Forum

I propose this version, more powerful in time of realization (with bigger code, i know). A nice comparison of size vs speed optimizing. Apropos speed, I was able to to make the PB-1000/2000C assembly version slightly faster: @00: ADW $24,#1 LDL $8,$0,L8 ADBL $0,$8,L8 ANC $8,#1 JR NZ,@01 ADBL $8,$0,...

Salut,
the only difference is the ROM type. The 850VS contains a flash ROM instead of the mask ROM in the 850V.

C.Ret a écrit : ↑

29 mai 2017 11:24

I just check your example and I get the correct answer with the last code on my SPARP PC-1211

Yes, your latest code is correct. I've overlooked a small bug in the ported code.
Merci for this nice MPO.

A closer look at the sequence of the distances between the numbers of the f^k(n) sequence shows, that it's in general the summation sequence of round(k/2) with an offset. So we need the difference of the total sum minus the sum below the offset, to get the distance to the offset. f^20(1) = 111 1 + 1...

So, i correct my code to this One-Liner version : 1:"F" INPUT N,K : R=INT (.5+√N , E=INT .5K , F=E-(RR>=N)+(K>2E : PRINT N+KR+EF;"^=";K;N:END Dear C.Ret, I'm really sorry, but apparently there is still a bug in your formula, if I made no mistake while porting it. Please check f^-123(12345)=2536, wh...

So, I restart computations on my SHARP PC-1211 with this code: 1:" " AREAD N:K=1:INPUT "^";K // Saisie de n et facultativement de k 2:R=INT √N,R=R+(√N>.5+R),E=INT .5K,F=E-R+INT √N+(K>2E),F=N+NK+EF // Calculs de R,E puis F 3:"=" PRINT F;" =^";K;N:END // Affichage peut être affiné I've transferred yo...

Sorry, but there seems to be a misunderstanding. I've noticed some wrong results in your examples, that made me curious to find out, if it's possible to have a formula with correct results for any argument. My first attempt was very similar to yours, but I noticed that the results are not correct in...

Calculating f^k(n) for k>0:

d=floor(sqrt(n)*2)-1
if frac(sqrt(n))=0 then d=d-1
n=n+round((k^2+2dk+2k)/4)


Calculating f^k(n) for k<0:

d=floor(sqrt(n)*2)-1
n=n-round((-k^2+2dk+2k)/4)

I have to modify my codes in order to track for perfect square in interim results when using k<0 in sequence. If it's possible to calculate f^k(n) without iterations, it would be more practical to check the result for k<0 with f^-k(result) for plausibility. I wrote "if possible", because I noticed ...

Faire attention cependant, ce code peu donner avec les composées inverses (k négatifs) des résultats aberrants. En particulier dans le cas de détermination du rang d'un entier carré parfait: 25[shft]-1[ENTER] affiche 20.=^-1.25. Ce qui ne signifie pas que 25 est le 20-ième entier non carré parfait....

Shorter version for the FX-180P using the square function:

The algorithm on the FX-180P for f(n): 01 1 02 M+ 03 2 04 Kin+1 05 Kout1 06 Kin+2 07 Kout2 08 x<=M 09 MR Usage example: 2017 Min 1 Kin1 Kin2 P1 And the inverse of f(n): 01 1 02 Kin-1 03 Kout3 04 Kin-2 05 2 06 Kin+3 07 Kout2 08 x>0 09 Kout1 Usage example: 2062 Kin1 Kin2 1 Kin-2 3 Kin3 P1

Ok, I've also tried an assembly language friendly way, using increment and decrement only. Maybe there is a shorter way, but here the algorithm in PASCAL for f(n): readln(n); s:=1; i:=1; repeat inc(n); inc(i,2); inc(s,i); until s>n; writeln(n); And the inverse of f(n): readln(n); s:=pred(n); i:=3; r...

Salut,

If I've understand the problem correcly, n+rnd(sqrt(n)) gives the n-th element of the sequence and n-int(sqrt(n)) gives the position of a non square number in the sequence.

All the pockets or calculators using a LC or RC circuit as oscillator can have a tolerance of some percent. But many
of them using ceramic resonators as oscillator, what is usually precise enough to have no significant speed differences.
The FX-850P and FX-880P using a 1.228 MHz ceramic resonator.