Add regression test for #613

[vchuravy]

Adds a regression test for #613: launching a kernel with a scalar workgroupsize over a 2D ndrange pads the first dimension of the iteration space to a multiple of the workgroupsize. On the POCL backend the padded work items were not masked, so a linear @index(Global) reached past prod(ndrange) and @inbounds writes stochastically segfaulted.

The original MWE only fails intermittently, so the test detects the failure mode deterministically instead: it writes linear indices into a flat array with room for the overshoot and checks that the tail stays untouched. Running the same kernel with unsafe_indices=true (masking skipped) trips the check, confirming it catches unmasked padded work items.

The reproducer no longer fails on current main with the vendored POCL (both at the test shape and the original 300×5000 / workgroupsize 256 shape from the issue), so this guards against regressions; #613 can likely be closed.

🤖 Generated with Claude Code

[github-actions]

Benchmark Results

Show table

	main	c54509b...	main / c54509b...
saxpy/default/Float32/1024	0.086 ± 0.0084 ms	0.0864 ± 0.01 ms	0.996 ± 0.15
saxpy/default/Float32/1048576	0.472 ± 0.026 ms	0.485 ± 0.026 ms	0.974 ± 0.076
saxpy/default/Float32/16384	0.0676 ± 0.035 ms	0.0702 ± 0.034 ms	0.963 ± 0.68
saxpy/default/Float32/2048	0.0834 ± 0.029 ms	0.0838 ± 0.03 ms	0.994 ± 0.5
saxpy/default/Float32/256	0.0867 ± 0.0098 ms	0.0867 ± 0.012 ms	1 ± 0.18
saxpy/default/Float32/262144	0.17 ± 0.032 ms	0.176 ± 0.032 ms	0.965 ± 0.25
saxpy/default/Float32/32768	0.0722 ± 0.032 ms	0.0746 ± 0.032 ms	0.968 ± 0.59
saxpy/default/Float32/4096	0.0784 ± 0.031 ms	0.0821 ± 0.031 ms	0.954 ± 0.52
saxpy/default/Float32/512	0.0857 ± 0.0094 ms	0.0867 ± 0.01 ms	0.989 ± 0.16
saxpy/default/Float32/64	0.0867 ± 0.0096 ms	0.0869 ± 0.011 ms	0.997 ± 0.17
saxpy/default/Float32/65536	0.0869 ± 0.031 ms	0.0911 ± 0.032 ms	0.954 ± 0.48
saxpy/default/Float64/1024	0.0855 ± 0.018 ms	0.0862 ± 0.02 ms	0.992 ± 0.31
saxpy/default/Float64/1048576	0.581 ± 0.074 ms	0.572 ± 0.072 ms	1.01 ± 0.18
saxpy/default/Float64/16384	0.0683 ± 0.031 ms	0.0674 ± 0.03 ms	1.01 ± 0.64
saxpy/default/Float64/2048	0.0824 ± 0.032 ms	0.0821 ± 0.032 ms	1 ± 0.55
saxpy/default/Float64/256	0.0858 ± 0.011 ms	0.0862 ± 0.012 ms	0.995 ± 0.19
saxpy/default/Float64/262144	0.197 ± 0.036 ms	0.195 ± 0.038 ms	1.01 ± 0.27
saxpy/default/Float64/32768	0.0774 ± 0.03 ms	0.0795 ± 0.03 ms	0.973 ± 0.52
saxpy/default/Float64/4096	0.0798 ± 0.03 ms	0.0808 ± 0.029 ms	0.988 ± 0.52
saxpy/default/Float64/512	0.0857 ± 0.01 ms	0.0861 ± 0.011 ms	0.995 ± 0.17
saxpy/default/Float64/64	0.0871 ± 0.012 ms	0.0864 ± 0.013 ms	1.01 ± 0.21
saxpy/default/Float64/65536	0.0943 ± 0.031 ms	0.0993 ± 0.031 ms	0.95 ± 0.43
saxpy/static workgroup=(1024,)/Float32/1024	0.0836 ± 0.0093 ms	0.0839 ± 0.01 ms	0.997 ± 0.16
saxpy/static workgroup=(1024,)/Float32/1048576	0.427 ± 0.033 ms	0.434 ± 0.031 ms	0.982 ± 0.1
saxpy/static workgroup=(1024,)/Float32/16384	0.0637 ± 0.032 ms	0.0659 ± 0.031 ms	0.967 ± 0.67
saxpy/static workgroup=(1024,)/Float32/2048	0.0811 ± 0.029 ms	0.082 ± 0.03 ms	0.989 ± 0.51
saxpy/static workgroup=(1024,)/Float32/256	0.084 ± 0.012 ms	0.0843 ± 0.012 ms	0.996 ± 0.2
saxpy/static workgroup=(1024,)/Float32/262144	0.152 ± 0.031 ms	0.157 ± 0.031 ms	0.965 ± 0.27
saxpy/static workgroup=(1024,)/Float32/32768	0.0679 ± 0.029 ms	0.0703 ± 0.03 ms	0.966 ± 0.59
saxpy/static workgroup=(1024,)/Float32/4096	0.0798 ± 0.03 ms	0.0793 ± 0.03 ms	1.01 ± 0.54
saxpy/static workgroup=(1024,)/Float32/512	0.0835 ± 0.011 ms	0.0844 ± 0.0099 ms	0.989 ± 0.17
saxpy/static workgroup=(1024,)/Float32/64	0.084 ± 0.015 ms	0.0843 ± 0.014 ms	0.996 ± 0.24
saxpy/static workgroup=(1024,)/Float32/65536	0.0799 ± 0.031 ms	0.0837 ± 0.032 ms	0.955 ± 0.51
saxpy/static workgroup=(1024,)/Float64/1024	0.0833 ± 0.012 ms	0.0834 ± 0.018 ms	0.999 ± 0.26
saxpy/static workgroup=(1024,)/Float64/1048576	0.531 ± 0.089 ms	0.499 ± 0.064 ms	1.06 ± 0.22
saxpy/static workgroup=(1024,)/Float64/16384	0.0654 ± 0.029 ms	0.0642 ± 0.028 ms	1.02 ± 0.63
saxpy/static workgroup=(1024,)/Float64/2048	0.0804 ± 0.031 ms	0.08 ± 0.032 ms	1 ± 0.56
saxpy/static workgroup=(1024,)/Float64/256	0.0835 ± 0.011 ms	0.0839 ± 0.012 ms	0.995 ± 0.19
saxpy/static workgroup=(1024,)/Float64/262144	0.192 ± 0.034 ms	0.182 ± 0.038 ms	1.05 ± 0.29
saxpy/static workgroup=(1024,)/Float64/32768	0.0745 ± 0.027 ms	0.0756 ± 0.027 ms	0.985 ± 0.5
saxpy/static workgroup=(1024,)/Float64/4096	0.0776 ± 0.03 ms	0.0765 ± 0.029 ms	1.01 ± 0.55
saxpy/static workgroup=(1024,)/Float64/512	0.0836 ± 0.011 ms	0.0837 ± 0.012 ms	0.999 ± 0.19
saxpy/static workgroup=(1024,)/Float64/64	0.0842 ± 0.014 ms	0.084 ± 0.013 ms	1 ± 0.23
saxpy/static workgroup=(1024,)/Float64/65536	0.0938 ± 0.028 ms	0.0931 ± 0.028 ms	1.01 ± 0.43
time_to_load	1.05 ± 0.01 s	1.12 ± 0.022 s	0.936 ± 0.021

Benchmark Plots

A plot of the benchmark results have been uploaded as an artifact to the workflow run for this PR.
Go to "Actions"->"Benchmark a pull request"->[the most recent run]->"Artifacts" (at the bottom).