forums.silverfrost.com Forum Index forums.silverfrost.com
Welcome to the Silverfrost forums
 
 FAQFAQ   SearchSearch   MemberlistMemberlist   UsergroupsUsergroups   RegisterRegister 
 ProfileProfile   Log in to check your private messagesLog in to check your private messages   Log inLog in 

Search found 1953 matches
forums.silverfrost.com Forum Index
Author Message
  Topic: Calculating the outer vector product
JohnCampbell

Replies: 21
Views: 299

PostForum: Support   Posted: Fri Jun 22, 2018 12:42 pm   Subject:
Paul,

Your suggestion of FTN95 may be treating SPREAD as "elemental" is certainly consistent with the timing performance. I hope it can be fixed.

Imagine if Dot_Product had the same pr ...
  Topic: Calculating the outer vector product
JohnCampbell

Replies: 21
Views: 299

PostForum: Support   Posted: Fri Jun 22, 2018 2:09 am   Subject:
Paul,

As I posted yesterday: See the linked test program that shows SPREAD is very slow.

https://www.dropbox.com/s/4r5kzn5onefvkka/kl2.f90?dl=0

Download it and run this test. I put write ...
  Topic: Calculating the outer vector product
JohnCampbell

Replies: 21
Views: 299

PostForum: Support   Posted: Thu Jun 21, 2018 12:05 pm   Subject:
The link I provided demonstrates a number of loop alternatives, including array syntax instead of the inner loop, which was the fastest option with FTN95.
As Klaus reported, it also shows that FTN95 ...
  Topic: Calculating the outer vector product
JohnCampbell

Replies: 21
Views: 299

PostForum: Support   Posted: Thu Jun 21, 2018 7:57 am   Subject: Re:
FTN95 does require an explicit call to MATMUL and this seems right to me. Paul, wouldn't this be a different calculation ?

oops ! The following adaptation still has a delay:
Call elapse_Tim ...
  Topic: Calculating the outer vector product
JohnCampbell

Replies: 21
Views: 299

PostForum: Support   Posted: Thu Jun 21, 2018 6:34 am   Subject:
Klaus,

I would replace CPU_Time with Elapse_Time, which calls System_Clock or RDTSC_VAL@(). You won't get a meaningful result with CPU_Time for your first test.

I am trying m=n=1000, which certa ...
  Topic: Start using debuggers, people
JohnCampbell

Replies: 50
Views: 3628

PostForum: 64-bit   Posted: Sun Jun 03, 2018 6:52 am   Subject: Re:
To get a proof that /undef is more powerful then "implicit none" take any large code made by someone which never used /undef but used only "implicit none" and make a fun of its au ...
  Topic: Start using debuggers, people
JohnCampbell

Replies: 50
Views: 3628

PostForum: 64-bit   Posted: Sun Jun 03, 2018 3:25 am   Subject:
Dan,

Are you going to make the same mistake 100 times a day for the rest of your life ? ...
Imaging how much time you are wasting having a coffee each day; and all the time you have spent moving a ...
  Topic: Stack Size
JohnCampbell

Replies: 16
Views: 1010

PostForum: 64-bit   Posted: Fri May 25, 2018 11:38 pm   Subject:
davidb & mecej4,

I have now reread your posts on this thread. To me, what you have written is just excuses for what is a very poor/lazy system design.
If the stack is to overflow, there shoul ...
  Topic: changing the stack at runtime?
JohnCampbell

Replies: 5
Views: 249

PostForum: Support   Posted: Fri May 25, 2018 11:16 pm   Subject:
Can we have a new page in ftn95.chm something like "Stack size for dummies", or more "How to manage the stack"

We need actual examples of:
* how to reset the stack when using S ...
  Topic: Side by side versions?
JohnCampbell

Replies: 10
Views: 754

PostForum: General   Posted: Mon May 14, 2018 6:49 am   Subject:
John,

%1 is the first option provided to the batch file.
So assuming the batch file is called set_ftn95.bat
I want to select the version in directory ftn95_8.20F
I would issue the command " ...
  Topic: Error Argument of ichar at (1) must be of length one
JohnCampbell

Replies: 8
Views: 460

PostForum: Support   Posted: Sun May 13, 2018 12:45 pm   Subject:
The use of ICHAR appears to test if the first character of the string read starts with a number. An alternative could be to read the string as a number and use iostat=iostat or test if the character f ...
  Topic: Why we have to declare random@ ?
JohnCampbell

Replies: 15
Views: 823

PostForum: Support   Posted: Wed May 09, 2018 1:21 am   Subject:
Dan,

You should review this above example of using inconsistent kind definitions for reals.
Looks like another piece of devilry waiting to happen when converting to /64
  Topic: Why we have to declare random@ ?
JohnCampbell

Replies: 15
Views: 823

PostForum: Support   Posted: Tue May 08, 2018 3:25 pm   Subject:
Paul,

Thanks for considering these "special" functions.

I am not sure how extensive a longer list could be, but could the following be included:
real*10 function CPU_CLOCK@()
integer ...
  Topic: Why we have to declare random@ ?
JohnCampbell

Replies: 15
Views: 823

PostForum: Support   Posted: Mon May 07, 2018 7:23 am   Subject: Re:
RANDOM@ ... which clashes with the documented value of 2 in FTN95.HLP.
I was struggling to find where the kind for random@ is now documented. I finally found it listed as RANDOM.

The following pro ...
  Topic: Why we have to declare random@ ?
JohnCampbell

Replies: 15
Views: 823

PostForum: Support   Posted: Sun May 06, 2018 8:50 am   Subject:
I tried a few options to get it to work in PLATO with "Release x64"
I needed "real*8, external :: random@" to get it to work

program trand
! implicit none
! implicit real &# ...
 
Page 1 of 131 Goto page 1, 2, 3 ... 129, 130, 131  Next
All times are GMT + 1 Hour
Jump to:  


Powered by phpBB © 2001, 2005 phpBB Group