| From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
|---|---|
| To: | Tomas Vondra <tomas(dot)vondra(at)enterprisedb(dot)com> |
| Cc: | PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org>, Masahiko Sawada <sawada(dot)mshk(at)gmail(dot)com>, John Naylor <johncnaylorls(at)gmail(dot)com> |
| Subject: | Re: suspicious valgrind reports about radixtree/tidstore on arm64 |
| Date: | 2024-06-19 23:51:52 |
| Message-ID: | 248275.1718841112@sss.pgh.pa.us |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
I wrote:
> I hypothesize that the reason we're not seeing equivalent failures
> on x86_64 is one of
> 1. x86_64 valgrind is stupider than aarch64, and fails to track that
> the contents of the SIMD registers are only partially defined.
> 2. x86_64 valgrind is smarter than aarch64, and is able to see
> that the "mask off invalid entries" step removes all the
> potentially-uninitialized bits.
Hah: it's the second case. If I patch radixtree.h as attached,
then x86_64 valgrind complains about
==00:00:00:32.759 247596== Conditional jump or move depends on uninitialised value(s)
==00:00:00:32.759 247596== at 0x52F668: local_ts_node_16_search_eq (radixtree.h:1018)
showing that it knows that the result of vector8_highbit_mask is
only partly defined. Kind of odd though that aarch64 valgrind
is getting the hard part right and not the easy(?) part.
regards, tom lane
| Attachment | Content-Type | Size |
|---|---|---|
| make-radixtree-failure-visible-on-x86_64.patch | text/x-diff | 900 bytes |
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Michael Paquier | 2024-06-20 00:46:30 | Re: Pluggable cumulative statistics |
| Previous Message | Tom Lane | 2024-06-19 23:33:21 | Re: suspicious valgrind reports about radixtree/tidstore on arm64 |