Forum Replies Created
-
AuthorPosts
-
Keith
KeymasterDid you try running the Check -> Check CellView command when the symbol is open for edit? It should clean up issues like this.
This could be caused, for example, if you created a pin without entering a name, or with a non-printing character as the name, and then deleted the pin shape.
Keith
KeymasterAll Glade does on WIndows is set PYTHONHOME (note not PYTHON_HOME) to Python27. You'll see a message along the lines of
# PYTHONHOME is: Python27
if the python home is set up, and if so what its pointing to. There is no reason you shouldn't change this – just that most Windows installations don't come with Python, whereas Mac/Linux normally does.
So try setting PYTHONHOME and PYTHONPATH based on your Python installation with the ttfquery/glyphquery modules.
Keith
KeymasterHave you tried setting PYTHONPATH to include a path to the directory that holds the ttfquery python module?
Keith
KeymasterThere were 2 problems.
The inv1 symbol had 3 shapes on the layer 'pin drawing'. Only 2 of these were pins with net info, the third was an ellipse object that could not be seen, but if you made only the pin layer visible/selectable and selected all, using the query dialog you could find this shape and delete it.
There was a bug in the code in that if a symbol had shapes on the pin layer with no net info, a crash would occur in the Check CellView cmd. I have fixed this – will be in 4.5.24. Meanwhile if you fix the symbol, you should not experience a crash.
Keith
KeymasterThanks – I tried placing a single inv symbol in an empty schematic, then creating a wire from the output terminal with a floating end. Check ran through OK and reported the floating wire, so maybe there is something different in your testcase.
Keith
KeymasterWhat version of Glade?
Keith
KeymasterNo problem, thanks for reporting it – the issue appears to have been there for some time, but as infix mode is the default it has been overlooked. It also only occurs with shortcut key commands that require a shift or ctrl modifier key.
I hope to have the fix in 4.5.17 for release this weekend.
Keith
KeymasterAh. I see what you mean, when infix mode is turned off, and a modifier key has been pressed (shift or ctrl) then it fails to complete the command.
There is a simple fix, but that may have side effects. I need to determine why key events are sometimes not getting handled as they should.
I'll update when a proper fix is available.
Keith
KeymasterStrange, I cannot reproduce this with Ubuntu 16.04 on a VMware virtual machine.
This is with the most recent build from the website.When you say 'don't work', do you mean it crashes or the commands do nothing?
Keith
KeymasterA similar bug was found this morning and a patch has been done, please try re-downloading the ubuntu build and see if this fixes the problem for you?
Keith
KeymasterTry setting:
export LC_ALL=C
before running Glade, as a temporary workaround.
For some reason the locale default is not set, which is a bit odd, as I fixed this for a similar problem about a year ago. I can only assume the fix was not properly checked in… I'll try and put this in the 4.5.17 release.
Keith
KeymasterGlade uses LVS (Layout vs Schematic) derived from Gemini. It is a standalone executable for comparing two netlists, and is integrated into the Glade environment such that an extracted view can be netlisted, run Gemini and differences highlighted in the extracted view.
Extensions to Gemini allow reading of CDL compatible netlists, and knowledge of MOS devices, resistors, capacitors, inductors, diodes and bipoolar transistors. Device properties (such as W/L for MOS devices, R/C/L values for resistors/capacitors/inductors, area for bipolar devices and area/perimeter for diodes can be compared. MOS series and parallel devices can be reduced (or not).
A number of bugs and enhancements have been made including a fairly fundamental one in the local matching algorithm that gave rise to false negatives and would increase runtime.
Gemini does not require initial correspondence points unlike some LVS tools, although they can make results easier to read. A correspondence point file can however be specified which can sometimes help with symmetrical circuits to resolve matching.
When a mismatch occurs, Gemini will list the mismatched nets and devices for both netlists. For nets with a large number of devices (e.g. power/ground) the device list is suppressed if large; this can be controlled by the 'net size limit' option. Bear in mind that ALL devices are listed that connect to an unmatched net; many or most may be correctly connected. When debugging, look for commonality of patterns rather than focus on individual errors.
Series connected devices are reduced (unless the option to reduce is turned off) to a single device. This means that for example if devices M4, M5 and M6 are series connected, they will be reduced into a single 'device' with multiple gates. The device will end up with the name of one of the original devices; the remainder are no longer present in the comparison reports. It is possible to report if the series connection order of devices remains the same. Parallel or fingerd devices are also reduced unless the option is turned off; a check is made that the devices are the same length, and the widths are summed into the remaining device.
Don't forget that to run Gemini within Glade, you will need an extracted view i.e. one with devices and connectivity. If you try and run LVS on a layout view it will probably fail due to missing device definitions or connectivity.
Keith
KeymasterThere is a known problem when running DRC or any scripts using the geom… layer processing functions on the Mac, when starting Glade from the icon.
Glade creates temporary files for derived layers, these files are called file0001.dat etc. To find a location where to write these files, Glade first looks for the env var 'GLADE_DRC_WORK_DIR'. If this is set, then that directory is used. Else Glade tries to write the temp files to the current working directory.
Unfortunately on the Mac, when starting from an icon, env vars are not written from the user's shell. They can be set however using the method described in the README.txt file in the Mac distribution.
When starting from an icon, the 'current working directory' on the Mac is '/' i.e. the filesystem root, which usually won't be writable and hence causes writing temp files to fail. An exception will be thrown later in the script when the files are attempted to be read.
In version 4.5.6 this will be changed so that the fallback temp file location (if GLADE_DRC_WORK_DIR and current work dir are not writeable) will be to use the application's directory. This should hopefully be writable. Note that for the Mac it will be in glade.app/Contents/MacOS, which is where the actual executable resides.
Please set GLADE_DRC_WORK_DIR wherever possible.
Keith
KeymasterOn Windows (64 bit), PCs with Dell's Backup and Recovery software causes a crash on Qt5 applications (such as Glade).
See http://en.community.dell.com/support-forums/software-os/f/3526/t/19634253 for more details.
The only solution is to uninstall the Dell Backup and Recovery software until Dell can fix the problem.
Keith
KeymasterIn version 4.4.70 the first version of netlist driven layout has been implemented. The idea is that given a SPICE/CDL netlist (e.g. used for simulation), layout can be created with the correctly sized PCells and connectivity info.
The flow requires PCells for each device in the netlist e.g. MOS P/N devices, resistors, capacitors etc. The PCells need pin shapes e.g. poly/diffusion pins for S/G/D/B nodes. Examples can be found in the distribution.
The Tools->Netlist View menu opens a text window, and from this you can open a netlist. The netlist is shown in this window, with syntax highlighting. It can be edited here if required, saved etc. But also you can use the 'Gen Layout' menu item to create a layout cellView containing the instances from the netlist and their connectivity. Instances are placed purely in clusters according to device type; they can then be manually moved. With the Selction Options 'Show Connectivity' option enabled, flightlines are shown between device pins.
Cross-probing and cross-selection between the netlist and layout is possible, either for instances, nets or devices. M factor MOS devices are supported (again the distribution contains an example netlist featuring devices with w/l/m parameters).
Manual routing has been enhanced so that with the cursor over an instance pin, the starting/ending point is snapped to the pin centre and takes the layer of the pin, and the net the pin is connected to.
Various other capabilities will be added in future, as will documentation on using this flow.
-
AuthorPosts