PB3.50 Win32 - Better optimized (-6k of Ed) code!?

BackupUser · Post by **BackupUser** » Mon Jan 06, 2003 4:46 am

Restored from previous forum. Originally posted by MrVainSCL.

Hi @ all,
the docs of PB3.50 Win32 are saying for example following:

"Changed: The assembler (FASM), resulting of better optimized code (PB editor is 6 kb smaller for example)"

Mhhhh... i have tested and compiled some own sources and there is not only one executeable created with v3.50 that is small as compiled with the old version

So, how could it be possible that the editor is now smaller Fred? Maybe you have splitted a lib used in the source or replaced any stuff via API or just removed any old not needed stuff

Anyway it would be nice to have better optimized and smaller executeables... (esp. faster innerloops)

PIII450, 256MB Ram, 80GB HD + 6,4 GB, RivaTNT, DirectX9.0, SB AWE64, Win2000 + all Updates...

greetz
MrVainSCL! aka Thorsten

BackupUser · Post by **BackupUser** » Mon Jan 06, 2003 8:44 am

Restored from previous forum. Originally posted by fred.

First, you exec must be big. If you try to get some kb on a 16 kb exec, it's no way. FASM optimize innerloop by using short jump automatically which makes the code smaller. PureBasic 3.50 introduces new optimisation about constant Chr() and changed some op to smaller/faster so it will always result to smaller exec. Keep in mind than Win32 exec are 512 bytes boundary, so because the size changes, it could have to save 512 bytes.

Fred - AlphaSND

BackupUser · Post by **BackupUser** » Mon Jan 06, 2003 10:23 am

Restored from previous forum. Originally posted by PB.

> Mhhhh... i have tested and compiled some own sources and there is
> not only one executeable created with v3.50 that is small as
> compiled with the old version

Sadly, I have to agree.

Here's some examples of my own exes recompiled with v3.50 and how
much they differ from being compiled under v3.40 (all sources <10k):

Code: Select all

    v3.40           v3.50
    -----           -----
 5,664 bytes     6,176 bytes
18,976 bytes    19,488 bytes
30,752 bytes    31,264 bytes

Don't know what's going on, but there's definitely a new-found bloat
happening somewhere. I use UPX on my exes afterwards anyway, so
it doesn't really bother me too much.

BackupUser · Post by **BackupUser** » Mon Jan 06, 2003 10:48 am

Restored from previous forum. Originally posted by fred.

May be the editor is may be a special case. I've added several command in some libs, so may be you don't benefit of the optims. Anyway, they exists for sure

.

Fred - AlphaSND

BackupUser · Post by **BackupUser** » Mon Jan 06, 2003 10:55 am

Restored from previous forum. Originally posted by PB.

> I've added several command in some libs, so may be you don't
> benefit of the optims. Anyway, they exists for sure

.

I believe you. Maybe some specific commands, common to all my sources,
are causing the extra code... but it's hard to tell at this stage.

BackupUser · Post by **BackupUser** » Mon Jan 06, 2003 7:27 pm

Restored from previous forum. Originally posted by Hi-Toro.

Aw, Jeez, get over it -- it's 1K!

--
See ya,
James L Boyd.
http://www.hi-toro.com/
--

BackupUser · Post by **BackupUser** » Mon Jan 06, 2003 9:59 pm

Restored from previous forum. Originally posted by PB.

> Aw, Jeez, get over it -- it's 1K!

Yeah, yeah... but that's just 1k for those particular small apps.
Naturally we are/were worried about the effect on larger apps,
that's all.

BackupUser · Post by **BackupUser** » Mon Jan 06, 2003 10:45 pm

Restored from previous forum. Originally posted by tinman.

Originally posted by PB

> Aw, Jeez, get over it -- it's 1K!

Yeah, yeah... but that's just 1k for those particular small apps.
Naturally we are/were worried about the effect on larger apps,
that's all.

OK, how about:

3.40: 153,120 bytes
3.50: 144,928 bytes

It's the biggest source I think I have. Complaining about 512 bytes on each of your example executables was rather pointless. With executables that size, the libraries would have made up almost as much or more than your own code, so any increase in library size would have offset any gains by compiler optimisations.

Fred - good job!

--
It's not minimalist - I'm increasing efficiency by reducing input effort.
(Win98first ed. + SP1, PB3.40)

BackupUser · Post by **BackupUser** » Mon Jan 06, 2003 11:32 pm

Restored from previous forum. Originally posted by Justin.

I tested it with my PBWindowEditor app (that i hope to release soon), and droped from

92192 -> 90656

looks very good.

BackupUser · Post by **BackupUser** » Mon Jan 06, 2003 11:38 pm

Restored from previous forum. Originally posted by fred.

At least it proves I'm not a liar

. Thanks guys !

Fred - AlphaSND

BackupUser · Post by **BackupUser** » Tue Jan 07, 2003 12:07 am

Restored from previous forum. Originally posted by PB.

Tinman wrote: "Complaining about 512 bytes on each of your example
executables was rather pointless."

It was just a knee-jerk reaction of a first impression. I have no
doubt that some executables will be optimised, but I got a bit scared
because my samples got larger. Don't worry, I'm okay with it.

Fred wrote: "it proves I'm not a liar"

I know you're not a liar. As Tinman said, Good Job!