Angband-4.0.1

[alkis]

Source:https://drive.google.com/file/d/0B5H...ew?usp=sharing
Executable+datafiles: https://drive.google.com/file/d/0B5H...ew?usp=sharing

Well, I had a look to angband-4.0.1 and I was curious how it would run on amiga. Meh, A1200 (emulated) is painfully slow, unplayable. A4000/040 (emulated) makes it barely playable.

The resources needed are way up for a game like this. Compiled with m68k-amigaos-gcc (GCC) 3.4.0
Needs:
- >=68020
- 8Mb fast
- stack 30000
- does not need ixemul.library

I took graphics and sound out. I upped the screen to 5bits depth so all the 28 colors of Angband can be seen

Cheers.

[Locutus]

I love roguelikes, i'll try it on a real 060 and lets see how it does on real hardware

[alkis]

Originally Posted by Locutus

I love roguelikes, i'll try it on a real 060 and lets see how it does on real hardware

Let me know how it runs there

[Thorham]

Cool

Seems to run properly on a 50 mhz 68030. Seems playable, too, but initialization takes a while.

What are the slow parts? Worth optimizing in assembly language?

[Havie]

Had no idea what this was and had a look on YouTube. Does this really need high specs - there must be a lot going under the hood...

[ajk]

All of the more advanced rogue-likes are highly complex, apart from the graphics

[modrobert]

Originally Posted by Thorham

Cool

Seems to run properly on a 50 mhz 68030. Seems playable, too, but initialization takes a while.

What are the slow parts? Worth optimizing in assembly language?

Works fine with my A1200 (see specs in sig) as well. Took roughly two minutes to "init" the game, and like half a minute after character generation is finished to load the map and start the game.

After this there is no lag, the game runs fine.

BTW: I copied the contents of the zip to RAM:, and played it from there, so that probably improves performance (ACA cards have fast memory access).

[Thorham]

Originally Posted by modrobert

After this there is no lag, the game runs fine.

Just try going down some stairs. Takes ages. What's worse is that levels aren't persistent. Each time you visit a level it's regenerated. This game seriously needs a 68060

[modrobert]

Originally Posted by Thorham

Just try going down some stairs. Takes ages. What's worse is that levels aren't persistent. Each time you visit a level it's regenerated. This game seriously needs a 68060

Yes, it took 35 seconds when you are in the town level, to go down the staircase. At this point the whole dungeon is randomly generated, this takes time, but changing between levels within the dungeon is not randomized (AFAIK), and is quicker (5-10 seconds). I think this is OK, it's a bit like loading a level in any game, but would be nice with a "please wait..." text or something, instead of the confusing "--more--" sign.

BTW: Just got killed by a yellow centipede.

[alexsperi]

Cool! Good idea!

I tried it in FS-Uae but during init screen it throws a guru. I think it's related to the facts that it does not find some kind of preference file. Some hints? I don't want to wait until the weekend when I can return home to play it on real hardware. :-)

[alkis]

Originally Posted by alexsperi

Cool! Good idea!

I tried it in FS-Uae but during init screen it throws a guru. I think it's related to the facts that it does not find some kind of preference file. Some hints? I don't want to wait until the weekend when I can return home to play it on real hardware. :-)

You forgot to type "stack 30000" before typing "angband", didn't you?

(That, or you run it on a 68000 and not a 68020)

[alkis]

Originally Posted by Thorham

Cool

Seems to run properly on a 50 mhz 68030. Seems playable, too, but initialization takes a while.

What are the slow parts? Worth optimizing in assembly language?

Nice to know it runs on real hardware. Thanks.

Now about the slow parts, I can assume the following:

• The thing runs internally with wide characters (w_char) for every string. Btw, this bugs the 'recall' information. (if you look 'l' on a monster and hit 'r' you should see the 'recall' information, which you don't now)
• Has a space optimization on the terrain that has multiple bytes (char) hold the features of each map cell. So, the code has to always compute the byte offset say flag with the number 18 (18/8) should be and also the binary mask (3). I am pretty sure that hurts speed (I did a couple of optimizations there on the C level)

The above observations were made on the linux side (on linux executable) with valgrind and kcachegrind. I assume these observations are trasnferable to the amiga.

Now, how COOL it would be to be able to remote profile the exe?

[alexsperi]

You are right,
the first time I tried it without entering the stack command from the cli, but reading better your first post I added it, and the game started the initialization procedure sadly failing after the "reading prefs" message.

For sure I'm doing something wrong so this is a screenshot of the software failure



Hope it helps

[Locutus]

It runs on my 4k, opens up a AGA screen (would have been cool if screenmode was selectable. but no worries really!).

The only issue i have is that on program exit it crashes with a #80000008 failure.

Edit: turned out to be stack issue ofcourse.

[alkis]

Originally Posted by alexsperi

You are right,
the first time I tried it without entering the stack command from the cli, but reading better your first post I added it, and the game started the initialization procedure sadly failing after the "reading prefs" message.

The error message is "normal". Don't worry about it.
Hmmm, how much fast memory does that setup have?

[alexsperi]

I usually setup fs-uae with 2Megs of chip ram, 8Megs of Fast ram, and 128Megs of Zorro III ram (this is a setup that is quite similar to my real machine).

[alkis]

Originally Posted by alexsperi

I usually setup fs-uae with 2Megs of chip ram, 8Megs of Fast ram, and 128Megs of Zorro III ram (this is a setup that is quite similar to my real machine).

Weird, that's the setup I am using for testing! (FS-UAE VERSION 2.4.3 on linux 64bit)

Also, in my log file:
...
Known ROM 'KS ROM v3.0 (A4000)' loaded
...
=KS ROM v3.0 (A4000) rev 39.106 (512k)

[alkis]

Bug fix: Recall,Examine information now correctly displayed.
(had to hook a mbstowcs() implementation, I guess gcc's one was buggy)

Just the executable: https://drive.google.com/file/d/0B5H...ew?usp=sharing

[alexsperi]

Originally Posted by alkis

Weird, that's the setup I am using for testing! (FS-UAE VERSION 2.4.3 on linux 64bit)

Also, in my log file:
...
Known ROM 'KS ROM v3.0 (A4000)' loaded
...
=KS ROM v3.0 (A4000) rev 39.106 (512k)

I'm using v2.4.1 on a macbook pro, but I don't think it's a problem. Maybe it's a kickstart-related issue? Mine is 3.1 (with wb 3.1). After dinner I will try with e fresh 3.0 install.

[modrobert]

Originally Posted by alexsperi

I'm using v2.4.1 on a macbook pro, but I don't think it's a problem. Maybe it's a kickstart-related issue? Mine is 3.1 (with wb 3.1). After dinner I will try with e fresh 3.0 install.

I use kickstart 3.1 on my A1200 and 'angband' loads fine after 'stack 300000' (yes, realize now I added another zero by mistake, but perhaps that was a good thing).

BTW: There is a Mac build of Angband.

http://rephial.org/