On Thu, 2005-09-08 at 12:18 +0200, Nicolas Huillard wrote:
> Laurence Abbott a écrit :
> > It is still sitting at about 20-30% CPU usage so I'm not sure where the
> > problem lies. It looks like the framebuffer is only accepting frames
> > about half as often as it should, causing the buffer overload.
> >
> > Any clues for what I should test next?
>
> Did you check frame-buffer memory size (in the BIOS, kernel boot, etc.) ?
> This may be related to simple/souble/triple buffering : check once more
> fbset -s and /etc/fb.modes (the last you published was OK, but many days
> old).
Here's what I'm getting from dmesg when the viafb module is loaded:
viafb: VIA UNICHROME framebuffer 1.0 initializing
viafb: viafb : CC900000
viafb: framebuffer size = 64 Mb
viafb: Found Device Rev:0
mtrr: MTRR 4 not used
mtrr: 0xd8000000,0x4000000 overlaps existing 0xd8000000,0x800000
viafb: X:720 Y:576
viafb: detected VT1622A (7)
viafb: mode=720 bpp=32 refresh=50 TVon=1 TVtype=2
viafb: VQ start:3FC0000 end:3FFFFFF size:40000
viafb: Cursor start:3FBF000 end:3FBFFFF size:1000
viafb: detected VT1622A (7)
viafb: mode=720 bpp=32 refresh=50 TVon=1 TVtype=2
setmode x:720 y:576 bpp:32
setmode x:720 y:576 bpp:32
Console: switching to colour frame buffer device 90x36
viafb: irq handler installed, IRQ(0x200) = 80080c00
fb0: UNICHROME frame buffer device
(and later on when vdr starts:)
setmode x:720 y:576 bpp:32
fbset -s (and /etc/fb.modes) both give:
mode "720x576-50"
# D: 28.000 MHz, H: 31.250 kHz, V: 50.000 Hz
geometry 720 576 720 1152 32
timings 35714 32 8 46 0 136 3
rgba 8/16,8/8,8/0,8/24
endmode
vt1622 gives a dump of:
TVRegs
0x00: 04 00 00 04 00 00 20 4f 84 2a 5a 00 49 44 00 0f
0x10: 00 00 a7 1e 63 1c 24 8b 09 2a ee 10 02 80 00 99
0x20: 17 0c 6f 79 48 64 30 93 49 5f 15 a5 23 77 ff 00
0x30: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
0x40: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 42
0x50: 5f d0 23 70 02 5f d0 7e 23 94 d0 08 a8 16 00 00
0x60: 00 00 00 5f 03 00 00 00 00 00 00
CRTCRegs
0x00: 67 59 58 8b 5b 83 6f f0 00 60 1e 00 00 00 00 00
0x10: 5a 03 3f 68 00 40 70 e3 ff
CLK1: 26.999997
CLK2: 26.999997
CRTC_HTOTAL : 00000067 103
CRTC_HEND : 00000059 89
CRTC_HSTART_BLAN: 00000058 88
CRTC_HEND_BLANK : 0000006b 107
CRTC_HSTART_SYNC: 0000005b 91
CRTC_HEND_SYNC : 00000003 3
CRTC_VTOTAL : 0000026f 623
CRTC_VEND : 0000023f 575
CRTC_VSTART_BLAN: 00000040 64
CRTC_VEND_BLANK : 00000070 112
CRTC_VSTART_SYNC: 0000025a 602
CRTC_VEND_SYNC : 00000003 3
CRTC_VMAX_SCAN_L: 00000060 96
PLL_N : 00000063 99
PLL_D : 0000001e 30
PLL_P : 00000007 7
PLL_P2 : 00000007 7
PLL_R : 00000005 5
SUB_CARRIER : 2a098b24 705268516
SUB_CARRIER_PHAS: 0000026f 623
GH_TOTAL : 0000035f 863
GH_ACTIVE : 000002d0 720
GV_TOTAL : 00000270 624
TH_TOTAL : 0000035f 863
TH_ACTIVE : 000002d0 720
START_VIDEO : 0000004f 79
START_HORIZ : 00000084 132
START_VERT : 0000012a 298
TVIDEO_START : 00000108 264
TVIDEO_END : 000006a8 1704
THSYNC_WIDTH : 0000007e 126
TBURST_START : 00000394 916
TBURST_END : 000000d0 208
TV_BURST_STEP : 0000000c 12
VSCALE_FAC : 00000000 0
HSCALE_FAC : 00000000 0
SH_TOTAL : 0000035f 863
CR_AMP : 0000005a 90
BLACK_LEVEL : 00000000 0
Y_AMP : 00000049 73
CB_AMP : 00000044 68
BLANK_LEVEL : 00000079 121
HUE : 00000000 0
INPUT_THRESHOLD : 00000000 0
INPUT_DELTA : 00000000 0
Both viafb (patcher2k) and DirectFB have patches from Mark Adams
applied.
I'm using vdr-1.3.30 and softdevice-1.3.0. I've tried removing all other
plugins and get the same results. Same for both live TV and replay
modes, also with all pixel formats. Changing the buffer mode doesn't
change it, although if it is set to 'save', I get lots of '+' output,
presumably showing the output buffer filling up!
Cheers,
Laz
_______________________________________________
directfb-users mailing list
[email protected]
http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-users
