**** BEGIN LOGGING AT Tue Jul 15 02:59:59 2014 Jul 15 03:09:10 * mranostay is back Jul 15 04:05:03 karki: Yep :) Jul 15 06:08:04 panto : would it be better to use "switch case" or "if else" in the opcode handling fn in the PRU? Jul 15 06:10:42 The thing is there are many cases, switch case would look messy. "if else" can handle ranges so the code will look cleaner (e.g. case 1,2,3 are all SET DIOs. so one if statement will take care of all 3). Jul 15 06:11:22 but the thing is, with the use of "switch case" the compiler optimizes it in form of a jump table - so this is faster. Jul 15 06:11:46 so there are pros and cons of either choice. Jul 15 06:12:00 alexanderhiam, jkridner : whats your call on this ^ Jul 15 10:46:53 jkridner : ping Jul 15 11:13:58 * Abhishek____ checks in from the airport :) Jul 15 11:15:00 * Abhishek____ looks at the extra underscores in his nick Jul 15 11:51:19 I got my BB delivered today. Jul 15 11:51:23 :D Jul 15 11:52:00 weirdly I got 4 BBB's :') Jul 15 12:24:48 karki_: lol :)) Jul 15 13:50:20 karki_: What do you mean? Jul 15 13:52:17 :) Jul 15 14:05:33 I got a package with 4 BBB in it :D Jul 15 14:05:36 :') Jul 15 14:06:07 wow, calls for a party :D Jul 15 14:12:01 KumarAbhishek_ : I think the package was meant for all of us :p me, you, rekha and praveen! Jul 15 14:12:47 * KumarAbhishek wonders how the underscore got added to my nick again :P Jul 15 14:13:17 * karki hates the extra underscores :/ Jul 15 14:15:13 karki: Did you try socket.io ? Jul 15 14:28:43 karki_: switch/case is more readable to me. Jul 15 14:32:10 jkridner : switch case it is :) Jul 15 14:32:27 KumarAbhishek : not yet! Jul 15 16:32:53 alexanderhiam : my new compiler-interpreter architecture can take care of all the old commands :) Step 1 complete. Now to support the new commands :D Jul 15 16:33:12 Basically my PRU firmware got an update :p Jul 15 16:33:47 awesome Jul 15 16:35:25 about the switch/case in the interpreter: it might be worth reworking your commnad mapping a bit to make it a lot easier/more efficient to interpret Jul 15 16:36:52 e.g. always the same bit to indicate a 64-bit command, you could have 2 bits to indicate the number of args and have all the args in the same place across all instructions Jul 15 16:37:30 that way you could extract all the arguments before even looking at the op code Jul 15 16:39:44 alexanderhiam : that's not possible I'm afraid. If I keep two bits for no. and type of args, many well fit 32 bit inst will become 64 bit! :/ Jul 15 16:40:00 but I have grouped similar instructions Jul 15 16:40:10 there are about 3-4 groups Jul 15 16:40:30 single operands, 2 operands, etc Jul 15 16:40:45 so in those inst. I have done what you have mentioned :) Jul 15 16:41:48 i.e. similar encoding for all inst of one type, so the operand extraction code will be small and fast :D Jul 15 16:42:08 *of one group Jul 15 16:56:08 karki: hey, my internet cut out as I was making a suggestion, did you get that? Jul 15 16:57:52 yeah Jul 15 16:57:56 i replied too Jul 15 16:58:03 http://logs.nslu2-linux.org/livelogs/beagle-gsoc.txt Jul 15 16:58:21 alexanderhiam : ^ Jul 15 16:59:42 I think it could work if you convert array indexes to their absolute address in the variable memory and treat them exactly as integer variables Jul 15 17:01:00 yeah, but what about arr[v] conditions? that has to be handled in the PRU Jul 15 17:01:32 I'm treating arr[constant] as a var even now :p Jul 15 17:02:27 ah, good point. you could potentially turn that into two instructions Jul 15 17:02:30 the problem is with the "same place" across all instructions..... Jul 15 17:03:12 but don't worry. you can review my code tomorrow before/after the meeting and suggest stuff if you don't find it clean Jul 15 17:03:17 of efficient Jul 15 17:03:46 alexanderhiam : I'll push everything on github soon. Jul 15 17:03:49 :) Jul 15 17:03:54 ok cool Jul 15 18:33:36 alexanderhiam, jkridner, panto : I'm able to run BS scripts now with a *much* cleaner frontend API, + I can include variables in scripts and change them on the fly :D Jul 15 18:35:13 so first phase of the interpreter is a success :); I'll continue tomorrow and try to finish it off! By this week I'll finish the complete language architecture including inputs and return values :) Jul 15 18:35:50 PRU Speak V1.0.0 could be out by next monday, with complete instruction set et all!! **** ENDING LOGGING AT Wed Jul 16 02:59:59 2014