[maleadt]

Split out of #906.

The CPU backend gives PoCL as many threads as Julia has, and the main CI job uses one, so work-groups only ran in parallel in the POCL session test's 4-thread subprocess. This adds a job running the suite with JULIA_NUM_THREADS=4 on Linux and macOS, with Julia 1.10 and 1.13.

[github-actions]

Benchmark results

Compared with c42b1f458, the merge base.
Times are the minimum time of every benchmark. A change is unusual when it exceeds both three times the benchmark's typical change between consecutive runs on the default branch and a minimum per category (performance 3%, latency 5%, applications 3%, other 3%). A benchmark that starts or stops allocating is listed however few the allocations. A single run can still be a fluke, so rerun before drawing conclusions.

	Geometric mean	Unusual: slower	Unusual: faster	Compared
Performance	−0.4%	0	0	32
Latency	+0.5%	0	0	13
Applications	+0.5%	0	0	5

No unusual changes.

All benchmarks

Benchmark	Baseline	This PR	Change
Performance
julia/serial/arithmetic/muladd-reference	655 µs	655 µs	−0.0%	noise unknown
julia/serial/math/exp-f64-reference	292 µs	284 µs	−2.8%	noise unknown
julia/serial/narrow/masked-reference	714 µs	715 µs	+0.2%	noise unknown
pocl/parallel/atomics/histogram	5.14 ms	5.16 ms	+0.3%	noise unknown
pocl/parallel/groupreduce/add-f64	1.65 ms	1.66 ms	+0.3%	noise unknown
pocl/parallel/groupreduce/fallback	19.0 ms	18.1 ms	−4.8%	noise unknown
pocl/parallel/groupscan/add-i32	2.52 ms	2.52 ms	+0.1%	noise unknown
pocl/parallel/indexing/cartesian3d	78.5 µs	77.7 µs	−0.9%	noise unknown
pocl/parallel/indexing/linear3d	80.3 µs	82.9 µs	+3.4%	noise unknown
pocl/parallel/indexing/ntuple3d	78.7 µs	69.8 µs	−11.3%	noise unknown
pocl/parallel/indexing/ntuple4d	157 µs	152 µs	−3.2%	noise unknown
pocl/parallel/localmem/transpose-naive	456 µs	449 µs	−1.6%	noise unknown
pocl/parallel/localmem/transpose-tiled	92.9 µs	102 µs	+9.7%	noise unknown
pocl/parallel/private/across-barrier	165 µs	164 µs	−0.2%	noise unknown
pocl/parallel/scheduling/default	182 µs	182 µs	+0.0%	noise unknown
pocl/parallel/scheduling/explicit	182 µs	181 µs	−0.3%	noise unknown
pocl/parallel/stream/copy/cache	781 ns	672 ns	−14.0%	noise unknown
pocl/parallel/stream/copy/dram	6.41 ms	6.24 ms	−2.6%	noise unknown
pocl/parallel/stream/triad/cache	7.17 µs	7.59 µs	+5.8%	noise unknown
pocl/parallel/stream/triad/dram	9.42 ms	9.30 ms	−1.3%	noise unknown
pocl/parallel/util/copyto-small	24.0 ns	24.0 ns	+0.0%	noise unknown
pocl/parallel/util/zeros-large	1.01 ms	997 µs	−0.9%	noise unknown
pocl/parallel/util/zeros-small	619 ns	640 ns	+3.4%	noise unknown
pocl/serial/arithmetic/contract	1.91 ms	1.91 ms	−0.2%	noise unknown
pocl/serial/arithmetic/muladd	756 µs	767 µs	+1.4%	noise unknown
pocl/serial/const/marked	728 µs	731 µs	+0.3%	noise unknown
pocl/serial/const/unmarked	910 µs	920 µs	+1.1%	noise unknown
pocl/serial/math/exp-f64	154 µs	153 µs	−0.4%	noise unknown
pocl/serial/math/sin-f64	194 µs	193 µs	−0.5%	noise unknown
pocl/serial/narrow/masked	807 µs	830 µs	+2.8%	noise unknown
pocl/serial/narrow/unmasked	766 µs	804 µs	+4.9%	noise unknown
pocl/serial/stream/triad/dram	8.62 ms	8.64 ms	+0.3%	noise unknown
Latency
host/import/ka	501 ms	503 ms	+0.3%	noise unknown
host/precompile/ka	44.4 s	44.8 s	+0.8%	noise unknown
host/startup	128 ms	129 ms	+0.3%	noise unknown
pocl/parallel/launch/dynamic-default	4.37 µs	4.39 µs	+0.6%	noise unknown
pocl/parallel/launch/groupreduce-f64	6.59 µs	6.62 µs	+0.5%	noise unknown
pocl/parallel/launch/many-args	11.2 µs	11.8 µs	+5.3%	noise unknown
pocl/parallel/launch/partial-workgroup	4.25 µs	4.14 µs	−2.7%	noise unknown
pocl/serial/compile/groupreduce	1.39 s	1.39 s	−0.2%	noise unknown
pocl/serial/compile/localmem-barrier	697 ms	693 ms	−0.5%	noise unknown
pocl/serial/compile/new-kernel	117 ms	126 ms	+7.7%	noise unknown
pocl/serial/compile/new-workgroupsize	14.1 ms	13.5 ms	−4.5%	noise unknown
pocl/serial/first-kernel/cold-cache	1.95 s	1.94 s	−0.1%	noise unknown
pocl/serial/first-kernel/warm-cache	279 ms	279 ms	+0.1%	noise unknown
Applications
pocl/parallel/dgsem/fluxdiff-node	291 µs	288 µs	−1.0%	noise unknown
pocl/parallel/dgsem/weak-element	159 µs	159 µs	−0.1%	noise unknown
pocl/parallel/dgsem/weak-node-localmem	119 µs	118 µs	−0.2%	noise unknown
pocl/parallel/stencil/timestep	7.75 ms	7.80 ms	+0.7%	noise unknown
pocl/serial/stencil/timestep	11.3 ms	11.6 ms	+3.1%	noise unknown

History: https://juliagpu.org/perf/?pkg=kernelabstractions

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 83.95%. Comparing base (c42b1f4) to head (5714ac0).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #940      +/-   ##
==========================================
- Coverage   84.80%   83.95%   -0.85%     
==========================================
  Files          31       31              
  Lines        2593     2543      -50     
==========================================
- Hits         2199     2135      -64     
- Misses        394      408      +14     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.