summaryrefslogtreecommitdiff
path: root/Makefile
diff options
context:
space:
mode:
authorTrent Piepho <xyzzy@speakeasy.org>2007-03-02 14:42:21 -0800
committerTrent Piepho <xyzzy@speakeasy.org>2007-03-02 14:42:21 -0800
commitabcb50a0ccaaec6003d034878db63683bc059b2a (patch)
treea0e467927f8ca007a79f5b66e28245d045a3006a /Makefile
parentad944c387a5d1c2a38e15a3bd48e65c2e3800707 (diff)
downloadmediapointer-dvb-s2-abcb50a0ccaaec6003d034878db63683bc059b2a.tar.gz
mediapointer-dvb-s2-abcb50a0ccaaec6003d034878db63683bc059b2a.tar.bz2
or51132: refactor i2c code, improve error resilience
From: Trent Piepho <xyzzy@speakeasy.org> The code the i2c transactions was leftover from the old V4L-based ATSC driver. It did too little with too much code. It is re-written to remove unnecessary parameters and be more efficient. A demod register can now be read with one function call, instead of repeating a dozen line block of code each time. There were msleep()'s, which appear to be unnecessary, spread around all the I2C transactions. These have been removed. Reading SNR used to take about 130 ms, now it's down to 1.8 ms. Reads from the demodulator's registers do not return correct results sometimes. Adding or removing the delays in the I2C transactions did not appear to effect the probability of failure. If anything, the transactions without delays were less likely to fail, but since far more transactions could be made per second the number of failures per hour was greater. To increase reliability, the SNR and get_params functions will now retry once if they get bad data back. This appears to have reduced the probability of failure to effectively zero. Some error messages are cleaned up or given KERN_* levels when they were missing. or51132_setmode() wasn't returning correct error codes, which is fixed as well. Signed-off-by: Trent Piepho <xyzzy@speakeasy.org> CC: Rusty Scott <rustys@ieee.org>
Diffstat (limited to 'Makefile')
0 files changed, 0 insertions, 0 deletions