feat[gpu]: for kernel and framework for running pipeline

[joseph-isaacs]

Looks like we need to JIT

benchmark runs:

gpu_decompress_kernel_only/u32/1GB
                        time:   [5.6860 ms 5.6878 ms 5.6892 ms]
                        thrpt:  [175.77 GiB/s 175.81 GiB/s 175.87 GiB/s]

gpu_decompress_kernel_only/u32/10GB
                        time:   [56.376 ms 56.384 ms 56.392 ms]
                        thrpt:  [177.33 GiB/s 177.35 GiB/s 177.38 GiB/s]
          
gpu_for_decompress_kernel_only/u32/1GB
                        time:   [19.657 ms 19.674 ms 19.694 ms]
                        thrpt:  [50.778 GiB/s 50.827 GiB/s 50.871 GiB/s]

gpu_for_decompress_kernel_only/u32/10GB
                        time:   [203.45 ms 203.81 ms 204.21 ms]
                        thrpt:  [48.969 GiB/s 49.064 GiB/s 49.151 GiB/s]

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 87.63%. Comparing base (e604d22) to head (1f66e5e).
⚠️ Report is 15 commits behind head on develop.

☔ View full report in Codecov by Sentry.
📢 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.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[robert3005]

You can add launch.record_kernel_launch(sys::CUevent_flags::CU_EVENT_DEFAULT) to the launch arguments and then the launch returns a tuple of before and after events

[joseph-isaacs]

Yep. but I want to know the time to run both kernels

[codspeed-hq]

CodSpeed Performance Report

Merging #4857 will not alter performance

_{Comparing ji/bench-gpu-scan (1f66e5e) with develop (61ef9b6)¹}

Summary

✅ 1172 untouched

Footnotes

1. No successful run was found on develop (159c8a1) during the generation of this report, so 61ef9b6 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

[robert3005]

Some suggestions to simplify time measurement

[joseph-isaacs]

Its not clear how the simplification works when multiple launches?