Split out level 3 gemm tests

[kshyatt]

Testing locally, the level 3 and split-out level 3 GEMM-y tests seem to take the same amount of time. Should help with parallelization. Also removed an extraneous comment.

[github-actions]

CUDA.jl Benchmarks

Benchmark suite	Current: ddac187	Previous: a0c2f4b	Ratio
latency/precompile	45404021171 ns	45297385295 ns	1.00
latency/ttfp	6388606471 ns	6375596178 ns	1.00
latency/import	3040221123 ns	3036561495 ns	1.00
integration/volumerhs	9565588 ns	9567419 ns	1.00
integration/byval/slices=1	146893 ns	146746 ns	1.00
integration/byval/slices=3	425698 ns	425517.5 ns	1.00
integration/byval/reference	144994 ns	145010 ns	1.00
integration/byval/slices=2	286459 ns	286216 ns	1.00
integration/cudadevrt	103520 ns	103513 ns	1.00
kernel/indexing	14218 ns	14419 ns	0.99
kernel/indexing_checked	15449 ns	15499 ns	1.00
kernel/occupancy	728.203125 ns	748.2734375 ns	0.97
kernel/launch	2109.4 ns	2194.6666666666665 ns	0.96
kernel/rand	14991 ns	17335 ns	0.86
array/reverse/1d	19743 ns	19412 ns	1.02
array/reverse/2d	25140 ns	24576 ns	1.02
array/reverse/1d_inplace	11280 ns	11029 ns	1.02
array/reverse/2d_inplace	13218 ns	13223 ns	1.00
array/copy	20708 ns	20740 ns	1.00
array/iteration/findall/int	158377 ns	158179 ns	1.00
array/iteration/findall/bool	138671 ns	138583 ns	1.00
array/iteration/findfirst/int	153822.5 ns	153423 ns	1.00
array/iteration/findfirst/bool	154577.5 ns	154821 ns	1.00
array/iteration/scalar	75766.5 ns	77451 ns	0.98
array/iteration/logical	213464 ns	216735 ns	0.98
array/iteration/findmin/1d	41245 ns	41556.5 ns	0.99
array/iteration/findmin/2d	94061 ns	94128 ns	1.00
array/reductions/reduce/1d	35422 ns	42013 ns	0.84
array/reductions/reduce/2d	41072.5 ns	51911 ns	0.79
array/reductions/mapreduce/1d	33410 ns	39275 ns	0.85
array/reductions/mapreduce/2d	41182.5 ns	49505.5 ns	0.83
array/broadcast	21668 ns	21668 ns	1
array/copyto!/gpu_to_gpu	13525 ns	11569 ns	1.17
array/copyto!/cpu_to_gpu	211822 ns	211873 ns	1.00
array/copyto!/gpu_to_cpu	245148.5 ns	245423 ns	1.00
array/accumulate/1d	108822 ns	108388.5 ns	1.00
array/accumulate/2d	79771 ns	79823 ns	1.00
array/construct	1117.05 ns	1208.35 ns	0.92
array/random/randn/Float32	43263 ns	43873.5 ns	0.99
array/random/randn!/Float32	26118 ns	25937 ns	1.01
array/random/rand!/Int64	27149 ns	27271 ns	1.00
array/random/rand!/Float32	8683.333333333334 ns	8766.666666666666 ns	0.99
array/random/rand/Int64	29975 ns	29637 ns	1.01
array/random/rand/Float32	12857 ns	12723 ns	1.01
array/permutedims/4d	66803 ns	66923 ns	1.00
array/permutedims/2d	56890 ns	56439 ns	1.01
array/permutedims/3d	59106 ns	58867 ns	1.00
array/sorting/1d	2919614 ns	2933352 ns	1.00
array/sorting/by	3499898 ns	3500830 ns	1.00
array/sorting/2d	1084142 ns	1085059 ns	1.00
cuda/synchronization/stream/auto	1028.3 ns	1038.4 ns	0.99
cuda/synchronization/stream/nonblocking	6556.8 ns	6432 ns	1.02
cuda/synchronization/stream/blocking	794 ns	807.5918367346939 ns	0.98
cuda/synchronization/context/auto	1212.5 ns	1194.1 ns	1.02
cuda/synchronization/context/nonblocking	6758.6 ns	6649.8 ns	1.02
cuda/synchronization/context/blocking	901.3823529411765 ns	886.6415094339623 ns	1.02

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