reala-misaki
65112b9305
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
54 lines
2.0 KiB
Lua
54 lines
2.0 KiB
Lua
M = manager.machine
|
|
CPU = M.devices[":maincpu"]
|
|
SP = CPU.spaces["program"]
|
|
FLAG,STAT,NREAD,BUF = 0x18000,0x18004,0x18008,0x20000
|
|
TOTAL = 16*65536
|
|
local f=io.open("bench.bin","rb"); local d=f:read("a"); f:close()
|
|
CODE={} ; for i=1,#d do CODE[i]=string.byte(d,i) end
|
|
print("[BENCH] loaded bench.bin bytes="..#d)
|
|
STATE,T0,NFIRE = "wait",nil,0
|
|
local function T() local t=M.time return t.seconds + t.attoseconds/1e18 end
|
|
local function P(s) print("[BENCH] "..s) end
|
|
|
|
local function tick()
|
|
NFIRE = NFIRE + 1
|
|
if NFIRE==1 then P("notifier firing OK") end
|
|
local t=T()
|
|
if STATE=="wait" then
|
|
if t < 5.0 then return end
|
|
for i=1,#CODE do SP:write_u8(0x10000+i-1, CODE[i]) end
|
|
SP:write_u32(FLAG,0); SP:write_u32(STAT,0); SP:write_u32(NREAD,0); SP:write_u32(BUF,0)
|
|
CPU.state["SR"].value=0x2000
|
|
CPU.state["SP"].value=0x8000
|
|
CPU.state["PC"].value=0x10000
|
|
STATE="run"
|
|
P(string.format("injected t=%.3f frames=%d",t,NFIRE))
|
|
return
|
|
end
|
|
if STATE=="run" then
|
|
local fl=SP:read_u32(FLAG)
|
|
if fl==1 and not T0 then T0=t; P(string.format("started t=%.4f",t)) end
|
|
if fl==0xFF then
|
|
STATE="done"
|
|
local dt=t-(T0 or t)
|
|
P(string.format("DONE reads=%d status=%08X",SP:read_u32(NREAD),SP:read_u32(STAT)))
|
|
P(string.format("elapsed=%.4f s bytes=%d",dt,TOTAL))
|
|
if dt>0 then P(string.format("THROUGHPUT = %.1f KB/s",TOTAL/1024/dt)) end
|
|
P(string.format("buf=%08X %08X",SP:read_u32(BUF),SP:read_u32(BUF+4)))
|
|
M:exit()
|
|
elseif fl==0xEE then
|
|
STATE="done"; P(string.format("READ FAILED status=%08X reads=%d",SP:read_u32(STAT),SP:read_u32(NREAD))); M:exit()
|
|
elseif t>25 then
|
|
STATE="done"
|
|
P(string.format("TIMEOUT flag=%08X status=%08X reads=%d PC=%08X SR=%04X",
|
|
fl,SP:read_u32(STAT),SP:read_u32(NREAD),CPU.state["PC"].value,CPU.state["SR"].value))
|
|
M:exit()
|
|
end
|
|
end
|
|
end
|
|
SUB = emu.add_machine_frame_notifier(function()
|
|
local ok,err = pcall(tick)
|
|
if not ok then print("[BENCH] LUA ERROR: "..tostring(err)); M:exit() end
|
|
end)
|
|
print("[BENCH] subscription="..tostring(SUB))
|