doesn't say how to contribute

[lulcat]

Your results are very iffy vs. local runs here. By that I don't mean the run times as I am using 24T on a different CPU (Ryzen 7th gen);
but intra results on these so called program test results.

Anyway, spectral-norm, 1.d , for multi threading, a simple edit on a line should make it MT 1-m.d:
(fs, don't even know how to make quoted code here :p)

--- 1.d.orig	2025-03-24 09:45:54.198352668 +0100
+++ 1.d	2025-03-24 09:47:15.757138282 +0100
@@ -34,7 +34,7 @@
 
 void times(double[] v, double[] u, int n, bool reverse)
 {
-    for (auto i = 0; i < n; i++)
+    foreach(i;parallel(iota(n)))
     {
         auto sum = 0.0;
         for (auto j = 0; j < n; j++)

times here for 8000:

c++ im = 158-160ms
c m = 160-162ms
d m = 236ms (+50%)

I know from experience this can be brought down into about 200ms most likely, but then again, changing just one line says somehting about production time vs. runtime :) over far more complex changes in said other languages.

[iamkisly]

doesn't say how to contribute

#457 (comment)

[lulcat]

ye, worth noting when I did another test, binarytrees, java which is so 'fast' vs. D again, ST. I use for all c++ , c , d the following flags for my system: -march=x86-64-v3 -O3 for java I didnd[' tknow how to optimise but understand it seems JIT will auto optimise?

Anyway, for smaller sets, again the website paints an awful picture... java here took 234ms for 15 or so? howevcer had a very flat timeline on the exponential curve up towards 20 (15, 18 both were at 235 ms. that's nice).

D on the other hand, took only for 15 or so, like c++ iirc or so, a fraction like 20-30ms? but ion the other hand much quickler went up to 1100ms on 18 or 20. but again this is not at all how the site portrays results (to the the point it seems like one has ulterior motives here!).

[cyrusmsk]

I know from experience this can be brought down into about 200ms most likely, but then again, changing just one line says somehting about production time vs. runtime :) over far more complex changes in said other languages.

Several solutions could be improved actually.
But the repo author is not very active with merging PRs