ppc: mark regress-419922463.wasm as slow on sim [v8/v8 : main]
Milad Farazmand (Gerrit)
Milad Farazmand added 1 comment![]( "Open in Gerrit")
Related details
- Clemens Backes
Code-Owners
Code-Review
Review-Enforcement
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Clemens Backes (Gerrit)
Clemens Backes voted and added 1 comment![]( "Open in Gerrit")
Votes added by Clemens Backes
| Code-Review | +1 |
1 comment
If you ever want to use that fuzzer on ppc, this is actually something you should fix. E.g. by lowering the maximum number of executed instructions: https://source.chromium.org/chromium/chromium/src/+/main:v8/test/common/wasm/fuzzer-common.h;l=31;drc=a9f09a1411a697ace066a84efd018f49c28cc8a3
It's already lowered to 16k on simulators, if the test still takes that long there's something wrong.
Related details
- Milad Farazmand
Code-Owners
Code-Review
Review-Enforcement
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Milad Farazmand (Gerrit)
Milad Farazmand added 1 comment![]( "Open in Gerrit")
If you ever want to use that fuzzer on ppc, this is actually something you should fix. E.g. by lowering the maximum number of executed instructions: https://source.chromium.org/chromium/chromium/src/+/main:v8/test/common/wasm/fuzzer-common.h;l=31;drc=a9f09a1411a697ace066a84efd018f49c28cc8a3
It's already lowered to 16k on simulators, if the test still takes that long there's something wrong.
PPC simulator is known to be slow and we do have plans to eventually refactor it to speed it up. On a native PPC machine test runs at `0m0.481s` as is. Changing the value you mentioned to `4k` brings the Simulator run down to `0m42.173s` and `2k` makes it run at `0m0.036s` which I'm not sure if it's valid anymore?
Related details
Code-Owners
Code-Review
Review-Enforcement
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Clemens Backes (Gerrit)
Clemens Backes added 1 comment![]( "Open in Gerrit")
Milad FarazmandIf you ever want to use that fuzzer on ppc, this is actually something you should fix. E.g. by lowering the maximum number of executed instructions: https://source.chromium.org/chromium/chromium/src/+/main:v8/test/common/wasm/fuzzer-common.h;l=31;drc=a9f09a1411a697ace066a84efd018f49c28cc8a3
It's already lowered to 16k on simulators, if the test still takes that long there's something wrong.
PPC simulator is known to be slow and we do have plans to eventually refactor it to speed it up. On a native PPC machine test runs at `0m0.481s` as is. Changing the value you mentioned to `4k` brings the Simulator run down to `0m42.173s` and `2k` makes it run at `0m0.036s` which I'm not sure if it's valid anymore?
42 seconds for 4k instructions seems super slow. Your choice if you want to profile this or leave it for later. Do you even run that fuzzer anywhere?
2k is probably making the test run into that limit before hitting the slow thing. If that limit is hit during the Liftoff "reference run" then we skip Turbofan compilation and running in Turbofan. Still 2k steps would get executed in Liftoff, so I don't know why it's that much faster.
In any case, feel free to land this CL and investigate later.
Related details
- Milad Farazmand
Code-Owners
Code-Review
Review-Enforcement
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Milad Farazmand (Gerrit)
Milad Farazmand added 1 comment![]( "Open in Gerrit")
Milad FarazmandIf you ever want to use that fuzzer on ppc, this is actually something you should fix. E.g. by lowering the maximum number of executed instructions: https://source.chromium.org/chromium/chromium/src/+/main:v8/test/common/wasm/fuzzer-common.h;l=31;drc=a9f09a1411a697ace066a84efd018f49c28cc8a3
It's already lowered to 16k on simulators, if the test still takes that long there's something wrong.
Clemens BackesPPC simulator is known to be slow and we do have plans to eventually refactor it to speed it up. On a native PPC machine test runs at `0m0.481s` as is. Changing the value you mentioned to `4k` brings the Simulator run down to `0m42.173s` and `2k` makes it run at `0m0.036s` which I'm not sure if it's valid anymore?
42 seconds for 4k instructions seems super slow. Your choice if you want to profile this or leave it for later. Do you even run that fuzzer anywhere?
2k is probably making the test run into that limit before hitting the slow thing. If that limit is hit during the Liftoff "reference run" then we skip Turbofan compilation and running in Turbofan. Still 2k steps would get executed in Liftoff, so I don't know why it's that much faster.
In any case, feel free to land this CL and investigate later.
Thank you for the information. We only run the checked‑in fuzzer regression tests via `run-tests.py --variant=exhaustive` (not sure if this is what you are referring to?). I've created an issue to investigate this further later tho I suspect once the simulator is refactored this should get solved.
Related details
Code-Owners
Code-Review
Review-Enforcement
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Milad Farazmand (Gerrit)
Milad Farazmand voted Commit-Queue+2![]( "Open in Gerrit")
| Commit-Queue | +2 |
Related details
Code-Owners
Code-Review
Review-Enforcement
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
V8 LUCI CQ (Gerrit)
V8 LUCI CQ submitted the change![]( "Open in Gerrit")
Change information
ppc: mark regress-419922463.wasm as slow on sim
Test takes 1m3.452s on the simulator.
- M test/fuzzer/fuzzer.status
Code-Review: +1 by Clemens Backes
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |