Re: [U-Boot] [PATCH,V2] JFFS2: accelerate scanning.
Hi Baidu, > Syncs up with jffs2 in the linux kernel: > 1/ Change DEFAULT_EMPTY_SCAN_SIZE from 4KB to 256 Bytes. > 2/ If the 1KB data is 0xFF after the cleanmarker, skip > and scan the next sector. > 3/ Change the buffer size from 4KB to 128KB which is the > common size of erase block. There is no "common size of erase block". Looking into the Linux code, it uses "max(erase block size, 128k)" for its buffer to speed up reading from NAND and the 128k seem to be a kmalloc limit. So maybe a "increase buffer size from 4KiB to 128KiB to reduce number of read operations" would be more fitting. By the way, does this change contribute to the performance increase at all, or is the increase simply due to DEFAULT_EMPTY_SCAN_SIZE? Also as for the other patch, can you split the commit into the individual changes corresponding to the list items? In this way, one could also easily measure which change really speeds up the operation... Thanks! Detlev -- It is practically impossible to teach good programming to students that have had a prior exposure to BASIC: as potential programmers they are mentally mutilated beyond hope of regeneration.-- Edsger Dijkstra -- DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-40 Fax: (+49)-8142-66989-80 Email: d...@denx.de ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Re: [U-Boot] [PATCH,V2] JFFS2: accelerate scanning.
Hi,Detlev : 2011/4/19 Detlev Zundel : > Hi Baidu, > >> Syncs up with jffs2 in the linux kernel: >> 1/ Change DEFAULT_EMPTY_SCAN_SIZE from 4KB to 256 Bytes. >> 2/ If the 1KB data is 0xFF after the cleanmarker, skip >> and scan the next sector. >> 3/ Change the buffer size from 4KB to 128KB which is the >> common size of erase block. > > There is no "common size of erase block". Looking into the Linux code, > it uses "max(erase block size, 128k)" for its buffer to speed up reading > from NAND and the 128k seem to be a kmalloc limit. > > So maybe a "increase buffer size from 4KiB to 128KiB to reduce number of > read operations" would be more fitting. By the way, does this change > contribute to the performance increase at all, or is the increase simply > due to DEFAULT_EMPTY_SCAN_SIZE? > Yes, I think it is useful to speed up the scanning. I have split these changes to individual patch. Thans. ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Re: [U-Boot] [PATCH,V2] JFFS2: accelerate scanning.
Hi Baidu, > Hi,Detlev : > > 2011/4/19 Detlev Zundel : >> Hi Baidu, >> >>> Syncs up with jffs2 in the linux kernel: >>> 1/ Change DEFAULT_EMPTY_SCAN_SIZE from 4KB to 256 Bytes. >>> 2/ If the 1KB data is 0xFF after the cleanmarker, skip >>> and scan the next sector. >>> 3/ Change the buffer size from 4KB to 128KB which is the >>> common size of erase block. >> >> There is no "common size of erase block". Looking into the Linux code, >> it uses "max(erase block size, 128k)" for its buffer to speed up reading >> from NAND and the 128k seem to be a kmalloc limit. >> >> So maybe a "increase buffer size from 4KiB to 128KiB to reduce number of >> read operations" would be more fitting. By the way, does this change >> contribute to the performance increase at all, or is the increase simply >> due to DEFAULT_EMPTY_SCAN_SIZE? >> > > Yes, I think it is useful to speed up the scanning. Don't get me wrong, but I was not asking whether you "think" it speeds up the scanning. When it comes to performance, I learnt to trust numbers olnly. This may in part be because I myself occassionally was completely wrong in predicting performance issues. So I am still eager to see actual numbers if this _really_ speeds up scanning. Cheers Detlev -- Die meisten schaetzen nicht, was sie verstehen; aber was sie nicht fassen koennen, verehren sie. Um geschaetzt zu werden, muessen die Sachen Muehe kosten; daher wird geruehmt, wer nicht verstanden wird. --- Baltasar Gracian -- DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-40 Fax: (+49)-8142-66989-80 Email: d...@denx.de ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot