UPGRADE YOUR BROWSER

We have detected your current browser version is not the latest one. Xilinx.com uses the latest web technologies to bring you the best online experience possible. Please upgrade to a Xilinx.com supported browser:Chrome, Firefox, Internet Explorer 11, Safari. Thank you!

cancel
Showing results for 
Search instead for 
Did you mean: 
Highlighted
Observer p.hayk
Observer
1,115 Views
Registered: ‎11-21-2013

Vivado 2015.4 vs 2018.2, 2018.2 loses in all aspects

 

Recently we moved our project from 2015.4 version to 2018.2 and we were really shocked with the bad results that 2018.2 gives in comparison to 2015.4.

 

The implementation time of 2018.2 increased around 6 times

So Implementation time:

2015.4 : 00:19:56

2018.2:  01:52:06

 

We used the same conditions i.e. same strategy, same machine, same design, same IPs. And the tools were run in parallel on same machine with same amount of threads. No other resource was using our machine, only these 2 vivados.

 

Than we checked the timing, slightly better results for 2018.2.

WNS:

2015.4: -4.638

2018.2: -4.777

 

TNS:

2015.4: -1703.105

2018.2: -548.562

 

WNS:

2015.4: -0.064

2018.2L -0.052

 

Than resource utilization:

2018.2

--------

LUT        22024    134600  16.362555

LUTRAM              4441      46200    9.612555

FF           33019    269200  12.265602

BRAM    277.5     365        76.0274

IO           186        400        46.5

MMCM 2             10           20.0

PLL         1             10           10.0

 

2015.4

--------

LUT        23818    134600  17.695395

LUTRAM              4443      46200    9.616883

FF           31999    269200  11.886701

BRAM    278        365        76.16438

IO           220        400        55.0

BUFG     18           32           56.25

MMCM 2             10           20.0

PLL         1             10           10.0

 

 

If you see very small, almost negligible benefit.

 

Can someone please explain how is this possible?

Like we updated almost 3 years older tool version and got worse results with the new tool. Is it expected?

 

P.S. 2018.2 has issue with internal text editor, when we modify the code outside of tool, the tool sometimes text-editor does not recognize our edits.

0 Kudos
5 Replies
Scholar u4223374
Scholar
1,100 Views
Registered: ‎04-26-2015

Re: Vivado 2015.4 vs 2018.2, 2018.2 loses in all aspects

Why are your two designs using different numbers of I/O pins? Anything else I can believe, but missing I/O pins means that something is very messed-up.

 

The timing results are largely irrelevant - both tools work until they meet timing (or decide that they cannot meet timing) and then stop. It's quite possible that one of them could have met tighter timing requirements if it had to, but the fact that both did meet timing doesn't really tell us anything.

 

Apart from that, was this just implementation? Or did you re-run synthesis too? It occurs to me that if you opened a Vivado 2015.4 project in Vivado 2018.2 and run implementation immediately, it might decide to re-run synthesis first - which could easily account for a lot of time.

Scholar drjohnsmith
Scholar
1,059 Views
Registered: ‎07-09-2009

Re: Vivado 2015.4 vs 2018.2, 2018.2 loses in all aspects

re the text editor problem you mention,

 

is it possible different files are on different machines with different clock / times ?

 

Ive not seen it with the editor, but other parts of vivado this seems to be a feature,

   ensure all your machines where files are located and you run vivado on are locked to the same time master 

 

 

<== If this was helpful, please feel free to give Kudos, and close if it answers your question ==>
0 Kudos
Observer p.hayk
Observer
1,039 Views
Registered: ‎11-21-2013

Re: Vivado 2015.4 vs 2018.2, 2018.2 loses in all aspects

Regarding to IOs, really strange. Still debugging

I re-run synthesis and than implementation

 

0 Kudos
Observer p.hayk
Observer
1,038 Views
Registered: ‎11-21-2013

Re: Vivado 2015.4 vs 2018.2, 2018.2 loses in all aspects

Nope, same machine, same user
0 Kudos
Xilinx Employee
Xilinx Employee
1,022 Views
Registered: ‎03-21-2008

Re: Vivado 2015.4 vs 2018.2, 2018.2 loses in all aspects

As this design appears to be failing timing, it may be that the older release simply gave up earlier while the newer release tried much harder to meet timing. You seem to report WNS twice - can you explain the difference here - is one the hold slack? Either way, it would be worth re running with a design where the timing can actually be met as that gives a much more realistic view of actual runtime.