1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
|
[[!meta copyright="Copyright © 2013 Free Software Foundation, Inc."]]
[[!meta license="""[[!toggle id="license" text="GFDL 1.2+"]][[!toggleable
id="license" text="Permission is granted to copy, distribute and/or modify this
document under the terms of the GNU Free Documentation License, Version 1.2 or
any later version published by the Free Software Foundation; with no Invariant
Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license
is included in the section entitled [[GNU Free Documentation
License|/fdl]]."]]"""]]
[[!tag open_issue_gnumach]]
# IRC, freenode, #hurd, 2013-04-02
<psockali> hi again, maybe someone has some metrics
<psockali> is mprotect / munprotect faster in hurd then in linux ?
<pinotree> ?
<psockali> can i protected a memory page against write access in hurd
<psockali> and if so, is it a fast operation ?
<youpi> you can, I never measured, but it's probably the same cost as in
linux
<youpi> I don't see why it would be different, as it boils down to the same
x86 trick
<psockali> but i suppose it is part of the mach kernel doing the protection
and not part of the unix layer ?
<youpi> it is
<youpi> the unix layer doesn't have mm control
<youpi> it has to ask the kernel
<braunr> it's slower on mach, as it's less optimized because of historical
reasons
<braunr> but that's about it
<youpi> less optimized, how so?
<braunr> well, more entry fragmentation
<youpi> in the end you mark the page table and flush the tlb
<braunr> yes
<braunr> the high level virtual memory layer is a bit slower
<youpi> but fragmentation doesn't come into play it you just have one
memory object, does it?
<braunr> it does, as it's about memory areas, not objects
<braunr> the object is merely a backing store
<braunr> protection, inheritance, copy on write are all area (vm_map_entry)
attributes
<braunr> also, some operations affect all the address spaces where a
physical page is mapped
<braunr> although i think linux does the same thing as mach/bsd now
<youpi> but mprotect/munprotect doesn't, does it?
<braunr> no
<braunr> or perhaps by side effect, in some situations, i'm not sure
<braunr> i think it depends if the memory is shared between processes, but
i don't remember the details and can't think of a proper example right
now
<braunr> but anyway, "slower" here is negligible unless address spaces are
really huge and filled with lots of map entries
<braunr> psockali: why do you ask ?
<psockali> can i post a link here ?
<braunr> about what ?
<psockali> it's regarding azul / managed runtime initiative
<psockali> a GC for java
<braunr> why not
<braunr> although i don't see the point for now :)
<psockali> they have a custom MM management module for their GC as linux
kernel modul
<psockali> and i was wondering if mach would be any faster then linux in
that aspect
<psockali>
http://stackoverflow.com/questions/3358545/whats-actually-in-the-managed-runtime-initiatives-kernel-patches-and-jvm
<braunr> psockali: generally speaking, mach is slower than linux because of
its age and the fact it didn't receive as much attention and
microoptimization as linux did
<braunr> psockali: about this article, there is no reason mach would be
faster
|
