Why the hoop jumping?

Why the hoop jumping?
by Oziphantom on 2018-04-14 (#216850)

So I'm digging around Ogre Battle and it has this function that it uses a lot at $001199, and it takes a Long stored after the jsr. It then calls said long. Here is a trace of its execution

Code:

$05/B19E 20 99 11    JSR $1199  [$05:1199]   A:FFD2 X:0650 Y:07C4 P:envMxdizc
$05/B1A1 66 8E 05
$05/119A 5C 8E 9F 00 JMP $009F8E[$00:9F8E]   A:03E2 X:0650 Y:01C4 P:envmxdizc
$00/9F8E 8B          PHB                     A:03E2 X:0650 Y:01C4 P:envmxdizc
$00/9F8F F4 9D 11    PEA $119D               A:03E2 X:0650 Y:01C4 P:envmxdizc
$00/9F92 0B          PHD                     A:03E2 X:0650 Y:01C4 P:envmxdizc
$00/9F93 4B          PHK                     A:03E2 X:0650 Y:01C4 P:envmxdizc
$00/9F94 08          PHP                     A:03E2 X:0650 Y:01C4 P:envmxdizc
$00/9F95 C2 20       REP #$20                A:03E2 X:0650 Y:01C4 P:envmxdizc
$00/9F97 0B          PHD                     A:03E2 X:0650 Y:01C4 P:envmxdizc
$00/9F98 48          PHA                     A:03E2 X:0650 Y:01C4 P:envmxdizc
$00/9F99 3B          TSC                     A:03E2 X:0650 Y:01C4 P:envmxdizc
$00/9F9A 5B          TCD                     A:1FDD X:0650 Y:01C4 P:envmxdizc
$00/9F9B D4 0D       PEI ($0D)  [$B1A0]      A:1FDD X:0650 Y:01C4 P:envmxdizc
$00/9F9D A5 0B       LDA $0B    [$00:1FE8]   A:1FDD X:0650 Y:01C4 P:envmxdizc
$00/9F9F 85 0D       STA $0D    [$00:1FEA]   A:0505 X:0650 Y:01C4 P:envmxdizc
$00/9FA1 68          PLA                     A:0505 X:0650 Y:01C4 P:envmxdizc
$00/9FA2 85 0C       STA $0C    [$00:1FE9]   A:B1A0 X:0650 Y:01C4 P:eNvmxdizc
$00/9FA4 E6 0C       INC $0C    [$00:1FE9]   A:B1A0 X:0650 Y:01C4 P:eNvmxdizc
$00/9FA6 A7 0C       LDA [$0C]  [$05:B1A1]   A:B1A0 X:0650 Y:01C4 P:eNvmxdizc
$00/9FA8 E6 0C       INC $0C    [$00:1FE9]   A:8E66 X:0650 Y:01C4 P:eNvmxdizc
$00/9FAA 3A          DEC A                   A:8E66 X:0650 Y:01C4 P:eNvmxdizc
$00/9FAB 85 06       STA $06    [$00:1FE3]   A:8E65 X:0650 Y:01C4 P:eNvmxdizc
$00/9FAD E6 0C       INC $0C    [$00:1FE9]   A:8E65 X:0650 Y:01C4 P:eNvmxdizc
$00/9FAF E2 20       SEP #$20                A:8E65 X:0650 Y:01C4 P:eNvmxdizc
$00/9FB1 A7 0C       LDA [$0C]  [$05:B1A3]   A:8E65 X:0650 Y:01C4 P:eNvMxdizc
$00/9FB3 85 08       STA $08    [$00:1FE5]   A:8E00 X:0650 Y:01C4 P:envMxdiZc
$00/9FB5 48          PHA                     A:8E00 X:0650 Y:01C4 P:envMxdiZc
$00/9FB6 AB          PLB                     A:8E00 X:0650 Y:01C4 P:envMxdiZc
$00/9FB7 C2 20       REP #$20                A:8E00 X:0650 Y:01C4 P:envMxdiZc
$00/9FB9 68          PLA                     A:8E00 X:0650 Y:01C4 P:envmxdiZc
$00/9FBA 2B          PLD                     A:03E2 X:0650 Y:01C4 P:envmxdizc
$00/9FBB 28          PLP                     A:03E2 X:0650 Y:01C4 P:envmxdiZc
$00/9FBC 6B          RTL                     A:03E2 X:0650 Y:01C4 P:envmxdizc
$00/8E66 29 FF 03    AND #$03FF              A:03E2 X:0650 Y:01C4 P:envmxdizc
$00/8E69 0A          ASL A                   A:03E2 X:0650 Y:01C4 P:envmxdizc
$00/8E6A A8          TAY                     A:07C4 X:0650 Y:01C4 P:envmxdizc
$00/8E6B B9 6F 8E    LDA $8E6F,y[$00:9633]   A:07C4 X:0650 Y:07C4 P:envmxdizc
$00/8E6E 60          RTS                     A:FFD2 X:0650 Y:07C4 P:eNvmxdizc
$00/119E AB          PLB                     A:FFD2 X:0650 Y:07C4 P:eNvmxdizc
$00/119F 6B          RTL                     A:FFD2 X:0650 Y:07C4 P:envmxdizc
$05/B1A4 85 8A       STA $8A    [$00:008A]   A:FFD2 X:0650 Y:07C4 P:envmxdizc
$05/B1A6 E2 20       SEP #$20                A:FFD2 X:0650 Y:07C4 P:envmxdizc
$05/B1A8 60          RTS                     A:FFD2 X:0650 Y:07C4 P:envMxdizc

This seems a lot of work and fancing around to do JSL $008e66, looks like it probably clears any DP movement/data page when doing the jump, so you call the function without changing the DP/Data bank. But also seems like overkill for such functionality. Am I missing something?

Re: Why the hoop jumping?
by psycopathicteen on 2018-04-14 (#216863)

I should post this code on wiki.superfamicom.org as an example of non-optimized code.

Re: Why the hoop jumping?
by Oziphantom on 2018-04-14 (#216864)

If you do be sure to call it not optimized code

or verbose code.

If you really want to see code that is badly optimized, have a look at tetris attack... oh boy, bad code for days.... even the tables are bad... hell they couldn't be bothered to make a random number generator so just threw half a bank at random number table

Re: Why the hoop jumping?
by calima on 2018-04-16 (#216935)

Maybe they needed lots of random numbers, fast

Re: Why the hoop jumping?
by 93143 on 2018-04-16 (#216936)

In a puzzle game?

I was going to use a random number table for my bullet hell port, since it's going to be squeezed for performance and I figured it'd be faster. But it turns out that accessing data outside the current bank is a pain in the ass on Super FX, and I managed to hunt up a reasonably high-quality 16-bit version of xorshift that's nearly as fast as the table lookup would have been, while having twice the period...

Re: Why the hoop jumping?
by tepples on 2018-04-16 (#216939)

Random numbers in Puzzle League also have to avoid self-clearing patterns with three aligned like panels. Do the random numbers in that game look pre-processed to meet this constraint?

Re: Why the hoop jumping?
by Sumez on 2018-04-16 (#216940)

93143 wrote:

I was going to use a random number table for my bullet hell port

What game are you porting?

Re: Why the hoop jumping?
by tepples on 2018-04-16 (#216941)

My guess is that naming the original game would violate some NDA with its publisher and cause 93143 to lose the license to port it. And even if the publisher has a policy to tolerate non-commercial fan works, like Team Shanghai Alice's Touhou Project games, announcing the title might put an undue burden on 93143 of fielding incessant "are we there yet" requests from its fan base.

Re: Why the hoop jumping?
by Sumez on 2018-04-16 (#216945)

Oh, I was just curious...

Re: Why the hoop jumping?
by Oziphantom on 2018-04-16 (#216947)

It is a puzzle game so avoiding a match 3 is probably idea, but from memory I think it has set starting patterns or at least a few possible starting patterns.

Code:

a048290                          .word $C52C,$2405,$F5B7,$AD2A,$224F,$70DE,$C297,$F421
                                 .word $45C2,$96B0,$E4B1,$FD3B,$AB54,$C1E5,$CA3F,$57FD
                                 .word $188B,$40FB,$E8F0,$72BE,$1120,$0605,$3866,$242B
                                 .word $617F,$2C27,$B6B9,$26C4,$964A,$6C1D,$E1F3,$965B
                                 .word $83D5,$A6B3,$3F93,$759A,$BDAA,$5E4B,$D70D,$287C
                                 .word $2004,$395C,$B647,$FAD0,$4A56,$85EF,$6D1B,$96BC
                                 .word $1DC3,$B023,$90D9,$8E36,$3FA5,$CEA8,$36C4,$D98A
                                 .word $9B09,$1645,$7E93,$4FD9,$E02F,$6355,$06F0,$2D96
                                 .word $FF0C,$B543,$73FA,$9609,$AFCA,$AF14,$EEC4,$0ECE
                                 .word $EA44,$19DC,$A3D5,$FE55,$6F8D,$5E45,$43A8,$3562
                                 .word $4068,$0D0D,$812C,$638C,$24CF,$5986,$9744,$9EC0
                                 .word $246F,$9B17,$C046,$E0BD,$1028,$CDB8,$BD7D,$8598
                                 .word $F990,$0F78,$52A9,$CF37,$B66E,$25F8,$C87C,$63D8
                                 .word $6241,$F5EF,$6B1D,$CB89,$DAB9,$0AA6,$0BA7,$F4AF
                                 .word $423B,$157B,$7DA8,$B082,$7F5F,$6960,$19A4,$348D
                                 .word $BB73,$7D5B,$3C92,$9931,$E6F7,$6C06,$C65C,$5C21
                                 .word $3221,$760F,$9B62,$E0E5,$9459,$7FB7,$23F5,$E859
                                 .word $48BB,$8C55,$CCDE,$202C,$4C9A,$4BD2,$85D6,$9464
                                 .word $E1FA,$8A2C,$430E,$35D7,$1014,$BDF5,$7EA6,$59B2
                                 .word $21D4,$7BD3,$B338,$3AF4,$235B,$FF00,$E14D,$74F1
                                 .word $697F,$AACA,$0EE4,$89D1,$0948,$7D12,$C1F0,$5E11
                                 .word $5D73,$A1CE,$88D8,$BEFF,$84F1,$E189,$72F7,$D440
                                 .word $DF67,$2CE0,$931C,$B34B,$97AE,$1605,$850A,$CFEE
                                 .word $1452,$553E,$E3F7,$83C6,$8614,$4765,$CF0E,$8CC9
                                 .word $5E6A,$6967,$6BEF,$8ABD,$D3FD,$DFC7,$612C,$84C0
                                 .word $5F28,$F11B,$5ECC,$63C1,$417D,$8A8C,$90C9,$7403
                                 .word $FC41,$B958,$2E94,$E89F,$D4ED,$3251,$EE8E,$5601
                                 .word $56AD,$CC5C,$8E8F,$3567,$CFE3,$02F5,$4F61,$6467
                                 .word $D1A3,$75A9,$7343,$A469,$B436,$6698,$C469,$1B27
                                 .word $109A,$3EFB,$0E78,$D132,$46FE,$089A,$A10D,$346E
                                 .word $F648,$F352,$D235,$9692,$8891,$D397,$7AF4,$ACAB
                                 .word $A6A6,$9FEE,$74C0,$0F98,$BE87,$F271,$2105,$BD8D
                                 .word $83E9,$8C4D,$E5A1,$9693,$6AB6,$D145,$A867,$E105
                                 .word $2F89,$472E,$589E,$C712,$5035,$1A73,$6581,$D53F
                                 .word $8F3D,$9990,$42BF,$7CE4,$725B,$B899,$E8FB,$93AC
                                 .word $C4FC,$8EB3,$534B,$D118,$14C0,$D697,$05BA,$55FB
                                 .word $33FC,$7115,$81C8,$21FD,$B83A,$E08C,$CFE7,$981A
                                 .word $7DB6,$CC74,$FDFE,$0522,$21E0,$80D7,$9AE7,$1638
                                 .word $86DF,$6BD2,$3BF3...

if that is massaged or not, you guess is a lot better than mine

Re: Why the hoop jumping?
by 93143 on 2018-04-16 (#216955)

tepples wrote:

Well, mainly it's closer to the latter. I've had bad experiences with overpromising and underdelivering in the past - not so much that people bug me, but that I disappoint them. I still plan to finish all my projects, but when I start something it's more likely to end up like Duke Nukem Forever than Haunted: Halloween '85.

I mentioned it on this forum a few years ago because I needed help getting started and figuring things out - it's my first SNES project. But maybe I should stop gratuitously referencing it when it doesn't add much to the conversation...

Sumez wrote:

Oh, I was just curious...

Sorry. This happens every time, and I never learn...

Oziphantom wrote:

massaged

That could be a good point.

On the other hand, it shouldn't be impossible to "massage" a real PRNG, say by rejecting results that don't meet a certain standard of playability, or by tying probability ranges to history when using the result - doesn't Tetris do this?

I suppose if you've got half a bank free anyway, a table might be the simplest approach...

Re: Why the hoop jumping?
by Sumez on 2018-04-16 (#216957)

Yes, NES Tetris does that. Strangely every Nintendo published Tetris game since that doesn't, and they really suffer from it.

Re: Why the hoop jumping?
by lidnariq on 2018-04-16 (#216958)

Oziphantom wrote:

if that is massaged

600 bytes isn't really enough to do any statistical analysis, but a histogram of those looks less flat than I'd expect.

Re: Why the hoop jumping?
by unregistered on 2018-04-16 (#216961)

Sumez wrote:

Yes, NES Tetris does that. Strangely every Nintendo published Tetris game since that doesn't, and they really suffer from it.

Does "every" include Famicom's "Tetris 2 (+BomBliss)"? I love that game!

Re: Why the hoop jumping?
by tepples on 2018-04-16 (#216962)

Tetris 2 + Bombliss is BPS published, though its Game Boy port Tetris Blast was published by Nintendo in North America.

The RNG in Tetris since 2001 is heavily massaged. The "bag" RNG in most games deals all of a set of 7 distinct Tetriminos without replacement, and only once the set is empty is it refilled. The "history" RNG in Arika's Tetris The Grand Master is heavily biased against the four most recently dealt Tetriminos. (For more information, see Hard Drop Wiki.) Both of these paradigms improve the player's quality of life somewhat, with fewer SZ gluts and fewer I droughts.

I've used something conceptually similar to history in Thwaite for target building choice, RHDE for piece choice, and my NES port of robotfindskitten for NKI choice. The RNG in these games considers only those choices, roughly one-third to two-thirds of all available choices, that were least recently dealt. This has the advantage of a more even distribution while also letting me use raw numbers from the RNG with a power of 2 range, eliminating the need to multiply.

Puzzle League, on the other hand, needs a bit more massaging than a simple bag or LRU scheme. Not only must pieces be non-matching horizontally, but they also have to be non-matching vertically against the previously generated row. It'd be tricky to make a constant-time algorithm for that.

Re: Why the hoop jumping?
by Oziphantom on 2018-04-16 (#216969)

Back to my original post for a second...
Having looked at a couple of more cases where they use the above "system".. the penny dropped.

The 65816 unlike every other CPU/System that lets me move the Z/B/DP and the SP is byte aligned. The others are Page aligned. Hence the TSC TCD part was confusing me, as in why would you set the DP to be the same page as the SP?? It doesn't, it sets it to be the current location of the bottom of the stack so LDA $04 reads the 4th thing on the stack... I mean I also assumed if one was to do such a thing on a 65816 you would use lda ($04,s ).. to have stack relative and to not use it is criminal... So now it is senseless but makes sense

Re: Why the hoop jumping?
by tepples on 2018-04-16 (#216971)

Use of D as a frame pointer would allow direct page addressing modes other than dd and (dd),Y. Consider it like BP on x86.

Re: Why the hoop jumping?
by Oziphantom on 2018-04-16 (#216972)

Ah there is no [d,s] only (d,s)...