Hi,
a few small simplifications and optimizations to StringConcatFactory:
- refactor makeConcat/makeConcatWithArguments methods to remove the
generateRecipe argument
- fold back-to-back constants into one when parsing recipe, which
allows the prepender logic to be simplified
- do input checking in the parse pass
- get rid of the Recipe and RecipeElement classes by representing the
parsed recipe as a list of Strings which are either null (argument slot)
or not (constant)
- eagerly rebind on setup of those method handles that we'll always
transform (transforming will call rebind() anyway which will now be a no-op)
- simplify filtering logic by picking up the erased type, so that we can
do c == Object.class rather than !c.isPrimitive
Also adding a microbenchmark I used to step by step verify the fixes
having a real and measurable effect[1], while verifying a few percent
improvement on a few existing startup tests[2].
Bug: https://bugs.openjdk.java.net/browse/JDK-8246152
Webrev: http://cr.openjdk.java.net/~redestad/8246152/open.00/
Testing: tier1
Thanks!
/Claes
[1]
java -jar build/linux-x64/images/test/micro/benchmarks.jar
StringConcatFactoryBootstraps -f 2 -w 5 -r 5 -prof gc -jvmArgs=-Xmx1g
Before:
Benchmark Mode Cnt Score Error Units
StringConcatFactoryBootstraps.makeConcatWithConstants avgt 10 3046.855 ±
62.981 ns/op
StringConcatFactoryBootstraps.makeConcatWithConstants:·gc.alloc.rate.norm
avgt 10 4596.214 ± 121.121 B/op
After:
Benchmark Mode Cnt Score Error Units
StringConcatFactoryBootstraps.makeConcatWithConstants avgt 10 2764.634 ±
56.244 ns/op
StringConcatFactoryBootstraps.makeConcatWithConstants:·gc.alloc.rate.norm
avgt 10 4136.187 ± 0.013 B/op
[2]
http://cr.openjdk.java.net/~redestad/scratch/erase_scf_types.00/ObjStringCombos.java
Before:
3,661,949,568 instructions # 1.09 insn per
cycle ( +- 1.04% )
728,280,499 branches # 492.540 M/sec
( +- 1.06% )
30,496,964 branch-misses # 4.19% of all
branches ( +- 0.93% )
0.301502707 seconds time elapsed
( +- 1.04% )
After:
3,485,679,714 instructions # 1.09 insn per
cycle ( +- 0.90% )
692,247,529 branches # 491.822 M/sec
( +- 0.92% )
28,939,534 branch-misses # 4.18% of all
branches ( +- 0.90% )
0.291642667 seconds time elapsed
( +- 0.93% )
