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: 10dbaec	Previous: 1e35ff7	Ratio
latency/precompile	56839170229.5 ns	57073264350 ns	1.00
latency/ttfp	8293961507.5 ns	8361242115.5 ns	0.99
latency/import	4493983266 ns	4520871512 ns	0.99
integration/volumerhs	9608403 ns	9609145.5 ns	1.00
integration/byval/slices=1	146729 ns	146784 ns	1.00
integration/byval/slices=3	425688 ns	425930 ns	1.00
integration/byval/reference	144944 ns	144913 ns	1.00
integration/byval/slices=2	286200 ns	286275 ns	1.00
integration/cudadevrt	103408 ns	103477 ns	1.00
kernel/indexing	14162 ns	14088 ns	1.01
kernel/indexing_checked	14910 ns	14920 ns	1.00
kernel/occupancy	674.2215189873418 ns	670.5283018867924 ns	1.01
kernel/launch	2167 ns	2192.1111111111113 ns	0.99
kernel/rand	15573 ns	18597.5 ns	0.84
array/reverse/1d	20062 ns	19990 ns	1.00
array/reverse/2dL_inplace	66968 ns	66851 ns	1.00
array/reverse/1dL	70236 ns	70214 ns	1.00
array/reverse/2d	22059 ns	21764 ns	1.01
array/reverse/1d_inplace	11396 ns	9644 ns	1.18
array/reverse/2d_inplace	13382 ns	11083 ns	1.21
array/reverse/2dL	74226.5 ns	73680.5 ns	1.01
array/reverse/1dL_inplace	66779 ns	66780 ns	1.00
array/copy	20954.5 ns	20656 ns	1.01
array/iteration/findall/int	159452.5 ns	157234 ns	1.01
array/iteration/findall/bool	141317 ns	139637.5 ns	1.01
array/iteration/findfirst/int	162645 ns	161491 ns	1.01
array/iteration/findfirst/bool	163137.5 ns	161981.5 ns	1.01
array/iteration/scalar	73710 ns	72914 ns	1.01
array/iteration/logical	218890.5 ns	215503 ns	1.02
array/iteration/findmin/1d	54171.5 ns	52893.5 ns	1.02
array/iteration/findmin/2d	97228 ns	96673.5 ns	1.01
array/reductions/reduce/Int64/1d	44423 ns	43374 ns	1.02
array/reductions/reduce/Int64/dims=1	44937 ns	44924.5 ns	1.00
array/reductions/reduce/Int64/dims=2	61859 ns	61289 ns	1.01
array/reductions/reduce/Int64/dims=1L	89117 ns	89013 ns	1.00
array/reductions/reduce/Int64/dims=2L	88328.5 ns	88275 ns	1.00
array/reductions/reduce/Float32/1d	37481 ns	37043 ns	1.01
array/reductions/reduce/Float32/dims=1	42641 ns	43018 ns	0.99
array/reductions/reduce/Float32/dims=2	59997 ns	59774 ns	1.00
array/reductions/reduce/Float32/dims=1L	52520 ns	52409 ns	1.00
array/reductions/reduce/Float32/dims=2L	72293 ns	72278 ns	1.00
array/reductions/mapreduce/Int64/1d	44124 ns	43540 ns	1.01
array/reductions/mapreduce/Int64/dims=1	55556.5 ns	45057.5 ns	1.23
array/reductions/mapreduce/Int64/dims=2	62182 ns	61470 ns	1.01
array/reductions/mapreduce/Int64/dims=1L	89096.5 ns	88923 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	88416.5 ns	88349 ns	1.00
array/reductions/mapreduce/Float32/1d	36661.5 ns	36698 ns	1.00
array/reductions/mapreduce/Float32/dims=1	41802 ns	41442 ns	1.01
array/reductions/mapreduce/Float32/dims=2	60068 ns	59908 ns	1.00
array/reductions/mapreduce/Float32/dims=1L	52820 ns	52585 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	72324 ns	72014 ns	1.00
array/broadcast	19976 ns	20078 ns	0.99
array/copyto!/gpu_to_gpu	12947 ns	12908 ns	1.00
array/copyto!/cpu_to_gpu	217098 ns	213437 ns	1.02
array/copyto!/gpu_to_cpu	284144 ns	283206 ns	1.00
array/accumulate/Int64/1d	125267 ns	124198 ns	1.01
array/accumulate/Int64/dims=1	83680 ns	83165 ns	1.01
array/accumulate/Int64/dims=2	158512 ns	157631 ns	1.01
array/accumulate/Int64/dims=1L	1709548 ns	1709733 ns	1.00
array/accumulate/Int64/dims=2L	967025 ns	966057.5 ns	1.00
array/accumulate/Float32/1d	109355 ns	108414 ns	1.01
array/accumulate/Float32/dims=1	80542 ns	79731.5 ns	1.01
array/accumulate/Float32/dims=2	147497 ns	146657 ns	1.01
array/accumulate/Float32/dims=1L	1619041 ns	1616606.5 ns	1.00
array/accumulate/Float32/dims=2L	698387 ns	697417 ns	1.00
array/construct	1279.5 ns	1271.5 ns	1.01
array/random/randn/Float32	45369 ns	45612 ns	0.99
array/random/randn!/Float32	24921 ns	24822 ns	1.00
array/random/rand!/Int64	27323 ns	27264 ns	1.00
array/random/rand!/Float32	8827 ns	8854 ns	1.00
array/random/rand/Int64	30118 ns	29823 ns	1.01
array/random/rand/Float32	13144.5 ns	13073 ns	1.01
array/permutedims/4d	60086 ns	59525 ns	1.01
array/permutedims/2d	53908 ns	53919 ns	1.00
array/permutedims/3d	54858.5 ns	54583 ns	1.01
array/sorting/1d	2758934.5 ns	2757051 ns	1.00
array/sorting/by	3345603 ns	3344047 ns	1.00
array/sorting/2d	1081937 ns	1080794 ns	1.00
cuda/synchronization/stream/auto	1060.8 ns	1034 ns	1.03
cuda/synchronization/stream/nonblocking	8238.4 ns	8105 ns	1.02
cuda/synchronization/stream/blocking	811.7555555555556 ns	796.4842105263158 ns	1.02
cuda/synchronization/context/auto	1182.3 ns	1198.2 ns	0.99
cuda/synchronization/context/nonblocking	6948.4 ns	8018.6 ns	0.87
cuda/synchronization/context/blocking	905.9090909090909 ns	918.6428571428571 ns	0.99

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

[codecov]

Codecov Report

❌ Patch coverage is 60.00000% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 89.46%. Comparing base (b30cae9) to head (8f8a3da).
⚠️ Report is 1 commits behind head on master.

Files with missing lines	Patch %	Lines
lib/cublas/linalg.jl	60.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2958      +/-   ##
==========================================
+ Coverage   89.30%   89.46%   +0.15%     
==========================================
  Files         150      150              
  Lines       13084    13087       +3     
==========================================
+ Hits        11685    11708      +23     
+ Misses       1399     1379      -20     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.