{"id":367,"date":"2012-02-27T12:29:56","date_gmt":"2012-02-27T11:29:56","guid":{"rendered":"http:\/\/blog.blockos.org\/?p=367"},"modified":"2015-11-30T10:50:52","modified_gmt":"2015-11-30T09:50:52","slug":"alices-bubbles-part-3","status":"publish","type":"post","link":"https:\/\/blog.blockos.org\/?p=367","title":{"rendered":"Alice\u2019s bubbles (Part 3)"},"content":{"rendered":"

Here comes level names. This was the most painful part of the whole translation hacking. First, they are made of sprites. Second, there is a limited number of them. And last but not least, the table that tells which sprites to use for a given level name can’t be expanded.
\nThe LSB of the pattern indices for the first level are stored at $5196 (file offset) and $519C for the second. After putting a read breakpoint on $*4f96 (remember that there’s usually a 512 bytes header to PC Engine roms) and $*4f9C, I ended up at $8E3F. A quick look at the code show me there is a address table stored at $8F81 where each address starts 1 byte before the LSB pattern index list. The first byte indicates the number of sprites. Here’s the complete index list:<\/p>\n

8F95 : 05 00 02 04 06 08
\n 8F9B : 05 0E 02 12 06 08
\n 8FA1 : 05 18 0A 16 06 08
\n 8FA7 : 05 14 00 16 06 08
\n 8FAD : 05 1C 02 20 06 08
\n 8FB3 : 04 18 1A 06 08
\n 8FB8 : 04 22 24 06 08
\n 8FBD : 05 02 10 18 06 08
\n 8FC3 : 04 26 28 06 08<\/code><\/p>\n

At this point, I managed to avoid using compressed data. Unfortunatelly the whole sprites were compressed. In fact, there are 4 “decompression” schemes. The first one is a direct transfer to VRAM using the tia<\/em> instruction. The second is using a simple RLE<\/a>. The third one and the most tricky mixes RLE<\/a> and some XOR. The last one simply copies the data to VRAM using a handcrafted loop.<\/p>\n

f796:   LDA <$7d\r\n        ORA <$7e\r\n        BEQ f7e1\r\n        JSR f888\r\n        CMP #$00\r\n        BNE f7ac\r\n                ; #1 Transfer to VRAM using tia\r\n                TIA $3700, $0002, $0020\r\n                BRA f7d2\r\nf7ac:   CMP #$02\r\n        BNE f7bc\r\n                ; #2 RLE\r\n                JSR f817\r\n                TIA $3700, $0002, $0020\r\n                BRA f7d2\r\nf7bc:   CMP #$03\r\n        BNE f7cf\r\n                ; #3 RLE + XOR\r\n                JSR f817\r\n                JSR f857\r\n                TIA $3700, $0002, $0020\r\n                BRA f7d2\r\nf7cf: ; #4 Loop transfer\r\n        JSR f803\r\n\r\nf7d2:   LDA <$7d\r\n        SBC #$01\r\n        STA <$7d\r\n        LDA <$7e\r\n        SBC #$00\r\n        STA <$7e\r\n        BRA f796\r\n\r\nf7e1:<\/pre>\n
The RLE<\/a> routine:<\/p>\n
f817:   LDA [$78]\r\n        STA <$83\r\n        LDY #$04\r\n        LDA #$09\r\n        STA <$81\r\n        STZ <$80\r\n        CLX\r\n        LDA #$20\r\n        STA <$82\r\n       \r\nf828:   DEC <$81\r\n        BNE f83a\r\n        PHY\r\n        LDA #$08\r\n        STA <$81\r\n        INC <$80\r\n        LDY <$80\r\n        LDA [$78], Y\r\n        STA <$83\r\n        PLY\r\n\r\nf83a:   ROR <$83 ; the rle counter\r\n        BCS f849\r\n        STZ $3740, X\r\n        INX\r\n        DEC <$82\r\n        BNE f828\r\n        JMP f7f6\r\n\r\nf849:   LDA [$78], Y\r\n        STA $3740, X\r\n        INY\r\n        INX\r\n        DEC <$82\r\n        BNE f828\r\n        JMP f7f6<\/pre>\n
The “XOR” routine:<\/p>\n
f857:   PHX\r\n        PHY\r\n        LDY #$07\r\n        CLX\r\n       \r\nf85c:   LDA $3740, X\r\n        EOR $3742, X\r\n        STA $3742, X\r\n        LDA $3741, X\r\n        EOR $3743, X\r\n        STA $3743, X\r\n        LDA $3750, X\r\n        EOR $3752, X\r\n        STA $3752, X\r\n        LDA $3751, X\r\n        EOR $3753, X\r\n        STA $3753, X\r\n       \r\n        INX\r\n        INX\r\n       \r\n        DEY\r\n        BNE f85c\r\n       \r\n        PLY\r\n        PLX\r\n        RTS<\/pre>\n
The XOR is used here to maximize the number of repetitions in order to make the RLE<\/a> more efficient. Well, that’s how I understood it. I didn’t make any benchmark to verify if it was really the case. The encoder was written pretty quickly. I reworked the font so that it more or less matches the ASCII set. This simplified the script encoder and the display routine.
\n<\/p>\n
Back to level names! I got help from a bunch of people to translate them. The last character of each level can be translated to country<\/em>. Unfortunately that was far too big so I switched to “land”<\/em>.<\/p>\n