Mirror of forum.nim-lang.org

23 :: Wow I am impressed

• adrianveith (orginal) [2012-05-28T15:59:37+02:00] view original

  my first Nimrod program runs circles around D

  the implementation of my Lazy-Btree outperforms the same algorithm in D by a magnitude (all build as release code)

  BTREE - inserting 1_000_000 string, int pairs
  
  DMD. 26926 ms (first run)  - 7140 ms (second run with GC turned off)
  Nimrod: 1203 ms
  
  BTREE - searching 1_000_000 string, int pairs
  
  DMD: 1650 ms
  Nimrod: 484 ms
  
  and this is even quite good compared to the internal HASH in D and Nimrod
  
  HASH - inserting 1_000_000 string, int pairs
  
  DMD: 1887 ms
  Nimrod: 1266 ms
  
  HASH - searching 1_000_000 string, int pairs
  
  DMD: 1309 ms
  Nimrod: 468 ms

---

• llm (orginal) [2012-05-28T19:14:06+02:00] view original

  but didn't use Nimrod gcc as its backend? so you should better compare gdc (dmd-frontend with gcc backend) everyone knows that dmd codegen isn't good as gcc's

---

• adrianveith (orginal) [2012-05-29T09:17:31+02:00] view original

  I did this test with gdc and ldc some time ago, and performance was not much better than dmd (in this special test case). But currently the source does not compile with neither gdc nor ldc.

---

• Veritas (orginal) [2015-01-22T02:24:34+01:00] view original

  Is the source code for this benchmark available?

---

• fadg44a3w4fe (orginal) [2015-01-22T03:04:04+01:00] view original

  @Veritas

  This is quite old, so it's likely it's no longer applicable (D has probably gotten much faster). Regardless, some research turned up https://github.com/AdrianV/nimrod-tools/blob/master/lazybtree.nim, though I am unable to find the D benchmark.

---

• Veritas (orginal) [2015-01-22T04:38:06+01:00] view original

  @fadg44a3w4fe, Thanks!

  I should have thought to check the dates. I wonder why it was appearing in the recent forum discussions even prior to my post.

---