lab2中的2.5第三个benchmark soplex_sparc运行时挂了……怀疑程序问题?

5 views
Skip to first unread message

杨笛一

unread,
May 28, 2011, 1:26:56 PM5/28/11
to mips-a...@googlegroups.com
求解答!
鹅鹅鹅

谢谢

Qiang Ruixin

unread,
May 29, 2011, 1:56:18 AM5/29/11
to MIPS
同,那个magic break之后就走不下去了
我是先把magic disable之后再跑100_000_000,之后再测的。


On May 29, 1:26 am, 杨笛一 <huyaoyang...@gmail.com> wrote:
> 求解答!
> 鹅鹅鹅
>
> 谢谢

杨笛一

unread,
May 29, 2011, 1:59:27 AM5/29/11
to mips-a...@googlegroups.com
对!
后面3.1的话也遇到一些貌似的问题,我现在的解决方法是把那个
magic-break-enable
然后istc-disable和dstc-disable去掉之后,先让程序预热,然后再清空,拿数据,就对了……
我也不知道这样对不对,但貌似结果蛮正常的。如果没有上述的操作,就要跪了。
求解释。

王敏捷

unread,
May 29, 2011, 6:46:35 AM5/29/11
to mips-a...@googlegroups.com
看了下,貌似和ic中的read penalty有关,如果值为0就不会有问题。而且其他benchmark也都没问题。
 
目前没有什么好的解决方案。推荐qrx的方法。在break point那里magic-break-disable。
 
对于istc和dstc引用user guide中的一段话:
When simulating an instruction cache, g-cache is able to use the Instruction STC tospeed up the simulation and report the number of instruction misses, but it won’t
report the correct number of total transactions.  If you wish to have a correct total
amount of instruction fetches, you need to disable ISTC usage at the command-line
with the istc-disable command.
所以如果不disable掉,会带来一些统计上的误差吧。
 
 

杨笛一

unread,
May 29, 2011, 6:53:23 AM5/29/11
to mips-a...@googlegroups.com
这样啊!!!

谢谢助教~

Shou Heming

unread,
May 29, 2011, 6:57:13 AM5/29/11
to MIPS
其实3.1的问题是:
那个gemm-3.1.c里面有两个MAGIC-BREAK...但是到magic-break之后再运行一下并不会continue XX条语句,
我是c 1了走出这一段代码再预热100_000_000,然后再刷100_000_000条抓的数据...

On May 29, 6:46 pm, 王敏捷 <wmjlyjema...@gmail.com> wrote:
> 看了下,貌似和ic中的read penalty有关,如果值为0就不会有问题。而且其他benchmark也都没问题。
>
> 目前没有什么好的解决方案。推荐qrx的方法。在break point那里magic-break-disable。
>
> 对于istc和dstc引用user guide中的一段话:
>

> *When simulating an instruction cache, g-cache is able to use the


> Instruction STC tospeed up the simulation and report the number of

> instruction misses, but it won't*
> *report the correct number of total transactions. If you wish to have a
> correct total*
> *amount of instruction fetches, you need to disable ISTC usage at the
> command-line*
> *with the istc-disable command.*
>
> 所以如果不disable掉,会带来一些统计上的误差吧。
> * *


>
> 在 2011年5月29日 下午1:59,杨笛一 <huyaoyang...@gmail.com>写道:
>
>
>
>
>
>
>
> > 对!
> > 后面3.1的话也遇到一些貌似的问题,我现在的解决方法是把那个
> > magic-break-enable

> > 然后istc-disable和dstc-disable去掉之后,先让程序预热,然后再清空,拿数据,就对了......


> > 我也不知道这样对不对,但貌似结果蛮正常的。如果没有上述的操作,就要跪了。
> > 求解释。
>

东东

unread,
May 29, 2011, 7:01:33 AM5/29/11
to mips-a...@googlegroups.com
真聪明,赞!

2011/5/29 Shou Heming <sho...@gmail.com>
Reply all
Reply to author
Forward
0 new messages