O HAI THIS BLOG PURPZIEZ 2 B UZED AZ MAH PLESIOUS MEM. :)

2007/05/11

oldlibc.float.karma--; netdev.watchdog.karma--;

TSSIのRSSIへの回り込みはAGCのステートマシンをムニョって誤摩化した.
が,新たに二つ問題が発生.一つは解決済み,一つは原因を調査中.
[T3h TSSI/RSSI mixing issue, I worked around it by modifed FSM of AGC.
Meh, 2 new PITAs commin'. One of these were fixed, another one's progress
is still incestigating.]

前者は古いlibcをSH4なんつーマイナーなアーキテクチャで使用している為,
floatな演算が全部死んでいると言う悪夢.struct timevalの精度がusなので,
適当に10^6倍して整数で計算すると言うチキンな逃げもあったが,SoftFloat
をstdint.hでportしたブツでモニョった.Softfloatのライセンスに限ったモノ
でもないけれど,ライセンス云々はWTFPL以外ヨクワカンランし,詳細を
理解する気もあんまり無いので,後で叩かれない様にportしたブツはこの辺
に晒しておく.
[Former is due to old libc on SH4, a minor arch. It results nightmate of totally
invalid floating point arithmetic operation. As you know, resolution of struct
timeval is us, so in my case, to x10^6 w/ integer operation is enough to work
around this crap, how chicken though. To be no chicken is better I thought,
then I've poked it w/ SoftFloat portion w/ stdint.h. Any license is not easy to
understand its detail, except WTFPL IMHO. To prevant license violation, t3h
portion of SoftFloat is stuffed here.]

後者はSoftMACがnetdev watchdogなtransmission timeoutになる事.
多分,PCI DMA関連がハードもソフトも腐っているからだろうけど,
取り敢えず,要調査.
[Latter is a SoftMAC issue about netdev watchdog transmission faults
due to its timeout. Guess b/c PCI DMA's software/hardware suck0rz.
Anyway I'll fix ASAP.]

0 件のコメント: