Remove diagm in favour of GPUArrays

[kshyatt]

Needs to wait for a new GPUArrays to be tagged (assuming tests pass)

[kshyatt]

Let's bump this on top of GPUArrays and I'l remove the [sources]

[github-actions]

CUDA.jl Benchmarks

Benchmark suite	Current: adf22c4	Previous: 0048e5a	Ratio
latency/precompile	56646452642.5 ns	57035764507 ns	0.99
latency/ttfp	8197390132.5 ns	8371031413.5 ns	0.98
latency/import	4356410400 ns	4496982792 ns	0.97
integration/volumerhs	9607159 ns	9624750 ns	1.00
integration/byval/slices=1	147011 ns	147060 ns	1.00
integration/byval/slices=3	425973.5 ns	426247.5 ns	1.00
integration/byval/reference	145125 ns	145079 ns	1.00
integration/byval/slices=2	286464 ns	286395 ns	1.00
integration/cudadevrt	103679 ns	103630 ns	1.00
kernel/indexing	14403 ns	14142 ns	1.02
kernel/indexing_checked	15133 ns	15085 ns	1.00
kernel/occupancy	670.2955974842768 ns	683.7784810126582 ns	0.98
kernel/launch	2217.777777777778 ns	2192.3333333333335 ns	1.01
kernel/rand	15392 ns	15556 ns	0.99
array/reverse/1d	20426 ns	19955 ns	1.02
array/reverse/2dL_inplace	66940 ns	67142 ns	1.00
array/reverse/1dL	70544 ns	70143 ns	1.01
array/reverse/2d	21812 ns	21759 ns	1.00
array/reverse/1d_inplace	11592 ns	9799 ns	1.18
array/reverse/2d_inplace	13472 ns	11305 ns	1.19
array/reverse/2dL	73704.5 ns	73707 ns	1.00
array/reverse/1dL_inplace	67011 ns	66873 ns	1.00
array/copy	20911 ns	20703 ns	1.01
array/iteration/findall/int	158488 ns	157587.5 ns	1.01
array/iteration/findall/bool	140466 ns	140083 ns	1.00
array/iteration/findfirst/int	161938.5 ns	161900 ns	1.00
array/iteration/findfirst/bool	162344.5 ns	162539 ns	1.00
array/iteration/scalar	73105 ns	73928 ns	0.99
array/iteration/logical	216945 ns	216280.5 ns	1.00
array/iteration/findmin/1d	50782.5 ns	53048.5 ns	0.96
array/iteration/findmin/2d	97085 ns	96623.5 ns	1.00
array/reductions/reduce/Int64/1d	43853 ns	43822 ns	1.00
array/reductions/reduce/Int64/dims=1	44765 ns	45024.5 ns	0.99
array/reductions/reduce/Int64/dims=2	61662 ns	61470 ns	1.00
array/reductions/reduce/Int64/dims=1L	89195 ns	88941 ns	1.00
array/reductions/reduce/Int64/dims=2L	88335 ns	87825.5 ns	1.01
array/reductions/reduce/Float32/1d	37556 ns	37058 ns	1.01
array/reductions/reduce/Float32/dims=1	51787 ns	50073 ns	1.03
array/reductions/reduce/Float32/dims=2	59944 ns	59949 ns	1.00
array/reductions/reduce/Float32/dims=1L	52540 ns	52439 ns	1.00
array/reductions/reduce/Float32/dims=2L	72277 ns	72064 ns	1.00
array/reductions/mapreduce/Int64/1d	43791.5 ns	43941 ns	1.00
array/reductions/mapreduce/Int64/dims=1	51253.5 ns	44815 ns	1.14
array/reductions/mapreduce/Int64/dims=2	61633 ns	61592 ns	1.00
array/reductions/mapreduce/Int64/dims=1L	89031.5 ns	88958 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	88191.5 ns	88240 ns	1.00
array/reductions/mapreduce/Float32/1d	37365 ns	37401 ns	1.00
array/reductions/mapreduce/Float32/dims=1	41903 ns	43167 ns	0.97
array/reductions/mapreduce/Float32/dims=2	60007.5 ns	59977 ns	1.00
array/reductions/mapreduce/Float32/dims=1L	52904 ns	52531 ns	1.01
array/reductions/mapreduce/Float32/dims=2L	72377 ns	72397 ns	1.00
array/broadcast	20128 ns	20122 ns	1.00
array/copyto!/gpu_to_gpu	13159 ns	11368 ns	1.16
array/copyto!/cpu_to_gpu	215657 ns	215772 ns	1.00
array/copyto!/gpu_to_cpu	283085.5 ns	282327 ns	1.00
array/accumulate/Int64/1d	125130 ns	124672 ns	1.00
array/accumulate/Int64/dims=1	83335 ns	83443 ns	1.00
array/accumulate/Int64/dims=2	158509.5 ns	157725 ns	1.00
array/accumulate/Int64/dims=1L	1720548 ns	1710398.5 ns	1.01
array/accumulate/Int64/dims=2L	967847 ns	966565 ns	1.00
array/accumulate/Float32/1d	110322 ns	108966.5 ns	1.01
array/accumulate/Float32/dims=1	80302 ns	80321 ns	1.00
array/accumulate/Float32/dims=2	147411 ns	148101 ns	1.00
array/accumulate/Float32/dims=1L	1627874.5 ns	1619028 ns	1.01
array/accumulate/Float32/dims=2L	701566 ns	698667 ns	1.00
array/construct	1265.4 ns	1281.2 ns	0.99
array/random/randn/Float32	45829.5 ns	48183.5 ns	0.95
array/random/randn!/Float32	25511 ns	24910 ns	1.02
array/random/rand!/Int64	27510 ns	27271 ns	1.01
array/random/rand!/Float32	8750.333333333334 ns	8866.333333333334 ns	0.99
array/random/rand/Int64	30122 ns	37929.5 ns	0.79
array/random/rand/Float32	13078 ns	13112 ns	1.00
array/permutedims/4d	56328.5 ns	55650 ns	1.01
array/permutedims/2d	54109 ns	54104.5 ns	1.00
array/permutedims/3d	54980 ns	54918 ns	1.00
array/sorting/1d	2757574.5 ns	2757756 ns	1.00
array/sorting/by	3356022 ns	3344340.5 ns	1.00
array/sorting/2d	1088159 ns	1081498 ns	1.01
cuda/synchronization/stream/auto	1042 ns	1053.2 ns	0.99
cuda/synchronization/stream/nonblocking	7619.5 ns	7607.6 ns	1.00
cuda/synchronization/stream/blocking	818.5913978494624 ns	872.9166666666666 ns	0.94
cuda/synchronization/context/auto	1174.8 ns	1195.3 ns	0.98
cuda/synchronization/context/nonblocking	7280.299999999999 ns	7954 ns	0.92
cuda/synchronization/context/blocking	921.6279069767442 ns	921.375 ns	1.00

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

[kshyatt]

Failure looks related (to bumping the GPUArrays version)