We might want to clean up the String.hashCode() implementation.

Let the patch speak for itself:
 http://cr.openjdk.java.net/~shade/8058643/webrev.01/

It improves the readability, and also does not affect performance much:
 http://cr.openjdk.java.net/~shade/8058643/StringHashCode.java

The improvement in nonCached case seems to be due to slightly better code generation, if one can closely examine the assembly:
 http://cr.openjdk.java.net/~shade/8058643/baseline.perfasm
 http://cr.openjdk.java.net/~shade/8058643/patched.perfasm

It seems to be because we have one less branch checking for value.length.