Flash Flood

tails · Post by **tails** » Thu Dec 04, 2008 3:02 pm

tog wrote:How small has it really got to be?

Adum says his code runs about 10% faster on the Warmup. So, I think 17190 ( = 19100 * 0.9 ) can be a good aim.

tails · Post by **tails** » Thu Dec 04, 2008 3:48 pm

I found my code also runs 10% faster on the Warmup one.

I tried how many extra cycles I could add to my code, and that was about 1200. It means my original code runs in about 17900 ( = 19100 - 1200 ) cycles.

16004 / 17900 = 0.894...

gfoot · Post by **gfoot** » Thu Dec 04, 2008 7:02 pm

I was still failing Flash Flood with a warmup count of 17085. I made an optimisation, though, and got my warmup count down to 16666, and finally passed Flash Flood.

I was tempted to set up a search process to work out exactly what the board looks like, but that would be abuse.

You do still get a descriptive error message if you fail to flood correctly. Maybe adum changed it back?

I'd also like to suggest maybe a version with the same map and same 19100 cycle limit, but a maximum program size of 128 instructions...

the_impaler · Post by **the_impaler** » Thu Dec 04, 2008 7:11 pm

As I understand now if you run out of cycles then python HVM throws exception and it is treated as any other exceptions with message "failed to run correctly". To show a different message hackvm.py should have exception handler for max-cycles and that would be overkill . After all, "user friendly" is not in hacker dictionary, it would be hard to put it in design specs.
And I am not yet passed Warmup stage...

tog · Post by **tog** » Fri Dec 05, 2008 9:55 am

Gosh, I *finally* made it. My program was accepted at the step from 17095 to 17028. On the way there it got really ugly. It seems totally crazy to me to get it below 17000!

tails · Post by **tails** » Fri Dec 05, 2008 11:03 am

Congratulations.

Your 'Surely Smallest Mouse' was as crazy as this one to me

homeas · Post by **homeas** » Wed Dec 31, 2008 12:54 am

it wouldn't be bad to post at least the cycles for those programs, that manage this test map. My behemoth program with 480 ops needs 14855 cycles for the warm up map which was apparently good enought for this job -- but it needs over 22700 for a complete empty map (this could be further improved).

arthur · Post by **arthur** » Fri Sep 17, 2010 1:55 am

homeas wrote:it wouldn't be bad to post at least the cycles for those programs, that manage this test map. My behemoth program with 480 ops needs 14855 cycles for the warm up map which was apparently good enought for this job -- but it needs over 22700 for a complete empty map (this could be further improved).

hey, 14413 cycles for the warm up
but the Tsunami challenge is really crazy

how many cycles for warm up will be ok for Tsunami?

EDIT: totally no idea for Aazhi Peralai challenge

lvhao945 · Post by **lvhao945** » Sat Dec 24, 2011 4:14 pm

Challenge 'Aazhi Peralai' .........I still insist it's impossible.......= =!

Hippo · Post by **Hippo** » Wed May 21, 2014 9:06 pm

No idea for Aazhi as well

[edit]I got it

[/edit]

Carefull reading of challenge textations helps to pass.

Hippo · Post by **Hippo** » Thu May 22, 2014 2:55 pm

tog wrote:Gosh, I *finally* made it. My program was accepted at the step from 17095 to 17028. On the way there it got really ugly. It seems totally crazy to me to get it below 17000!

I were around 15300 on flood challenge with a very large code (loops are slow) (not all cycles were expanded

).