1.05
kernel/indexing_checked	13528 ns	13078 ns	1.03
kernel/launch	2173.8888888888887 ns	2032.3333333333333 ns	1.07
kernel/occupancy	685.8343949044586 ns	667.8466257668712 ns	1.03
kernel/rand	13777 ns	14903 ns	0.92
latency/import	4092841703 ns	4116528672 ns	0.99
latency/precompile	4782383781 ns	4814427609 ns	0.99
latency/ttfp	4955136012 ns	5198634459 ns	0.95

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