FFT Multiplication (HP-48G/GX/G+)

[Gerson W. Barbosa]

(02-14-2014 10:37 PM)Raymond Del Tondo Wrote:  

Code:


* Interface  :    [%] [%]'  -->  [%]''

Hello, Raymond!

On my HP-48GX (and on the emulator) things go like this: [%] [%]' --> ENTER --> [%]". Is this behavior normal?

Here is the running times table, times in seconds:

Code:


# elements     HP-71B     hp 49g      HP-48GX     hp 50g   48GX(SysRPL)
     16         24.11      10.42         9.96       5.57        4.4  (*)
     32         53.41      22.58        21.91      11.90        9.2
     64        116.74      47.86        47.64      25.16       20.6
    128        269.89     110.14       106.45      47.54       43.6
    256          --       255.87       254.00     104.42       95.4

(*) SysRPL times have been measured with a chronometer because I wasn't able to do it programmatically.

The HP-48GX times have decreased because now I have 54 K RAM free (previously I had only 11 K). The 49/50g programs are faster because I am using ...DOLIST... instead of HADAMARD:

Code:


%%HP: T(3)A(R)F(.);
\<< DUP2 SIZE OBJ\-> DROP SWAP SIZE OBJ\-> DROP DUP2 MAX UNROT + 4. ROLLD LN 2. LN / DUP FP NOT NOT SWAP IP + 2. SWAP ^ DUP + DUP ROT SWAP 1. \->LIST RDM UNROT 1. \->LIST RDM FFT OBJ\-> 1. GET \->LIST SWAP FFT OBJ\-> 1. GET \->LIST
  \<< *
  \>> DOLIST OBJ\-> 1. \->LIST \->ARRY IFFT 1. ROT SUB RE 0. RND DUP SIZE OBJ\-> DROP 1. - 1.
  FOR i i GETI 10000. MOD PICK3 ROT DUP UNROT GET ROT + PUT i GETI 10000. / IP PICK3 ROT GET 10000. / IP + i SWAP PUTI OVER OVER GET 10000. MOD PUT -1.
  STEP
\>>

The previous version using HADAMARD is more compact (299.5 bytes), but gets slower as the arrays grow larger:

Code:


# elements      hp 49g     hp 50g
     16          11.45       5.35
     32          25.04      12.45
     64          54.01      26.47
    128         131.35      65.35
    256         367.35     148.84

Cheers,

Gerson.