Ok, great! For divisions may I suggest division/multiplication with a constant? That allows you to calibrate the full scale of the ADC or a termometer to your needed value with your imposed error and code length. Higher error meaning lower code lenght, lower error creates higher code lenght. The old routine (constdivmul) was written by a russian named Golovnichenko, a great piece of software adapted separately by Kyle and Stef Mientky (if I recall well). Thank you for your work, I have appreciated and learned new things from these libraries!
On Wed, Dec 23, 2020 at 12:01 AM 'Oliver Seitz' via jallib < [email protected]> wrote: > Hi Vasile, > > of course! In some way, you can check it right now: Compare the format > library, which still uses divisions, to the print library (integer > routines). My intension is not primarily to add bells and whistles, but to > save resources. You can compile a program using the print library for the > 10f200 chip if you limit the decimal conversion to 1 byte. > > When I did my own print library, I mainly wanted it to be more flexible. I > was impressed by the possibility of using byte*256 sized numbers, and > wanted to convert those to decimal. When I was finished however, we > discussed both libraries. My new library could output bigger numbers, > worked remarkably faster, and used less both code and data space. > Therefore, as there was not a single advantage on the old library, it was > decided to replace it. > > I did in fact measure the instruction cycles both libraries needed to > convert every single number that can be stored in a dword. The old library > was indeed faster - when converting "0". For all of the other 4294967294 > numbers, my library was faster ;-) > > And Rob, as usual I had some extras in mind, you guessed some of them :-) > > I was, however, thinking about constants with defaults that one could > define, or use variables, if during runtime different output styles are > needed. > About like this: > > FORMAT_THOUSANDS="," -- or "'", or " ", or 0 if none wanted > FORMAT_DECIMAL="." > FORMAT_PLUS_SIGN=0 -- or " " to always reserve space for the sign, or > "+" to always display a sign > FORMAT_LEADER="0" -- char with wich a right-aligned number is padded, > usually "0" or " " > > > In general, I really do want format to work the way it does now, > displaying a decimal separator wherever I want it. I like to use > fixed-point math. Right now, my thermometer has to transmit the > measurements in millicelsius. I really would prefer having a decimal point > (or comma). There is a special procedure in format.jal which is there for > outputting certain digits out of a number. That's helpful for my > thermometer, which can do the math internally in millicelsius but can > easily be written to a display with only one digit after the point. Again, > without using divisions. I hate divisions, they're using so much resources. > ;-) > > Greets, > Kiste > > > Am Dienstag, 22. Dezember 2020, 18:40:22 MEZ hat vsurducan < > [email protected]> Folgendes geschrieben: > > > > > > Hi Oliver, > Can we check your improvements first for the resources used? The > experience of the last 20 years on jal shows that biggest improvements can > not be used anymore on small pics... > The length of the number (in my way of understanding) depends on the > output device type. On some is useless to have too many numbers. On a LCD > display I do not see any requiring a 9 digit (except perhaps displaying a > frequency) . On the other hand, on a serial console it might be useful for > something... > thx, > Vasile > > On Tue, Dec 22, 2020 at 1:59 PM 'Oliver Seitz' via jallib < > [email protected]> wrote: > > Hi all :-) > > > > It's been nine days and nine years since the original print library was > finally replaced by my version. I remember, I've said then, that I would > also do the format library soon. I feel like it's about time now... > > > > So, usually when I start thinking about improvement, I'm focused on > mimicking the current behavior as close as possible. Yet... This is what it > currently does: > > > > format_sdword(output,12345678, 6,0) -> 5678 > > format_sdword(output,12345678, 7,0) -> �45678 > > format_sdword(output,12345678, 7,1) -> �4567.8 > > format_sdword(output,12345678, 8,0) -> C345678 > > format_sdword(output,12345678, 8,1) -> C34567.8 > > format_sdword(output,12345678, 8,2) -> C3456.78 > > format_sdword(output,12345678, 9,0) -> 12345678 > > format_sdword(output,12345678, 9,1) -> 1234567.8 > > format_sdword(output,12345678, 9,2) -> 123456.78 > > format_sdword(output,12345678,10,0) -> 12345678 > > format_sdword(output,12345678,10,1) -> 1234567.8 > > format_sdword(output,12345678,10,2) -> 123456.78 > > format_sdword(output,12345678,11,0) -> 12345678 > > format_sdword(output,12345678,11,1) -> 1234567.8 > > format_sdword(output,12345678,11,2) -> 123456.78 > > format_sdword(output,12345678,12,0) -> 12345678 > > format_sdword(output,12345678,12,1) -> 1234567.8 > > format_sdword(output,12345678,12,2) -> 123456.78 > > format_sdword(output,12345678,13,0) -> 12345678 > > format_sdword(output,12345678,13,1) -> 1234567.8 > > format_sdword(output,12345678,13,2) -> 123456.78 > > > > The third parameter, said to give the field length to be used, is quite > limited: > > > > 1) The actual length differs, depending on the presence of a decimal > point > > 2) Length greater than 11 is treated as 11 > > 3) Too short a field leads to corruption of the number, always reserving > places for sign and decimal point > > > > 3) probably can't be of any use - here's the question if the number > should be truncated or replaced by error signs to show that something's > wrong. > > > > For 1) and 2), there's a possibility that existing programs rely on this > behavior. But, as those are 'undocumented "features" ', I feel like they > can be replaced by more useful things. Like, fields can be as long as 255 > characters, and the number is truly right-aligned, if it has decimals or > not. > > > > If the number of decimals is greater than the field lengt-1, it is > treated as zero. Here's another option to fill: what should be done if both > numbers are nearly-equal, equal or the decimals number greater than the > field length? Left-aligned, ignoring the field length? > > > > Any preferences anyone? > > > > Greets, > > Kiste > > > > -- > > You received this message because you are subscribed to the Google > Groups "jallib" group. > > To unsubscribe from this group and stop receiving emails from it, send > an email to [email protected]. > > To view this discussion on the web visit > https://groups.google.com/d/msgid/jallib/2110995783.3397177.1608638352475%40mail.yahoo.com > . > > > > > -- > You received this message because you are subscribed to the Google Groups > "jallib" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To view this discussion on the web visit > https://groups.google.com/d/msgid/jallib/CAM%2Bj4qv_pVn2XgeJ5cLMSv%2BJY%3Dm-Tn9BvyKfF3%3DfWDRr9FA16Q%40mail.gmail.com > . > > -- > You received this message because you are subscribed to the Google Groups > "jallib" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To view this discussion on the web visit > https://groups.google.com/d/msgid/jallib/1676376410.3617961.1608674508106%40mail.yahoo.com > . > -- You received this message because you are subscribed to the Google Groups "jallib" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/jallib/CAM%2Bj4quqrAQMkzDfQNXhpN%2BkY_bp_OoqgO%3Dpy46hmDwnBfCbEQ%40mail.gmail.com.
