don't promote to i64 unnecessarily in unsafe_cached_load

[simeonschaub]

Might be more of a micro optimization, since eventually we will promote to i64 in add_ptr anyways, but it doesn't hurt

[github-actions]

CUDA.jl Benchmarks

Details

Benchmark suite	Current: 3cf1981	Previous: fbb9098	Ratio
array/accumulate/Float32/1d	98213 ns	98020 ns	1.00
array/accumulate/Float32/dims=1	74516 ns	73968 ns	1.01
array/accumulate/Float32/dims=1L	1599568 ns	1598577 ns	1.00
array/accumulate/Float32/dims=2	140041 ns	139829 ns	1.00
array/accumulate/Float32/dims=2L	660634 ns	660479 ns	1.00
array/accumulate/Int64/1d	118249 ns	118236 ns	1.00
array/accumulate/Int64/dims=1	78433 ns	78818 ns	1.00
array/accumulate/Int64/dims=1L	1716517 ns	1717142 ns	1.00
array/accumulate/Int64/dims=2	151796 ns	151913 ns	1.00
array/accumulate/Int64/dims=2L	986037 ns	987639 ns	1.00
array/broadcast	18384 ns	18185 ns	1.01
array/construct	904.5945945945946 ns	900.1904761904761 ns	1.00
array/copy	15938 ns	16081 ns	0.99
array/copyto!/cpu_to_gpu	206292 ns	210623 ns	0.98
array/copyto!/gpu_to_cpu	239640 ns	242406 ns	0.99
array/copyto!/gpu_to_gpu	9140 ns	8883.666666666666 ns	1.03
array/iteration/findall/bool	131421 ns	132352 ns	0.99
array/iteration/findall/int	145321 ns	146447 ns	0.99
array/iteration/findfirst/bool	67876 ns	67784 ns	1.00
array/iteration/findfirst/int	69264 ns	69220 ns	1.00
array/iteration/findmin/1d	62583 ns	62612 ns	1.00
array/iteration/findmin/2d	99262 ns	99990 ns	0.99
array/iteration/logical	185326 ns	186900 ns	0.99
array/iteration/scalar	61249 ns	62246 ns	0.98
array/permutedims/2d	48693 ns	49143 ns	0.99
array/permutedims/3d	49703 ns	49912 ns	1.00
array/permutedims/4d	49895 ns	50252 ns	0.99
array/random/rand/Float32	11079 ns	10722 ns	1.03
array/random/rand/Int64	21819 ns	20792 ns	1.05
array/random/rand!/Float32	7699.25 ns	7728 ns	1.00
array/random/rand!/Int64	17854 ns	20553 ns	0.87
array/random/randn/Float32	32432 ns	33301 ns	0.97
array/random/randn!/Float32	23109 ns	24939 ns	0.93
array/reductions/mapreduce/Float32/1d	31245 ns	31290 ns	1.00
array/reductions/mapreduce/Float32/dims=1	37163 ns	37312 ns	1.00
array/reductions/mapreduce/Float32/dims=1L	49870 ns	50036 ns	1.00
array/reductions/mapreduce/Float32/dims=2	54061 ns	54292 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	65854 ns	65880 ns	1.00
array/reductions/mapreduce/Int64/1d	38275 ns	39323 ns	0.97
array/reductions/mapreduce/Int64/dims=1	40093 ns	40145 ns	1.00
array/reductions/mapreduce/Int64/dims=1L	87973 ns	87779 ns	1.00
array/reductions/mapreduce/Int64/dims=2	56786 ns	57093 ns	0.99
array/reductions/mapreduce/Int64/dims=2L	82454 ns	82500 ns	1.00
array/reductions/reduce/Float32/1d	31048 ns	31207 ns	0.99
array/reductions/reduce/Float32/dims=1	36970 ns	37284 ns	0.99
array/reductions/reduce/Float32/dims=1L	49841 ns	50034 ns	1.00
array/reductions/reduce/Float32/dims=2	54388 ns	54677 ns	0.99
array/reductions/reduce/Float32/dims=2L	67393 ns	67287 ns	1.00
array/reductions/reduce/Int64/1d	38151 ns	38695 ns	0.99
array/reductions/reduce/Int64/dims=1	39825 ns	39938 ns	1.00
array/reductions/reduce/Int64/dims=1L	87949 ns	87933 ns	1.00
array/reductions/reduce/Int64/dims=2	56735 ns	56801 ns	1.00
array/reductions/reduce/Int64/dims=2L	82050 ns	82417 ns	1.00
array/reverse/1d	16163 ns	16212 ns	1.00
array/reverse/1dL	68883 ns	69012 ns	1.00
array/reverse/1dL_inplace	66853 ns	66891 ns	1.00
array/reverse/1d_inplace	7932 ns	9703.666666666666 ns	0.82
array/reverse/2d	19361 ns	19393 ns	1.00
array/reverse/2dL	72690 ns	72842 ns	1.00
array/reverse/2dL_inplace	66692 ns	66766 ns	1.00
array/reverse/2d_inplace	9394 ns	10134 ns	0.93
array/sorting/1d	2656716 ns	2658542 ns	1.00
array/sorting/2d	1037676 ns	1039211 ns	1.00
array/sorting/by	3193106 ns	3193897 ns	1.00
cuda/synchronization/context/auto	1054.4 ns	1040 ns	1.01
cuda/synchronization/context/blocking	816.1827956989247 ns	795.5444444444445 ns	1.03
cuda/synchronization/context/nonblocking	5678.333333333333 ns	5738.666666666667 ns	0.99
cuda/synchronization/stream/auto	908.1025641025641 ns	889.2083333333334 ns	1.02
cuda/synchronization/stream/blocking	687.3790849673203 ns	676.8387096774194 ns	1.02
cuda/synchronization/stream/nonblocking	5476.833333333333 ns	5562.285714285715 ns	0.98
integration/byval/reference	147285 ns	147385 ns	1.00
integration/byval/slices=1	149332 ns	149634 ns	1.00
integration/byval/slices=2	292303 ns	292176 ns	1.00
integration/byval/slices=3	434863 ns	435096 ns	1.00
integration/cudadevrt	104266 ns	104391 ns	1.00
integration/volumerhs	9316225 ns	9304645 ns	1.00
kernel/indexing	12324 ns	12267 ns	1.00
kernel/indexing_checked	13221 ns	13063 ns	1.01
kernel/launch	1999.3 ns	1980.4 ns	1.01
kernel/occupancy	641.8323353293413 ns	638.797619047619 ns	1.00
kernel/rand	15262 ns	14927 ns	1.02
latency/import	3888686071 ns	3890391684 ns	1.00
latency/precompile	4659548273 ns	4668709973 ns	1.00
latency/ttfp	4884258734 ns	4883527794 ns	1.00

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