Exclude Int128 from testing on linalg/core

[luraess]

Possibly workaround #1066 according to #1066 (comment)

[github-actions]

AMDGPU.jl Benchmarks

Details

Benchmark suite	Current: fa224c7	Previous: 300ec59	Ratio
amdgpu/synchronization/context/device	547.5 ns	555 ns	0.99
amdgpu/synchronization/stream/blocking	235 ns	242.5 ns	0.97
amdgpu/synchronization/stream/nonblocking	317.5 ns	317.5 ns	1
applications/bitonic_sort	1643136.75 ns	1647958.75 ns	1.00
applications/convolution	107869.25 ns	108124 ns	1.00
applications/floyd_warshall	11738634 ns	11798317.25 ns	0.99
applications/histogram	790251.75 ns	777591.25 ns	1.02
applications/prefix_sum	253338.75 ns	249503.5 ns	1.02
array/accumulate/Float32/1d	73911.25 ns	74683.5 ns	0.99
array/accumulate/Float32/dims=1	276241.5 ns	272806.5 ns	1.01
array/accumulate/Float32/dims=1L	79898.75 ns	81016.25 ns	0.99
array/accumulate/Float32/dims=2	70828.75 ns	70953.5 ns	1.00
array/accumulate/Float32/dims=2L	2613862 ns	2758164.75 ns	0.95
array/accumulate/Int64/1d	77486.25 ns	64993.5 ns	1.19
array/accumulate/Int64/dims=1	243443.5 ns	242320.75 ns	1.00
array/accumulate/Int64/dims=1L	84238.75 ns	84236.25 ns	1.00
array/accumulate/Int64/dims=2	85056.25 ns	84286.25 ns	1.01
array/accumulate/Int64/dims=2L	2894876.25 ns	2895096.75 ns	1.00
array/broadcast	50925.75 ns	73328.5 ns	0.69
array/construct	2397.5 ns	2292.5 ns	1.05
array/copy	37863.25 ns	37818 ns	1.00
array/copyto!/cpu_to_gpu	112006.75 ns	110564.25 ns	1.01
array/copyto!/gpu_to_cpu	118849.25 ns	109994 ns	1.08
array/copyto!/gpu_to_gpu	59546 ns	59358.25 ns	1.00
array/iteration/findall/bool	135467.25 ns	131911.75 ns	1.03
array/iteration/findall/int	149899.75 ns	150767.25 ns	0.99
array/iteration/findfirst/bool	185240.25 ns	186207.75 ns	0.99
array/iteration/findfirst/int	163475 ns	149192.25 ns	1.10
array/iteration/findmin/1d	110144.25 ns	109899.25 ns	1.00
array/iteration/findmin/2d	96564 ns	109099.25 ns	0.89
array/iteration/logical	243713.75 ns	247116.25 ns	0.99
array/iteration/scalar	299417.25 ns	295864.25 ns	1.01
array/permutedims/2d	71906 ns	72066 ns	1.00
array/permutedims/3d	71273.5 ns	71476 ns	1.00
array/permutedims/4d	74228.75 ns	74108.5 ns	1.00
array/random/rand/Float32	39398 ns	45305.75 ns	0.87
array/random/rand/Int64	54508.25 ns	54538.25 ns	1.00
array/random/rand!/Float32	44880.75 ns	65118.5 ns	0.69
array/random/rand!/Int64	73201 ns	73168.5 ns	1.00
array/random/randn/Float32	80978.75 ns	78696.25 ns	1.03
array/random/randn!/Float32	81148.5 ns	81231.25 ns	1.00
array/reductions/mapreduce/Float32/1d	87341.25 ns	87606.25 ns	1.00
array/reductions/mapreduce/Float32/dims=1	81848.75 ns	83026.25 ns	0.99
array/reductions/mapreduce/Float32/dims=1L	831635.25 ns	834374.5 ns	1.00
array/reductions/mapreduce/Float32/dims=2	83916.25 ns	83986 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	144247.25 ns	143689.5 ns	1.00
array/reductions/mapreduce/Int64/1d	86698.75 ns	86853.75 ns	1.00
array/reductions/mapreduce/Int64/dims=1	82311.25 ns	82863.5 ns	0.99
array/reductions/mapreduce/Int64/dims=1L	830432.75 ns	835519.5 ns	0.99
array/reductions/mapreduce/Int64/dims=2	83491.25 ns	83591 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	143584.75 ns	144489.5 ns	0.99
array/reductions/reduce/Float32/1d	87074 ns	87313.75 ns	1.00
array/reductions/reduce/Float32/dims=1	82003.75 ns	82811 ns	0.99
array/reductions/reduce/Float32/dims=1L	831452.75 ns	830319.5 ns	1.00
array/reductions/reduce/Float32/dims=2	83638.75 ns	83711.25 ns	1.00
array/reductions/reduce/Float32/dims=2L	144077.25 ns	144469.75 ns	1.00
array/reductions/reduce/Int64/1d	86926.5 ns	86816.25 ns	1.00
array/reductions/reduce/Int64/dims=1	82546.5 ns	84218.75 ns	0.98
array/reductions/reduce/Int64/dims=1L	830077.75 ns	826649.75 ns	1.00
array/reductions/reduce/Int64/dims=2	83056.25 ns	84171.25 ns	0.99
array/reductions/reduce/Int64/dims=2L	144757.25 ns	144622 ns	1.00
array/reverse/1d	45633.25 ns	38853 ns	1.17
array/reverse/1dL	73123.5 ns	73531.25 ns	0.99
array/reverse/1dL_inplace	67683.75 ns	63698.5 ns	1.06
array/reverse/1d_inplace	61461 ns	61103.25 ns	1.01
array/reverse/2d	50550.75 ns	50528 ns	1.00
array/reverse/2dL	82076.25 ns	82581 ns	0.99
array/reverse/2dL_inplace	92214 ns	91988.75 ns	1.00
array/reverse/2d_inplace	63693.5 ns	63808.5 ns	1.00
array/sorting/1d	334447.5 ns	333074.75 ns	1.00
gemm/tiled	1899070.75 ns	1893852.25 ns	1.00
gemm/tiled_unbounded	1886625.5 ns	1906635 ns	0.99
integration/byval/reference	39311 ns	39531 ns	0.99
integration/byval/slices=1	40501 ns	40971 ns	0.99
integration/byval/slices=2	157682 ns	131722 ns	1.20
integration/byval/slices=3	234474 ns	236344 ns	0.99
integration/volumerhs	4924503 ns	4910111 ns	1.00
kernel/indexing	35765.5 ns	57398.25 ns	0.62
kernel/indexing_checked	58926 ns	38100.5 ns	1.55
kernel/launch	1422.5 ns	1365 ns	1.04
kernel/rand	97244 ns	97473.75 ns	1.00
latency/import	1750315705 ns	1749876538 ns	1.00
latency/precompile	40018867955 ns	40117363625 ns	1.00
latency/ttfp	2333251104 ns	2335417989 ns	1.00
stencil/diffusion3d	1635223.75 ns	1624928.5 ns	1.01
stencil/diffusion3d_checked	1655274 ns	1658789.25 ns	1.00

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

[simeonschaub]

Why this check of the LLVM version? It's broken everywhere atm, so I don't see a reason to restrict this to only LLVM >= 19

[luraess]

I thought the Yggdrasil PR fixed it for < v"19". But can remove the check

[luraess]

Seems to be the case @simeonschaub given the test d actually not fail on Julia 1.10 and 1.11 (e.g. https://buildkite.com/julialang/amdgpu-dot-jl/builds/4144#01a07d8e-68c7-4453-adbf-707fcffb6239/L1248)

[simeonschaub]

Suggested change

) = filter(!=(Int128), eltypes)

[simeonschaub]

Hmm, looks like it did not apply correctly. I actually meant to just delete lines 87 to 100

[simeonschaub]

Thanks!