|
|
|
|
systemc-forum - RE: [systemc-forum] TLM2.0 timing diagram w/ for multiple burst transac
|
Message Thread:
Previous |
Next
|
- To: <systemc-forum@xxxxxxxxxxxxxxxxx>
- From: "Bradley, Mike" <mike_bradley@xxxxxxxxxx>
- Date: Thu, 21 Oct 2010 08:55:21 -0600
- Send Email to systemc-forum@lists.systemc.org:
- Send new message
- Reply to this message
|
Hi,
I don't think you can have very good "AT" performance without taking
burst transfers into account. Indeed most CPU data transfers and
instruction read will be cache line accesses, which are bursts.
Would not TLM 2.0 be able to timing-model this? If not, and you have to
go down to RTL and back, seems like a kludge.
-Mike
-----Original Message-----
From: Bernard Deadman [mailto:bdeadman@xxxxxxxxxx]
Sent: Thursday, October 21, 2010 10:45 AM
To: A, Nizamudheen
Cc: Bradley, Mike; systemc-forum@xxxxxxxxxxxxxxxxx
Subject: Re: [systemc-forum] TLM2.0 timing diagram w/ for multiple burst
transaction
Hi Nizam,
TLM basically models memory reads / writes and most of the TLM models
achieve a compromise between performance and accuracy without
considering how block transfers are broken down into word reads & writes
on a bus, so there's no standardization for this level of precision. As
an example we deliver transactors that interface / convert TLM to RTL
buses like AXI & OCP that support burst mode transfers. The transactors
are fully protocol compliant at the cycle-level but the TLM still
operates with a byte array & length payload.
The disadvantage of modeling the individual transfers in TLM is, of
course, performance. Taken to extremes the TLM models are going to
replication what the RTL does to get the accuracy at the word transfer
level (almost cycle-accurate in most cases) but then you lose the
performance advantage from using TLM so why not stick with the RTL and
avoid the cost of creating the second set of models?
Perhaps we can help you better if you can explain your goals in modeling
at this level of precision in a TLM environment?
Regards
Bernard
A, Nizamudheen wrote:
>
> Hi Mike,
>
> Thanks for your response.
>
> I was trying to create an extension to the base-protocol and I
> couldn't find any example or statement in LRM regarding multi-burst
> transaction.
>
> Regards,
>
> Nizam
>
>
------------------------------------------------------------------------
>
> *From:* systemc-forum@xxxxxxxxxxxxxxxxx
> [mailto:systemc-forum@xxxxxxxxxxxxxxxxx] *On Behalf Of *Bradley, Mike
> *Sent:* Thursday, October 21, 2010 7:26 PM
> *To:* systemc-forum@xxxxxxxxxxxxxxxxx
> *Subject:* RE: [systemc-forum] TLM2.0 timing diagram w/ for multiple
> burst transaction
>
> Hi Nizam,
>
> I Looked around but could not find one. Best I could find is in the
> TLM2.0 LRM. There are commercial tools that implement TLM2.0 burst
> transactions for you (only one I know of is Vista from Mentor, but I'm
> sure there are others). Have you looked at commercial solutions, or
> just wanting to create your own?
>
> -Mike
>
> *From:* systemc-forum@xxxxxxxxxxxxxxxxx
> [mailto:systemc-forum@xxxxxxxxxxxxxxxxx] *On Behalf Of *A, Nizamudheen
> *Sent:* Tuesday, October 19, 2010 7:06 AM
> *To:* systemc-forum@xxxxxxxxxxxxxxxxx
> *Subject:* [systemc-forum] TLM2.0 timing diagram w/ for multiple burst
> transaction
>
> Hi,
>
> Can some-one point me to a TLM2.0 transaction diagram for a
> multi-burst read/write transaction?
>
> Regards,
>
> Nizam
>
|
|
By Date:
Previous |
Next
|
Current Thread
|
By Thread:
Previous |
Next
|
- TLM2.0 timing diagram w/ for multiple burst transaction, A, Nizamudheen 11:06 GMT
- RE: [systemc-forum] TLM2.0 timing diagram w/ for multiple burst transaction, Bradley, Mike 13:56 GMT
- RE: [systemc-forum] TLM2.0 timing diagram w/ for multiple burst transaction, A, Nizamudheen 14:16 GMT
- Re: [systemc-forum] TLM2.0 timing diagram w/ for multiple burst transaction, Bernard Deadman 14:45 GMT
- RE: [systemc-forum] TLM2.0 timing diagram w/ , Bradley, Mike (you are here)
- Re: [systemc-forum] TLM2.0 timing diagram w/ for multiple burst transaction, Bernard Deadman 15:18 GMT
- RE: [systemc-forum] TLM2.0 timing diagram w/ for multiple burst transaction, Bradley, Mike 16:01 GMT
- Re: [systemc-forum] TLM2.0 timing diagram w/ for multiple burst transaction, Bernard Deadman 16:52 GMT
- RE: [systemc-forum] TLM2.0 timing diagram w/ for multiple burst transaction, Bradley, Mike 20:34 GMT
- taf! ?, Bradley, Mike 15:52 GMT
|
|