Back in 2001 or so, in the PS2 days, I used to have a sign pinned to my desk with the words "A GAME IS NOT A FILM" that I could point to. I see that little has changed.
Remote
Steven Tattersall
@tattlemuss@mastodon.gamedev.place
mastodon 4.7.3home: #atari 16-bit things (demoscene)
work: PlayStation Advanced Technology Group
0 Followers
0 Following
14 Posts
Joined November 04, 2022
Website:
Github:
Location:
Cambridge UK
Demozoo:
Seems a bit unfair. That's not what bobsleighs are designed for.
Scene guru @ggn@mastodon.gamedev.place has started a tutorial on cracking ST games: https://beyondbrown.mooo.com/post/cracking_getting_started/ which has prompted me to add a "stop on bootsector" option to the hrdb debugger.
Replying to
the purpose of this exercise is to compress my decode table -- I don't want a 64K array of decoder function indices. The previous image showed that there is 2D structure in the bit pattern -- every 512 values there is repetition.
In this image I swizzle the opcode values using bitshifts, and making each row 64 values wide. It turns out that 75% of the 1024 rows always have the same value in every column.
This should allow me to shrink the table to about 17.4K with a simple lookup.
Replying to
And there is an added bonus: after switching over to this new scheme, I noticed that a lot of the "unpackable" blocks of 64 entries were actually the same.
In fact, of the 244 original unpacked blocks there are only 72 really unique ones. By tweaking my table generator to remap these repetitions, my table data now occupies 7.6K rather than the original 64K. The runtime cost is some bit swizzles and an extra table lookup when looking in an "unpacked" block. Win! #atari
I thank John Mastodon again for his provision of the mute option.
Replying to
@evilpaul_atebit@icosahedron.website I'm hoping there is a "catalog" in there somewhere
Replying to
@goto80 I thought it was fantastic, thankyou for creating this
Replying to
@syllopsium @congusbongus yes, this was the situation in 2001. I once worked on a game where the player did no meaningful interaction for the first 10-15 minutes (due to unskippable cutscenes)
Replying to
@questor it was very satisfying :) From the Motorola PRM I knew there must be patterns in there, I just couldn't "see" them.
There is probably a way to encode this very efficiently as a tree of decisions, but I can't see that yet.
Replying to
Actual game post: here's one of the roles our group is hiring for, a DL engineer: https://careers.playstation.com/deep-learning-engineer-for-game-technology/job/5831515004 #jobs
If they just renamed it to the Straight of Hormuz they wouldn't have had so much difficulty.