| From: | Andreas Karlsson <andreas(at)proxel(dot)se> |
|---|---|
| To: | Xing Guo <higuoxing(at)gmail(dot)com>, pgsql-hackers(at)lists(dot)postgresql(dot)org |
| Subject: | Re: JIT: The nullness of casetest.value can be determined at the JIT compile time. |
| Date: | 2024-09-03 12:09:04 |
| Message-ID: | abeaa792-1db5-4438-b72f-fdc7a1c444ab@proxel.se |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On 8/31/24 10:04 AM, Xing Guo wrote:
> The nullness of casetest.value can be determined at the JIT compile
> time. We can emit fewer codes by utilizing this property. The attached
> patch is trying to fix it.
I have not reviewed the code yet but the idea seems good.
But I wonder if we shouldn't instead simplify the code a bit by
specializing these steps when generating them instead of doing the work
runtime/while generating machine code. Yes, I doubt the performance
benefits matter but I personally think the code is cleaner before my
patch than after it.
Long term it would be nice to get rid off
caseValue_datum/domainValue_datum as mentioned by Andres[1] but that is
a bigger job so think that either your patch or my patch would make
sense to apply before that.
Andreas
| Attachment | Content-Type | Size |
|---|---|---|
| v1-0001-Specialize-EEOP_-_TESTVAL-steps.patch | text/x-patch | 9.0 KB |
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Daniel Gustafsson | 2024-09-03 12:18:07 | Re: [PATCH] Avoid mixing custom and OpenSSL BIO functions |
| Previous Message | Pavel Stehule | 2024-09-03 11:41:31 | Re: proposal: schema variables |