How can I get the SDF from xilinx2primetime have values that reflect the temperature and voltage prorating that was done in implementation?
You have brought up a very important topic here. In order to generate the
proper SDF file reflecting a specific T & V (Temperature, Voltage), we need to
run the NGDANNO with -p switch. The -p switch reads in the PCF file which
contain the desired T and V.
This means that you need a specific set of SDF, Verilog Netlist for each pair of
T V. I will ask Raj to create for you two sets of Netlist and SDF for the
following V and T :
T = 85C
V = 1.70V
T = 40C
V = 1.90V
Raj, please proceed to modify PCF file with each set of T an V and send the
resultant Verilog and SDF file to John.
Shiv, please make sure this feature is placed into the regression suite.
Chris, please add a Solution Record to cover this.
"Vargas, John B." wrote:
> I was analyzing the jtag paths more in detail today and I was able to
> compare point by point with trace. I can see that the output from the 5.1i
> ngdtover and xilinx2primetime which is annotated into primetime is
> completely incorrect, but the one you sent has values that are more similar
> to the trace values.
> 4) This means that we can't do anything with primetime except with your
> converted files. Do you know when we can get a version of our own where we
> can do the conversion?
> 5) Even with the files you sent, the data is still not the same, though.
> For example for the following path in the files you sent back:
> DefineEndpoints FromFFs
> DefineEndpoints ToFFs
> TRACE 85C/1.70V = 10.635ns
> TRACE -40C/1.90V = 5.487ns
> However Primeime (-max and -min) = 14.01ns
> If I do a report_lib on the simprims library I get
> Operating Conditions:
> Name Process Temp Voltage Tree Type
> WCCOM 1.00 70.00 4.75 worst_case
> So, how do I run the Primetime analysis on a specific set of
> temperature/voltage conditions that match the trace values?
> -John V.
> -----Original Message-----
> From: Vargas, John B.
> Sent: Thursday, February 13, 2003 12:55 PM
> To: 'Raj Klair'
> Subject: RE: Primetime translation
> 1) Thanks very much for the translation. I have been looking into it and now
> I don't see the hold errors in the Jtag area that was in the previous one.
> This is good news, but not the end of the story because this ncd had a
> working jtag. I have attached one more set of files for a version of this
> design (same ngd) that was simply re-placed and routed and ended up with
> failing jtag. If you could translate this one, then we will have a good
> sample of the same design and one that fails and one that passes.
> 2) I tried to use the same constraints file (sdc) you provided but with an
> .sdf and .v file that I generated with 5.1i. I get different results from
> the same source files. With the 5.1i ngd2ver and xilinx2primetime I do get
> hold errors on the jtag and with your translation I don't.
> Could you explain what is different about the new SDF and V files that you
> are generating?
> 3) I do see some setup errors that I can't explain, I am trying to time the
> same path
> with primetime and trace. I will send you more details on that soon.
> -John V.
> -----Original Message-----
> From: Raj Klair [mailto:firstname.lastname@example.org]
> Sent: Tuesday, February 11, 2003 6:39 PM
> To: Vargas, John B.
> Subject: Re: Primetime translation
> Please find enclosed the design with the files (unisys.tar.gz).
> The files are built from the fpga_top_pins.ngd and the c4ft....ucf
> 1. For TIMEGRP spi_inputs OFFSET OUT .. AFTER has been applied on the
> INPUT ports (you may use OFFSET IN .. BEFORE for these ports).
> These input ports are :-
> 2. PERIOD is defined on internal nets. These are :-
> fsynci_pz (changed manually to fsynci_pz in fpga_top_pins.sdc)
> mp_jt_tckz ( changed manually to mp_jt_tck in fpga_top_pins.sdc)
> Please feel free to contact me if you need more assistance in this
> "Vargas, John B." wrote:
> > Raj,
> > I wondered if have an update on the translation. Are you running into
> > problems? Anything I can help with?
> > Also any news on the differences between trace and primetime for the last
> > netlist with respect to the hold violations?
> > -John V.
> > -----Original Message-----
> > From: Hamid Agah [mailto:Hamid.Agah@xilinx.com]
> > Sent: Sunday, February 09, 2003 1:05 AM
> > To: Vargas, John B.
> > Cc: Raj Klair; Chris Zeh
> > Subject: Re: Primetime translation
> > Importance: High
> > Hi John,
> > 1. Raj Klair is working on your designs and should be able to provide you
> > with the new netlist,
> > SDC, SDF files hopefully by Monday. Raj, please validate the files and
> > forward them to
> > John. PLease also include the .twr (PAR report) as well.
> > 2. While I am away, please contact Chris Zeh @ (email@example.com) for
> > further follow up on
> > PrimeTime issues.
> > I will return on 2/20.
> > HAmid
> > "Vargas, John B." wrote:
> > > Hamid,
> > >
> > > I am finally starting to time with Primetime and I am seeing different
> > > results than the trace tool. I would like to do an effort to correlate
> > > designs by timing the same paths and analyzing the segment delays in
> > detail.
> > > However, we have changed the design and the constraints significantly
> > > the one you processed for us.
> > >
> > > Attached is a new constraints file. Could you have this translated?
> > >
> > > Also, can I perform the SDF file creation with the translator version I
> > have
> > > and then use your updated constraints? Or do I need you to do that too?