Palindromic Numbers Consisting of the Product of Two 3-digit Numbers

[altenbach]

---

@Oligarlicky wrote:

I'm noticing some weird stuff depending on how I benchmark. If I stick it all in a for loop I can hit ~2 ns/ itteration. To be fair, I'm only doing 256 palindrome checks in my search (everything else gets filtered out).

---

Watch out for folding artifacts. Hard to tell without seing the code?

Try to caclulate how many clock cycles it takes per iteration. 😄 (e.g. 2ns on a 4GHz prcessor is only 8 clock cycles)

[jcarmody]

---

@Oligarlicky wrote:

[...] all 6 digit palindromes are multiples of 11 [...]

---

That got me down to 0.3 mS.  Thanks. 

---

@Oligarlicky wrote:

I'm noticing some weird stuff depending on how I benchmark. If I stick it all in a for loop I can hit ~2 ns/ itteration. To be fair, I'm only doing 256 palindrome checks in my search (everything else gets filtered out).

---

 I checked 1,388. 😄

[altenbach]

---

altenbach wrote: OK, if I include that test, mine drops to 0.12ms (Still using string compare, though)  😄

---

No longer doing the "11" check, but average time after some optimization is now 0.035ms (35us).

[karryli]

Thanks so much!

[altenbach]

Let us know when the assignment is over so we can compare our versions. Maybe we can all learn something... 😄

Often combining the various "tricks" by the top performers allows the design of new code that is faster than all of them.

[karryli]

It was over last Thurseday. I actually tried my best to understand your codes and other top performers' ones by asking my TA and instructors for Labview workshop on campus. Hopefully, I can progress a lot faster with you guys' help.

[altenbach]

For various reasons I love "code miniatures", code that achieves the goal with the smallest amount of code.

For example, my ~35µs code only contains the following elements (+ controls/indicators, the benchmark skeleton, and  a few diagram constants that are not shown ;)).

Can you do it? Can you do it with less?

The code needs to solve the problem from first principles (i.e. you cannot just use an indicator wired to a diagram constant of the solution 🐵

(Note that no additional copies of these 13 functions can be added. e.g. you must use exactly two subtraction functions (or less if you dare!))

 

[Oligarlicky]

One while loop? Madness

Square? That's gotta be reducing the search space somehow

Also, I couldn't use the min max because I needed the boolean for my case structure.

Surely constant folding is allowed. It's not my fault the compiler is so aggresive 😉

[altenbach]

---

@Oligarlicky wrote:

Surely constant folding is allowed. It's not my fault the compiler is so aggresive 😉

---

My code does not do constant folding. e.g. if I replace any of the diagram constant wired to the outer loop with a control, the speed is about the same.

In any case, we could continue the technical discussion in this new, more recreational thread instead. 😄