Adapt to LLVMDowngrader_jll 0.11 and GPUCompiler 2.8

[maleadt]

Bumps the compat bounds to LLVMDowngrader_jll 0.11 (JuliaPackaging/Yggdrasil#14758, built against LLVM 23) and GPUCompiler 2.8 (JuliaGPU/GPUCompiler.jl#931). The downgrader C API is unchanged, so this is compat-only.

[github-actions]

AMDGPU.jl Benchmarks

Details

Benchmark suite	Current: d1a2cf9	Previous: 4f0ab1a	Ratio
amdgpu/synchronization/context/device	567.5 ns	562.5 ns	1.01
amdgpu/synchronization/stream/blocking	240 ns	235 ns	1.02
amdgpu/synchronization/stream/nonblocking	337.5 ns	320 ns	1.05
applications/bitonic_sort	1137386.25 ns	1098082.75 ns	1.04
applications/convolution	104711.25 ns	104229 ns	1.00
applications/floyd_warshall	8868676 ns	8878742.25 ns	1.00
applications/histogram	807982.25 ns	815134 ns	0.99
applications/prefix_sum	232332.75 ns	231900.75 ns	1.00
array/accumulate/Float32/1d	75291 ns	80493.75 ns	0.94
array/accumulate/Float32/dims=1	263556 ns	272491.25 ns	0.97
array/accumulate/Float32/dims=1L	79846 ns	88573.75 ns	0.90
array/accumulate/Float32/dims=2	70281 ns	81833.75 ns	0.86
array/accumulate/Float32/dims=2L	2751418.75 ns	2754656.25 ns	1.00
array/accumulate/Int64/1d	77256.25 ns	79688.75 ns	0.97
array/accumulate/Int64/dims=1	241468.25 ns	244706 ns	0.99
array/accumulate/Int64/dims=1L	83508.75 ns	84231.25 ns	0.99
array/accumulate/Int64/dims=2	83876.25 ns	86993.75 ns	0.96
array/accumulate/Int64/dims=2L	2892030.75 ns	3133859.25 ns	0.92
array/broadcast	71978.5 ns	71851 ns	1.00
array/construct	2115 ns	2385 ns	0.89
array/copy	37240.5 ns	38018 ns	0.98
array/copyto!/cpu_to_gpu	110716.5 ns	110996.5 ns	1.00
array/copyto!/gpu_to_cpu	120734 ns	111031.5 ns	1.09
array/copyto!/gpu_to_gpu	48610.75 ns	59370.75 ns	0.82
array/iteration/findall/bool	133906.75 ns	135379.5 ns	0.99
array/iteration/findall/int	140154.25 ns	150062.25 ns	0.93
array/iteration/findfirst/bool	181530 ns	152017.25 ns	1.19
array/iteration/findfirst/int	160627.5 ns	142527 ns	1.13
array/iteration/findmin/1d	113356.75 ns	106111.5 ns	1.07
array/iteration/findmin/2d	107201.5 ns	108821.5 ns	0.99
array/iteration/logical	240265.75 ns	243736 ns	0.99
array/iteration/scalar	292254.25 ns	300659.25 ns	0.97
array/permutedims/2d	70518.5 ns	71656 ns	0.98
array/permutedims/3d	70326 ns	71226 ns	0.99
array/permutedims/4d	72571 ns	73858.75 ns	0.98
array/random/rand/Float32	44275.5 ns	45950.75 ns	0.96
array/random/rand/Int64	52868.25 ns	54290.75 ns	0.97
array/random/rand!/Float32	64280.75 ns	65176 ns	0.99
array/random/rand!/Int64	70811 ns	66563.5 ns	1.06
array/random/randn/Float32	66636 ns	81206.25 ns	0.82
array/random/randn!/Float32	80228.75 ns	80368.5 ns	1.00
array/reductions/mapreduce/Float32/1d	79508.75 ns	95054 ns	0.84
array/reductions/mapreduce/Float32/dims=1	85838.75 ns	86063.75 ns	1.00
array/reductions/mapreduce/Float32/dims=1L	839962 ns	843989.5 ns	1.00
array/reductions/mapreduce/Float32/dims=2	81936.25 ns	75981 ns	1.08
array/reductions/mapreduce/Float32/dims=2L	139104.5 ns	138774.5 ns	1.00
array/reductions/mapreduce/Int64/1d	90881.5 ns	94306.5 ns	0.96
array/reductions/mapreduce/Int64/dims=1	84831.25 ns	86678.75 ns	0.98
array/reductions/mapreduce/Int64/dims=1L	844937 ns	846329.5 ns	1.00
array/reductions/mapreduce/Int64/dims=2	80881.25 ns	81706.25 ns	0.99
array/reductions/mapreduce/Int64/dims=2L	139254.5 ns	140432 ns	0.99
array/reductions/reduce/Float32/1d	82521.25 ns	94821.25 ns	0.87
array/reductions/reduce/Float32/dims=1	85478.75 ns	87331.25 ns	0.98
array/reductions/reduce/Float32/dims=1L	840351.75 ns	834664.5 ns	1.01
array/reductions/reduce/Float32/dims=2	82363.75 ns	76153.5 ns	1.08
array/reductions/reduce/Float32/dims=2L	138911.75 ns	139254.5 ns	1.00
array/reductions/reduce/Int64/1d	92781.25 ns	94776.5 ns	0.98
array/reductions/reduce/Int64/dims=1	84693.5 ns	86181 ns	0.98
array/reductions/reduce/Int64/dims=1L	843539.5 ns	839047 ns	1.01
array/reductions/reduce/Int64/dims=2	82026 ns	83326 ns	0.98
array/reductions/reduce/Int64/dims=2L	139549.5 ns	140177 ns	1.00
array/reverse/1d	43498 ns	42600.5 ns	1.02
array/reverse/1dL	56203.25 ns	72488.5 ns	0.78
array/reverse/1dL_inplace	78098.75 ns	78666 ns	0.99
array/reverse/1d_inplace	50593.25 ns	56155.75 ns	0.90
array/reverse/2d	48898.25 ns	49858.25 ns	0.98
array/reverse/2dL	81691 ns	82311.25 ns	0.99
array/reverse/2dL_inplace	89221.25 ns	89706.25 ns	0.99
array/reverse/2d_inplace	61108.5 ns	61980.75 ns	0.99
array/sorting/1d	333424.75 ns	332449.75 ns	1.00
gemm/tiled	1925307 ns	1904269.25 ns	1.01
gemm/tiled_unbounded	1928027.25 ns	1909711.75 ns	1.01
integration/byval/reference	39030 ns	39271 ns	0.99
integration/byval/slices=1	39330 ns	41411 ns	0.95
integration/byval/slices=2	135922 ns	132252 ns	1.03
integration/byval/slices=3	238752 ns	238713 ns	1.00
integration/volumerhs	4918610 ns	4997480 ns	0.98
kernel/indexing	55646 ns	28585.25 ns	1.95
kernel/indexing_checked	56660.75 ns	57423.5 ns	0.99
kernel/launch	1285 ns	1255 ns	1.02
kernel/rand	97281.5 ns	98873.75 ns	0.98
latency/import	1705064958 ns	1700382856 ns	1.00
latency/precompile	39927733725 ns	40071892794 ns	1.00
latency/ttfp	2307133138 ns	2305806765 ns	1.00
stencil/diffusion3d	1614670.25 ns	1623510.75 ns	0.99
stencil/diffusion3d_checked	1658811 ns	1662718.75 ns	1.00

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

[christiangnrd]

Close #1080

[luraess]

Can you rebase on main to see that CI passes and we could merge