AnsweredAssumed Answered

Why does ADS's Dataset File exhaust large memory ?

Question asked by TNAKA on Oct 6, 2006
Latest reply on Oct 10, 2006 by FKUSIDLO

Like • Show 0 Likes0
Comment • 5

Hi.

I noticed that ADS's Data Set File exhausts very large memory.

I have binary data file which is composed of signed 16bits integer pairs(I and Q). Â
Includeded I/Q are 1638350 points.
So original binary file size is
(2+2)*1638350=6553400 bytes(about 6Mbytes)

Then I converted this binary file to generic MDIF(ascii) like followings.

BEGIN
% index(int) idata(int) qdata(int)
.............
END

Size of Generic MDIF is more than 23Mbytes.
I can understand this size of Generic MDIF.

Next I converted this Generic MDIF to ADS's Data Set File.
To my suprise, resultant Dataset File is more than 130Mbytes.

Why does ADS's Dataset File exhaust such large memory although it is binary format ?

???

No one else has this question

Outcomes

FKUSIDLO

Oct 9, 2006 9:23 PM

The reason for this is:
ADS works with extended floating point format (64 Bit). So if you generate a Dataset (the result file of ADS) I and Q get both extended to 64 Bit which gives you the about 130 MByte of result

Actions

TNAKA Oct 10, 2006 5:06 AM

Hi.

I set data type as "int" in Generic MDIF definition.
Actually all variavles are identified as "int" since "variable information" shows they are "int".

You mean ADS data set treat all variables as double precision regardless of data type which "variable infomation" shows ?

Even so, it can't occupy 130Mbytes.

My examples is
I/Q points=1638350
(2+2)*1638350=6553400 bytes.

If this I/Q points are casted to 8bytes instead of 2bytes and we add independent variable(index) also as 8bytes, expected file size will be

(8+8+8)*1638350=37.5Mbytes.

???

Actions

FKUSIDLO

Oct 10, 2006 5:47 PM

Your calculation is correct, but the ds-file does not just contain the raw simulation results but also some sort of link information, means how the data is alined with each other. If you use a hex editor you'll find such information in the file.
But the fileformat itself is proprietary and therefore no information that is saved in the ds-file is available.

Actions

TNAKA Oct 10, 2006 7:44 PM

Hi.

Thanks for response.

I've felt that ADS Data Display is very very slow for large data, for example, from transient analysis.
Maybe other people also feel same as me.

Compared to PSF data file of Cadence, ADS's Data set is extraordinary large.

This large size file could cause very slowness of ADS Data Display ?

Actions

FKUSIDLO

Oct 10, 2006 7:56 PM

Hi,
again you are right.
But there are several possibilities to represent the simulation data, that is not available with Cadence.
Dependent on how you represent the data you can improve the drawing speed and you can get the information out of the data that you are interested in.
With the later versions (i.e. 2005 and 2006 version)of ADS we also have worked to improve the drawing speed of the Data Display Window.

Actions

5 Replies

FKUSIDLO Oct 9, 2006 9:23 PM
Mark Correct
Correct Answer
The reason for this is:
ADS works with extended floating point format (64 Bit). So if you generate a Dataset (the result file of ADS) I and Q get both extended to 64 Bit which gives you the about 130 MByte of result
Like • Show 0 Likes0
Actions
TNAKA Oct 10, 2006 5:06 AM
Mark Correct
Correct Answer
Hi.

I set data type as "int" in Generic MDIF definition.
Actually all variavles are identified as "int" since "variable information" shows they are "int".

You mean ADS data set treat all variables as double precision regardless of data type which "variable infomation" shows ?

Even so, it can't occupy 130Mbytes.

My examples is
I/Q points=1638350
(2+2)*1638350=6553400 bytes.

If this I/Q points are casted to 8bytes instead of 2bytes and we add independent variable(index) also as 8bytes, expected file size will be

(8+8+8)*1638350=37.5Mbytes.

???
Like • Show 0 Likes0
Actions
FKUSIDLO Oct 10, 2006 5:47 PM
Mark Correct
Correct Answer
Your calculation is correct, but the ds-file does not just contain the raw simulation results but also some sort of link information, means how the data is alined with each other. If you use a hex editor you'll find such information in the file.
But the fileformat itself is proprietary and therefore no information that is saved in the ds-file is available.
Like • Show 0 Likes0
Actions
TNAKA Oct 10, 2006 7:44 PM
Mark Correct
Correct Answer
Hi.

Thanks for response.

I've felt that ADS Data Display is very very slow for large data, for example, from transient analysis.
Maybe other people also feel same as me.

Compared to PSF data file of Cadence, ADS's Data set is extraordinary large.

This large size file could cause very slowness of ADS Data Display ?
Like • Show 0 Likes0
Actions
FKUSIDLO Oct 10, 2006 7:56 PM
Mark Correct
Correct Answer
Hi,
again you are right.
But there are several possibilities to represent the simulation data, that is not available with Cadence.
Dependent on how you represent the data you can improve the drawing speed and you can get the information out of the data that you are interested in.
With the later versions (i.e. 2005 and 2006 version)of ADS we also have worked to improve the drawing speed of the Data Display Window.
Like • Show 0 Likes0
Actions

Why does ADS's Dataset File exhaust large memory ?

Outcomes

Related Content

Recommended Content