RE: [rohc] Sigcomp Torture Test discrepancy

Abbie,

Please address these corrections in the RFC AUTH48 stage, with
cc to WG chairs and ADs.

Thanks,
/L-E

----Original Message----
From: rohc-bounces@ietf.org [mailto:rohc-bounces@ietf.org] On
Behalf Of Surtees, Abigail
Sent: den 23 december 2005 11:21
To: John Barber; rohc@ietf.org; Lars-Erik Jonsson (LU/EAB)
Subject: RE: [rohc] Sigcomp Torture Test discrepancy

> Hi John,
> 
> You are right - I've just run the test on our vm and it does
> take 66 cycles - this appears to be a typo or relic.  Your
> other comments are also correct.
> 
> However, this draft has now been approved by the IESG for
> publication as an informational RFC (see
> http://www1.ietf.org/mail-archive/web/rohc/current/msg04084.html).
> 
> Lars-Erik, if there is no objection within the working group,
> is it possible to take account of these comments before RFC
> publication or is it too late?
> 
> Best regards and Merry Christmas,
> 
> Abbie
> 
> 	-----Original Message-----
> 	From: rohc-bounces@ietf.org
> [mailto:rohc-bounces@ietf.org] On Behalf Of John Barber
> 	Sent: Wednesday, December 21, 2005 10:10 PM
> 	To: rohc@ietf.org
> 	Subject: [rohc] Sigcomp Torture Test discrepancy
> 
> 
> 	Hi all,
> 
> 	I have a disagreement with the expected result of one
> of the Torture Tests
> (http://www.ietf.org/internet-drafts/draft-ietf-rohc-sigcomp-t
> orture-tests-03.txt). 
> 
> 	For the test A.3.5, Bytecode state creation, execution
> of the first message is said to consume 67 UDVM cycles. My
> figure is 66.
> 
> 	The instructions executed and the number of cycles are
> as follows:
> 
> 	COPY (length 8)   1 + 8  = 9 cycles
> 	STATE-CREATE (state_length 8) 1 + 8 = 9 cycles
> 	LOAD = 1 cycle
> 	COPY (length 8) 1 + 8 = 9 cycles
> 	STATE-CREATE (state_length 8) 1 + 8 = 9 cycles
> 	STATE-CREATE (state_length 8) 1 + 8 = 9 cycles
> 	JUMP = 1 cycle
> 	INPUT-BYTES (length 1) 1 + 1 = 2 cycles
> 	OUTPUT (output_length 2) 1 + 2 = 3 cycles (note:
> output_length is 3 in the assembler code, but the code is
> modified to give an output_length of 2 at the time the
> instruction is executed, so this is the value that should be used)
> 	END-MESSAGE (state_length 13) 1 + 13 = 14 cycles
> 
> 	Total of above = 66 cycles.
> 
> 	If I have made a mistake in the above, please could you
> let me know where I have gone wrong? Otherwise if I am right,
> perhaps this could be logged as a correction for the next
> version of the document...
> 
> 	Whilst I'm on the subject, two other incredibly minor issues:
> 	1. Section 2.5, page 12, quotes offsets of 178 for
> overlap_end and 163 for overlap_start. I am seeing values of
> 177 and 162 instead, whether running my own assembled code or
> the supplied bytecode. This does not affect the result of the test,
> however. 
> 	2. Section 2.16, page 29, there seems to be a missing
> "at (512)" statement immediately before the label ":state_start"
> 
> 	Regards, John
> 
> 	P.S. I would like to add that the Torture Test document
> has been an incredibly useful resource during my Sigcomp
> implementation. Many thanks to all involved for producing
> such a detailed set of tests.

_______________________________________________
Rohc mailing list
Rohc@ietf.org
https://www1.ietf.org/mailman/listinfo/rohc