50g User RPL Micro-Challenge: Fraction to a Power

[Joe Horn]

The 50g in exact mode takes a very long time to raise a fraction (ratio of integers) to integer powers. Example: 11/12 to the 115th power:

11/12
115
<< ^ EVAL >>

takes about 40 seconds to return the exact answer.

Micro-Challenge: Generate that same answer in less than 1 second, using only vanilla User RPL (no System RPL or SYSEVALs or FLASHEVALs etc). The fastest program wins.

[Gerald H]

My suggestion:

0.2197s

Code:

« SWAP OBJ→ 5.
ROLLD DROP SWAP
PICK3 ^ SWAP ROT ^
ROT 3. →ALG
»

[Werner]

The date on my IPOW.RPL file reads 2006 ;-)
Edit: but it can't get it done in less than 1 second, so Gerald's idea is the way to go!
Werner

[John Keith]

While Gerald H's program is probably unbeatable, I'm guessing this

Code:

\<<SWAP OBJ\-> DROP2 UNROT OVER ^ UNROT ^ / \>>

is closer to what Joe has in mind, since it takes around 945ms on my HP 50.

John

[Gerald H]

With John's better stack manipulation:

: .2186

Code:

« SWAP OBJ→ 5.
ROLLD DROP UNROT
OVER ^ UNROT ^ ROT
3. →ALG
»

[Gerald H]

I don't know where the face came from - I entered ":s:" without quotes.

[Gerson W. Barbosa]

(10-09-2017 12:48 PM)Gerald H Wrote:  I don't know where the face came from - I entered ":s:" without quotes.

Just check “Disable Smilies” in the Post Options.

:s:

[Gerald H]

Faster:

:s: .1831

Code:

« SWAP COMP→ 5.
ROLLD 5. ROLLD
UNROT OVER ^ UNROT
^ 4. ROLL 4. ROLL
→ALG
»

[Gerald H]

Thank you, Gerson, it worked.

[Gerald H]

Faster?

:s: .1825

Code:

« SWAP COMP→ 5.
ROLLD 4. ROLLD
UNROT OVER ^ UNROT
^ ROT 4. ROLL →ALG
»

[Gilles59]

Code:

 << SWAP FXND UNROT OVER ^ UNROT ^ / >>

0.9282s

Gerald solution is very fast ! (but needs development Library ;D ) I wonder why it s so fast ?

[Gerald H]

It's the division that's so slow.

[Gerald H]

This version is about as fast as my last try:

Code:

« SWAP FXND UNROT
OVER ^ UNROT ^ { /
} OBJ→ DROP 3. →ALG
»

[Gilles59]

(10-09-2017 03:30 PM)Gerald H Wrote:  It's the division that's so slow.

Because a direct '/' tries to simplify the fraction?
and ->ALG does not try ?

[Gerald H]

Average over 100 runs

.182s

Code:

« SWAP COMP→ DROP 4
ROLLD UNROT OVER ^
UNROT ^ ROT 3. →ALG
»

[Arno K]

Replacing ->ALG by EVAL increases the time my program Needs (.2163s) to .9723s.
Arno

[Joe Horn]

(10-09-2017 12:15 PM)John Keith Wrote:  While Gerald H's program is probably unbeatable, I'm guessing this

Code:

\<<SWAP OBJ\-> DROP2 UNROT OVER ^ UNROT ^ / \>>

is closer to what Joe has in mind, since it takes around 945ms on my HP 50.

John

I had these two in mind (both contain 8 commands):

<< SWAP OBJ-> DROP ->LIST SWAP ^ EVAL / >>

<< SWAP FXND UNROT OVER ^ UNROT ^ / >>

... but both suffer from the division at the end (yes, Gilles, division simplifies fractions, which isn't needed here). That's why Gerald's program is so fast.

[brickviking]

(10-09-2017 10:26 AM)Joe Horn Wrote:  The 50g in exact mode takes a very long time to raise a fraction (ratio of integers) to integer powers. Example: 11/12 to the 115th power:

11/12
115
<< ^ EVAL >>

takes about 40 seconds to return the exact answer.

Micro-Challenge: Generate that same answer in less than 1 second, using only vanilla User RPL (no System RPL or SYSEVALs or FLASHEVALs etc). The fastest program wins.

On my calculator, putting that onto the stack and using this: [RSHIFT] →NUM gives me the answer in less than a second.
(Although in saying that, if I hit EVAL instead, that takes a while, and gives me a fraction with really REALLY long numbers.)

I don't know if my answer counts as RPL or not.

(whistles innocently)

(Post 113)

[ttw]

I tried

<< (11^115)/(12^115) EVAL>> TEVAL

which took 2.5 sec. Seems like a lack of optimization in the expression parser.

[Joe Horn]

(10-09-2017 10:37 PM)brickviking Wrote:  

(10-09-2017 10:26 AM)Joe Horn Wrote:  The 50g in exact mode takes a very long time to raise a fraction (ratio of integers) to integer powers. Example: 11/12 to the 115th power:

11/12
115
<< ^ EVAL >>

takes about 40 seconds to return the exact answer.

Micro-Challenge: Generate that same answer in less than 1 second, using only vanilla User RPL (no System RPL or SYSEVALs or FLASHEVALs etc). The fastest program wins.

On my calculator, putting that onto the stack and using this: [RSHIFT] →NUM gives me the answer in less than a second.

Not the exact answer, which is what was specified.