Yuku free message boards

What would be the point over the currently-used pucrunch/ttpack compression?

Quote:

---

Now that I think about it, I've used XPak, which claims to be better that ttpack.

---

"claims" != "is". It's a stripped-down version of ttpack (less memory use) to support on-calc compression, so it doesn't compress as well.

Quote:

---

After a quick search, pucrunch seems to be limited to Z80 (correct me if that's not true).

---

That's complete nonsense, you must have been on the page for a Z80 port of pucrunch. The pucrunch algorithm is portable C, and it's what ttpack is derived from. www.cs.tut.fi/~albert/Dev/pucrunch/. We also have optimized 68k assembly implementations of the decompressor now, thanks to Samuel Stearly.

Quote:

---

UPX has the advantage that the executable is self-extracting, that is, it does not need anything else to decompress it. This is why UPX is great for other platforms as well.

---

And which is why it can never work that way on the calculator. There's a reason ttpack uses a separate launcher. By the way, XPak tries to do that too. It doesn't work well:
* Archived compressed programs are copied to RAM, then decompressed to RAM again, using more RAM than if a launcher decompresses them directly from archive. XPak uses a really ugly hack to get rid of the compressed copy in RAM, which we did not consider acceptable for TIGCC. The hack also needs extra memory (you need to explicitly free the copy).
* You can't work around the 8 or 24 KB limit that way, which is one of the main reasons to use TIGCC's compression feature in the first place! Well, it does help because you're now limited to 8 or 24 KB of compressed size, but that's still not anywhere near as useful as the separate-launcher method which is only limited by 64 KB uncompressed size (a limit which XPak also has, of course, you can't create a handle for the decompressed program larger than that). That can be twice as much as 24 KB compressed, and it's a lot more than 8 KB compressed (which matters if you care about AMS 2.00-2.03 compatibility).

Quote:

---

XPak is basically a archiving program for the calculator (which, in my opinion, is pointless).

---

No, there's also the XPak compression for executables I mentioned in the previous paragraph. But that one is built into Paul Froissart's proprietary compiler, there's no version which works with TIGCC programs that I know of. Not that you'd want to use it anyway; as I said, ttpack is better.

Quote:

---

ttpack compresses uses another machine to compress, but the resulting files need to have the ttunpack routines on the calculator.

---

That's why we automatically generate a 1 KB launcher which contains the decompression routine, and also works around the 8 or 24 KB limits. Using that launcher, running a compressed program is as easy as running an uncompressed one, the only difference is that there's that big PPG "data" file which is actually the real executable in compressed form.

Quote:

---

However, I do not know good the ttpack algorithm is at compression, and there is no way to know how well UPX will compress these executables yet.

---

If you want an algorithm with better compression, look at LZMA (which is already supported by the launcher used in TIGCC, you just need to toggle a compile-time option), not UPX. UPX is optimized for fast decompression, not high compression rate. But LZMA is about 20 times slower to decompress than ttunpack is, which is why most people don't like it. (They don't want to wait up to 20 seconds for their game to start up. I don't really see the problem as this doesn't affect gameplay at all, and I'm willing to accept that wait if I can put more stuff into the calculators' small archive memory, but tastes are different.)

Quote:

---

I couldn't find any links to download ttpack. Do you have a link? I'd like to see how well it can compress.

---

It's part of TIGCC, just check the compression option in the IDE!

So, to sum up, there's no advantage at all for using UPX with TIGCC, so there's no point in porting it (well, if the UPX authors want to waste their time, they're free, but I won't help you or them in any way on that silly quest), the only reason you even suggested it is that you missed the obvious, i.e. that TIGCC already supports program compression and that most TIGCC programs are already using it!