This is interesting. I did know that there even is a native fbdev driver for the neomagic. However the driver says that it's based on the XFree86 code. So I don't expect it to do some magic we don't do.
The speedup may have another reason: the mtrr setup may not work correctly. With the fbdev driver in the kernel mtrr should be enabled from this driver.
I need to check this once I find the time. I discovered a problem with mtrr on an IBM Thinkpad using a Neomagic just the other day.
Egbert.
You may be right about mtrr. When xv is fast:
# cat mtrr reg00: base=0x00000000 ( 0MB), size= 128MB: write-back, count=1 reg01: base=0x08000000 ( 128MB), size= 64MB: write-back, count=1 reg02: base=0xfd000000 (4048MB), size= 16MB: write-combining, count=2
When it's slow: # cat /proc/mtrr reg00: base=0x00000000 ( 0MB), size= 128MB: write-back, count=1 reg01: base=0x08000000 ( 128MB), size= 64MB: write-back, count=1 reg02: base=0xfd000000 (4048MB), size= 4MB: write-combining, count=1
For more details about the slowness, see this thread:
http://www.mail-archive.com/[EMAIL PROTECTED]/msg00452.html http://www.mail-archive.com/[EMAIL PROTECTED]/msg00454.html http://www.mail-archive.com/[EMAIL PROTECTED]/msg00455.html
I can send dmesg and XFree86.0.log when xv is fast if you need. I narrowed it down a bit more: all I need to do is "modprobe neofb" before starting X and then xv is fast.
Matt
_______________________________________________ XFree86 mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xfree86
