[etherlab-users] EtherCAT data size compatibility issue

[Gavin Lambert]

There aren’t any official standards (as far as I know) for how portability of floats works between machines with different endianness.

 

Usually however (AFAIK) they follow the same swapping rules as integers, so typically the best approach is to receive the possibly-wrong-endian value into a byte array or integer variable (never a float variable), then apply the conditional byte swap as if it were an integer, then reinterpret the bit pattern as a float (via a pointer cast, not a value cast nor union).  So the Etherlab macros should work fine for this purpose.  (Note that some compilers prefer that you perform an intermediate cast from integer to byte array then to float, rather than going directly from integer to float.)

 

Usually this will do the trick, assuming that the master and slave use compatible definitions of how floating point are represented (while rare, there are some processors around that don’t use IEEE floats).  The main thing is to never put data that might be byte-swapped into a float/double variable (not even temporarily), because you’ll run into trouble with special interpretation of particular bit patterns.

 

From: Eshwar Lal [mailto:eshwarkhatri at gmail.com] 
Sent: Tuesday, 14 February 2017 06:06
To: Gavin Lambert <gavin.lambert at compacsort.com>
Subject: Re: [etherlab-users] EtherCAT data size compatibility issue

 

Hi,

 

Thanks for your response, yes there are macros available for 64 bit read/write in ecrt.h.

 

one question, is it valid only for integer values or I could use them for floating point data types as well?

 

On Fri, Feb 10, 2017 at 12:22 AM, Gavin Lambert <gavinl at compacsort.com <mailto:gavinl at compacsort.com> > wrote:

64-bit macros are already defined in ecrt.h, so just use them.

 

In general though, if you want to deal with other data sizes, have a look at how the macros are defined, and define your own versions.  They’re just wrappers around standard conditional byte-swapping calculations.

 

Data on EtherCAT is little-endian by convention (although occasionally a slave will do something different), so the byte-swapping is only needed if your master code might run on a big-endian processor.

 

From: Eshwar Lal
Sent: Friday, 10 February 2017 02:26
To: etherlab-users at etherlab.org <mailto:etherlab-users at etherlab.org> 
Subject: [etherlab-users] EtherCAT data size compatibility issue

 

 

hello all,

 

I am using EtherCAT master with ESC FC1100 for performing HIL simulations, I am using ecrt.h api functions to send a receive data variables,

 

EC_READ_U32 and EC_WRITE_U32, which support only 32bit unsigned values, however my process variable is 64bit in size.

 

is there any possible solution to this problem? so that I can transfer the variables 64bit in size over etherCAT and perform calculation for my simulation or any other possible tip to convert the 64bit variable to 32bit just for transmit and receive purpose. Or a method to transfer and receive 64bit values over EtherCAT.

it would be great to hear a positive solution to rectify the issue.

 

Note: I am using Linux kernel with RTAI.


_______________________________________________
etherlab-users mailing list
etherlab-users at etherlab.org <mailto:etherlab-users at etherlab.org> 
http://lists.etherlab.org/mailman/listinfo/etherlab-users





 

-- 

Best Regards.
Eshwar Lal 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.etherlab.org/pipermail/etherlab-users/attachments/20170214/a50f9672/attachment-0004.htm>