ROT ROT - oh GROB!
in [HP48]
|
Prev: HP28S (ROM)
Next: Duct Sizing Program
From: Steen Schmidt on 26 Jan 2006 08:29 Michael wrote: > for those critical bits. The cool thing, is that in SRPL, null > locals are quite fast, so you have the best of both, er, triple > worlds: speed, readability, flexibility ;-) That's the first first time I've seen NULLLAMs associated with readability ;-) Here's a small tidbit from the uncompressed source of my port of FluidHP on hpcalc.org: ******** 7GETLAM %SQRT %SQRT DUP 6PUTLAM 18GETLAM SWAP 21GETLAM OVER BINT24 GETLAM %* %+ %* %+ 5PUTLAM 17GETLAM 6GETLAM 20GETLAM OVER BINT23 GETLAM %* %+ %* %+ DUP 4PUTLAM 5GETLAM DUPDUP %* 22GETLAM 6GETLAM %+ 6GETLAM %* 19GETLAM %+ 4GETLAM %* %4 %* %+ %SQRT %+ %/ 15GETLAM %+ DUP 2PUTLAM 14GETLAM OVER %+ 2GETLAM %* 13GETLAM %+ %SQRT %- 1PUTLAM 10GETLAM BINT25 GETLAM 9GETLAM %- 1GETLAM 10GETLAM %- %* 7GETLAM 9GETLAM %- %/ %+ ******** I use up to 28 NULLLAMs in nested environments, so readability is pretty much a goner (but code size is down and performance is up, so.. :-). To make a fair point the rest of the code isn't much better - another tidbit: ******** %0 %0 % 1489.50410795 %0 %1 % 743.07798314 %0 %2 % -97.7083187978 %0 %12 % 2.47424647057 %0 %18 % -0.63281320016 %0 %24 % 1.13859521297 %0 %7 %4 %* % -0.478118636486 %0 %10 %4 %* % 8.52081234315E-3 %1 %0 % 0.937471473779 %1 %2 % 3.35931186049 %1 %6 % 3.38093556015 %1 %12 % 0.168445396719 %1 %18 % 0.738757452367 %1 %24 % -0.471287374362 %1 TWENTYFOUR PICK % 0.150202731397 %1 TWENTYFOUR PICK % -2.17641142198E-3 %2 %2 % -2.18107553248E-2 %2 %8 % -0.108297844037 %2 %18 % -4.63333246358E-2 %2 TWELVE PICK % 7.12803519596E-5 %3 %1 % 1.103283179E-4 %3 %2 % 1.89552483879E-4 %3 %12 % 3.08915411605E-3 %3 %24 % 1.35555045549E-3 %4 %2 % 2.86402374774E-7 %4 %12 % -1.07798573575E-5 %4 %18 % -7.64627124548E-5 %4 %24 ******** All the code is like that, I didn't even have to search for special obfuscated parts of it :-/ I have forgotten how many stack levels are used, but it's close to hundred. The Debug4x source is included in the FluidHP 2.0 zip-file on hpcalc.org if anyone wants to take a peek. Regards Steen
From: Cyrille de Br?bisson on 26 Jan 2006 11:16 hello, on ROT ROT.. try using UNROT, faster and easier to read. yes, locals are easier to read, but much slower (storage part, memory movements to create and destroy the local stack, name matching each time you access a local, need to store in locals (again name matching)... yes, using locals in a sircumspect way can improve speed on COMPLEX algorythms, but this is rare. Normally, you never use Locals for critical loops... No stack diagrams, use your head, this is a great concentration exercise and helps fight brain deadness :-) SysRPL is faster, and nullname Locals in SysRPL are FASTER than stack manipulation (cause there is no need to move memory along when manipulating stack), and are REAdABLE! when used properly: ie in MASD: {{ LocalVar1 LocalVar2 LocalVar3 .... }}(where localVarx is whatever name you want) and then use LocalVarx to replace xGETLAM and LocalVarx! for xPUTLAM ABND when working with the HPTOOLS, use a DEFINE LocalVarx xGETLAM and DEFINE LocalVarx! xPUTLAM to get the readability... regards, cyrille "Steen Schmidt" <sschmidt(a)nospam.dk> wrote in message news:xn0ehozp29cvh6000(a)news.tele.dk... > Michael wrote: > >> for those critical bits. The cool thing, is that in SRPL, null >> locals are quite fast, so you have the best of both, er, triple >> worlds: speed, readability, flexibility ;-) > > That's the first first time I've seen NULLLAMs associated with > readability ;-) > > Here's a small tidbit from the uncompressed source of my port of > FluidHP on hpcalc.org: > > ******** > 7GETLAM %SQRT %SQRT DUP 6PUTLAM 18GETLAM SWAP 21GETLAM OVER > BINT24 GETLAM %* %+ %* %+ 5PUTLAM 17GETLAM 6GETLAM 20GETLAM > OVER BINT23 GETLAM %* %+ %* %+ DUP 4PUTLAM 5GETLAM DUPDUP %* > 22GETLAM 6GETLAM %+ 6GETLAM %* 19GETLAM %+ 4GETLAM %* %4 %* > %+ %SQRT %+ %/ 15GETLAM %+ DUP 2PUTLAM 14GETLAM OVER %+ > 2GETLAM %* 13GETLAM %+ %SQRT %- 1PUTLAM 10GETLAM BINT25 GETLAM > 9GETLAM %- 1GETLAM 10GETLAM %- %* 7GETLAM 9GETLAM %- %/ %+ > ******** > > I use up to 28 NULLLAMs in nested environments, so readability is > pretty much a goner (but code size is down and performance is up, so.. > :-). > > To make a fair point the rest of the code isn't much better - another > tidbit: > > ******** > %0 %0 % 1489.50410795 %0 %1 % 743.07798314 %0 %2 % -97.7083187978 > %0 %12 % 2.47424647057 %0 %18 % -0.63281320016 %0 %24 > % 1.13859521297 %0 %7 %4 %* % -0.478118636486 %0 %10 %4 %* > % 8.52081234315E-3 %1 %0 % 0.937471473779 %1 %2 % 3.35931186049 > %1 %6 % 3.38093556015 %1 %12 % 0.168445396719 %1 %18 > % 0.738757452367 %1 %24 % -0.471287374362 %1 TWENTYFOUR PICK > % 0.150202731397 %1 TWENTYFOUR PICK % -2.17641142198E-3 %2 > %2 % -2.18107553248E-2 %2 %8 % -0.108297844037 %2 %18 > % -4.63333246358E-2 %2 TWELVE PICK % 7.12803519596E-5 %3 %1 > % 1.103283179E-4 %3 %2 % 1.89552483879E-4 %3 %12 % 3.08915411605E-3 > %3 %24 % 1.35555045549E-3 %4 %2 % 2.86402374774E-7 %4 %12 > % -1.07798573575E-5 %4 %18 % -7.64627124548E-5 %4 %24 > ******** > > All the code is like that, I didn't even have to search for special > obfuscated parts of it :-/ I have forgotten how many stack levels are > used, but it's close to hundred. The Debug4x source is included in the > FluidHP 2.0 zip-file on hpcalc.org if anyone wants to take a peek. > > Regards > Steen
From: Steen Schmidt on 26 Jan 2006 11:03 Cyrille de Br?bisson wrote: > when working with the HPTOOLS, use a DEFINE LocalVarx xGETLAM and > DEFINE LocalVarx! xPUTLAM to get the readability... I didn't want to do that in the porting of FluidHP, as I use the same data both within and outside of an inform box - this means that my stored data suddenly gets offset (when I in one environment need to recall NULLLAM 14, I in the next environment need to recall NULLLAM 18 to get to the same data for instance). Defining names for the NULLLAMs would in this case only confuse matters I think. I agree with (and practice) your point of view to make defines to make code more readable - FluidHP was just the result of a macrochallenge between me and TW to get the fastest and smallest port :-) I have thorough comments throughout the source though, so it should at all times be quite easy to determine which NULLLAM contain what data (the comments inlcude elaborate lists of the current data structure in memory). Regards Steen
From: Arnaud Amiel on 26 Jan 2006 12:17 This nulllam with a name in MASD is wonderful for me. One of the best features of the whole metakernel. Thanks Cyrille. Well I guess you did it?? Arnaud
From: timwessman on 26 Jan 2006 12:53
> I agree with (and practice) your point of view to make defines to make > code more readable - FluidHP was just the result of a macrochallenge > between me and TW to get the fastest and smallest port :-) And I was soundly thrashed due to my schedule suddenly getting very busy. Hail the mighty Steen!!! TW (maybe I'll give it a go sometime. . . ;-) |