Re: Fastcode MM B&V 0.48
- From: "Dennis" <marianndkc@xxxxxxxxxxxxxxx>
- Date: Thu, 28 Jul 2005 20:20:18 +0200
Hi All
I think that the "Block downsize test" benchmark does not work as intended.
It was meant to
Allocate a lot of strings of size = 20000.
Downsize them to Size=1
The benchmark did not run for long enough and somebody added a loop
for n := 1 to 50 do // loop added to have more than 1000 MTicks for this
benchmark
This changed the benchmark to do this?
Allocate a lot of strings of size = 20000.
Downsize them to Size=1
Upsize to Size = 20000
Downsize them to Size=1
etc.
What do you think?
procedure TDownsizeTest.RunBenchmark;
var
i, n, LOffset: integer;
begin
inherited;
for n := 1 to 50 do // loop added to have more than 1000 MTicks for
this benchmark
begin
{Allocate a lot of strings}
SetLength(FStrings, 50000);
for i := 0 to high(FStrings) do
begin
{Grab a 20K block}
SetLength(FStrings[i], 20000);
{Touch memory}
LOffset := 1;
while LOffset <= 20000 do
begin
FStrings[i][LOffset] := #1;
Inc(LOffset, 4096);
end;
{Reduce the size to 1 byte}
SetLength(FStrings[i], 1);
end;
{Update the peak address space usage}
UpdateUsageStatistics;
end;
end;
Regards
Dennis
.
- Follow-Ups:
- Re: Fastcode MM B&V 0.48
- From: Dennis
- Re: Fastcode MM B&V 0.48
- References:
- Fastcode MM B&V 0.48
- From: Dennis
- Fastcode MM B&V 0.48
- Prev by Date: Re: Fastcode MM B&V 0.48
- Next by Date: Re: Fastcode CPUID Unit
- Previous by thread: Re: Fastcode MM B&V 0.48
- Next by thread: Re: Fastcode MM B&V 0.48
- Index(es):
Relevant Pages
|