Enzyme: fix propagation of runtime activity

[wsmoses]

Required for chmy if differentiating an existing differentiated kernel

[github-actions]

Benchmark Results

	main	e679d50...	main/e679d50b98b05b...
saxpy/default/Float16/1024	2.78 ± 0.2 μs	2.79 ± 0.19 μs	0.994
saxpy/default/Float16/1048576	2.08 ± 0.0064 ms	2.08 ± 0.0046 ms	1
saxpy/default/Float16/16384	0.0328 ± 0.00014 ms	0.0328 ± 0.00014 ms	1
saxpy/default/Float16/2048	5.21 ± 0.041 μs	5.22 ± 0.038 μs	0.997
saxpy/default/Float16/256	1 ± 0.11 μs	0.963 ± 0.074 μs	1.04
saxpy/default/Float16/262144	0.524 ± 0.0095 ms	0.524 ± 0.0095 ms	1
saxpy/default/Float16/32768	0.065 ± 0.00017 ms	0.0651 ± 0.00018 ms	1
saxpy/default/Float16/4096	10.1 ± 0.05 μs	10.1 ± 0.06 μs	0.998
saxpy/default/Float16/512	1.57 ± 0.16 μs	1.57 ± 0.05 μs	0.994
saxpy/default/Float16/64	0.622 ± 0.016 μs	0.618 ± 0.016 μs	1.01
saxpy/default/Float16/65536	0.129 ± 0.00033 ms	0.129 ± 0.00034 ms	1
saxpy/default/Float32/1024	1.02 ± 0.012 μs	1.03 ± 0.016 μs	0.993
saxpy/default/Float32/1048576	0.963 ± 0.0075 ms	0.966 ± 0.0081 ms	0.997
saxpy/default/Float32/16384	15.4 ± 0.12 μs	15.5 ± 0.13 μs	0.997
saxpy/default/Float32/2048	1.71 ± 0.019 μs	1.74 ± 0.024 μs	0.986
saxpy/default/Float32/256	0.532 ± 0.12 μs	0.53 ± 0.13 μs	1
saxpy/default/Float32/262144	0.238 ± 0.0093 ms	0.238 ± 0.0098 ms	0.998
saxpy/default/Float32/32768	30.3 ± 0.15 μs	30.4 ± 0.18 μs	0.999
saxpy/default/Float32/4096	3.03 ± 0.024 μs	3.02 ± 0.025 μs	1
saxpy/default/Float32/512	0.695 ± 0.11 μs	0.696 ± 0.12 μs	1
saxpy/default/Float32/64	0.427 ± 0.0058 μs	0.414 ± 0.0048 μs	1.03
saxpy/default/Float32/65536	0.0602 ± 0.00039 ms	0.0601 ± 0.00028 ms	1
saxpy/default/Float64/1024	1.06 ± 0.018 μs	1.06 ± 0.02 μs	1
saxpy/default/Float64/1048576	1.02 ± 0.028 ms	1.02 ± 0.024 ms	0.997
saxpy/default/Float64/16384	15.8 ± 0.21 μs	15.8 ± 0.28 μs	0.997
saxpy/default/Float64/2048	1.75 ± 0.023 μs	1.76 ± 0.029 μs	0.995
saxpy/default/Float64/256	0.526 ± 0.0087 μs	0.515 ± 0.012 μs	1.02
saxpy/default/Float64/262144	0.243 ± 0.01 ms	0.243 ± 0.01 ms	0.999
saxpy/default/Float64/32768	30.9 ± 0.61 μs	31 ± 0.43 μs	0.997
saxpy/default/Float64/4096	3.06 ± 0.045 μs	3.05 ± 0.041 μs	1
saxpy/default/Float64/512	0.7 ± 0.11 μs	0.7 ± 0.11 μs	0.999
saxpy/default/Float64/64	0.397 ± 0.0066 μs	0.393 ± 0.0081 μs	1.01
saxpy/default/Float64/65536	0.0626 ± 0.0022 ms	0.0613 ± 0.00077 ms	1.02
saxpy/static workgroup=(1024,)/Float16/1024	2.12 ± 0.21 μs	2.11 ± 0.21 μs	1
saxpy/static workgroup=(1024,)/Float16/1048576	0.16 ± 0.0098 ms	0.17 ± 0.017 ms	0.945
saxpy/static workgroup=(1024,)/Float16/16384	4.36 ± 0.24 μs	4.35 ± 0.22 μs	1
saxpy/static workgroup=(1024,)/Float16/2048	2.15 ± 0.23 μs	2.13 ± 0.22 μs	1.01
saxpy/static workgroup=(1024,)/Float16/256	2.66 ± 0.037 μs	2.63 ± 0.037 μs	1.01
saxpy/static workgroup=(1024,)/Float16/262144	0.043 ± 0.0026 ms	0.0437 ± 0.0037 ms	0.985
saxpy/static workgroup=(1024,)/Float16/32768	6.6 ± 0.28 μs	6.62 ± 0.25 μs	0.996
saxpy/static workgroup=(1024,)/Float16/4096	2.43 ± 0.034 μs	2.44 ± 0.033 μs	0.998
saxpy/static workgroup=(1024,)/Float16/512	3.18 ± 0.061 μs	3.14 ± 0.086 μs	1.01
saxpy/static workgroup=(1024,)/Float16/64	2.27 ± 0.021 μs	2.24 ± 0.022 μs	1.01
saxpy/static workgroup=(1024,)/Float16/65536	12.7 ± 0.67 μs	12.5 ± 0.44 μs	1.02
saxpy/static workgroup=(1024,)/Float32/1024	1.97 ± 0.026 μs	1.96 ± 0.025 μs	1
saxpy/static workgroup=(1024,)/Float32/1048576	0.268 ± 0.025 ms	0.264 ± 0.03 ms	1.02
saxpy/static workgroup=(1024,)/Float32/16384	4.12 ± 0.93 μs	4.12 ± 0.9 μs	1
saxpy/static workgroup=(1024,)/Float32/2048	2.31 ± 0.22 μs	2.28 ± 0.22 μs	1.01
saxpy/static workgroup=(1024,)/Float32/256	2.65 ± 0.42 μs	2.65 ± 0.45 μs	1
saxpy/static workgroup=(1024,)/Float32/262144	0.0654 ± 0.0059 ms	0.0594 ± 0.0083 ms	1.1
saxpy/static workgroup=(1024,)/Float32/32768	7.08 ± 0.38 μs	7.16 ± 0.57 μs	0.988
saxpy/static workgroup=(1024,)/Float32/4096	2.59 ± 0.2 μs	2.57 ± 0.2 μs	1.01
saxpy/static workgroup=(1024,)/Float32/512	2.48 ± 0.23 μs	2.47 ± 0.22 μs	1.01
saxpy/static workgroup=(1024,)/Float32/64	2.43 ± 0.051 μs	2.42 ± 0.052 μs	1
saxpy/static workgroup=(1024,)/Float32/65536	16.8 ± 1.4 μs	16.5 ± 2.8 μs	1.02
saxpy/static workgroup=(1024,)/Float64/1024	2.06 ± 0.028 μs	2.06 ± 0.029 μs	1
saxpy/static workgroup=(1024,)/Float64/1048576	0.594 ± 0.065 ms	0.578 ± 0.071 ms	1.03
saxpy/static workgroup=(1024,)/Float64/16384	7.06 ± 1.3 μs	7.2 ± 1.4 μs	0.981
saxpy/static workgroup=(1024,)/Float64/2048	2.54 ± 0.26 μs	2.54 ± 0.25 μs	1
saxpy/static workgroup=(1024,)/Float64/256	2.41 ± 0.055 μs	2.41 ± 0.055 μs	0.998
saxpy/static workgroup=(1024,)/Float64/262144	0.108 ± 0.013 ms	0.127 ± 0.013 ms	0.849
saxpy/static workgroup=(1024,)/Float64/32768	16.9 ± 1.7 μs	16.5 ± 2.4 μs	1.02
saxpy/static workgroup=(1024,)/Float64/4096	3.02 ± 0.35 μs	2.99 ± 0.35 μs	1.01
saxpy/static workgroup=(1024,)/Float64/512	2.39 ± 0.041 μs	2.41 ± 0.046 μs	0.994
saxpy/static workgroup=(1024,)/Float64/64	2.37 ± 0.077 μs	2.38 ± 0.082 μs	0.995
saxpy/static workgroup=(1024,)/Float64/65536	0.0318 ± 0.0045 ms	0.0337 ± 0.0041 ms	0.944
time_to_load	0.313 ± 0.0023 s	0.316 ± 0.0022 s	0.993

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).

[vchuravy]

We have to figure out the compat issue, currently there is no real reason for KA to stop supporting 1.6, besides the Enzyme extension

[wsmoses]

I think the compat issue was introduced previously (presumably by me on the 0.13 enzyme PR). We should definitely fix it, but I presume this diff shouldn’t change things for it