Commit 09decf56 authored by A. Unique TensorFlower's avatar A. Unique TensorFlower Committed by TensorFlower Gardener
Browse files

Microoptimizations of graph construction code.

Before:
Run on *********** (72 X 2993 MHz CPUs); 2018-12-13T16:09:43.471855971-08:00
CPU: Intel Skylake Xeon with HyperThreading (36 cores) dL1:32KB dL2:1024KB dL3:24MB
Benchmark                 Time(ns)        CPU(ns)     Iterations
----------------------------------------------------------------
BM_GraphCreation/10/2        67142          86268           8252
BM_GraphCreation/64/2       138640         163264           4262
BM_GraphCreation/512/2      801036         837092            838
BM_GraphCreation/4k/2      7670132        7719032             89
BM_GraphCreation/32k/2    87954443       88133128              8
BM_GraphCreation/10/4        85895         106133           6589
BM_GraphCreation/64/4       176924         202943           3445
BM_GraphCreation/512/4     1092235        1124801            620
BM_GraphCreation/4k/4     10167172       10242199             68
BM_GraphCreation/32k/4   116535329      116863022              6
BM_GraphCreation/10/8       128276         152347           4595
BM_GraphCreation/64/8       290808         322147           2167
BM_GraphCreation/512/8     1995712        2040134            349
BM_GraphCreation/4k/8     17648175       17725397             39
BM_GraphCreation/32k/8   201791945      202232200              3
BM_GraphCreation/10/16      212183         240520           2909
BM_GraphCreation/64/16      474982         506036           1000
BM_GraphCreation/512/16    3590180        3641964            195
BM_GraphCreation/4k/16    32178292       32265093             22
BM_GraphCreation/32k/16  359809818      360593206              2

After:
Run on *********** (72 X 2993 MHz CPUs); 2018-12-13T16:48:26.030782518-08:00
CPU: Intel Skylake Xeon with HyperThreading (36 cores) dL1:32KB dL2:1024KB dL3:24MB
Benchmark                 Time(ns)        CPU(ns)     Iterations
----------------------------------------------------------------
BM_GraphCreation/10/2        65638          84729           8276
BM_GraphCreation/64/2       130192         154173           4579
BM_GraphCreation/512/2      766354         802899            881
BM_GraphCreation/4k/2      6966973        7019842             98
BM_GraphCreation/32k/2    82443771       82643748              8
BM_GraphCreation/10/4        82697         102636           6743
BM_GraphCreation/64/4       171184         197236           3574
BM_GraphCreation/512/4     1000612        1030750            676
BM_GraphCreation/4k/4      9268842        9346867             74
BM_GraphCreation/32k/4   110080002      110330854              7
BM_GraphCreation/10/8       161076         181417           4764
BM_GraphCreation/64/8       300977         331782           2081
BM_GraphCreation/512/8     1781437        1829938            387
BM_GraphCreation/4k/8     16062834       16148914             44
BM_GraphCreation/32k/8   188352170      188727906              4
BM_GraphCreation/10/16      201874         229188           3049
BM_GraphCreation/64/16      445487         479042           1462
BM_GraphCreation/512/16    3173224        3224053            218
BM_GraphCreation/4k/16    29365146       29457557             24
BM_GraphCreation/32k/16  326978055      327510864              2
PiperOrigin-RevId: 225466082
parent 2a675156
Loading
Loading
Loading
Loading
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Please to comment