Fix pointer types in cublasHgemm

[lpawela]

In the current main branch this fails

N = 256
A = CUDA.rand(Float16, N, N)
B = CUDA.rand(Float16, N, N)
C = CUDA.zeros(Float16, N, N)
CUDA.CUBLAS.gemm!('N', 'N', one(Float16), A, B, zero(Float16), C)

with

ERROR: MethodError: no method matching unsafe_convert(::Type{Ptr{Float16}}, ::Float16)
The function `unsafe_convert` exists, but no method is defined for this combination of argument types.

Closest candidates are:
  unsafe_convert(::Type{Cwstring}, ::Any)
   @ Base strings/cstring.jl:101
  unsafe_convert(::Type{Ptr{T}}, ::Array{T}) where T
   @ Base pointer.jl:65
  unsafe_convert(::Type{Ptr{T}}, ::Ptr{NTuple{N, T}}) where {N, T}
   @ Base refpointer.jl:188

This PR fixes this issue.

[github-actions]

CUDA.jl Benchmarks

Details

Benchmark suite	Current: f7b7855	Previous: 54c7586	Ratio
array/accumulate/Float32/1d	99736 ns	100040 ns	1.00
array/accumulate/Float32/dims=1	75957 ns	75598 ns	1.00
array/accumulate/Float32/dims=1L	1585719 ns	1585923 ns	1.00
array/accumulate/Float32/dims=2	141680 ns	140396 ns	1.01
array/accumulate/Float32/dims=2L	653773 ns	652952 ns	1.00
array/accumulate/Int64/1d	117344 ns	116760 ns	1.01
array/accumulate/Int64/dims=1	79011 ns	78966 ns	1.00
array/accumulate/Int64/dims=1L	1698717 ns	1697893 ns	1.00
array/accumulate/Int64/dims=2	150749 ns	150510 ns	1.00
array/accumulate/Int64/dims=2L	959443 ns	959254 ns	1.00
array/broadcast	18299 ns	18315 ns	1.00
array/construct	1198.3 ns	1307.5 ns	0.92
array/copy	16803 ns	16659 ns	1.01
array/copyto!/cpu_to_gpu	214558 ns	211984 ns	1.01
array/copyto!/gpu_to_cpu	281252 ns	280422 ns	1.00
array/copyto!/gpu_to_gpu	10590 ns	10380 ns	1.02
array/iteration/findall/bool	132091 ns	131500 ns	1.00
array/iteration/findall/int	145310 ns	145666 ns	1.00
array/iteration/findfirst/bool	68732 ns	68920 ns	1.00
array/iteration/findfirst/int	70960 ns	71508 ns	0.99
array/iteration/findmin/1d	67253 ns	66419 ns	1.01
array/iteration/findmin/2d	101147 ns	101084 ns	1.00
array/iteration/logical	190663 ns	190475 ns	1.00
array/iteration/scalar	65429 ns	65525 ns	1.00
array/permutedims/2d	49770 ns	49765 ns	1.00
array/permutedims/3d	50309 ns	50702 ns	0.99
array/permutedims/4d	50743 ns	51020 ns	0.99
array/random/rand/Float32	11657 ns	11632 ns	1.00
array/random/rand/Int64	22230 ns	22338 ns	1.00
array/random/rand!/Float32	8009.333333333333 ns	7935.333333333333 ns	1.01
array/random/rand!/Int64	18413 ns	17826 ns	1.03
array/random/randn/Float32	36449 ns	36306 ns	1.00
array/random/randn!/Float32	24057 ns	23975 ns	1.00
array/reductions/mapreduce/Float32/1d	33655 ns	33733 ns	1.00
array/reductions/mapreduce/Float32/dims=1	37928 ns	38085 ns	1.00
array/reductions/mapreduce/Float32/dims=1L	50402 ns	50507 ns	1.00
array/reductions/mapreduce/Float32/dims=2	55582 ns	55668 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	67451 ns	67497 ns	1.00
array/reductions/mapreduce/Int64/1d	39612 ns	40096 ns	0.99
array/reductions/mapreduce/Int64/dims=1	41044 ns	40849 ns	1.00
array/reductions/mapreduce/Int64/dims=1L	86458 ns	86591 ns	1.00
array/reductions/mapreduce/Int64/dims=2	57784 ns	58177 ns	0.99
array/reductions/mapreduce/Int64/dims=2L	82871 ns	83232 ns	1.00
array/reductions/reduce/Float32/1d	33492 ns	33766 ns	0.99
array/reductions/reduce/Float32/dims=1	38230 ns	38178 ns	1.00
array/reductions/reduce/Float32/dims=1L	50431 ns	50557 ns	1.00
array/reductions/reduce/Float32/dims=2	55368 ns	55534 ns	1.00
array/reductions/reduce/Float32/dims=2L	67964 ns	68199 ns	1.00
array/reductions/reduce/Int64/1d	39467 ns	39982 ns	0.99
array/reductions/reduce/Int64/dims=1	40521 ns	41049 ns	0.99
array/reductions/reduce/Int64/dims=1L	86277 ns	86662 ns	1.00
array/reductions/reduce/Int64/dims=2	57786 ns	58182 ns	0.99
array/reductions/reduce/Int64/dims=2L	82512 ns	83392 ns	0.99
array/reverse/1d	16797 ns	17096 ns	0.98
array/reverse/1dL	67649 ns	67978 ns	1.00
array/reverse/1dL_inplace	65328 ns	65357 ns	1.00
array/reverse/1d_inplace	8854 ns	8332.666666666666 ns	1.06
array/reverse/2d	20034 ns	20040 ns	1.00
array/reverse/2dL	71800 ns	71744 ns	1.00
array/reverse/2dL_inplace	65102 ns	65037 ns	1.00
array/reverse/2d_inplace	10259 ns	9712 ns	1.06
array/sorting/1d	2724690 ns	2725346 ns	1.00
array/sorting/2d	1063501 ns	1061549 ns	1.00
array/sorting/by	3269235 ns	3268510 ns	1.00
cuda/synchronization/context/auto	1115.3 ns	1148.8 ns	0.97
cuda/synchronization/context/blocking	920 ns	947 ns	0.97
cuda/synchronization/context/nonblocking	5948.8 ns	6099.8 ns	0.98
cuda/synchronization/stream/auto	966.578947368421 ns	1006 ns	0.96
cuda/synchronization/stream/blocking	828.156626506024 ns	859.8142857142857 ns	0.96
cuda/synchronization/stream/nonblocking	5904.833333333333 ns	5966.2 ns	0.99
integration/byval/reference	143161 ns	143152 ns	1.00
integration/byval/slices=1	145336 ns	145205 ns	1.00
integration/byval/slices=2	283738 ns	283811 ns	1.00
integration/byval/slices=3	422228 ns	422282 ns	1.00
integration/cudadevrt	101648 ns	101717 ns	1.00
integration/volumerhs	8883509 ns	8896793 ns	1.00
kernel/indexing	12728 ns	12625 ns	1.01
kernel/indexing_checked	13354 ns	13467 ns	0.99
kernel/launch	2109.5555555555557 ns	2103.8888888888887 ns	1.00
kernel/occupancy	735.5757575757576 ns	692.2105263157895 ns	1.06
kernel/rand	14893 ns	14754 ns	1.01
latency/import	3846061458 ns	3848826206 ns	1.00
latency/precompile	4624359073 ns	4629101097 ns	1.00
latency/ttfp	4491854168 ns	4482824517 ns	1.00

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

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 16.32%. Comparing base (54c7586) to head (f7b7855).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #3157   +/-   ##
=======================================
  Coverage   16.32%   16.32%           
=======================================
  Files         124      124           
  Lines        9875     9875           
=======================================
  Hits         1612     1612           
  Misses       8263     8263           

☔ 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.