plutus
plutus copied to clipboard
IntersectMBO
SCP-2289: Try making a special "for evaluation" term type
Another one of Ed's suggestions. I got rid of annotations and replaced
names with Uniques directly, which we can even unpack into the term.
This seems to have better Core: indeed, it gets rid of some wrapping and unwrapping when looking up or inserting variable mappings.
But the result is overall much worse.
Again, I am baffled. We do a traversal before, and potentially a traversal at the end or when we make errors, but all of that should be lazy. So I don't understand how we can be causing such massive slowdowns. Ideas welcome, again!
crowdfunding/1 733.6 μs 1.094 ms +49.1%
crowdfunding/2 733.7 μs 1.093 ms +49.0%
crowdfunding/3 733.2 μs 1.092 ms +48.9%
crowdfunding/4 299.0 μs 491.4 μs +64.3%
crowdfunding/5 298.9 μs 491.2 μs +64.3%
future/1 355.0 μs 501.0 μs +41.1%
future/2 876.1 μs 1.236 ms +41.1%
future/3 875.2 μs 1.236 ms +41.2%
future/4 1.111 ms 1.892 ms +70.3%
future/5 1.384 ms 2.133 ms +54.1%
future/6 1.121 ms 1.902 ms +69.7%
future/7 1.385 ms 2.133 ms +54.0%
multisigSM/1 737.7 μs 1.433 ms +94.3%
multisigSM/2 734.7 μs 1.435 ms +95.3%
multisigSM/3 753.4 μs 1.472 ms +95.4%
multisigSM/4 769.4 μs 1.476 ms +91.8%
multisigSM/5 937.7 μs 1.671 ms +78.2%
multisigSM/6 735.8 μs 1.443 ms +96.1%
multisigSM/7 730.4 μs 1.443 ms +97.6%
multisigSM/8 754.6 μs 1.466 ms +94.3%
multisigSM/9 773.3 μs 1.475 ms +90.7%
multisigSM/10 942.9 μs 1.667 ms +76.8%
vesting/1 715.6 μs 1.087 ms +51.9%
vesting/2 591.0 μs 947.8 μs +60.4%
vesting/3 714.7 μs 1.083 ms +51.5%
marlowe/trustfund/1 2.137 ms 4.004 ms +87.4%
marlowe/trustfund/2 1.539 ms 3.484 ms +126.4%
marlowe/zerocoupon/1 2.223 ms 4.124 ms +85.5%
marlowe/zerocoupon/2 1.333 ms 3.284 ms +146.4%
Pre-submit checklist:
- Branch
- [ ] Commit sequence broadly makes sense
- [ ] Key commits have useful messages
- [ ] Relevant tickets are mentioned in commit messages
- [ ] Formatting, materialized Nix files, PNG optimization, etc. are updated
- PR
- [ ] Self-reviewed the diff
- [ ] Useful pull request description
- [ ] Reviewer requested
Pre-merge checklist:
- [ ] Someone approved it
- [ ] Commits have useful messages
- [ ] Review clarifications made it into the code
- [ ] History is moderately tidy; or going to squash-merge
Or maybe I just can't get my head around the idea that a simple recursive traversal of the term could take 136% of the time to evaluate it. Surely not???
BTW, your stats look too shaky, are you sure there's no crazy variance in there?
Doesn't look lazy to me. Everything is spine-strict, so forcing to WHNF is forcing to NF.
Yeah, but we always return them in a lazy Just or similar. Also I checked by replacing etermToTerm with const Error () and it made no difference.
Previously, we'd only once pattern match on every AST node of a Term and the rest of the time deal with CekValues.
That's not quite true. CekValue contains terms, and we'll compute them multiple times. If we apply a lambda to 4 different arguments we will compute the term inside the CekValue 4 times.
Maybe Edward meant that we should change the on-chain representation? Then it would take us less time to construct the original Term with all the junk in it and we wouldn't need one more traversal.
If it's truly the case that it's the initial traversal that causes this then yeah, we'd have to do something like that.
BTW, your stats look too shaky, are you sure there's no crazy variance in there?
I ran it both yesterday and today on the benchmarking machine and got the same results. They're just quite weird!
So overall what I thought we might get is:
- Pattern matches are smaller: no pointless annotations
- Working with
Uniques monomorphically means GHC can just pass them directly, can unpack them etc.
I thought this might be a small improvement at best, I'm just baffled that it's instead much slower!
Yeah, but we always return them in a lazy Just or similar. Also I checked by replacing etermToTerm with const Error () and it made no difference.
I thought you meant termToETerm rather than etermToTerm.
That's not quite true. CekValue contains terms, and we'll compute them multiple times. If we apply a lambda to 4 different arguments we will compute the term inside the CekValue 4 times.
Yeah, you're right.
I thought this might be a small improvement at best, I'm just baffled that it's instead much slower!
How about adding termToTerm on top of master that just traverses its argument and returns it back to see how much the traversal costs us?
I tried the Marlowe benchmarks on my laptop and got this:
marlowe/trustfund/1 4.033 ms 7.354 ms +82.3%
marlowe/trustfund/2 2.933 ms 6.381 ms +117.6%
marlowe/zerocoupon/1 4.361 ms 7.483 ms +71.6%
marlowe/zerocoupon/2 2.566 ms 6.055 ms +136.0%
and then I removed the exclamation marks from the ETerm constructors apart from EConst and I got this
marlowe/trustfund/1 4.033 ms 5.465 ms +35.5%
marlowe/trustfund/2 2.933 ms 4.075 ms +38.9%
marlowe/zerocoupon/1 4.361 ms 5.606 ms +28.5%
marlowe/zerocoupon/2 2.566 ms 3.656 ms +42.5%
(master on the left in both cases), so making it strict seems harmful (I tried that with the existing Term type a while ago and also got bad results). I suppose the conversion isn't just traversing the term, but also allocating memory, which will slow things down too. We did get a pretty decent speedup when we removed the traversal that was adding memory annotations to the AST, even though just before we removed it it was just annotating everything with 0.
I suggest moving the ETerm type into a separate file and doing the conversion before the benchmarking, and seeing what happens then (although you'll have to make sure the conversion really does get done before Criterion gets hold of the AST, and I'm never totally confident that I'm doing that kind of thing correctly).
Maybe only trigger the benchmarks if that command is the only thing appearing in the comment?
Haha, amazing. You do have to say it with the slash, so it should mostly be safe if you don't literally quote someone else saying it :p
crowdfunding/crowdfunding-success-1 696.1 μs 627.9 μs -9.8%
crowdfunding/crowdfunding-success-2 696.2 μs 631.3 μs -9.3%
crowdfunding/crowdfunding-success-3 694.8 μs 631.0 μs -9.2%
currency/currency-1 626.5 μs 570.2 μs -9.0%
currency/currency-2 805.6 μs 780.6 μs -3.1%
escrow/escrow-redeem_1-1 962.7 μs 933.2 μs -3.1%
escrow/escrow-redeem_1-2 962.0 μs 934.1 μs -2.9%
escrow/escrow-redeem_2-1 1.104 ms 1.070 ms -3.1%
escrow/escrow-redeem_2-2 1.104 ms 1.070 ms -3.1%
escrow/escrow-redeem_2-3 1.102 ms 1.069 ms -3.0%
escrow/escrow-refund-1 477.0 μs 440.5 μs -7.7%
future/future-increase-margin-1 625.3 μs 568.0 μs -9.2%
future/future-increase-margin-2 808.5 μs 776.4 μs -4.0%
future/future-increase-margin-3 1.396 ms 1.349 ms -3.4%
future/future-increase-margin-4 1.402 ms 1.351 ms -3.6%
future/future-increase-margin-5 1.306 ms 1.250 ms -4.3%
future/future-increase-margin-6 1.936 ms 1.729 ms -10.7%
future/future-pay-out-1 627.5 μs 566.1 μs -9.8%
future/future-pay-out-2 810.4 μs 776.5 μs -4.2%
future/future-pay-out-3 1.394 ms 1.350 ms -3.2%
future/future-pay-out-4 1.392 ms 1.350 ms -3.0%
future/future-pay-out-5 1.809 ms 1.723 ms -4.8%
future/future-settle-early-1 626.5 μs 567.3 μs -9.4%
future/future-settle-early-2 808.3 μs 777.1 μs -3.9%
future/future-settle-early-3 1.396 ms 1.353 ms -3.1%
future/future-settle-early-4 1.402 ms 1.357 ms -3.2%
future/future-settle-early-5 1.435 ms 1.383 ms -3.6%
game-sm/game-sm-success-1 1.004 ms 967.0 μs -3.7%
game-sm/game-sm-success-2 537.5 μs 586.4 μs +9.1%
game-sm/game-sm-success-3 1.446 ms 1.409 ms -2.6%
game-sm/game-sm-success-4 750.9 μs 702.3 μs -6.5%
game-sm/game-sm-success_2-1 1.004 ms 969.7 μs -3.4%
game-sm/game-sm-success_2-2 537.7 μs 584.4 μs +8.7%
game-sm/game-sm-success_2-3 1.447 ms 1.406 ms -2.8%
game-sm/game-sm-success_2-4 751.4 μs 700.9 μs -6.7%
game-sm/game-sm-success_2-5 1.466 ms 1.381 ms -5.8%
game-sm/game-sm-success_2-6 749.6 μs 695.0 μs -7.3%
multisig-sm/multisig-sm-01 1.059 ms 1.026 ms -3.1%
multisig-sm/multisig-sm-02 1.068 ms 1.029 ms -3.7%
multisig-sm/multisig-sm-03 1.084 ms 1.045 ms -3.6%
multisig-sm/multisig-sm-04 1.100 ms 1.057 ms -3.9%
multisig-sm/multisig-sm-05 1.299 ms 1.251 ms -3.7%
multisig-sm/multisig-sm-06 1.157 ms 1.025 ms -11.4%
multisig-sm/multisig-sm-07 1.067 ms 1.030 ms -3.5%
multisig-sm/multisig-sm-08 1.083 ms 1.138 ms +5.1%
multisig-sm/multisig-sm-09 1.102 ms 1.062 ms -3.6%
multisig-sm/multisig-sm-10 1.305 ms 1.260 ms -3.4%
ping-pong/ping-pong-1 833.0 μs 772.6 μs -7.3%
ping-pong/ping-pong-2 834.2 μs 771.5 μs -7.5%
ping-pong/ping-pong_2-1 529.5 μs 478.0 μs -9.7%
prism/prism-1 499.2 μs 457.0 μs -8.5%
prism/prism-2 1.089 ms 1.038 ms -4.7%
prism/prism-3 991.4 μs 950.5 μs -4.1%
pubkey/pubkey-1 443.1 μs 401.9 μs -9.3%
stablecoin/stablecoin_1-1 2.100 ms 2.034 ms -3.1%
stablecoin/stablecoin_1-2 633.3 μs 574.8 μs -9.2%
stablecoin/stablecoin_1-3 2.369 ms 2.295 ms -3.1%
stablecoin/stablecoin_1-4 687.1 μs 626.7 μs -8.8%
stablecoin/stablecoin_1-5 2.871 ms 2.955 ms +2.9%
stablecoin/stablecoin_1-6 719.9 μs 807.0 μs +12.1%
stablecoin/stablecoin_2-1 2.091 ms 2.045 ms -2.2%
stablecoin/stablecoin_2-2 631.8 μs 577.4 μs -8.6%
stablecoin/stablecoin_2-3 2.359 ms 2.323 ms -1.5%
stablecoin/stablecoin_2-4 685.5 μs 626.9 μs -8.5%
token-account/token-account-1 594.0 μs 540.0 μs -9.1%
token-account/token-account-2 695.9 μs 636.6 μs -8.5%
token-account/token-account-3 885.2 μs 849.9 μs -4.0%
uniswap/uniswap-1 698.8 μs 630.6 μs -9.8%
uniswap/uniswap-2 1.114 ms 1.072 ms -3.8%
uniswap/uniswap-3 756.0 μs 697.2 μs -7.8%
uniswap/uniswap-4 831.2 μs 794.6 μs -4.4%
uniswap/uniswap-5 3.882 ms 3.750 ms -3.4%
uniswap/uniswap-6 1.128 ms 1.075 ms -4.7%
uniswap/uniswap-7 2.799 ms 2.816 ms +0.6%
uniswap/uniswap-8 1.070 ms 1.073 ms +0.3%
vesting/vesting-1 900.4 μs 868.2 μs -3.6%
marlowe/trustfund/trustfund-1 2.101 ms 2.015 ms -4.1%
marlowe/trustfund/trustfund-2 1.503 ms 1.473 ms -2.0%
marlowe/zerocoupon/zerocoupon-1 2.173 ms 2.116 ms -2.6%
marlowe/zerocoupon/zerocoupon-2 1.303 ms 1.283 ms -1.5%
[kwxm: edited to add backticks for better legibility]
stablecoin/stablecoin_1-6 719.9 μs 807.0 μs +12.1%
AGAIN???? https://input-output-rnd.slack.com/archives/G01EE4NQ9U0/p1624872008442400?thread_ts=1624872002.442300&cid=G01EE4NQ9U0
So what's going on here? The original benchmarks for this PR showed that it made things much worse, but now it's an improvement and there have been no other commits in the interim.
So what's going on here? The original benchmarks for this PR showed that it made things much worse, but now it's an improvement and there have been no other commits in the interim.
No, I squashed some stuff in. In particular I pulled the conversion to ETerm out of the part that gets benchmarked. I think that's fair because in a "proper" version of this we could e.g. push that right into deserialization. We can't quite do that now because of the intervening conversion from de Bruijn indices, but once we're not doing that, I even included an instance for Flat (ETerm uni fun) such that you can basically deserialize something you serialized as a Term. So that should really be cheap.
I pulled the conversion to
ETermout of the part that gets benchmarked.
Oh, right. I should have re-read the other comments!
Also some in the game state machine. Quite odd.
Those are exactly the same scripts that were anomalous in your frames PR last week. I'd like to know what's going on there. I spent about 3 hours talking to @bezirg about the similar thing with stablecoin_1-6 yesterday and we couldn't come to any sensible conclusion (partly because we spent a lot of time talking past each other: it eventually appeared to be the case that we'd done two completely different things that both caused slowdowns in the same script). The only possibility that we were able to come up with was that the contents of TxInfo had been re-ordered and it was taking longer to look something up. The slowdown was something like 17% though, so it would be surprising if looking stuff up was causing that (although other scripts had largish speedups that we didn't examine which could be due to the same thing). Maybe built-in Data will help.
Random guess: the ones that regress are ones that are quite fast already. If they're scripts that do something like (say) look through the list of outputs to verify that someone is paid appropriately, then maybe reordering so that the output they're looking for is later could have an effect.
crowdfunding/crowdfunding-success-1 694.3 μs 629.9 μs -9.3%
crowdfunding/crowdfunding-success-2 697.8 μs 633.6 μs -9.2%
crowdfunding/crowdfunding-success-3 697.0 μs 633.8 μs -9.1%
currency/currency-1 629.2 μs 568.7 μs -9.6%
currency/currency-2 811.4 μs 779.0 μs -4.0%
escrow/escrow-redeem_1-1 967.8 μs 930.6 μs -3.8%
escrow/escrow-redeem_1-2 967.2 μs 931.8 μs -3.7%
escrow/escrow-redeem_2-1 1.110 ms 1.065 ms -4.1%
escrow/escrow-redeem_2-2 1.110 ms 1.065 ms -4.1%
escrow/escrow-redeem_2-3 1.110 ms 1.064 ms -4.1%
escrow/escrow-refund-1 479.5 μs 437.9 μs -8.7%
future/future-increase-margin-1 629.8 μs 567.7 μs -9.9%
future/future-increase-margin-2 811.0 μs 778.1 μs -4.1%
future/future-increase-margin-3 1.398 ms 1.356 ms -3.0%
future/future-increase-margin-4 1.398 ms 1.356 ms -3.0%
future/future-increase-margin-5 1.302 ms 1.254 ms -3.7%
future/future-increase-margin-6 1.936 ms 1.743 ms -10.0%
future/future-pay-out-1 631.6 μs 568.3 μs -10.0%
future/future-pay-out-2 814.5 μs 779.9 μs -4.2%
future/future-pay-out-3 1.402 ms 1.355 ms -3.4%
future/future-pay-out-4 1.404 ms 1.357 ms -3.3%
future/future-pay-out-5 1.826 ms 1.730 ms -5.3%
future/future-settle-early-1 630.3 μs 569.2 μs -9.7%
future/future-settle-early-2 811.9 μs 780.0 μs -3.9%
future/future-settle-early-3 1.404 ms 1.355 ms -3.5%
future/future-settle-early-4 1.403 ms 1.356 ms -3.3%
future/future-settle-early-5 1.436 ms 1.378 ms -4.0%
game-sm/game-sm-success-1 1.003 ms 965.6 μs -3.7%
game-sm/game-sm-success-2 537.3 μs 583.5 μs +8.6%
game-sm/game-sm-success-3 1.441 ms 1.405 ms -2.5%
game-sm/game-sm-success-4 748.9 μs 697.8 μs -6.8%
game-sm/game-sm-success_2-1 1.001 ms 964.4 μs -3.7%
game-sm/game-sm-success_2-2 536.6 μs 583.5 μs +8.7%
game-sm/game-sm-success_2-3 1.451 ms 1.400 ms -3.5%
game-sm/game-sm-success_2-4 753.9 μs 698.1 μs -7.4%
game-sm/game-sm-success_2-5 1.471 ms 1.386 ms -5.8%
game-sm/game-sm-success_2-6 750.9 μs 698.1 μs -7.0%
multisig-sm/multisig-sm-01 1.065 ms 1.033 ms -3.0%
multisig-sm/multisig-sm-02 1.074 ms 1.035 ms -3.6%
multisig-sm/multisig-sm-03 1.090 ms 1.052 ms -3.5%
multisig-sm/multisig-sm-04 1.106 ms 1.063 ms -3.9%
multisig-sm/multisig-sm-05 1.305 ms 1.258 ms -3.6%
multisig-sm/multisig-sm-06 1.163 ms 1.030 ms -11.4%
multisig-sm/multisig-sm-07 1.072 ms 1.028 ms -4.1%
multisig-sm/multisig-sm-08 1.087 ms 1.136 ms +4.5%
multisig-sm/multisig-sm-09 1.105 ms 1.059 ms -4.2%
multisig-sm/multisig-sm-10 1.299 ms 1.254 ms -3.5%
ping-pong/ping-pong-1 830.0 μs 771.9 μs -7.0%
ping-pong/ping-pong-2 830.9 μs 776.4 μs -6.6%
ping-pong/ping-pong_2-1 528.3 μs 478.1 μs -9.5%
prism/prism-1 496.3 μs 457.7 μs -7.8%
prism/prism-2 1.093 ms 1.043 ms -4.6%
prism/prism-3 989.0 μs 952.0 μs -3.7%
pubkey/pubkey-1 444.3 μs 403.2 μs -9.3%
stablecoin/stablecoin_1-1 2.122 ms 2.048 ms -3.5%
stablecoin/stablecoin_1-2 634.2 μs 576.6 μs -9.1%
stablecoin/stablecoin_1-3 2.374 ms 2.305 ms -2.9%
stablecoin/stablecoin_1-4 688.4 μs 628.0 μs -8.8%
stablecoin/stablecoin_1-5 2.885 ms 2.941 ms +1.9%
stablecoin/stablecoin_1-6 721.4 μs 805.8 μs +11.7%
stablecoin/stablecoin_2-1 2.095 ms 2.040 ms -2.6%
stablecoin/stablecoin_2-2 632.9 μs 574.5 μs -9.2%
stablecoin/stablecoin_2-3 2.363 ms 2.315 ms -2.0%
stablecoin/stablecoin_2-4 687.3 μs 627.0 μs -8.8%
token-account/token-account-1 593.7 μs 540.8 μs -8.9%
token-account/token-account-2 693.7 μs 639.2 μs -7.9%
token-account/token-account-3 881.8 μs 853.1 μs -3.3%
uniswap/uniswap-1 699.8 μs 633.8 μs -9.4%
uniswap/uniswap-2 1.113 ms 1.076 ms -3.3%
uniswap/uniswap-3 757.3 μs 699.6 μs -7.6%
uniswap/uniswap-4 833.1 μs 796.6 μs -4.4%
uniswap/uniswap-5 3.897 ms 3.763 ms -3.4%
uniswap/uniswap-6 1.128 ms 1.073 ms -4.9%
uniswap/uniswap-7 2.801 ms 2.716 ms -3.0%
uniswap/uniswap-8 1.070 ms 1.016 ms -5.0%
vesting/vesting-1 901.1 μs 864.0 μs -4.1%
marlowe/trustfund/trustfund-1 2.103 ms 1.995 ms -5.1%
marlowe/trustfund/trustfund-2 1.504 ms 1.465 ms -2.6%
marlowe/zerocoupon/zerocoupon-1 2.174 ms 2.094 ms -3.7%
marlowe/zerocoupon/zerocoupon-2 1.304 ms 1.273 ms -2.4%
crowdfunding/crowdfunding-success-1 695.2 μs 628.5 μs -9.6% crowdfunding/crowdfunding-success-2 695.2 μs 634.5 μs -8.7% crowdfunding/crowdfunding-success-3 693.8 μs 637.2 μs -8.2% currency/currency-1 625.3 μs 572.0 μs -8.5% currency/currency-2 811.1 μs 783.7 μs -3.4% escrow/escrow-redeem_1-1 968.7 μs 934.9 μs -3.5% escrow/escrow-redeem_1-2 968.5 μs 935.3 μs -3.4% escrow/escrow-redeem_2-1 1.109 ms 1.071 ms -3.4% escrow/escrow-redeem_2-2 1.114 ms 1.070 ms -3.9% escrow/escrow-redeem_2-3 1.113 ms 1.070 ms -3.9% escrow/escrow-refund-1 481.0 μs 440.2 μs -8.5% future/future-increase-margin-1 629.7 μs 567.5 μs -9.9% future/future-increase-margin-2 813.2 μs 777.1 μs -4.4% future/future-increase-margin-3 1.401 ms 1.350 ms -3.6% future/future-increase-margin-4 1.402 ms 1.352 ms -3.6% future/future-increase-margin-5 1.306 ms 1.250 ms -4.3% future/future-increase-margin-6 1.932 ms 1.729 ms -10.5% future/future-pay-out-1 627.4 μs 565.0 μs -9.9% future/future-pay-out-2 808.9 μs 778.6 μs -3.7% future/future-pay-out-3 1.400 ms 1.351 ms -3.5% future/future-pay-out-4 1.402 ms 1.360 ms -3.0% future/future-pay-out-5 1.824 ms 1.736 ms -4.8% future/future-settle-early-1 632.0 μs 570.7 μs -9.7% future/future-settle-early-2 814.4 μs 781.8 μs -4.0% future/future-settle-early-3 1.407 ms 1.362 ms -3.2% future/future-settle-early-4 1.406 ms 1.360 ms -3.3% future/future-settle-early-5 1.437 ms 1.384 ms -3.7% game-sm/game-sm-success-1 1.004 ms 968.8 μs -3.5% game-sm/game-sm-success-2 537.1 μs 586.1 μs +9.1% game-sm/game-sm-success-3 1.443 ms 1.409 ms -2.4% game-sm/game-sm-success-4 749.4 μs 702.2 μs -6.3% game-sm/game-sm-success_2-1 1.002 ms 970.9 μs -3.1% game-sm/game-sm-success_2-2 541.0 μs 585.7 μs +8.3% game-sm/game-sm-success_2-3 1.452 ms 1.404 ms -3.3% game-sm/game-sm-success_2-4 749.0 μs 700.9 μs -6.4% game-sm/game-sm-success_2-5 1.463 ms 1.379 ms -5.7% game-sm/game-sm-success_2-6 747.7 μs 695.7 μs -7.0% multisig-sm/multisig-sm-01 1.057 ms 1.028 ms -2.7% multisig-sm/multisig-sm-02 1.069 ms 1.033 ms -3.4% multisig-sm/multisig-sm-03 1.090 ms 1.052 ms -3.5% multisig-sm/multisig-sm-04 1.106 ms 1.063 ms -3.9% multisig-sm/multisig-sm-05 1.305 ms 1.259 ms -3.5% multisig-sm/multisig-sm-06 1.165 ms 1.034 ms -11.2% multisig-sm/multisig-sm-07 1.074 ms 1.032 ms -3.9% multisig-sm/multisig-sm-08 1.092 ms 1.140 ms +4.4% multisig-sm/multisig-sm-09 1.108 ms 1.062 ms -4.2% multisig-sm/multisig-sm-10 1.304 ms 1.260 ms -3.4% ping-pong/ping-pong-1 833.0 μs 772.2 μs -7.3% ping-pong/ping-pong-2 832.7 μs 771.6 μs -7.3% ping-pong/ping-pong_2-1 529.1 μs 477.6 μs -9.7% prism/prism-1 498.8 μs 456.8 μs -8.4% prism/prism-2 1.088 ms 1.036 ms -4.8% prism/prism-3 985.1 μs 950.0 μs -3.6% pubkey/pubkey-1 440.2 μs 402.6 μs -8.5% stablecoin/stablecoin_1-1 2.093 ms 2.047 ms -2.2% stablecoin/stablecoin_1-2 631.8 μs 577.3 μs -8.6% stablecoin/stablecoin_1-3 2.361 ms 2.313 ms -2.0% stablecoin/stablecoin_1-4 689.7 μs 631.2 μs -8.5% stablecoin/stablecoin_1-5 2.892 ms 2.948 ms +1.9% stablecoin/stablecoin_1-6 721.9 μs 807.8 μs +11.9% stablecoin/stablecoin_2-1 2.101 ms 2.046 ms -2.6% stablecoin/stablecoin_2-2 635.1 μs 577.9 μs -9.0% stablecoin/stablecoin_2-3 2.371 ms 2.320 ms -2.2% stablecoin/stablecoin_2-4 689.3 μs 627.4 μs -9.0% token-account/token-account-1 596.3 μs 541.2 μs -9.2% token-account/token-account-2 696.0 μs 637.7 μs -8.4% token-account/token-account-3 884.0 μs 849.7 μs -3.9% uniswap/uniswap-1 696.4 μs 629.9 μs -9.5% uniswap/uniswap-2 1.110 ms 1.078 ms -2.9% uniswap/uniswap-3 754.7 μs 702.1 μs -7.0% uniswap/uniswap-4 827.4 μs 800.5 μs -3.3% uniswap/uniswap-5 3.878 ms 3.784 ms -2.4% uniswap/uniswap-6 1.124 ms 1.080 ms -3.9% uniswap/uniswap-7 2.787 ms 2.725 ms -2.2% uniswap/uniswap-8 1.074 ms 1.022 ms -4.8% vesting/vesting-1 903.2 μs 869.2 μs -3.8% marlowe/trustfund/trustfund-1 2.113 ms 2.011 ms -4.8% marlowe/trustfund/trustfund-2 1.509 ms 1.472 ms -2.5% marlowe/zerocoupon/zerocoupon-1 2.189 ms 2.105 ms -3.8% marlowe/zerocoupon/zerocoupon-2 1.311 ms 1.280 ms -2.4%
@michaelpj I'm going to close this as part of my campaign to get rid of zombie PRs. This one's been open since June 2021and it says "This branch has conflicts that must be resolved". Yeah, I bet it does.
Oh, it also says "Some checks haven’t completed yet"!