Longjumps considered inexpensive...until they aren't

[John Rose]

On May 28, 2008, at 2:12 AM, Charles Oliver Nutter wrote:

> What else can we explore? What tools should we use to investigate?

You are already using LogCompilation, and that is the best way to  
understand the inlining decisions of the JIT.  With these special  
patterns, you are perhaps verging on a need (often repeated by  
customers) to give advice to the JIT about inlining and other  
compilation decisions.  If you could give simple advice to the JIT  
about what should be inlined, what would you tell it?

The PrintAssembly feature (new in product builds) is your best view  
of the actual machine code.
   http://wikis.sun.com/display/HotSpotInternals/PrintAssembly

This is the best way to find what the bottleneck really is, instead  
of treating the JIT as a black box.

The catch with PrintAssembly is that you have to build (or borrow)  
the disassembly plugin.  They are not yet widely available.  You'd be  
an early adopter...

I see two paths here:

1. Force the JIT to inline the code patterns you care about.  (One  
logical Ruby method should be compiled in one JIT compilation task.)

2. Make the JVM process the exceptions you care about more  
efficiently, in the out-of-line case.

Both should be investigated.  Both are probably good graduate  
theses...  Anyone?

The easiest one to try first is #1, since we already have a compiler  
oracle, and could quickly wire up the right guidance statements (in a  
JVM tweak), if we could figure out what Ruby wants to say to the JVM  
about inlining.

-- John