timo

UserfaultFD with #Rakulang

This Gist has a super hacky ugly implementation of a part of the UserfaultFD feature of the linux kernel (awkwardly) using raku's NativeCall. This could be much improved with a little bit of care.

Check this recording of what happens:

you can see multiple worker threads (doods) trying to read from and write to a memory region that was initially write-protected and watched for access attempts by the UFFD thread.
The doods are stopped in their tracks when they first access some location, then the UFFD thread decides to let them continue after copying some data into the area and when they try to write to it, it just un-write-protects the page.

timo

#!/bin/bash

# Video Quality
# The range of the CRF scale is 0–51, where 0 is lossless, 23 is the default,
# and 51 is worst quality possible. A lower value generally leads to higher
# quality, and a subjectively sane range is 17–28
QUALITY=28

# check if slop command exists
if ! command -v slop &> /dev/null

timo

0x7ffff6f518b0 has     4 edges over     3 states;     0 small (0.00)
0x7ffff6f52388 has   237 edges over    95 states;   118 small (49.79)
0x7ffff6f523e0 has   230 edges over    97 states;   108 small (46.96)
0x7ffff6f526c8 has    10 edges over     6 states;     5 small (50.00)
0x7ffff6f529d0 has     5 edges over     3 states;     1 small (20.00)
Stage start      :   0.000
0x7ffff6f5abc8 has    87 edges over    56 states;    43 small (49.43)
0x7ffff6f5b460 has    52 edges over    30 states;    20 small (38.46)
0x7ffff6f5b4b8 has    44 edges over    28 states;    18 small (40.91)
0x7ffff6f97168 has     5 edges over     3 states;     0 small (0.00)

timo

making patch
will create seq
seq has which Seq|66864704
0
first user:
0.5
1
1.5
second user:
iterator already taken for this seq:

timo

grep -B1 "^  Frame_" /fileserver-data/new_stuff/work/core.c.setting.dump.txt | grep -A1 -- '^--' | grep -v '^--' | cut -b 1-6 | sort | uniq -c | sort -rn
   1953 00007 
   1351 00004 
    389 00008 
    344 00034 
    330 00043 
    289 00017 
    262 00020 
    250 00040 
    235 00012 

timo

// ==UserScript==
// @name         Raku-Github-Linkify
// @namespace    https://github.com/timo/
// @version      0.3.5.3
// @description  Turns paths to raku source (core setting, grammar, actions) and moarvm code to clickable links on github issue/PR pages as well as gists
// @downloadUrl  https://gist.githubusercontent.com/timo/7cfe71a667bbdfa0dbd1510431da45a4/raw/Raku-Github-Linkify.js
// @homepageURL  https://gist.github.com/timo/7cfe71a667bbdfa0dbd1510431da45a4
// @author       Timo 'timotimo' Paulssen
// @match        https://github.com/*/*/issues/*
// @match        https://github.com/*/*/pull/*

timo

my $l = Lock.new; my %h; for 1..10 { my $p = start { sleep rand; $l.protect: { say "-- ", $p.WHERE; say %h.elems; %h{$p.WHERE}:delete; say %h.elems } }; say "+", $p.WHERE; %h{$p.WHERE} = $p; }; await %h.values; note %h.elems

timo

-- for testing, completely reset highest_child_id columns
update calls set highest_child_id = NULL;

-- find rows where no other row has the row's id as its parent_id
-- those shall seed the progress with initial "highest child id"
-- values
with no_children as (select id from
	(select c.id as id, count(children.id) as childcount
		from calls c
			left join calls children on c.id == children.parent_id

timo

unit class HashyObject;

has %!storage;

submethod TWEAK(*%data) { %!storage = %data }

submethod FALLBACK($name, |c) is rw { %!storage{$name} };

method perl {
    self.^name ~ ".new(" ~ %!storage.pairs.map(*.perl).join(", ") ~ ")"

timo

  BB 14 (0x7f030419cdf8):
    line: 283 (pc 188)
    Instructions:
      [Annotation: Logged (bytecode offset 188)]
      sp_getlex_o      r6(20), lex(idx=0,outers=0,newval)
      [Annotation: INS Deopt One (idx 8 -> pc 196; line 283)]
      [Annotation: INS Deopt One (idx 9 -> pc 202; line 283)]
      sp_guardconc      r6(4),  r6(20), sslot(1), litui32(196)
      [Annotation: Logged (bytecode offset 354)]
      sp_getlex_o      r9(25), lex(idx=0,outers=0,newval)