@Zac Spitzer , I have investigated this ticket with Lucee version 6.2.0.230-SNAPSHOT. When I compared the performance time in the test lab with the default PreciseMath set to false, the performance was slightly increased. Now, in the latest version of Lucee, the performance has improved, therefore, I am going to close this ticket.

Updated most of the tests which were assume precise math is enabled to pass when it’s disabled

https://github.com/lucee/Lucee/commit/c6bc6c1c7609cae792444b2d3d5914158a17983b

there is a regression I found https://luceeserver.atlassian.net/browse/LDEV-5178

https://github.com/lucee/Lucee/commit/b61478ee6d1c7a0b86f2a146144e1c892a1b3ab6

Looking at the max times with loops, they are higher when precise math is enabled, I’m guessing there’s more GC going on

it’s on by default in 6.0, 6.1 and 6.2

here’s the spreadsheet-cfml benchmarks

Open

and with LUCEE_PRECISE_MATH=false

Open

here are the benchmarks

LUCEE_PRECISE_MATH=false

Open

https://github.com/lucee/lucee-testlab/actions/runs/11911949700/attempts/1#summary-33195596635

and with LUCEE_PRECISE_MATH=true (unset, i.e default)

Open

https://github.com/lucee/lucee-testlab/actions/runs/11911586638/attempts/1#summary-33194409458

loops and primes for example are really improved

Open

and with LUCEE_PRECISE_MATH=false

Open

interestingly, json seems to regress performance wise? https://luceeserver.atlassian.net/browse/LDEV-5162

Open

and with LUCEE_PRECISE_MATH=false

Open