MapD建表测试(使用复制已有表的形式)
Posted 老装老庄
tags:
篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了MapD建表测试(使用复制已有表的形式)相关的知识,希望对你有一定的参考价值。
在GPU上的测试(4个Tesla P100-SXM2-16GB)
1.1万行测试:
CREATE TABLE flights_2008_new AS (SELECT * FROM flights_2008_10k);
select count(*) from flights_2008_new;
返回:
EXPR$0
10000
drop table flights_2008_test;
2.20万行测试:
CREATE TABLE flights_2008_new AS (SELECT * FROM flights_2008_200k);
select count(*) from flights_2008_new;
返回:
EXPR$0
200000
drop table flights_2008_new;
3.700万行测试:
CREATE TABLE flights_2008_new AS (SELECT * FROM flights_2008_7M);
报错,GPU内存溢出:
E0627 18:37:56.306819 3148 MapDHandler.cpp:2384] Exception: Query couldn‘t keep the entire working set of columns in GPU memory
Exception: Query couldn‘t keep the entire working set of columns in GPU memory
在CPU上的测试(ppc64le 128核):
1.1万行测试:
CREATE TABLE flights_2008_new AS (SELECT * FROM flights_2008_10k);
select count(*) from flights_2008_new;
返回:
EXPR$0
10000
drop table flights_2008_test;
2.20万行测试:
CREATE TABLE flights_2008_new AS (SELECT * FROM flights_2008_200k);
select count(*) from flights_2008_new;
返回:
EXPR$0
200000
drop table flights_2008_new;
3.700万行测试:
select count(*) from flights_2008_new;
返回:
EXPR$0
7009728
剩下的明天测试
以上是关于MapD建表测试(使用复制已有表的形式)的主要内容,如果未能解决你的问题,请参考以下文章