01:51:00 | Sammy | hello BZFlag :) |
02:00:32 | BZFlag | hello. |
02:00:51 | BZFlag | I guess I woke erikm up. Things are happening. ;-) |
02:04:45 | Sammy | BZFlag :what's happening ? |
02:16:48 | BZFlag | I added a new platform and a memory dump routine. |
02:17:11 | BZFlag | looks like erikm is adding neponset (a flavor or assabet) |
02:17:30 | BZFlag | I have updated Russ' read-from-jffs2 patch as well. |
02:25:10 | Sammy | cool. |
02:25:25 | Sammy | what's new platform ? |
02:58:52 | BZFlag | mine? unreleased. CES in january. |
04:29:06 | Sammy | BZFlag: I think this maybe the problem you say ... |
04:29:35 | Sammy | Received 4194304 (0x00400000) bytes |
04:29:51 | Sammy | flash_write_region(): flashing 0x00100000 (1048576) words from 0xc0400000 to 0x00180000 |
04:29:56 | Sammy | erasing at 0x00180168... scanning down... resume writing at 0x00180000 |
04:29:56 | Sammy | erasing at 0x0029b964... scanning down... resume writing at 0x0027fff8 |
04:29:56 | Sammy | erasing at 0x0033e17c... scanning down... resume writing at 0x002ffff8 |
04:29:56 | Sammy | erasing at 0x00340de0... scanning down... resume writing at 0x0033fff8 |
04:30:01 | Sammy | erasing at 0x0053a594... scanning down... resume writing at 0x004ffff8 |
04:30:01 | Sammy | erasing at 0x00540000... scanning down... resume writing at 0x0053fff8 |
04:30:01 | Sammy | 1048576 words source image |
04:30:02 | Sammy | 393213 words written to flash |
04:30:04 | Sammy | 807938 words skipped |
04:30:05 | Sammy | 6 erase operations |
04:30:07 | Sammy | 152569 words scanned down |
04:33:32 | Sammy | do you knoe why only written 393213 words and 807938 words skipped ? |
04:35:30 | Sammy | like you say before it's must at least write to 0x00EF0000, but at you see , it's only write only 0x00540000 , so many block don't currect write , why ? |
04:35:41 | Sammy | any idea or suggest ? |
04:50:02 | BZFlag | it only got 4194304 bytes. |
04:52:05 | Sammy | so , change it from blob to get more byte ? |
04:52:15 | Sammy | right ? |
05:41:19 | BZFlag | right. look at RAMDISK_SIZE and RAMDISK_RAM_BASE |
05:41:32 | BZFlag | I think it needs to be contiguous memory at present. |
06:08:59 | Sammy | BZFlag: thanx a lot , working .... |
06:46:04 | skanda | hello amrutha |
07:03:08 | amrutha | hello .. |
07:03:23 | amrutha | howz work coming up? |
07:20:28 | skanda | pretty cool, glad you asked:-) |
10:11:31 | Sammy | dinner time ... |
10:13:02 | amrutha | ........ |
10:14:03 | amrutha | erik:hai! |
10:19:27 | erikm | morning, * |
10:21:32 | amrutha | eruk: i have a dsp processor from texas instr. i want to make use of that processor with my board(which is very similar to assabet)... any inputs? |
10:24:39 | erikm | amrutha: does it use any interrupts? |
10:24:58 | amrutha | the processor? |
10:25:44 | amrutha | yes... it has 2 external interruots. |
10:25:45 | erikm | amrutha: ehm, let me put it like this: can the dsp toggle an interrupt on the CPU? |
10:26:57 | amrutha | hmm..well..there are two interrupt pins brought out which can be connected |
10:27:54 | erikm | hmm |
10:28:28 | erikm | how is the dsp connected? with a couple of gpio pins or is it mapped in a memory area of the cpu? |
10:30:34 | amrutha | erikm: right now my dsp processor what i have is like another development board. and the board which i am using is very similar to assabet (but for the absence of jflash)... so was just thinking if i could integrate both of them.. |
10:32:03 | amrutha | well..i could map the dsp processor to some memory area..or i could use the gpio pins of my board to the dsp processor..both are possible. |
10:33:17 | erikm | amrutha: the problem is that all people tell their sa-1110 board is "like assabet", but that's only because it also has an sa1110 and sdram. variance in sa1110 boards is quite large, even though they "look like assabet" |
10:33:51 | erikm | amrutha: (iow: the devil is in the details) |
10:34:15 | erikm | amrutha: mapping the dsp in a memory region makes it very easy to develop a userland driver first |
10:40:46 | amrutha | erikm: can you be more specific please. I was trying to use the dsp to improve the efficiency of the processor...is it possible? |
10:41:35 | amrutha | okie..for example...in my audio processing examples..using a dsp will be more efficient that a sa1110? |
10:42:06 | erikm | depends. |
10:42:29 | erikm | the sa1110 is awfully good at audio processing |
10:43:10 | erikm | especially the multiply/accumulate in a single clock cycle (with some pipeline delays) makes it extremely fast |
10:43:17 | amrutha | so that means incorporation of a processor like TMS 320C30 is not going to make thimgs better?? |
10:43:28 | erikm | depends |
10:44:04 | erikm | chances are high that moving the data to the DSP takes more time than doing the whole operation with the sa1110 |
10:44:22 | amrutha | thatw as exactly what i was thinking right now. |
10:44:59 | amrutha | which other area of application do you think of where things are made better by the use of tghis oprocessor? |
10:45:20 | erikm | to give you an idea: a cache miss on an sa1100 costs about 60 cpu cycles. |
10:46:08 | erikm | that might be a little bit different on the sa1110, but not that much |
10:46:47 | amrutha | oh okie... |
10:47:43 | erikm | the latency of the dsp might be lower, so it might differ, but also remember that the CPU can do burst cycles to memory, but not to the dsp |
10:48:04 | amrutha | erikm: if i were to target my application of this board purely for handhelds.... do you have any extra suggestions? |
10:48:53 | erikm | if the application is for handhelds like the compaq ipaq, don't use a dsp |
10:49:01 | amrutha | why? |
10:49:18 | erikm | how would you connect a dsp to the ipaq? |
10:49:33 | erikm | or to the yopy? |
10:50:01 | erikm | or to the hp jornada? |
10:50:13 | amrutha | it's not actually ipaq... but my design of the board is flexible ..i mean at my will.. |
10:50:51 | erikm | oh, ok, I thought you wanted to target the handhelds market in general |
10:51:09 | amrutha | could be.... i have a sa1110 and just the dsp processor..next to each other.. |
10:52:59 | erikm | if you want to use a dsp, make pretty damn sure that you have a good reason to use it |
10:53:17 | Sammy | is back ... |
10:53:25 | erikm | IOW: the time needed for data load, processing on the dsp, and data offload should be lower than the time it takes to do everything on the sa1110 |
10:55:19 | erikm | the intel ipp libraries will give you a nice estimate how fast the sa1110 can be, but be careful: in contrast with what intel likes you to believe, their code is not optimal |
10:55:23 | amrutha | true..how do i check that out? |
10:55:58 | erikm | so there is still speed to gain with a properly designed application |
10:56:52 | erikm | to give you an example: intel claims their FFT code in the IPP library is "optimal". we coded our own version that was four to eight times faster. |
10:57:23 | amrutha | i tried the fft code on my dsp processor too.. |
10:57:56 | erikm | and if they can't code something simple and well understood as the FFT, it gives a hint about the code quality of the rest of the IPP |
10:57:59 | amrutha | btw... how did you check the speed for your code? |
10:58:10 | erikm | profiler |
10:59:35 | erikm | if you are optimising a piece of code, you can't do without a profiler |
10:59:36 | amrutha | okie...so the major appliction that i was thinking was to use the dsp for the audio processing part of it..so as you said i am not sure how better it's performance wud be..than the sa1110. |
11:00:09 | amrutha | how do i make use of the profiler? |
11:00:31 | erikm | see "man gprof" |
11:01:02 | erikm | and if you don't have room for that on your ramdisk/jffs2 fs, you can also use a couple of gettimeofday() calls to do the calculation yourself |
11:03:07 | erikm | SIGLUNCH |
11:15:21 | Sammy | later :) |
14:01:34 | | 0.125 |
14:01:34 | sammy_wms | ibot: 1 / 8 |
14:01:46 | | 0.113636363636364 |
14:01:46 | sammy_wms | ibot: (1 / 8.8) |
14:02:08 | | 1.5 |
14:02:08 | sammy_wms | ibot: (1 / 2*3) |
14:02:20 | | 3.5 |
14:02:20 | sammy_wms | ibot: (1 / 2+3) |
14:02:50 | | 10 |
14:02:50 | sammy_wms | ibot:( 1/ 2^10 ) |
14:03:03 | | 0.5 |
14:03:03 | sammy_wms | ibot:( 1/ 2 ) |
14:03:09 | | 5 |
14:03:09 | sammy_wms | ibot:( 1/ 2*10 ) |
14:03:21 | | 0.05 |
14:03:21 | sammy_wms | ibot:( 1/ (2*10) ) |
14:03:35 | | 0.05 |
14:03:35 | sammy_wms | ibot: 1/ (2*10) |
14:04:07 | | -0.037037037037037 |
14:04:07 | sammy_wms | ibot: 1/ (2*10*5/4+3-11/3*15) |
14:04:38 | sammy_wms | he he :P |
14:06:04 | | sammy_wms: huh? |
14:06:04 | sammy_wms | ibot: 1/ 5s |
14:06:10 | | 0.0181818181818182 |
14:06:10 | sammy_wms | ibot: 1/ 55 |
14:06:14 | | sammy_wms: what? |
14:06:14 | sammy_wms | ibot: 1/ s5 |
14:06:25 | | sammy_wms: huh? |
14:06:25 | sammy_wms | ibot: s / 10 |
14:06:33 | | sammy_wms: excuse me? |
14:06:33 | sammy_wms | ibot: s1 / 10 |
14:06:38 | | sammy_wms: i'm not following you... |
14:06:38 | sammy_wms | ibot: s 1 / 10 |
14:06:45 | | sammy_wms: huh? |
14:06:45 | sammy_wms | ibot: s (1 / 10) |
16:04:03 | sammy_wms | reboot .... |
17:51:56 | erikm | returns |
17:55:03 | BZFlag | heya erikm |
17:55:28 | BZFlag | you take a look at the updated jffs2 patch? |
17:55:38 | BZFlag | should I put it in CVS? |
17:55:52 | BZFlag | I wanted you to take a look at the file layout first |
17:57:26 | erikm | BZFlag: I looked at it quickly. I'll have a proper look in an hour or so |
19:33:16 | prpplague | BZFlag: hey what was the other item in uclibc you wanted to get andersee to add? |
19:39:11 | erikm | diagnoses a dead lart |
19:40:37 | prpplague | doh |
19:40:40 | erikm | ... and finds a broken Vcore power supply |
19:40:58 | prpplague | shall we have a 21 resistor pop salute? |
19:41:45 | erikm | not yet... |
19:42:38 | erikm | thinks the ltc1266 is dead |
19:44:37 | erikm | either that or the fet |
19:54:37 | erikm | guesses the ltc1266 |
20:17:55 | erikm | goes hunting for food |
20:18:00 | erikm | cya |