Fix ROCm discovery on Ubuntu 26

[luraess]

Ubuntu 26 ships ROCm to be installed with apt in a location AMDGPU does not yet expects and with modified LLD naming which causes issues #920 (comment).

[luraess]

Ubuntu 26.04 installs ROCm 7.1.0 into /usr/lib/x86_64-linux-gnu/ (Debian multiarch) rather than /opt/rocm/, causing all ROCm components to be reported as unavailable.

Two fixes:

• check_rocm_path: probe Debian/Ubuntu multiarch subdirectories (lib/<triplet>) so check_rocm_path("/usr") resolves correctly on Ubuntu 26.04.
• find_ld_lld: replace split-by-position version parsing with a regex, fixing a warning caused by Ubuntu's lld emitting "Ubuntu LLD 21.0.0 ..." instead of "AMD LLD X.Y.Z ...".

[github-actions]

AMDGPU.jl Benchmarks

Details

Benchmark suite	Current: 5da2d37	Previous: 756602c	Ratio
amdgpu/synchronization/context/device	670 ns	600 ns	1.12
amdgpu/synchronization/stream/blocking	270 ns	240 ns	1.13
amdgpu/synchronization/stream/nonblocking	390 ns	340 ns	1.15
array/accumulate/Float32/1d	86161 ns	86251 ns	1.00
array/accumulate/Float32/dims=1	405695 ns	393845 ns	1.03
array/accumulate/Float32/dims=1L	135282 ns	131681 ns	1.03
array/accumulate/Float32/dims=2	127641 ns	103022 ns	1.24
array/accumulate/Float32/dims=2L	2805630 ns	2827930 ns	0.99
array/accumulate/Int64/1d	101512 ns	96412 ns	1.05
array/accumulate/Int64/dims=1	287204 ns	285244 ns	1.01
array/accumulate/Int64/dims=1L	166992 ns	160812 ns	1.04
array/accumulate/Int64/dims=2	127752 ns	120772 ns	1.06
array/accumulate/Int64/dims=2L	2984092 ns	3014433 ns	0.99
array/broadcast	55101 ns	128932 ns	0.43
array/construct	1650 ns	1680 ns	0.98
array/copy	40031 ns	39371 ns	1.02
array/copyto!/cpu_to_gpu	114432 ns	114832 ns	1.00
array/copyto!/gpu_to_cpu	184063 ns	152432 ns	1.21
array/copyto!/gpu_to_gpu	66940 ns	88321 ns	0.76
array/iteration/findall/bool	180313 ns	181912 ns	0.99
array/iteration/findall/int	189253 ns	190933 ns	0.99
array/iteration/findfirst/bool	118742 ns	114451 ns	1.04
array/iteration/findfirst/int	115342 ns	116331 ns	0.99
array/iteration/findmin/1d	169232 ns	166203 ns	1.02
array/iteration/findmin/2d	155882 ns	156173 ns	1.00
array/iteration/logical	351165 ns	346025 ns	1.01
array/iteration/scalar	292754 ns	289864 ns	1.01
array/permutedims/2d	73431 ns	64761 ns	1.13
array/permutedims/3d	74201 ns	73791 ns	1.01
array/permutedims/4d	76741 ns	76481 ns	1.00
array/random/rand/Float32	50831 ns	51540 ns	0.99
array/random/rand/Int64	57701 ns	56210 ns	1.03
array/random/rand!/Float32	86451 ns	142162 ns	0.61
array/random/rand!/Int64	93101 ns	141832 ns	0.66
array/random/randn/Float32	99532 ns	86921 ns	1.15
array/random/randn!/Float32	167102 ns	152202 ns	1.10
array/reductions/mapreduce/Float32/1d	132822 ns	132902 ns	1.00
array/reductions/mapreduce/Float32/dims=1	94811 ns	95052 ns	1.00
array/reductions/mapreduce/Float32/dims=1L	773211 ns	777081 ns	1.00
array/reductions/mapreduce/Float32/dims=2	96132 ns	96731 ns	0.99
array/reductions/mapreduce/Float32/dims=2L	301244 ns	299584 ns	1.01
array/reductions/mapreduce/Int64/1d	132672 ns	133322 ns	1.00
array/reductions/mapreduce/Int64/dims=1	95232 ns	78081 ns	1.22
array/reductions/mapreduce/Int64/dims=1L	781741 ns	783471 ns	1.00
array/reductions/mapreduce/Int64/dims=2	96332 ns	96252 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	297824 ns	308254 ns	0.97
array/reductions/reduce/Float32/1d	132732 ns	132802 ns	1.00
array/reductions/reduce/Float32/dims=1	94522 ns	94832 ns	1.00
array/reductions/reduce/Float32/dims=1L	773801 ns	774621 ns	1.00
array/reductions/reduce/Float32/dims=2	96842 ns	96802 ns	1.00
array/reductions/reduce/Float32/dims=2L	294304 ns	307245 ns	0.96
array/reductions/reduce/Int64/1d	133122 ns	129672 ns	1.03
array/reductions/reduce/Int64/dims=1	94951 ns	78151 ns	1.21
array/reductions/reduce/Int64/dims=1L	782072 ns	781931 ns	1.00
array/reductions/reduce/Int64/dims=2	95571 ns	96192 ns	0.99
array/reductions/reduce/Int64/dims=2L	298104 ns	298414 ns	1.00
array/reverse/1d	44211 ns	44380 ns	1.00
array/reverse/1dL	75291 ns	74131 ns	1.02
array/reverse/1dL_inplace	168092 ns	108282 ns	1.55
array/reverse/1d_inplace	78201 ns	86471 ns	0.90
array/reverse/2d	51710 ns	50661 ns	1.02
array/reverse/2dL	101402 ns	100341 ns	1.01
array/reverse/2dL_inplace	103552 ns	117622 ns	0.88
array/reverse/2d_inplace	112932 ns	95391 ns	1.18
array/sorting/1d	349634 ns	341945 ns	1.02
integration/byval/reference	39080 ns	38830 ns	1.01
integration/byval/slices=1	40310 ns	40880 ns	0.99
integration/byval/slices=2	130932 ns	158462 ns	0.83
integration/byval/slices=3	238703 ns	238013 ns	1.00
integration/volumerhs	5040093 ns	4942659 ns	1.02
kernel/indexing	62441 ns	43630 ns	1.43
kernel/indexing_checked	61321 ns	128022 ns	0.48
kernel/launch	1290 ns	1290 ns	1
kernel/rand	197003 ns	106671 ns	1.85
latency/import	1489766868 ns	1501349912 ns	0.99
latency/precompile	11930072145 ns	12041117438 ns	0.99
latency/ttfp	10850312414 ns	10491950084 ns	1.03

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