User Tools

Site Tools


xilinx:ise

History

IIRC, the Xilinx lineage goes something like:

  • XACTStep
  • Alliance
  • ISE
  • Vivado

Pretty sure XACTStep was developed in house, but I've been told that ISE tools (not sure about the GUI) was actually from the acquisition of NeoCAD (not sure where Alliance fits in). NeoCAD had reverse engineered Xilinx's chips and made an independnet, presumably better toolchain. Evidently this is wher ethe “.ngc”, “.ngd”, etc extensions come from. For example, its evidently really NeoCAD Generic Database, but Xilinx made the backronym Native Generic Database to re-brand it.

I've also been told that Xilinx (previously?) contracted out the ISE GUI to another company.

VCS friendly

How to better separate log files and such: http://blog.jwcxz.com/?p=823

github official ISE gitignore: https://github.com/github/gitignore/blob/master/Global/XilinxISE.gitignore

  • NOTE: this will attempt to ignore *all* generated files including .bit. For large projects I reccomend you use a build server to generate official builds (ex: Jenkins) and don't check these in. However, for small projects this overhead may not be worth it and you may want to check in some of these files by removing lines or consider not using a gitignore entirely

Linux command line summary

Performance comparison

Project1

JM dfd4fd18

Uses “Total REAL time” as opposed to “Total CPU time”

The primary reason for this study was to upgrade our old build server (imb). I wanted to see if I should put it into a corporate datacenter VM (blackd) or a dedicated machine (imb2). The workstation ultimately won a clear victory with the datacenter VM being about 45% slower. Major contributions:

  • VMs are slower. Autunite showed that the VM was about 27% slower than a similar physical machine setup
  • The workstation CPU is 35% higher clock rate than the datacenter CPUs. Where most applications might be more IO bound, not conforming to the datacenter allowed me to select a CPU more appropriate to the task

Note: the machine was still rackmounted into a datacenter so it would have UPS etc. Obviously your milage may vary if you go against your IT department's standard offerings

Machine Model CPU
cpu-

bench-

mark
RAM (GB) Storage OS ISE xst ngdbuild map par net Note
autunite Lenovo W520 Intel(R) Core(TM) i7-2760QM CPU @ 2.40GHz 6666 12 SSD Ubuntu 12.04 14.5
109

1.00

47

1.00

429

1.00

187

1.00

772

1.00
Local VM on Windows 7 host
autunite Lenovo W520 Intel(R) Core(TM) i7-2760QM CPU @ 2.40GHz 6666 16 SSD Ubuntu 14.04 14.5
87

0.798

41

0.776

333

0.776

145

0.775

606

0.785

Physical machine (removed linux and put Windows on it)

Very close but not exactly the same project as the others
tobernite Dell M3800 Intel(R) Core(TM) i7-4712HQ CPU @ 2.30GHz 7645 16 SSD Ubuntu 14.04 14.5
75

0.688

37

0.787

334

0.779

142

0.759

588

0.762
imb Supermicro Intel(R) Core(TM) i7 CPU 950 @ 3.07GHz 5641 24 HDD Ubuntu 12.04 14.5
105

0.963

47

1.00

398

0.928

161

0.861

711

0.921
imb SSD 102 47 398 198
745

0.965


SSD did not improve performance. Given that others were about the same, suspect something else happened during PAR to steal some speed

Ran build on dedicated mounted SSD after trimming SSD and then formatting it with ext4. Inudstrial SLC SATAIII SSD that doesn't require trim anyway though
brannerite Custom Intel(R) Xeon(R) CPU E5607 @ 2.27GHz 3531 32 HDD Ubuntu 12.04 14.5
149

1.37
70

1.49
533

1.24
231

1.24

983

1.27
thucalite HP Z440 Intel(R) Xeon(R) CPU E5-1650 v3 @ 3.50GHz 13433 32 SSD Ubuntu 14.04 14.5 68

0.624
37

0.787
269

0.627

116

0.620

490

0.635
imb2 HP Z440 Intel(R) Xeon(R) CPU E5-1650 v3 @ 3.50GHz 13433 64 SSD Ubuntu 14.04 14.5
67

0.615

36

0.639

274

0.639

115

0.615

492

0.637

Very close but not exactly the same project as the others

Shows that the extra memory sticks (bandwidth + size) didn't significantly help given the already large amount of memory
blackd Blade Intel(R) Xeon(R) CPU E5-2640 v3 @ 2.60GHz 14377 8 Cloud Ubuntu 14.04 14.5 106 46 400 162
714

0.925

45% slower than dedicated Z440

Above shows some correlation between cpubenchmark.com faster CPUs and build time. IS is able to use multi-cores (but not very well) which probably skews vs the cpubenchmark results.

smartxplorer

Symptoms:

$ smartxplorer
-bash: /opt/Xilinx/14.5/ISE_DS/ISE/bin/lin64/smartxplorer: No such file or directory
$ smartxplorer
smartxplorer: error while loading shared libraries: libncurses.so.5: cannot open shared object file: No such file or directory

Problem: smartxplorer is a 32 bit app even on a 64 bit install (along with a few other apps). You may not have 32 bit libs installed, leading to these errors.

Solution:

This may not be a minimal solution, but is sufficient. Ubuntu 12.04:

sudo apt-get install -y gcc-multilib libc6:i386 libncurses5:i386 libstdc++6:i386

NOTE: ia32-libs was eliminated from repos (E: Package 'ia32-libs' has no installation candidate), use above instead

xilinx/ise.txt · Last modified: 2015/12/01 18:30 by mcmaster-guest