[ANN] V4RB mac 2.4 b18 uploaded
Russ Tyndall
fitzbew at nc.rr.com
Mon Jul 17 11:01:23 CDT 2006
On 7/17/06 10:28 AM, "Ruslan Zasukhin" <sunshine at public.kherson.ua> wrote:
> On 7/17/06 5:20 PM, "Russ Tyndall" <fitzbew at nc.rr.com> wrote:
>
>>> YES, windows works slower...reason as we see in File.Flush,
>>>
>>> This is very interesting and not clear...
>>> Computers use 100% the same HDD hardware and controllers.
>>>
>>> Why windows take so much longer time ?????
>>>
>>>
>>> Our own C++ test on mac take 3-4 minutes only,
>>> On Windows about 15-20 minutes.
>>
>> I'm afraid my testing shows the problem as much more severe.
>>
>> Using 23b10 on X.4.6, my db.open() call returns to me in .7 seconds the
>> *first* time I open a 240mb db after a restart. Subsequent db.open() calls
>> take only about .35 seconds. (DP 800mhz G4)
>>
>> Using 2.3 (release) on Win XP Home, my db.Open() call returns to me in 31.0
>> seconds (yes, 1/2 minute!) the first time I open the db after a restart.
>> Subsequent db.Open() calls take only .2 seconds! 155x faster! (2.1ghz
>> Celeron)
>>
>> The reason why the problem is so nefarious is that it only happens the
>> *first* time a db is opened in a session --- I never even noticed it during
>> development. Only when a customer complained. :-(
>>
>> Why does Windows take so much longer during the first opening? I'm afraid
>> greater minds than mine must answer that. My instinct is that one day
>> Ruslan is going to be looking at the Valentina code responsible for this and
>> suddenly see a bug or a glitch that is causing a massive slow-down.
>
> I already have look.
>
> I have told -- only way we see -- is do LAZY open of VarChar and index
> files. This require re-write logic of some methods, using new flags, check
> of these flags in some number of methods...must be accurate
>
Well, let me ask -- and if the answer is too complex or proprietary, just
say so: why does Win32 require special treatment? Is the "LAZY open" a
work-around for some unknown OS issue? Or even a processor-level issue?
Endian issue? I am merely curious. I sense (guess) that something about
this problem is still puzzling you. Of course, it's not important that I
understand this, I know that it will be fixed/improved at some point.
>> Interestingly, the problem does not seem to exist under Win98SE (at least
>> under VPC 6.x.). But I see it clearly on "real" Win 2000 machines and
>> emulated XP and Me machines.
>
> Yes, also interesting for me, write small stupid tests that do bench OS
> calls...
>
I know nothing about benching OS calls, but please keep me in mind for small
and stupid. :-)
--
Russ Tyndall
Wake Forest, NC
More information about the Valentina-beta
mailing list