Infinite Recursion(paulbarry.com)
paulbarry.com
Infinite Recursion
http://paulbarry.com/articles/2009/09/02/infinite-recursion
3 comments
He forgot C! It's not in the language spec, but gcc will optimize tail calls, with -O2 or above.
From: http://stronglytypedblog.blogspot.com/2009/08/scala-tail-rec...
If the recursion is indirect, for example, Scala cannot optimize tail calls, because of the limited JVM instruction set.
Why is that?
If the recursion is indirect, for example, Scala cannot optimize tail calls, because of the limited JVM instruction set.
Why is that?
Apparently the JVM goto instruction can only jump to an address within the same method:
http://java.sun.com/docs/books/jvms/second_edition/html/Inst...
http://java.sun.com/docs/books/jvms/second_edition/html/Inst...
I think the CLR team was able to learn from some of the JVM's shortcomings; this limitation is one example. .NET IL has the TAIL. instruction prefix that allows tail calls. It can be prepended to any call instruction outside of an exception handling block.
There's also the JMP instruction, which is similar and allows the current method's arguments to be passed to another function, discarding the current stack frame and returning to the current method's caller.
There's also the JMP instruction, which is similar and allows the current method's arguments to be passed to another function, discarding the current stack frame and returning to the current method's caller.
Even more interesting is that C# and VB.NET, the two main languages targeting the CLR, do not make use of it. Before posting this though, I did a quick search to make sure they hadn't turned around and implemented it for C# 4 or something, and found this link:
https://connect.microsoft.com/VisualStudio/feedback/ViewFeed...
Apparently there are some performance issues with the TAIL IL instruction, though I wonder if it's really as slow as pushing a new frame on the call stack.
https://connect.microsoft.com/VisualStudio/feedback/ViewFeed...
Apparently there are some performance issues with the TAIL IL instruction, though I wonder if it's really as slow as pushing a new frame on the call stack.
"The same mistake is done on the Computer Language Benchmark Game. There's an Array sorting benchmark where Haskell just blows away even hand-optimized C. The secret is that the benchmark never prints out the contents of the sorted array. The C compiler isn't smart enough to recognize that the array is never actually used anywhere, but Haskell is, and so the entire benchmark basically compiles down to the equivalent of "int main() { return 0; }"."