URL:
<https://savannah.gnu.org/support/?110729>
Summary: m4 processing of large grammars slow
Project: Bison
Submitter: anarazel
Submitted: Tue 20 Sep 2022 07:08:29 PM UTC
Category: None
Priority: 5 - Normal
Severity: 3 - Normal
Status: None
Privacy: Public
Assigned to: None
Originator Email:
Open/Closed: Open
Discussion Lock: Any
Operating System: None
_______________________________________________________
Follow-up Comments:
-------------------------------------------------------
Date: Tue 20 Sep 2022 07:08:29 PM UTC By: Andres Freund <anarazel>
Hi,
translating postgres' bison grammars these days is the slowest build step of
building postgres. Interestingly enough the time isn't even spent building the
state machine or such, it's nearly all in m4.
Here's an example with postgres' main grammar. While there are other grammars
that are slower to build, they require prior build steps.
curl -o gram.y
'https://git.postgresql.org/gitweb/?p=postgresql.git;a=blob_plain;f=src/backend/parser/gram.y;hb=HEAD'
$ /usr/bin/bison --trace=time -Wno-deprecated -o gram.c -d gram.y
bison (GNU Bison) 3.8.2
Execution times (seconds)
CPU user CPU system wall clock
reader 0.044 ( 2%) 0.000 ( 0%) 0.044988 ( 2%)
LR(0) 0.038 ( 2%) 0.000 ( 0%) 0.038423 ( 2%)
LALR(1) 0.083 ( 4%) 0.003 (16%) 0.086182 ( 4%)
parser action tables 0.149 ( 7%) 0.000 ( 0%) 0.149236 ( 7%)
outputting parser 0.075 ( 3%) 0.003 (16%) 0.119665 ( 6%)
running m4 1.518 (79%) 0.013 (66%) 1.454934 (76%)
total time 1.914 0.020 1.900101
On slow systems this often ends up > 10s.
I don't know anything about m4, but perhaps there's something that can be done
to speed this step up?
Regards,
Andres
_______________________________________________________
Reply to this item at:
<https://savannah.gnu.org/support/?110729>
_______________________________________________
Message sent via Savannah
https://savannah.gnu.org/
