Keith Randall
dd24b1098a
cmd/compile: improve tighten pass
...
Move a value to the block which is the lowest common ancestor in the
dominator tree of all of its uses. Make sure not to move a value into a
loop.
Makes the tighten pass on average (across go1 benchmarks) 40% slower.
Still not a big contributor to overall compile time.
Binary size is just a tad smaller.
name old time/op new time/op delta
BinaryTree17-12 2.77s ± 9% 2.76s ± 9% ~ (p=0.878 n=8+8)
Fannkuch11-12 2.75s ± 1% 2.74s ± 1% ~ (p=0.232 n=8+7)
FmtFprintfEmpty-12 48.9ns ± 9% 47.7ns ± 0% ~ (p=0.431 n=8+8)
FmtFprintfString-12 143ns ± 8% 142ns ± 1% ~ (p=0.257 n=8+7)
FmtFprintfInt-12 123ns ± 1% 122ns ± 1% -1.04% (p=0.026 n=7+8)
FmtFprintfIntInt-12 195ns ± 7% 185ns ± 0% -5.32% (p=0.000 n=8+8)
FmtFprintfPrefixedInt-12 194ns ± 4% 195ns ± 0% +0.81% (p=0.015 n=7+7)
FmtFprintfFloat-12 267ns ± 0% 268ns ± 0% +0.37% (p=0.001 n=7+6)
FmtManyArgs-12 800ns ± 0% 762ns ± 1% -4.78% (p=0.000 n=8+8)
GobDecode-12 7.67ms ± 2% 7.60ms ± 2% ~ (p=0.234 n=8+8)
GobEncode-12 6.55ms ± 0% 6.57ms ± 1% ~ (p=0.336 n=7+8)
Gzip-12 237ms ± 0% 238ms ± 0% +0.40% (p=0.017 n=7+7)
Gunzip-12 40.8ms ± 0% 40.2ms ± 0% -1.52% (p=0.000 n=7+8)
HTTPClientServer-12 208µs ± 3% 209µs ± 3% ~ (p=0.955 n=8+7)
JSONEncode-12 16.2ms ± 1% 17.2ms ±11% +5.80% (p=0.001 n=7+8)
JSONDecode-12 57.3ms ±12% 55.5ms ± 3% ~ (p=0.867 n=8+7)
Mandelbrot200-12 4.68ms ± 6% 4.46ms ± 1% ~ (p=0.442 n=8+8)
GoParse-12 4.27ms ±44% 3.42ms ± 1% -19.95% (p=0.005 n=8+8)
RegexpMatchEasy0_32-12 75.1ns ± 0% 75.8ns ± 1% +0.99% (p=0.002 n=7+7)
RegexpMatchEasy0_1K-12 963ns ± 0% 1021ns ± 6% +5.98% (p=0.001 n=7+7)
RegexpMatchEasy1_32-12 72.4ns ±11% 70.8ns ± 1% ~ (p=0.368 n=8+8)
RegexpMatchEasy1_1K-12 394ns ± 1% 399ns ± 0% +1.23% (p=0.000 n=8+7)
RegexpMatchMedium_32-12 114ns ± 0% 115ns ± 1% +0.63% (p=0.021 n=7+7)
RegexpMatchMedium_1K-12 35.9µs ± 0% 37.6µs ± 1% +4.72% (p=0.000 n=7+8)
RegexpMatchHard_32-12 1.93µs ± 2% 1.91µs ± 0% -0.91% (p=0.001 n=7+7)
RegexpMatchHard_1K-12 60.2µs ± 3% 61.2µs ±10% ~ (p=0.442 n=8+8)
Revcomp-12 404ms ± 1% 406ms ± 1% ~ (p=0.054 n=8+7)
Template-12 64.6ms ± 1% 63.5ms ± 1% -1.66% (p=0.000 n=8+8)
TimeParse-12 347ns ± 8% 309ns ± 0% -11.13% (p=0.000 n=8+7)
TimeFormat-12 343ns ± 4% 331ns ± 0% -3.34% (p=0.000 n=8+7)
Change-Id: Id6da1239ddd4d0cb074ff29cffb06302d1c6d08f
Reviewed-on: https://go-review.googlesource.com/28712
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
2016-09-20 22:49:48 +00:00
Keith Randall
4fa050024f
cmd/compile: enable constant-time CFG editing
...
Provide indexes along with block pointers for Preds
and Succs arrays. This allows us to splice edges in
and out of those arrays in constant time.
Fixes worst-case O(n^2) behavior in deadcode and fuse.
benchmark old ns/op new ns/op delta
BenchmarkFuse1-8 2065 2057 -0.39%
BenchmarkFuse10-8 9408 9073 -3.56%
BenchmarkFuse100-8 105238 76277 -27.52%
BenchmarkFuse1000-8 3982562 1026750 -74.22%
BenchmarkFuse10000-8 301220329 12824005 -95.74%
BenchmarkDeadCode1-8 1588 1566 -1.39%
BenchmarkDeadCode10-8 4333 4250 -1.92%
BenchmarkDeadCode100-8 32031 32574 +1.70%
BenchmarkDeadCode1000-8 590407 468275 -20.69%
BenchmarkDeadCode10000-8 17822890 5000818 -71.94%
BenchmarkDeadCode100000-8 1388706640 78021127 -94.38%
BenchmarkDeadCode200000-8 5372518479 168598762 -96.86%
Change-Id: Iccabdbb9343fd1c921ba07bbf673330a1c36ee17
Reviewed-on: https://go-review.googlesource.com/22589
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-05-05 15:58:59 +00:00
Keith Randall
3c26c0db39
[dev.ssa] cmd/compile: short-circuit empty blocks
...
Empty blocks are introduced to remove critical edges.
After regalloc, we can remove any of the added blocks
that are still empty.
Change-Id: I0b40e95ac3a6cc1e632a479443479532b6c5ccd9
Reviewed-on: https://go-review.googlesource.com/18833
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
2016-01-22 22:12:12 +00:00
