After a multi-year break, I am trying to get back into FPGAs.
I am sticking with Xilinx. In the past I've used ISE. I figured that I should get Vivado up, and was looking forward to seeing the progress made in the last 5 years or so. I am sad to see I didn't miss much - it's two steps backwards and a pretend step forward. After downloading 40 GB (!) of nonsense and spending a good chunk of the day installing it, it is slow as a snail on opioids. A trivial project - a button connected to an LED, takes minutes to build.
I spent the rest of the day getting ISE 14.7 running. Luckily I've done it a few times before, so it was not impossible. Also lucky that 14.7 supports Artix chips, which I am hoping to use for a project. Using my old makefile my build time is 12 seconds or so.
This is insanity. Waiting minutes to see what errors you forgot to fix is unforgivable. Does anyone actually use this toolchain?
Vivado incredibly slow
Vivado incredibly slow
In theory, there is no difference between theory and practice. In practice, there is. ...Jan van de Snepscheut
Re: Vivado incredibly slow
LMAO 40GB? It better have an oculus rift version that you can run around through the logic gates and edit it with your hands. Yikes. Sounds like a typical company sitting comfortably on a near monopoly. Many such cases, sad.
Re: Vivado incredibly slow
At least it's 40GB that does something. Compare that with https://www.balena.io/etcher/ at 85MB. It does a small subset of UNIX dd command. It writes a binary image of a file system to an SD drive (it is recommended by the Raspberry Pi Foundation). Apparently it takes 85MB of code to do that, and it is a very successful application, the company's cornerstone. They were hiring last time I looked, so their investors must be very impressed. Coming soon: a command line interface! You can't make this stuff up.
In theory, there is no difference between theory and practice. In practice, there is. ...Jan van de Snepscheut
-
jmthompson
- Posts: 127
- Joined: 30 Dec 2017
- Location: Detroit, Michigan, USA
- Contact:
Re: Vivado incredibly slow
ThisWayUp wrote:
LMAO 40GB? It better have an oculus rift version that you can run around through the logic gates and edit it with your hands.
Re: Vivado incredibly slow
Vivado GUI startup time alone is unacceptable at around 10 seconds. That is completely dead time after double-clicking the icon, before the window comes up. In my opinion, 10 seconds is around the upper bound of how long the entire build should take. Not minutes. Shame on them!
My ISE 14.7 (weighing in at only 8GB) works reasonably well - my old XC3S50 builds of trivial circuits are under 15 seconds.
I just got a Nexys A7 with an Artix-100T. It is substantially bigger than anything I've ever used, and I have yet to see if it's much slower on trivial builds. But at least I got it to work with both Vivado and ISE 14.7 (although to program it I have to go via the JTAG port as it does not recognize the USB programming port).
Notably, ISE supports Artix-50/100, Spartan6 and below. It's like it's made for me - normally I bottom-fish (to wit: XC3S50!) but occasionally I want to go all out with a big, modern FPGA, with all those juicy LUT6s in giant slices.
I've been exploring Symbiflow, the open source toolchain. Although no luck yet.
And there is FPGAsm, which needs some fixing to handle large chips.
My ISE 14.7 (weighing in at only 8GB) works reasonably well - my old XC3S50 builds of trivial circuits are under 15 seconds.
I just got a Nexys A7 with an Artix-100T. It is substantially bigger than anything I've ever used, and I have yet to see if it's much slower on trivial builds. But at least I got it to work with both Vivado and ISE 14.7 (although to program it I have to go via the JTAG port as it does not recognize the USB programming port).
Notably, ISE supports Artix-50/100, Spartan6 and below. It's like it's made for me - normally I bottom-fish (to wit: XC3S50!) but occasionally I want to go all out with a big, modern FPGA, with all those juicy LUT6s in giant slices.
I've been exploring Symbiflow, the open source toolchain. Although no luck yet.
And there is FPGAsm, which needs some fixing to handle large chips.
In theory, there is no difference between theory and practice. In practice, there is. ...Jan van de Snepscheut