3890 A place to share, learn, and grow... en-US Fri, 22 Jan 2016 00:16:00 +0000 http://bbpress.org/?v=1.0.2 q http://forums.leaflabs.com/search.php http://forums.leaflabs.com/topic.php?id=953#post-5928 Sat, 06 Aug 2011 12:45:03 +0000 Darth Maker 5928@http://forums.leaflabs.com/ <p>And some of us are even more familiar with mid-range 8-bit PICs, which have 14-bit instructions. :D</p> <p>Although, the 18F series does use 16bit instructions. </p> http://forums.leaflabs.com/topic.php?id=953#post-5915 Fri, 05 Aug 2011 19:10:36 +0000 gbulmer 5915@http://forums.leaflabs.com/ <p>Darth Maker</p> <blockquote><p>It also should be noted that even with the exact same number of assembly lines, Maple/(most 32bit processors) compiled code will be larger, since the instructions themselves are larger (32bit vs 8-14bit). </p></blockquote> <p>Many people are familiar with the 8-bit ATmega often because of Arduino. </p> <p>The instruction size for that is mostly 16 bits with a few 32bits (a form of call, jump, load and store)<br /> <a href="http://en.wikipedia.org/wiki/Atmel_AVR" rel="nofollow">http://en.wikipedia.org/wiki/Atmel_AVR</a><br /> <a href="http://www.atmel.com/dyn/resources/prod_documents/doc0856.pdf" rel="nofollow">http://www.atmel.com/dyn/resources/prod_documents/doc0856.pdf</a> </p> http://forums.leaflabs.com/topic.php?id=953#post-5907 Fri, 05 Aug 2011 17:01:27 +0000 mbolivar 5907@http://forums.leaflabs.com/ <p>Actually, that's not quite true -- Thumb 2 is a variable-length instruction set, so many instructions are 16 bits wide.</p> <p><a href="http://en.wikipedia.org/wiki/ARM_architecture#Thumb-2" rel="nofollow">http://en.wikipedia.org/wiki/ARM_architecture#Thumb-2</a> </p> http://forums.leaflabs.com/topic.php?id=953#post-5895 Fri, 05 Aug 2011 02:15:37 +0000 Darth Maker 5895@http://forums.leaflabs.com/ <p>It also should be noted that even with the exact same number of assembly lines, Maple/(most 32bit processors) compiled code will be larger, since the instructions themselves are larger (32bit vs 8-14bit). </p> http://forums.leaflabs.com/topic.php?id=953#post-5872 Wed, 03 Aug 2011 16:25:44 +0000 robodude666 5872@http://forums.leaflabs.com/ <p>I made my comment under the personal experience that commenting out the lines in the Makefile related to including libraries caused the total compiled size to decrease -- this was observed before I made a thread asking what the output of the size meant. </p> http://forums.leaflabs.com/topic.php?id=953#post-5870 Wed, 03 Aug 2011 11:06:59 +0000 mbolivar 5870@http://forums.leaflabs.com/ <p>poslathian:</p> <blockquote><p> However, AFAIK the build system should NOT include all the libs by default in the final binary. The linker ostensibly only links in what is referenced. I will check with mbolivar and iperry to correct me if I am wrong. </p></blockquote> <p>that should indeed be the case. we call gcc with -ffunction-sections, -fdata-sections, and the linker with --gc-sections, so unused pieces of libmaple should get pruned out </p> http://forums.leaflabs.com/topic.php?id=953#post-5866 Wed, 03 Aug 2011 10:05:07 +0000 poslathian 5866@http://forums.leaflabs.com/ <p>I recommend using integer or fixed point math whenever possible! </p> http://forums.leaflabs.com/topic.php?id=953#post-5865 Wed, 03 Aug 2011 10:04:48 +0000 poslathian 5865@http://forums.leaflabs.com/ <p>Oh, also, using floats will automagically pull in the float libs necessary to do floating point math (since there is no hardware for it). That is probably the source of most of your program code above whats used by the USB stack. AVR's software floating point support is admittedly better than what we have for stm32. </p> http://forums.leaflabs.com/topic.php?id=953#post-5864 Wed, 03 Aug 2011 10:03:10 +0000 poslathian 5864@http://forums.leaflabs.com/ <p>The only *big* fixed cost is the SerialUSB interface, which is compiled in by default and eats up 12KB. This seems like a waste, but as robodude points out, you have plenty of memory. Once you run out of memory, there are ways ways to be thrifty with program size. </p> <p>However, AFAIK the build system should NOT include all the libs by default in the final binary. The linker ostensibly only links in what is referenced. I will check with mbolivar and iperry to correct me if I am wrong. </p> http://forums.leaflabs.com/topic.php?id=953#post-5857 Tue, 02 Aug 2011 19:32:11 +0000 ahunter10@gmail.com 5857@http://forums.leaflabs.com/ <p>OK, that makes sense. Sounds like the libraries will amount to a constant overhead, i.e. double the program length won't double the compiled size. </p> <p>Thanks for the explanation. I should play around with libmaple. </p> http://forums.leaflabs.com/topic.php?id=953#post-5856 Tue, 02 Aug 2011 19:12:27 +0000 robodude666 5856@http://forums.leaflabs.com/ <p>When libmaple compiles your sketch, at least on libmaple -- not sure about the MapleIDE -- it includes all of the libraries which seems to inflate the codebase. If you are using libmaple, you can go into the Makefile and comment out the libraries you don't need.</p> <p>The 19k you're seeing also includes all of libmaple's lower-level c functionality and C++ wirish classes, allowing you to do a lot of things so don't be too sad to see it go. The Maple has 120kB+ FLASH space, so it may be shocking at first to see 19k being used compared to what you'd see on an Arduino. </p> http://forums.leaflabs.com/topic.php?id=953#post-5855 Tue, 02 Aug 2011 18:53:57 +0000 ahunter10@gmail.com 5855@http://forums.leaflabs.com/ <p>I jut started playing around with the maple IDE, and I'm surprised how big the compiled code size is. This program below, which is mostly just floating point arithmetic, compiles to over 19k. This seems very large, and I'm wondering why. On the arduino, for comparison, the same code complies to 3.8k.</p> <p>I'm on mac os x, with maple IDE 0.0.11. Here's the code:</p> <pre><code>#define N 3 float A[N][N]; float B[N][N]; void setup(){ pinMode(D0, OUTPUT); } void loop(){ populateMatrix((float*)A,N); digitalWrite(D0,HIGH); invertMatrix3x3((float*)A, (float*)B); digitalWrite(D0,LOW); delay(100); } int invertMatrix3x3(float *A, float *B){ float det; B[0] = (A[4] * A[8]) - (A[5] * A[7]); B[1] = (A[2] * A[7]) - (A[1] * A[8]); B[2] = (A[1] * A[5]) - (A[2] * A[4]); B[3] = (A[5] * A[6]) - (A[3] * A[8]); B[4] = (A[0] * A[8]) - (A[2] * A[6]); B[5] = (A[2] * A[3]) - (A[0] * A[5]); B[6] = (A[3] * A[7]) - (A[4] * A[6]); B[7] = (A[6] * A[1]) - (A[0] * A[7]); B[8] = (A[0] * A[4]) - (A[1] * A[3]); //calculate determinant, using some already calculated values det = (A[0] * B[0]) + (A[1] * B[3]) + (A[2] * B[6]); //check if determinant==0 if ( abs(det) &lt; .0001 ){ return 0; } det = 1.0/det; for (int i=0; i&lt;3; i++){ for (int j=0; j&lt;3; j++){ B[i*3+j] *= det; } } return 1; } void populateMatrix(float *A, int n){ for (int i=0; i&lt;n; i++){ for (int j=0; j&lt;n; j++){ A[i*n+j] = (float)random(100) / (float)random(1,100); } } }</code></pre>