Fix KA global linear index

[christiangnrd]

(do not squash)
It's like this in CUDA, I fixed it in Metal, I don't know why I got it wrong with the KI PR

[github-actions]

Benchmark Results

	main	541f205...	main / 541f205...
saxpy/default/Float32/1024	0.0503 ± 0.024 ms	0.0483 ± 0.024 ms	1.04 ± 0.73
saxpy/default/Float32/1048576	0.193 ± 0.017 ms	0.201 ± 0.02 ms	0.96 ± 0.13
saxpy/default/Float32/16384	0.0531 ± 0.022 ms	0.0518 ± 0.021 ms	1.03 ± 0.59
saxpy/default/Float32/2048	0.0516 ± 0.02 ms	0.0467 ± 0.023 ms	1.1 ± 0.69
saxpy/default/Float32/256	0.0539 ± 0.022 ms	0.0457 ± 0.027 ms	1.18 ± 0.84
saxpy/default/Float32/262144	0.0916 ± 0.022 ms	0.0903 ± 0.024 ms	1.01 ± 0.37
saxpy/default/Float32/32768	0.0578 ± 0.023 ms	0.0535 ± 0.023 ms	1.08 ± 0.63
saxpy/default/Float32/4096	0.0523 ± 0.02 ms	0.0511 ± 0.021 ms	1.02 ± 0.56
saxpy/default/Float32/512	0.0538 ± 0.022 ms	0.0479 ± 0.026 ms	1.12 ± 0.76
saxpy/default/Float32/64	0.0497 ± 0.027 ms	0.0452 ± 0.027 ms	1.1 ± 0.89
saxpy/default/Float32/65536	0.0633 ± 0.023 ms	0.0573 ± 0.023 ms	1.1 ± 0.6
saxpy/default/Float64/1024	0.0497 ± 0.022 ms	0.0432 ± 0.026 ms	1.15 ± 0.85
saxpy/default/Float64/1048576	0.277 ± 0.026 ms	0.268 ± 0.022 ms	1.03 ± 0.13
saxpy/default/Float64/16384	0.0535 ± 0.022 ms	0.0518 ± 0.022 ms	1.03 ± 0.61
saxpy/default/Float64/2048	0.054 ± 0.018 ms	0.0489 ± 0.023 ms	1.11 ± 0.64
saxpy/default/Float64/256	0.0533 ± 0.023 ms	0.0471 ± 0.026 ms	1.13 ± 0.79
saxpy/default/Float64/262144	0.104 ± 0.018 ms	0.102 ± 0.019 ms	1.02 ± 0.26
saxpy/default/Float64/32768	0.0588 ± 0.022 ms	0.0556 ± 0.022 ms	1.06 ± 0.58
saxpy/default/Float64/4096	0.0482 ± 0.021 ms	0.0467 ± 0.022 ms	1.03 ± 0.67
saxpy/default/Float64/512	0.0535 ± 0.02 ms	0.0472 ± 0.027 ms	1.13 ± 0.77
saxpy/default/Float64/64	0.0513 ± 0.026 ms	0.0442 ± 0.027 ms	1.16 ± 0.92
saxpy/default/Float64/65536	0.0671 ± 0.023 ms	0.062 ± 0.022 ms	1.08 ± 0.53
saxpy/static workgroup=(1024,)/Float32/1024	0.0509 ± 0.023 ms	0.0465 ± 0.025 ms	1.09 ± 0.77
saxpy/static workgroup=(1024,)/Float32/1048576	0.186 ± 0.016 ms	0.706 ± 0.015 ms	0.264 ± 0.023
saxpy/static workgroup=(1024,)/Float32/16384	0.0514 ± 0.022 ms	0.0607 ± 0.022 ms	0.847 ± 0.47
saxpy/static workgroup=(1024,)/Float32/2048	0.0495 ± 0.02 ms	0.0497 ± 0.024 ms	0.996 ± 0.63
saxpy/static workgroup=(1024,)/Float32/256	0.0507 ± 0.023 ms	0.0473 ± 0.026 ms	1.07 ± 0.76
saxpy/static workgroup=(1024,)/Float32/262144	0.0885 ± 0.022 ms	0.219 ± 0.018 ms	0.405 ± 0.11
saxpy/static workgroup=(1024,)/Float32/32768	0.0549 ± 0.022 ms	0.0686 ± 0.023 ms	0.799 ± 0.42
saxpy/static workgroup=(1024,)/Float32/4096	0.0534 ± 0.015 ms	0.0528 ± 0.021 ms	1.01 ± 0.49
saxpy/static workgroup=(1024,)/Float32/512	0.0503 ± 0.024 ms	0.0472 ± 0.025 ms	1.06 ± 0.75
saxpy/static workgroup=(1024,)/Float32/64	0.0495 ± 0.026 ms	0.0476 ± 0.024 ms	1.04 ± 0.75
saxpy/static workgroup=(1024,)/Float32/65536	0.0595 ± 0.023 ms	0.091 ± 0.022 ms	0.654 ± 0.3
saxpy/static workgroup=(1024,)/Float64/1024	0.0464 ± 0.024 ms	0.0459 ± 0.024 ms	1.01 ± 0.73
saxpy/static workgroup=(1024,)/Float64/1048576	0.272 ± 0.024 ms	0.746 ± 0.055 ms	0.365 ± 0.042
saxpy/static workgroup=(1024,)/Float64/16384	0.0526 ± 0.022 ms	0.0614 ± 0.022 ms	0.856 ± 0.47
saxpy/static workgroup=(1024,)/Float64/2048	0.05 ± 0.021 ms	0.0512 ± 0.024 ms	0.977 ± 0.61
saxpy/static workgroup=(1024,)/Float64/256	0.0486 ± 0.024 ms	0.0482 ± 0.024 ms	1.01 ± 0.72
saxpy/static workgroup=(1024,)/Float64/262144	0.103 ± 0.018 ms	0.225 ± 0.019 ms	0.458 ± 0.09
saxpy/static workgroup=(1024,)/Float64/32768	0.0567 ± 0.021 ms	0.0733 ± 0.022 ms	0.773 ± 0.37
saxpy/static workgroup=(1024,)/Float64/4096	0.0486 ± 0.021 ms	0.052 ± 0.022 ms	0.935 ± 0.57
saxpy/static workgroup=(1024,)/Float64/512	0.0503 ± 0.02 ms	0.0484 ± 0.025 ms	1.04 ± 0.68
saxpy/static workgroup=(1024,)/Float64/64	0.0489 ± 0.025 ms	0.0481 ± 0.024 ms	1.02 ± 0.73
saxpy/static workgroup=(1024,)/Float64/65536	0.0638 ± 0.022 ms	0.0958 ± 0.021 ms	0.666 ± 0.27
time_to_load	0.938 ± 0.015 s	0.951 ± 0.031 s	0.986 ± 0.036

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).