Adapt to KernelIntrinsics/KA 0.10

[christiangnrd]

No description provided.

[christiangnrd]

id for AMDGPU is used to identify whether we need to launch hostcalls, based on the specific id of the global variable. It is not needed for KA, so can be dropped as in https://github.com/JuliaGPU/AMDGPU.jl/pull/868/files#diff-082b94339c8f038178ee472ca9b6feec6f27f434c469138f168031f248f223f9R197.

(From JuliaGPU/KernelAbstractions.jl#666)

Is my changing static local code generation from LLVMExternalLinkage to LLVMInternalLinkage going to cause any unintended effects? I had to do so to prevent local memory allocations from being treated as the same one if there happen to be more than one in a kernel. Otherwise I think we’ll have to readd the id parameter.

[pxl-th]

Is my changing static local code generation from LLVMExternalLinkage to LLVMInternalLinkage going to cause any unintended effects?

I don't think. I think it is a remnant from pre-HIP times.

[github-actions]

AMDGPU.jl Benchmarks

Details

Benchmark suite	Current: 16f913a	Previous: 724d15f	Ratio
amdgpu/synchronization/context/device	585 ns	577.5 ns	1.01
amdgpu/synchronization/stream/blocking	235 ns	237.5 ns	0.99
amdgpu/synchronization/stream/nonblocking	320 ns	322.5 ns	0.99
array/accumulate/Float32/1d	74686 ns	72383.75 ns	1.03
array/accumulate/Float32/dims=1	290731.75 ns	278954.25 ns	1.04
array/accumulate/Float32/dims=1L	89386.25 ns	80043.75 ns	1.12
array/accumulate/Float32/dims=2	77988.5 ns	71466 ns	1.09
array/accumulate/Float32/dims=2L	2656294 ns	2620236.25 ns	1.01
array/accumulate/Int64/1d	80281 ns	76381 ns	1.05
array/accumulate/Int64/dims=1	245216.25 ns	243786 ns	1.01
array/accumulate/Int64/dims=1L	85278.75 ns	84088.75 ns	1.01
array/accumulate/Int64/dims=2	86359 ns	83453.75 ns	1.03
array/accumulate/Int64/dims=2L	2812319 ns	2864940.25 ns	0.98
array/broadcast	57720.75 ns	68581 ns	0.84
array/construct	2052.5 ns	2195 ns	0.94
array/copy	37343.25 ns	37048 ns	1.01
array/copyto!/cpu_to_gpu	110839 ns	104081.5 ns	1.06
array/copyto!/gpu_to_cpu	110551.5 ns	104479 ns	1.06
array/copyto!/gpu_to_gpu	47378.25 ns	37960.5 ns	1.25
array/iteration/findall/bool	140419.5 ns	134574.5 ns	1.04
array/iteration/findall/int	157469.75 ns	148672.25 ns	1.06
array/iteration/findfirst/bool	124711.75 ns	123486.75 ns	1.01
array/iteration/findfirst/int	125254.25 ns	124531.75 ns	1.01
array/iteration/findmin/1d	114296.75 ns	110069 ns	1.04
array/iteration/findmin/2d	110054.25 ns	96364 ns	1.14
array/iteration/logical	257353.75 ns	245316.25 ns	1.05
array/iteration/scalar	306312 ns	296437 ns	1.03
array/permutedims/2d	72583.5 ns	71361 ns	1.02
array/permutedims/3d	72288.75 ns	70816 ns	1.02
array/permutedims/4d	85658.75 ns	73823.5 ns	1.16
array/random/rand/Float32	46143.25 ns	45335.75 ns	1.02
array/random/rand/Int64	54838.25 ns	54068.25 ns	1.01
array/random/rand!/Float32	47498.25 ns	47823 ns	0.99
array/random/rand!/Int64	73826 ns	66071 ns	1.12
array/random/randn/Float32	79926 ns	77073.75 ns	1.04
array/random/randn!/Float32	79158.75 ns	81381 ns	0.97
array/reductions/mapreduce/Float32/1d	101011.75 ns	94514 ns	1.07
array/reductions/mapreduce/Float32/dims=1	87703.75 ns	84653.75 ns	1.04
array/reductions/mapreduce/Float32/dims=1L	782454.25 ns	776879 ns	1.01
array/reductions/mapreduce/Float32/dims=2	87241.25 ns	83853.75 ns	1.04
array/reductions/mapreduce/Float32/dims=2L	144677 ns	143094.75 ns	1.01
array/reductions/mapreduce/Int64/1d	101356.5 ns	94963.75 ns	1.07
array/reductions/mapreduce/Int64/dims=1	86533.75 ns	84003.75 ns	1.03
array/reductions/mapreduce/Int64/dims=1L	789941.75 ns	789344.25 ns	1.00
array/reductions/mapreduce/Int64/dims=2	86993.75 ns	83641.25 ns	1.04
array/reductions/mapreduce/Int64/dims=2L	144672.25 ns	144282.25 ns	1.00
array/reductions/reduce/Float32/1d	104094.25 ns	92824 ns	1.12
array/reductions/reduce/Float32/dims=1	91221.25 ns	84624 ns	1.08
array/reductions/reduce/Float32/dims=1L	783906.5 ns	782216.75 ns	1.00
array/reductions/reduce/Float32/dims=2	91411.5 ns	84249 ns	1.09
array/reductions/reduce/Float32/dims=2L	143917 ns	143379.75 ns	1.00
array/reductions/reduce/Int64/1d	100016.5 ns	87928.75 ns	1.14
array/reductions/reduce/Int64/dims=1	88551.25 ns	84403.75 ns	1.05
array/reductions/reduce/Int64/dims=1L	787971.5 ns	789071.75 ns	1.00
array/reductions/reduce/Int64/dims=2	87199 ns	83748.75 ns	1.04
array/reductions/reduce/Int64/dims=2L	143832.25 ns	142694.5 ns	1.01
array/reverse/1d	43875.75 ns	43903.25 ns	1.00
array/reverse/1dL	71961 ns	71013.5 ns	1.01
array/reverse/1dL_inplace	79903.5 ns	79038.5 ns	1.01
array/reverse/1d_inplace	60626 ns	59903.25 ns	1.01
array/reverse/2d	49493.25 ns	49135.75 ns	1.01
array/reverse/2dL	84543.75 ns	79608.5 ns	1.06
array/reverse/2dL_inplace	88061.5 ns	91884 ns	0.96
array/reverse/2d_inplace	41405.5 ns	55113.25 ns	0.75
array/sorting/1d	341835 ns	338222.5 ns	1.01
integration/byval/reference	39581 ns	39280 ns	1.01
integration/byval/slices=1	40710 ns	40360 ns	1.01
integration/byval/slices=2	140742 ns	150632 ns	0.93
integration/byval/slices=3	235184 ns	236914 ns	0.99
integration/volumerhs	5054664 ns	5050775 ns	1.00
kernel/indexing	48825.75 ns	57423.5 ns	0.85
kernel/indexing_checked	54053.25 ns	47713.25 ns	1.13
kernel/launch	1627.5 ns	1567.5 ns	1.04
kernel/rand	102489 ns	98961.5 ns	1.04
latency/import	1924337669 ns	1712872529 ns	1.12
latency/precompile	42136499710 ns	37407776408 ns	1.13
latency/ttfp	5811781268 ns	5749419423 ns	1.01

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

[luraess]

@christiangnrd I just merged #932 so feel free to transfer #933 to this one and close it.

[christiangnrd]

I don't have permission to close issues or PRs on this repo but you can close it and I'll remember to add it to this pr once the KA interface is merged