www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.bugs - [Issue 5198] New: Appender much slower when appending ranges of elements than individual elements

reply d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=5198

           Summary: Appender much slower when appending ranges of elements
                    than individual elements
           Product: D
           Version: D2
          Platform: Other
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: Phobos
        AssignedTo: nobody puremagic.com
        ReportedBy: schveiguy yahoo.com



11:23:42 PST ---
Given an appender Appender!string a, the following code should be roughly
equivalent performance-wise:

a.put(' ');
a.put(" ");

But as shown by this code, the version that puts an individual character is an
order of magnitude faster.

import std.stdio;
import std.date;
import std.array;

void f0()
{
    Appender!string a;
    a.reserve(100_000_000);

    foreach(i; 0 .. 100_000_000)
    {
        a.put(' ');
    }
}

void f1()
{
    Appender!string a;
    a.reserve(100_000_000);

    foreach(i; 0 .. 100_000_000)
    {
        a.put(" ");
    }
}


void main()
{
    auto r = benchmark!(f0)(1);
    writeln(r, "ms");
    r = benchmark!(f1)(1);
    writeln(r, "ms");
}

Results:
[2433]ms
[13276]ms

Swapping the order of testing does not change the situation.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Nov 10 2010
next sibling parent d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=5198


Andrei Alexandrescu <andrei metalanguage.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
                 CC|                            |andrei metalanguage.com



11:55:35 PST ---
Great catch! I'm marking this as "assigned" as I see you've already assign it
to yourself.

Long term I wonder whether we should include performance tests as part of build
acceptance tests.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Nov 10 2010
prev sibling parent d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=5198


Steven Schveighoffer <schveiguy yahoo.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED



11:20:47 PST ---
turns out the appender was not reserving any extra space when appending an
array, it was simply reallocating -- just enough to hold the existing data +
the new data (in this example, 1 more element).  I fixed it so the same
algorithm used to expand the array for one element is now used to append an
array of elements.

changeset http://www.dsource.org/projects/phobos/changeset/2237

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Dec 27 2010