Improve CSCS CI

[luraess]

• Add Julia 1.11 note: MI300 requires Julia 1.12 and later per LLVM compat
• Use UENV image with Spack-avail MIOpen

Fixes #948

[luraess]

While MIOpen works on CDNA3 arch, there is a workaround needed for MIOpen CK Group-Xdlops solvers segfaulting on gfx942 (MI300). This could be removed (together with test skip in conv) once fixed upstream: ROCm/rocm-libraries#9088

MIOPEN_DEBUG_CONV_IMPLICIT_GEMM_ASM_FWD_GTC_XDLOPS_NHWC: "0"
MIOPEN_DEBUG_GROUP_CONV_IMPLICIT_GEMM_HIP_FWD_XDLOPS: "0"
MIOPEN_DEBUG_GROUP_CONV_IMPLICIT_GEMM_HIP_WRW_XDLOPS: "0"
MIOPEN_DEBUG_3D_CONV_IMPLICIT_GEMM_HIP_FWD_XDLOPS: "0"
MIOPEN_DEBUG_3D_CONV_IMPLICIT_GEMM_HIP_WRW_XDLOPS: "0"

[github-actions]

AMDGPU.jl Benchmarks

Details

Benchmark suite	Current: f6b052e	Previous: fe32f5f	Ratio
amdgpu/synchronization/context/device	620 ns	600 ns	1.03
amdgpu/synchronization/stream/blocking	250 ns	260 ns	0.96
amdgpu/synchronization/stream/nonblocking	340 ns	360 ns	0.94
array/accumulate/Float32/1d	65751 ns	86591 ns	0.76
array/accumulate/Float32/dims=1	353425 ns	259673 ns	1.36
array/accumulate/Float32/dims=1L	135212 ns	134812 ns	1.00
array/accumulate/Float32/dims=2	127252 ns	130142 ns	0.98
array/accumulate/Float32/dims=2L	2810501 ns	2813814 ns	1.00
array/accumulate/Int64/1d	97451 ns	94751 ns	1.03
array/accumulate/Int64/dims=1	260954 ns	282423 ns	0.92
array/accumulate/Int64/dims=1L	167073 ns	167832 ns	1.00
array/accumulate/Int64/dims=2	128152 ns	124791 ns	1.03
array/accumulate/Int64/dims=2L	2990614 ns	2990025 ns	1.00
array/broadcast	92091 ns	59581 ns	1.55
array/construct	1750 ns	1750 ns	1
array/copy	37621 ns	39211 ns	0.96
array/copyto!/cpu_to_gpu	183323 ns	182872 ns	1.00
array/copyto!/gpu_to_cpu	183172 ns	183832 ns	1.00
array/copyto!/gpu_to_gpu	132312 ns	131611 ns	1.01
array/iteration/findall/bool	183583 ns	179313 ns	1.02
array/iteration/findall/int	198233 ns	187973 ns	1.05
array/iteration/findfirst/bool	122052 ns	116661 ns	1.05
array/iteration/findfirst/int	116622 ns	114442 ns	1.02
array/iteration/findmin/1d	172203 ns	170272 ns	1.01
array/iteration/findmin/2d	155112 ns	156212 ns	0.99
array/iteration/logical	351705 ns	356495 ns	0.99
array/iteration/scalar	296035 ns	288894 ns	1.02
array/permutedims/2d	75821 ns	73501 ns	1.03
array/permutedims/3d	75421 ns	74751 ns	1.01
array/permutedims/4d	78182 ns	77331 ns	1.01
array/random/rand/Float32	54301 ns	53490 ns	1.02
array/random/rand/Int64	46691 ns	57851 ns	0.81
array/random/rand!/Float32	62931 ns	90731 ns	0.69
array/random/rand!/Int64	86332 ns	146392 ns	0.59
array/random/randn/Float32	77571 ns	89341 ns	0.87
array/random/randn!/Float32	60891 ns	114881 ns	0.53
array/reductions/mapreduce/Float32/1d	135712 ns	133462 ns	1.02
array/reductions/mapreduce/Float32/dims=1	96161 ns	95111 ns	1.01
array/reductions/mapreduce/Float32/dims=1L	871462 ns	776539 ns	1.12
array/reductions/mapreduce/Float32/dims=2	97212 ns	96761 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	297844 ns	297864 ns	1.00
array/reductions/mapreduce/Int64/1d	136162 ns	134151 ns	1.01
array/reductions/mapreduce/Int64/dims=1	96461 ns	95251 ns	1.01
array/reductions/mapreduce/Int64/dims=1L	1183867 ns	783439 ns	1.51
array/reductions/mapreduce/Int64/dims=2	99041 ns	96221 ns	1.03
array/reductions/mapreduce/Int64/dims=2L	299365 ns	305404 ns	0.98
array/reductions/reduce/Float32/1d	131962 ns	133552 ns	0.99
array/reductions/reduce/Float32/dims=1	95411 ns	95321 ns	1.00
array/reductions/reduce/Float32/dims=1L	870343 ns	774738 ns	1.12
array/reductions/reduce/Float32/dims=2	98751 ns	96962 ns	1.02
array/reductions/reduce/Float32/dims=2L	298154 ns	297593 ns	1.00
array/reductions/reduce/Int64/1d	135872 ns	133832 ns	1.02
array/reductions/reduce/Int64/dims=1	97112 ns	94581 ns	1.03
array/reductions/reduce/Int64/dims=1L	1106956 ns	783728 ns	1.41
array/reductions/reduce/Int64/dims=2	96391 ns	96342 ns	1.00
array/reductions/reduce/Int64/dims=2L	300095 ns	302023 ns	0.99
array/reverse/1d	44301 ns	43701 ns	1.01
array/reverse/1dL	76381 ns	75351 ns	1.01
array/reverse/1dL_inplace	110802 ns	108852 ns	1.02
array/reverse/1d_inplace	141822 ns	141152 ns	1.00
array/reverse/2d	52170 ns	52531 ns	0.99
array/reverse/2dL	102752 ns	101671 ns	1.01
array/reverse/2dL_inplace	177572 ns	122921 ns	1.44
array/reverse/2d_inplace	121172 ns	54280 ns	2.23
array/sorting/1d	343255 ns	341084 ns	1.01
integration/byval/reference	39940 ns	39091 ns	1.02
integration/byval/slices=1	41291 ns	40090 ns	1.03
integration/byval/slices=2	147563 ns	147892 ns	1.00
integration/byval/slices=3	237604 ns	242013 ns	0.98
integration/volumerhs	5024443 ns	5024742 ns	1.00
kernel/indexing	131492 ns	130302 ns	1.01
kernel/indexing_checked	133201 ns	63091 ns	2.11
kernel/launch	1340 ns	1290 ns	1.04
kernel/rand	197353 ns	126652 ns	1.56
latency/import	1572267183 ns	1592323983 ns	0.99
latency/precompile	36229167017 ns	36361017031 ns	1.00
latency/ttfp	2141909203 ns	2167313359 ns	0.99

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