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: 2c7d7f3	Previous: 9bc35f1	Ratio
amdgpu/synchronization/context/device	610 ns	670 ns	0.91
amdgpu/synchronization/stream/blocking	250 ns	280 ns	0.89
amdgpu/synchronization/stream/nonblocking	330 ns	400 ns	0.82
array/accumulate/Float32/1d	85441 ns	87651 ns	0.97
array/accumulate/Float32/dims=1	246523 ns	290364 ns	0.85
array/accumulate/Float32/dims=1L	136122 ns	136182 ns	1.00
array/accumulate/Float32/dims=2	129102 ns	127982 ns	1.01
array/accumulate/Float32/dims=2L	2696728 ns	2812289 ns	0.96
array/accumulate/Int64/1d	97991 ns	98192 ns	1.00
array/accumulate/Int64/dims=1	327244 ns	243584 ns	1.34
array/accumulate/Int64/dims=1L	170392 ns	168293 ns	1.01
array/accumulate/Int64/dims=2	123442 ns	122592 ns	1.01
array/accumulate/Int64/dims=2L	2930642 ns	2987552 ns	0.98
array/broadcast	147992 ns	147162 ns	1.01
array/construct	1780 ns	1720 ns	1.03
array/copy	39681 ns	39821 ns	1.00
array/copyto!/cpu_to_gpu	115212 ns	115351 ns	1.00
array/copyto!/gpu_to_cpu	111522 ns	97151 ns	1.15
array/copyto!/gpu_to_gpu	131072 ns	131032 ns	1.00
array/iteration/findall/bool	186652 ns	183372 ns	1.02
array/iteration/findall/int	194493 ns	191073 ns	1.02
array/iteration/findfirst/bool	124372 ns	122112 ns	1.02
array/iteration/findfirst/int	116681 ns	116541 ns	1.00
array/iteration/findmin/1d	173792 ns	171772 ns	1.01
array/iteration/findmin/2d	159452 ns	156672 ns	1.02
array/iteration/logical	356025 ns	355875 ns	1.00
array/iteration/scalar	296214 ns	303814 ns	0.97
array/permutedims/2d	75961 ns	74731 ns	1.02
array/permutedims/3d	74371 ns	74861 ns	0.99
array/permutedims/4d	88041 ns	77161 ns	1.14
array/random/rand/Float32	54770 ns	55211 ns	0.99
array/random/rand/Int64	60271 ns	56591 ns	1.07
array/random/rand!/Float32	72341 ns	145892 ns	0.50
array/random/rand!/Int64	95311 ns	61691 ns	1.54
array/random/randn/Float32	94492 ns	87631 ns	1.08
array/random/randn!/Float32	109962 ns	108792 ns	1.01
array/reductions/mapreduce/Float32/1d	134172 ns	134592 ns	1.00
array/reductions/mapreduce/Float32/dims=1	95261 ns	95752 ns	0.99
array/reductions/mapreduce/Float32/dims=1L	776661 ns	776561 ns	1.00
array/reductions/mapreduce/Float32/dims=2	97412 ns	97761 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	298424 ns	296454 ns	1.01
array/reductions/mapreduce/Int64/1d	134762 ns	134982 ns	1.00
array/reductions/mapreduce/Int64/dims=1	95281 ns	95992 ns	0.99
array/reductions/mapreduce/Int64/dims=1L	783101 ns	781911 ns	1.00
array/reductions/mapreduce/Int64/dims=2	97261 ns	97291 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	299704 ns	295774 ns	1.01
array/reductions/reduce/Float32/1d	133562 ns	134512 ns	0.99
array/reductions/reduce/Float32/dims=1	95531 ns	95801 ns	1.00
array/reductions/reduce/Float32/dims=1L	776211 ns	775531 ns	1.00
array/reductions/reduce/Float32/dims=2	97581 ns	97771 ns	1.00
array/reductions/reduce/Float32/dims=2L	295884 ns	302224 ns	0.98
array/reductions/reduce/Int64/1d	134322 ns	135202 ns	0.99
array/reductions/reduce/Int64/dims=1	95111 ns	95691 ns	0.99
array/reductions/reduce/Int64/dims=1L	783552 ns	778971 ns	1.01
array/reductions/reduce/Int64/dims=2	95641 ns	96941 ns	0.99
array/reductions/reduce/Int64/dims=2L	299214 ns	296404 ns	1.01
array/reverse/1d	43831 ns	43920 ns	1.00
array/reverse/1dL	75441 ns	75771 ns	1.00
array/reverse/1dL_inplace	158712 ns	110181 ns	1.44
array/reverse/1d_inplace	136472 ns	77711 ns	1.76
array/reverse/2d	51991 ns	51191 ns	1.02
array/reverse/2dL	102241 ns	102061 ns	1.00
array/reverse/2dL_inplace	179252 ns	178402 ns	1.00
array/reverse/2d_inplace	138762 ns	78972 ns	1.76
array/sorting/1d	343375 ns	342554 ns	1.00
integration/byval/reference	39830 ns	39581 ns	1.01
integration/byval/slices=1	40440 ns	40771 ns	0.99
integration/byval/slices=2	130942 ns	157862 ns	0.83
integration/byval/slices=3	238634 ns	239313 ns	1.00
integration/volumerhs	5043852 ns	5025691 ns	1.00
kernel/indexing	57961 ns	131032 ns	0.44
kernel/indexing_checked	131002 ns	132972 ns	0.99
kernel/launch	1300 ns	1310 ns	0.99
kernel/rand	165183 ns	123892 ns	1.33
latency/import	1612663271 ns	1566079357 ns	1.03
latency/precompile	40134605517 ns	36502576967 ns	1.10
latency/ttfp	2194480087 ns	2134396602 ns	1.03

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