I need a confirmation in speed

[ErosOlmi]

Hi all.

I need a confirmation about a change I made in thinCore.dll
I changed the way local variables are deallocated when exiting from a function calling.
I need a confirmation if you can see any improve in script execution. I got a 15%/20% faster execution when using scripts with many function calling.

Attached a script example and new thinCore.dll
Please execute the script before replacing thinCore.dll. Take note of timing, replace thinCore.dll, take note again about timing.

Thanks a lot
Eros

UPDATED thinCore.dll

[Michael Hartlef]

And here are the results from the german jury... ITALY 10 points ;D

[ErosOlmi]

Not only professional graphics, professional sounds, professional programming. Also professional tester.
I cannot ask more :D Thanks

So it seems we will get some more speed in next release in case of functions having parameters and/or local variables executed many times.
I will test some more but seems quite safe right now. I suppose I will be able to get some more speed in this area ;)

Ciao
Eros

[kryton9]

Here are my results. I see significant improvement in the first 2 tests and a loss in performance in the third test:

Before
Test 1 Test 2 Test 3
Run 1 2.078 2.141 1.812
Run 2 1.625 1.609 1.610
Run 3 1.000 1.453 0.875
Average 1.568 1.734 1.432

After
Test 1 Test 2 Test 3
Run 1 1.531 1.766 1.512
Run 2 1.609 1.531 1.531
Run 3 1.344 1.172 1.344
Average 1.495 1.490 1.462

Improvement 0.046566022 0.141072458 -0.020944845

So test:
1) 4.6% improvement
2) 14.1% improvement
3) 2.0% loss in speed

Nice test. I will use this to test my computer, thanks Eros. Great job in squeezing even more speed out of already a very fast language!!

[ErosOlmi]

Wow, another professional tester. Thanks a lot Ken.
I didn't expect to lose performace in 3rd test.
In any case the change I've done will have more positive results the more parameters and local variables will be present.

I will check again but pretty sure I have some more work to do in this area. Code in memory handling is more than 1 year old.

Ciao
Eros

[kryton9]

Mike, got a good improvement on that test. So who knows what effects it. There is enough over all speed improvment that is for sure, incredible how you got so much improvement out of it!!

[ErosOlmi]

Update thinCore.dll in first post of this thread.
A little bit more speed.

[Petr Schreiber]

Grr,

such a nice reports,
I had to do it differentely ;D

The speed gain is great Eros,
hope you didn't hurt any animal during your optimizing black magic :)


Thanks a lot,
Petr

[Michael Hartlef]

Petr, you're a poser ;D

[kryton9]

Petr, you win with nice chart!! Great touch!!!

I am happy, the new core shaved time on all three and amazingly improved.
This is compared to the last update, not before the update. So WOW!!!

Second Update
After
Test 1 Test 2 Test 3
Run 1 1.390 1.266 0.875
Run 2 1.328 1.219 0.828
Run 3 1.469 1.328 0.641
Average 1.396 1.271 0.781

Improvement 0.066235504 0.146788991 0.465694096

Test 1 : 6.6% Improvement from last update!!
Test 2 : 14.6% Improvement from last update!!!
Test 3 : 46.5% Improvement from last update!!!!!!!!!!!!!!!!!

[ErosOlmi]

OK guy. Great job.

Thanks a lot to our thinBasic tester team for the time dedicated to test new Core.
Not easy to have such a great response. I'm very lucky!

Eros

[Michael Hartlef]

Just thought, I test your newest update regarding the for/next error with the updated thincore from the 4th of march. Test 1 and 2 faster, test 3 slower. Good job Eros!

[ErosOlmi]

Hey, I was not optimizing but re-coding :D
I'm really lucky to have such friends. Thanks Mike.

Anyhow, differences are so minimal in this new version that I cannot talk about it as an improve.
But I have a lot of new ideas for next preview version I will work on after releasing current preview as stable.

[kryton9]

Even more speed, wow. Nice chart Mike, great presentation.