Allow GPUToolbox v2 and v3

[vchuravy]

Bumps the GPUToolbox compat to allow v2 and v3.

The breaking changes in GPUToolbox v2/v3 are confined to the @memoize/LazyInitialized redesign (memoization and threading). AMDGPU only uses @gcsafe_ccall, which is unchanged through v3.0.0.

Verified that AMDGPU resolves and precompiles against GPUToolbox v3.0.0 in a clean environment.

🤖 Generated with Claude Code

[vchuravy]

x-ref: trixi-framework/Trixi.jl#2833 (comment)

[github-actions]

AMDGPU.jl Benchmarks

Details

Benchmark suite	Current: b87a62e	Previous: 790ffef	Ratio
amdgpu/synchronization/context/device	590 ns	600 ns	0.98
amdgpu/synchronization/stream/blocking	250 ns	250 ns	1
amdgpu/synchronization/stream/nonblocking	340 ns	330 ns	1.03
array/accumulate/Float32/1d	85511 ns	85181 ns	1.00
array/accumulate/Float32/dims=1	316974 ns	346495 ns	0.91
array/accumulate/Float32/dims=1L	137092 ns	135832 ns	1.01
array/accumulate/Float32/dims=2	126252 ns	133722 ns	0.94
array/accumulate/Float32/dims=2L	2811816 ns	2807449 ns	1.00
array/accumulate/Int64/1d	98151 ns	97751 ns	1.00
array/accumulate/Int64/dims=1	246143 ns	327884 ns	0.75
array/accumulate/Int64/dims=1L	168422 ns	167762 ns	1.00
array/accumulate/Int64/dims=2	124501 ns	128652 ns	0.97
array/accumulate/Int64/dims=2L	2988997 ns	2987542 ns	1.00
array/broadcast	94541 ns	85431 ns	1.11
array/construct	1780 ns	1640 ns	1.09
array/copy	39531 ns	37090 ns	1.07
array/copyto!/cpu_to_gpu	113931 ns	119302 ns	0.95
array/copyto!/gpu_to_cpu	182802 ns	183102 ns	1.00
array/copyto!/gpu_to_gpu	74081 ns	73881 ns	1.00
array/iteration/findall/bool	182782 ns	180732 ns	1.01
array/iteration/findall/int	189443 ns	199623 ns	0.95
array/iteration/findfirst/bool	117901 ns	121852 ns	0.97
array/iteration/findfirst/int	116471 ns	114861 ns	1.01
array/iteration/findmin/1d	170082 ns	170243 ns	1.00
array/iteration/findmin/2d	156442 ns	155382 ns	1.01
array/iteration/logical	350375 ns	355195 ns	0.99
array/iteration/scalar	296064 ns	290084 ns	1.02
array/permutedims/2d	74871 ns	75151 ns	1.00
array/permutedims/3d	73861 ns	74541 ns	0.99
array/permutedims/4d	77081 ns	77212 ns	1.00
array/random/rand/Float32	51421 ns	50610 ns	1.02
array/random/rand/Int64	56820 ns	57201 ns	0.99
array/random/rand!/Float32	92412 ns	95971 ns	0.96
array/random/rand!/Int64	146742 ns	86611 ns	1.69
array/random/randn/Float32	87631 ns	99462 ns	0.88
array/random/randn!/Float32	161902 ns	108421 ns	1.49
array/reductions/mapreduce/Float32/1d	134022 ns	133292 ns	1.01
array/reductions/mapreduce/Float32/dims=1	95511 ns	95621 ns	1.00
array/reductions/mapreduce/Float32/dims=1L	776171 ns	778531 ns	1.00
array/reductions/mapreduce/Float32/dims=2	97341 ns	97572 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	298194 ns	308705 ns	0.97
array/reductions/mapreduce/Int64/1d	134212 ns	134702 ns	1.00
array/reductions/mapreduce/Int64/dims=1	95491 ns	96291 ns	0.99
array/reductions/mapreduce/Int64/dims=1L	782411 ns	785001 ns	1.00
array/reductions/mapreduce/Int64/dims=2	96621 ns	96932 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	298904 ns	308475 ns	0.97
array/reductions/reduce/Float32/1d	133902 ns	133372 ns	1.00
array/reductions/reduce/Float32/dims=1	95251 ns	95181 ns	1.00
array/reductions/reduce/Float32/dims=1L	775511 ns	775611 ns	1.00
array/reductions/reduce/Float32/dims=2	97361 ns	97251 ns	1.00
array/reductions/reduce/Float32/dims=2L	298144 ns	297585 ns	1.00
array/reductions/reduce/Int64/1d	134952 ns	135242 ns	1.00
array/reductions/reduce/Int64/dims=1	93521 ns	95041 ns	0.98
array/reductions/reduce/Int64/dims=1L	785071 ns	781271 ns	1.00
array/reductions/reduce/Int64/dims=2	95202 ns	96491 ns	0.99
array/reductions/reduce/Int64/dims=2L	299954 ns	302104 ns	0.99
array/reverse/1d	43901 ns	42991 ns	1.02
array/reverse/1dL	75831 ns	75381 ns	1.01
array/reverse/1dL_inplace	86461 ns	167852 ns	0.52
array/reverse/1d_inplace	98451 ns	72911 ns	1.35
array/reverse/2d	51250 ns	51030 ns	1.00
array/reverse/2dL	101471 ns	101682 ns	1.00
array/reverse/2dL_inplace	101382 ns	107362 ns	0.94
array/reverse/2d_inplace	142072 ns	85981 ns	1.65
array/sorting/1d	343184 ns	342214 ns	1.00
integration/byval/reference	39851 ns	39040 ns	1.02
integration/byval/slices=1	40490 ns	40281 ns	1.01
integration/byval/slices=2	159722 ns	146872 ns	1.09
integration/byval/slices=3	237713 ns	238173 ns	1.00
integration/volumerhs	5032575 ns	5024071 ns	1.00
kernel/indexing	68231 ns	67700 ns	1.01
kernel/indexing_checked	98071 ns	64401 ns	1.52
kernel/launch	1320 ns	1350 ns	0.98
kernel/rand	125792 ns	198943 ns	0.63
latency/import	1588190447 ns	1577477738 ns	1.01
latency/precompile	36295592303 ns	36338831604 ns	1.00
latency/ttfp	2159175095 ns	2145912756 ns	1.01

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