Fixes zero-dim matmatmul & matvecmul

[tam724]

Closes #2952 and #2607.
The (m x 0) * (0 x n) matmatmul and the (m x 0) * (0) matvecmul edgecase should probably be tested in the GPUArrays.jl testsuite (for all GPU backends). I'll add a PR there (JuliaGPU/GPUArrays.jl#646).

[github-actions]

CUDA.jl Benchmarks

Benchmark suite	Current: c154079	Previous: f4c05e0	Ratio
latency/precompile	56440455535 ns	56743162658.5 ns	0.99
latency/ttfp	8158073880 ns	8292887489.5 ns	0.98
latency/import	4496920342 ns	4493784612 ns	1.00
integration/volumerhs	9614290.5 ns	9612835.5 ns	1.00
integration/byval/slices=1	147247 ns	146961 ns	1.00
integration/byval/slices=3	426238.5 ns	425977 ns	1.00
integration/byval/reference	145275 ns	145162 ns	1.00
integration/byval/slices=2	286621.5 ns	286531 ns	1.00
integration/cudadevrt	103695 ns	103664 ns	1.00
kernel/indexing	14623 ns	14225 ns	1.03
kernel/indexing_checked	15017.5 ns	14963.5 ns	1.00
kernel/occupancy	671.2961783439491 ns	712.5909090909091 ns	0.94
kernel/launch	2245.8888888888887 ns	2140.1111111111113 ns	1.05
kernel/rand	16277 ns	17014 ns	0.96
array/reverse/1d	20414 ns	19857 ns	1.03
array/reverse/2dL_inplace	66994 ns	66720 ns	1.00
array/reverse/1dL	70621 ns	70068 ns	1.01
array/reverse/2d	22093 ns	21721 ns	1.02
array/reverse/1d_inplace	9904.5 ns	11535 ns	0.86
array/reverse/2d_inplace	13574 ns	13153 ns	1.03
array/reverse/2dL	74199.5 ns	73755 ns	1.01
array/reverse/1dL_inplace	67050 ns	66862 ns	1.00
array/copy	21232 ns	20647 ns	1.03
array/iteration/findall/int	159783 ns	158235 ns	1.01
array/iteration/findall/bool	141590 ns	139770.5 ns	1.01
array/iteration/findfirst/int	161911 ns	161047 ns	1.01
array/iteration/findfirst/bool	162355 ns	162113 ns	1.00
array/iteration/scalar	75128 ns	73378 ns	1.02
array/iteration/logical	219323.5 ns	216537 ns	1.01
array/iteration/findmin/1d	51897 ns	50322 ns	1.03
array/iteration/findmin/2d	97022 ns	96281.5 ns	1.01
array/reductions/reduce/Int64/1d	43894 ns	43275 ns	1.01
array/reductions/reduce/Int64/dims=1	55441 ns	44878 ns	1.24
array/reductions/reduce/Int64/dims=2	61803 ns	61376 ns	1.01
array/reductions/reduce/Int64/dims=1L	89289 ns	89018 ns	1.00
array/reductions/reduce/Int64/dims=2L	88383 ns	87717 ns	1.01
array/reductions/reduce/Float32/1d	37343 ns	36706 ns	1.02
array/reductions/reduce/Float32/dims=1	47678 ns	41841.5 ns	1.14
array/reductions/reduce/Float32/dims=2	60266 ns	59890 ns	1.01
array/reductions/reduce/Float32/dims=1L	52643 ns	52369 ns	1.01
array/reductions/reduce/Float32/dims=2L	72547.5 ns	71845 ns	1.01
array/reductions/mapreduce/Int64/1d	43990 ns	43034 ns	1.02
array/reductions/mapreduce/Int64/dims=1	46234 ns	44568 ns	1.04
array/reductions/mapreduce/Int64/dims=2	61850 ns	61598 ns	1.00
array/reductions/mapreduce/Int64/dims=1L	89271.5 ns	88831 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	88676 ns	88197 ns	1.01
array/reductions/mapreduce/Float32/1d	37376 ns	36550 ns	1.02
array/reductions/mapreduce/Float32/dims=1	42040 ns	51845 ns	0.81
array/reductions/mapreduce/Float32/dims=2	60371 ns	60046 ns	1.01
array/reductions/mapreduce/Float32/dims=1L	53317.5 ns	52895 ns	1.01
array/reductions/mapreduce/Float32/dims=2L	72675 ns	72274 ns	1.01
array/broadcast	20631.5 ns	20228 ns	1.02
array/copyto!/gpu_to_gpu	13559 ns	12997 ns	1.04
array/copyto!/cpu_to_gpu	216920.5 ns	214588 ns	1.01
array/copyto!/gpu_to_cpu	283648.5 ns	283061 ns	1.00
array/accumulate/Int64/1d	125592 ns	124766 ns	1.01
array/accumulate/Int64/dims=1	84564 ns	83121 ns	1.02
array/accumulate/Int64/dims=2	158606 ns	157489 ns	1.01
array/accumulate/Int64/dims=1L	1710160 ns	1708744 ns	1.00
array/accumulate/Int64/dims=2L	967189 ns	966369 ns	1.00
array/accumulate/Float32/1d	109583 ns	109029 ns	1.01
array/accumulate/Float32/dims=1	81043 ns	80115 ns	1.01
array/accumulate/Float32/dims=2	148178.5 ns	147066 ns	1.01
array/accumulate/Float32/dims=1L	1618740.5 ns	1617852.5 ns	1.00
array/accumulate/Float32/dims=2L	698802.5 ns	697700.5 ns	1.00
array/construct	1289.7 ns	1284.9 ns	1.00
array/random/randn/Float32	49870 ns	44088.5 ns	1.13
array/random/randn!/Float32	25113 ns	24724 ns	1.02
array/random/rand!/Int64	27825 ns	27197 ns	1.02
array/random/rand!/Float32	8914.333333333334 ns	8847.666666666666 ns	1.01
array/random/rand/Int64	30447 ns	29769 ns	1.02
array/random/rand/Float32	13414.5 ns	13169 ns	1.02
array/permutedims/4d	60045 ns	60066.5 ns	1.00
array/permutedims/2d	54855 ns	53803 ns	1.02
array/permutedims/3d	55596 ns	54690 ns	1.02
array/sorting/1d	2759173 ns	2756717 ns	1.00
array/sorting/by	3371294.5 ns	3343987 ns	1.01
array/sorting/2d	1089235 ns	1080056.5 ns	1.01
cuda/synchronization/stream/auto	1020.5 ns	1028.4 ns	0.99
cuda/synchronization/stream/nonblocking	7551.4 ns	7619.4 ns	0.99
cuda/synchronization/stream/blocking	805.1063829787234 ns	806.3333333333334 ns	1.00
cuda/synchronization/context/auto	1183.5 ns	1172.7 ns	1.01
cuda/synchronization/context/nonblocking	7267.2 ns	7177 ns	1.01
cuda/synchronization/context/blocking	908.8 ns	911.1923076923077 ns	1.00

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