Lines Matching refs:I

5 I felt there was room for optimisation. I bashed the code for a few hours
8 Still I was not too happy as I felt there was additional room for improvement.
10 Bad! I was hooked.
11 I decided to annotate my steps in this file. Perhaps it is useful to someone
22 This is done by means of a Hamming code. I'll try to explain it in
23 laymans terms (and apologies to all the pro's in the field in case I do
25 years ago, and I must admit it was not one of my favourites).
27 As I said before the ecc calculation is performed on sectors of 256
74 The story now becomes quite boring. I guess you get the idea.
93 I detected after writing this that ST application note AN1823
95 nicer picture.(but they use line parity as term where I use row parity)
96 Oh well, I'm graphically challenged, so suffer with me for a moment :-)
97 And I could not reuse the ST picture anyway for copyright reasons.
251 I also introduced the parity lookup. I expected this to be the fastest
252 way to calculate the parity, but I will investigate alternatives later
286 And of course the performance might depend on alignment, but I expect
287 that the I/O buffers in the nand driver are aligned properly (and
380 examples I kinda deviated from my regular programming style by allowing
437 or so. I also tried on an eeePC (Celeron, clocked at 900 Mhz). Interesting
443 loop unrolling. This will eliminate a few if statements. I'll try
479 I decided to proceed with a four time unrolled loop anyway. It was my gut
480 feeling that in the next steps I would obtain additional gain from it.
486 eliminate rp5 (or rp4, but I already foresaw another optimisation).
498 Along the line I also removed the initialisation of rp0/1/2/3.
508 However, still I thought there was more. I didn't like all the if
549 While making the changes I also found that I could exploit that tmppar
552 I removed the rp6 = cur; statement and did rp6 ^= tmppar; on next
561 (using time to measure the performance). After this iteration I was back
562 to 0.075 sec. Actually I had to decide to start measuring over 10
625 Actually this made things worse. Not very much, but I don't want to move
630 more time will help. I'll keep the optimisations from 7 for now.
673 million iterations. So basically I've improved the performance by a
684 For correcting errors I again used the ST application note as a starter,
685 but I also peeked at the existing code.