Good work, I'll play around with it when I get a chance. I don't see any need for 2 separate libs. If this one is faster and indeed scalable, it should replace the existing.
Matt. On Oct 30, 12:22 pm, Oliver Seitz <[email protected]> wrote: > Don't read if you're in a hurry. > Nothing important in this mail, only a whole lot of waffle. > ;-) > > Here are some results of some calculations and some > experiments... > > Memory: > > Both data and code size can be a bit larger or smaller when > you switch from print.jal to print_scalable.jal. Difference > is usually small. It mainly depends, if dword divisions are > used within the program. If you do not need dwords and limit > the range to one or two bytes before including > print_scalable.jal, you most certainly will save ram and > flash. > > Speed: > > How long should a conversion from a binary number to a > decimal number take? Simple answer: Not too long. What is > too long? Let's take the fastest PIC and the slowest > interface. > > The slowest and most common interface would be RS232. It > can usually do up to 115200 bit per second. Counting start- > and stopbits, that's 11520 chars per second. > > The decimal representation of a dword has about 9.74 digits > in average. I'll only talk about average values here. > Divided the chars per second by this number gives 1182.6, > which is the number of dwords that can be transmitted via > RS232 per second. > > Lets take a real fast PIC. 64MHz seems appropriate, that is > 16MHz instruction clock. This, divided by the numbers per > second makes 13530 cycles. If a conversion routine takes > less cycles, the interface will set the speed. If it takes > more, the interface will be waiting for the converted data. > If you're thinking at either slower clock speeds, or at USB, > I2C or SPI, the conversion should preferably be much > faster. > > How fast is it, then? For print.jal, I can not tell yet. > The test program is still running. The result so far is > 42906 cycles per dword. This only counts the numbers 0 to > 766158238. That is how far the program made it after 38 > days. > > For print_scalable, I could do some simplification to get > the result faster, so here I got the average for all the > values from 0 to 4294967290 as 3840. That is more than ten > times faster, and, the slowest number (3999999999) for > scalable ist still faster than the so far fastest number for > "conventional" (200000000). (The really fastest number, > zero, is handled by an unfair exception ;-) ) > > Some more examples: at 20MHz, print yields 1135 chars per > second, at 4MHz 227. Print_scalable at 20MHz converts 12628 > chars per second, and 2536 at 4MHz. A 4MHz controller using > print.jal thus can only transmit decimal data at a rate like > 2400 baud, which is not what I would expect. Print_scalable > makes at least around 19200 baud, which really looks less > dull. > > In the attached graph, a minimum, maximum and average has > been computed for each interval of ten millions to make > straight lines visible. > > Greets, > Kiste > > P.S.: If someone is interested, I can send the raw data by > email. The list of decimal numbers makes a text of 20GByte, > but xz packs it to 15MByte. > > P.P.S.: > Here's the "Hitlist" for print.jal: > > Top values: | Bottom values: > 1.: 48160 @ 757777777 | 58 @ 0 > 2.: 48150 @ 737777777 | 5506 @ 200000000 > 3.: 48140 @ 377777777 | 5515 @ 100000000 > 4.: 48130 @ 697777777 | 5546 @ 300000000 > 5.: 48126 @ 757777773 | 5596 @ 700000000 > 6.: 48125 @ 757737777 | 5904 @ 200000 > 7.: 48124 @ 757776777 | 5908 @ 100000 > 8.: 48120 @ 357777777 | 5928 @ 2000000 > 9.: 48116 @ 737777773 | 5932 @ 1000000 > 10.: 48115 @ 737737777 | 5947 @ 20000000 > 11.: 48114 @ 737776777 | 5954 @ 300000 > 12.: 48110 @ 337777777 | 5956 @ 10000000 > 13.: 48106 @ 377777773 | 5968 @ 3000000 > 14.: 48105 @ 377737777 | 5987 @ 30000000 > 15.: 48104 @ 377776777 | 5992 @ 90000000 > 16.: 48100 @ 177777777 | 5994 @ 700000 > 17.: 48096 @ 697777773 | 6008 @ 7000000 > 18.: 48095 @ 697737777 | 6017 @ 70000000 > 19.: 48094 @ 697776777 | 6325 @ 1000 > 20.: 48092 @ 757777772 | 6326 @ 2000 > 21.: 48091 @ 757737773 | 6349 @ 10000 > 22.: 48090 @ 277777777 | 6350 @ 20000 > 23.: 48089 @ 757736777 | 6366 @ 3000 > 24.: 48086 @ 357777773 | 6390 @ 30000 > 25.: 48085 @ 357737777 | 6410 @ 7000 > > And here for print_scalable.jal: > > Top values: | Bottom values: > 1.: 5225 @ 3999999991 | 104 @ 0 > 2.: 5222 @ 3999999990 | 280 @ 10 > 3.: 5198 @ 2999999991 | 283 @ 11 > 4.: 5195 @ 2999999990 | 319 @ 20 > 5.: 5188 @ 3899999991 | 322 @ 21 > 6.: 5187 @ 3999989991 | 358 @ 30 > 7.: 5185 @ 3899999990 | 361 @ 31 > 8.: 5184 @ 3999989990 | 397 @ 40 > 9.: 5181 @ 3999999890 | 400 @ 41 > 10.: 5180 @ 3999998991 | 436 @ 50 > 11.: 5177 @ 3999998990 | 439 @ 51 > 12.: 5176 @ 3998999991 | 460 @ 100 > 13.: 5173 @ 3998999990 | 463 @ 101 > 14.: 5171 @ 1999999991 | 475 @ 60 > 15.: 5169 @ 3999899991 | 478 @ 61 > 16.: 5168 @ 1999999990 | 499 @ 110 > 17.: 5166 @ 3999899990 | 502 @ 111 > 18.: 5161 @ 2899999991 | 514 @ 70 > 19.: 5160 @ 2999989991 | 517 @ 71 > 20.: 5158 @ 2899999990 | 532 @ 300 > 21.: 5157 @ 2999989990 | 535 @ 301 > 22.: 5156 @ 3889999991 | 538 @ 120 > 23.: 5154 @ 2999999890 | 541 @ 121 > 24.: 5153 @ 2999998991 | 553 @ 80 > 25.: 5151 @ 3979999991 | 556 @ 81 > > graph.png > 34KViewDownload -- You received this message because you are subscribed to the Google Groups "jallib" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/jallib?hl=en.
