Session 1: hardware research, content measurement, codec decision, MAME harness
Verified GVRAM is one word-access per pixel in ALL color modes; chose 256-color 256x192 with movem.l bursts (page 1 sacrificed as double-buffer). Measured 8 scenes from the Blu-ray source: blit costs under 8% of the 12fps cycle budget, so I/O is the bottleneck, not CPU. Naive delta+RLE reaches only 3.2:1 (365 KB/s, 470MB) -> decision to use 4x4 vector quantization (~30 KB/s). "Shot on twos" assumption failed: the transfer has zero duplicate frames, so 12fps requires explicit decimation. Documents three false measurement results and their root causes (per-frame Floyd-Steinberg dithering, temporal denoise, exact-match dedupe on noisy source). MAME Lua injection harness works and is reusable for cycle-cost measurement; the IOCS _B_READ disk benchmark is blocked returning -1. Claude-Session: https://claude.ai/code/session_01194oWYW8DQXK1SZ2DnChW6
This commit is contained in:
@@ -0,0 +1,36 @@
|
||||
FLAG = $18000
|
||||
STAT = $18004
|
||||
NREAD = $18008
|
||||
BUF = $20000
|
||||
CHUNK = 65536 ; bytes per read
|
||||
RECS = CHUNK/256 ; 256-byte records per read
|
||||
NCHUNK = 16 ; 16 x 64KB = 1MB total
|
||||
|
||||
org $10000
|
||||
start:
|
||||
move.l #1,FLAG.l ; signal: started
|
||||
move.l #NCHUNK,d6
|
||||
moveq #0,d5 ; record position
|
||||
moveq #0,d4 ; completed count
|
||||
loop:
|
||||
movem.l d4-d6,-(sp)
|
||||
moveq #$46,d0 ; _B_READ
|
||||
move.l #$80000000,d1 ; PDA=$80 (SASI unit0), mode=0
|
||||
move.l d5,d2 ; position (records)
|
||||
move.l #CHUNK,d3 ; byte count
|
||||
lea BUF,a1
|
||||
trap #15
|
||||
movem.l (sp)+,d4-d6
|
||||
move.l d0,STAT.l ; last status
|
||||
tst.l d0
|
||||
bmi.s failed
|
||||
addq.l #1,d4
|
||||
move.l d4,NREAD.l
|
||||
add.l #RECS,d5
|
||||
subq.l #1,d6
|
||||
bne.s loop
|
||||
move.l #$FF,FLAG.l ; signal: done OK
|
||||
stop: bra.s stop
|
||||
failed:
|
||||
move.l #$EE,FLAG.l ; signal: error
|
||||
bra.s stop
|
||||
Reference in New Issue
Block a user