Huge performance difference (26x faster) when compiling for 32 and 64 bits

by

I can reproduce this on 4.5.2. No RyuJIT here. Both x86 and x64 disassemblies look reasonable. Range checks and so on are the same. The same basic structure. No loop unrolling.

x86 uses a different set of float instructions. The performance of these instructions seems to be comparable with the x64 instructions except for the division:

  1. The 32 bit x87 float instructions use 10 byte precision internally.
  2. Extended precision division is super slow.

The division operation makes the 32 bit version extremely slow. Uncommenting the division equalizes performance to a large degree (32 bit down from 430ms to 3.25ms).

Peter Cordes points out that the instruction latencies of the two floating point units are not that dissimilar. Maybe some of the intermediate results are denormalized numbers or NaN. These might trigger a slow path in one of the units. Or, maybe the values diverge between the two implementations because of 10 byte vs. 8 byte float precision.

Peter Cordes also points out that all intermediate results are NaN… Removing this problem (valueList.Add(i + 1) so that no divisor is zero) mostly equalizes the results. Apparently, the 32 bit code does not like NaN operands at all. Let’s print some intermediate values: if (i % 1000 == 0) Console.WriteLine(result);. This confirms that the data is now sane.

When benchmarking you need to benchmark a realistic workload. But who would have thought that an innocent division can mess up your benchmark?!

Try simply summing the numbers to get a better benchmark.

Division and modulo are always very slow. If you modify the BCL Dictionary code to simply not use the modulo operator to compute the bucket index performance measurable improves. This is how slow division is.

Here’s the 32 bit code:

enter image description here

64 bit code (same structure, fast division):

enter image description here

This is not vectorized despite SSE instructions being used.

More Related Contents:

  1. Make first letter of a string upper case (with maximum performance)
  2. Generating permutations of a set (most efficiently)
  3. How to pass values across the pages in ASP.net without using Session
  4. Is String.Contains() faster than String.IndexOf()?
  5. For vs. Linq – Performance vs. Future
  6. What is quicker, switch on string or elseif on type?
  7. Most efficient way to insert Rows into MySQL Database
  8. What is the real overhead of try/catch in C#?
  9. Why does the c# compiler emit Activator.CreateInstance when calling new in with a generic type with a new() constraint?
  10. File.Copy vs. Manual FileStream.Write For Copying File
  11. Performance of nested yield in a tree
  12. String Interpolation vs String.Format
  13. How performant is StackFrame?
  14. Single MSI to install correct 32 or 64 bit c# application
  15. True Unsafe Code Performance
  16. Why is Parallel.ForEach much faster then AsParallel().ForAll() even though MSDN suggests otherwise?
  17. Does String.GetHashCode consider the full string or only part of it?
  18. try catch performance
  19. Fastest way to check if string contains only digits in C#
  20. LINQ performance FAQ
  21. Is it better to execute many sql commands with one connection, or reconnect every time?
  22. Is the conditional operator slow?
  23. Options for using System.Data.SQLite in a 32bit and 64bit C# world
  24. Why is my C# program faster in a profiler?
  25. SQL Server: Dynamic where-clause
  26. C# FindAll VS Where Speed
  27. A fast hash function for string in C#
  28. Does any one know of a faster method to do String.Split()?
  29. Why does Math.Exp give different results between 32-bit and 64-bit, with same input, same hardware
  30. TableLayoutPanel responds very slowly to events

Leave a Comment