Philips Semiconductors
Image Co-Processor
File: icp.fm5, modified 7/26/99
PRELIMINARY INFORMATION
13-21
13.6.5
Priority Delay and ICP Minimum Bus
Bandwidth
The Priority Delay field in the Status register sets the
time the ICP will wait for SDRAM service before chang-
ing from a low priority bus request to a high priority re-
quest. The ICP normally requests SDRAM bus service at
the lowest priority level, since it is a background process-
ing device. In some cases, service to the ICP could be
continuously delayed by other background devices, such
as the VLD processor or by high priority requests from
the DSPCPU.
The PD field sets a timer on the currently active bus re-
quest. The timer is loaded with the PD value and started
each time a bus request is submitted. The timer is incre-
mented once each block time, the time required to load
one block of 64 byte. If the timer reaches 16 before the
request is serviced, the ICP changes its bus request pri-
ority from low to high.
The resulting time delay until the ICP changes to high pri-
ority is
timer delay = (16 - PD)*(block time)
One block time is 16 clock times.
Table 13-8. Line start and pixel time for linear model,
no other load on SDRAM
function
t/linestart
(
s)
t/pixel
(ns)
horizontal lter, one component
1.1
11
horizontal lter, 3 components YUV 4:2:2
3.2
22
vertical lter, one component
0.2
29
vertical lter,3 components YUV 4:2:2
0.7
58
yuv to rgb8a,pci output
3.2
30
yuv to rgb15a,pci output
3.3
30
yuv to rgb24,pci output
3.7
34
yuv to rgb24a,pci output
5.3
40
yuv to rgb8a,sdram output
3.4
30
yuv to rgb15a,sdram output
3.3
31
yuv to rgb24,sdram output
3.1
33
yuv to rgb24a,sdram output
3.4
36
yuv to rgb8a,bitmask,pci output
2.5
32
yuv to rgb8a,rgbl15a overlay,pci output
3.8
32
yuv to rgb8a,rgbl24a overlay,pci output
4.0
39
yuv to rgb8a,yuv422a overlay,pci output
3.8
32
yuv to rgb8a,422sequencing,pci output
3.2
20
Table 13-9. Line start and pixel time for linear model,
SDRAM loaded 95%, priority delay = 1
function
t/linestart
(
s)
t/pixel
(ns)
horizontal lter, one component
0.9
20
horizontal lter,3 components YUV 4:2:2
2.8
40
vertical lter, one component
0.2
29
vertical lter,3 components YUV 4:2:2
0.7
58
yuv to rgb8a,pci output
3.8
30
yuv to rgb15a,pci output
3.8
30
yuv to rgb24,pci output
4.5
34
yuv to rgb24a,pci output
6.0
39
yuv to rgb8a,sdram output
4.3
31
yuv to rgb15a,sdram output
4.9
36
yuv to rgb24,sdram output
4.6
47
yuv to rgb24a,sdram output
5.0
53
yuv to rgb8a,bitmask,pci output
3.2
34
yuv to rgb8a,rgbl15a overlay,pci output
5.5
42
yuv to rgb8a,rgbl24a overlay,pci output
5.8
63
yuv to rgb8a,yuv422a overlay,pci output
5.5
42
yuv to rgb8a,422sequencing,pci output
4.9
21
Table 13-10. Line start and pixel time for linear
model, SDRAM loaded 95%, priority delay = 16
function
t/linestart
(
s)
t/pixel
(ns)
horizontal lter, one component
2.9
77
horizontal lter, 3 components YUV422
8.7
154
vertical lter, one component
0.4
87
vertical lter,3 components YUV 4:2:2
1.2
174
yuv to rgb8a,pci output
13.9
82
yuv to rgb15a,pci output
13.8
82
yuv to rgb24,pci output
13.7
82
yuv to rgb24a,pci output
14.0
82
yuv to rgb8a,sdram output
15.8
115
yuv to rgb15a,sdram output
18.5
151
yuv to rgb24,sdram output
17.5
187
yuv to rgb24a,sdram output
16.6
233
yuv to rgb8a,bitmask,pci output
14.3
91
yuv to rgb8a,rgbl15a overlay,pci output
20.7
153
yuv to rgb8a,rgbl24a overlay,pci output
21.6
232
yuv to rgb8a,yuv422a overlay,pci out-
put
20.8
153
yuv to rgb8a,422sequencing,pci output
14.0
80