d060cb82af
typo
2023-11-17 13:57:39 +01:00
0735b8d0a5
fix previous commit on repeating error messages
2023-11-15 22:54:53 +01:00
22a10c0e44
typo
2023-11-15 22:54:06 +01:00
55ba927973
If nan occurs during op calculation, issue the warning
...
no more than 10 times.
2023-11-14 16:57:14 +01:00
4da3dc51fd
If compatmode ps or lt: add a series resistance (100µOhms) to
...
diodes, when RS is not given or 0:
Improves convergence in behavioural models for op evaluation.
2023-11-14 16:56:31 +01:00
29bfe94318
Exclude some code under a unified #undef REM_USED
2023-11-13 12:32:15 +01:00
2cf2693706
few code cleanings
2023-11-12 17:04:36 +01:00
2a005292db
Sparse 1.3 is the matrix solver if no option (sparse or klu) is given.
...
KLU is selected by 'option klu'
If both 'option klu' and 'option sparse' are given,
the last one executed determines the solver.
'option sparse=0' de-selects spares, i.e. selects klu.
'option klu=0' de-selects klu, selects sparse.
'option klu=1' has the same meaning as 'option klu'.
'option sparse=1' has the same meaning as 'option sparse'.
2023-11-12 16:32:40 +01:00
5645968510
compiling with klu is now standard
2023-11-12 15:40:04 +01:00
c87df54f24
Remove unused global variable cp_dol.
...
This also prevents a gcc 13 bug, as cp_dol has been set to '\0'
by -O2 optimization (see
https://stackoverflow.com/questions/77407156/s-0-being-optimized-out-possible-gcc-13-bug-or-some-undefined-behaviour )
2023-11-11 10:39:51 +01:00
02aa2c03f9
klu is now compiled-in as standard
2023-11-10 19:48:52 +01:00
359e3d4609
Fix Bug #467 - ".OPTION INTERP has accuracy issue".
2023-11-09 12:09:05 +00:00
10386ee3e3
Commit [35d338]ngspice partly reverted commit 423f9a9db to fix deletion
...
of source files if make was accidentally used in the source tree.
That caused the original bug to return. Fix the problem by deleting
the generated source files by name (no such files exist in source.
Also make the generated files depend on cmpp, making 'make clean'
unnecessary after a chenge to cmpp.
2023-11-09 12:08:02 +00:00
f6f7319792
Add null-pointer checks to some code that crashed when trying
...
to .print results from a non-existent analysis. Also remove the
troublesome .plot and .print lines from two examples.
2023-11-09 12:07:31 +00:00
4228ba6e8a
Ignore end-of-line comments in quoted strings.
2023-11-09 12:06:34 +00:00
0a087809d6
If both node names are the same, just print them once in the error message.
2023-11-08 13:12:06 +01:00
0555c5d9c7
Make previous KLU support check more transparent.
2023-11-08 11:49:28 +01:00
295f212de8
When CIDER is detected, until KLU works with CIDER, print Error: CIDER simulation is not (yet) supported with 'option klu'. Use 'option sparse' instead. and then call controlled exit(1).
2023-11-08 11:49:22 +01:00
35d338f256
partly revert commit 423f9a9db - don't remove xspice/icm .c files
2023-11-08 07:59:36 +01:00
d27b888f13
Noise simulation is buggy when using the KLU solver. Thus for now it is disabled.
...
Use 'option sparse' instead.
2023-11-06 15:03:09 +01:00
4e1452cdb7
rm compiler warnings
2023-11-04 19:36:31 +01:00
42f882e4e4
add osdinoise.c
2023-11-04 19:36:31 +01:00
e67a6a1297
add osdinoise
2023-11-04 19:36:30 +01:00
5223f4b27d
fix(hicuml2): ensure flicker noise power is always stored in state vec
2023-11-04 19:36:30 +01:00
c21f87051a
always compute osdi noise powers
...
ngspice will reuse operating points for noise analysis so we can not trust
the mode flags
2023-11-04 19:36:30 +01:00
826cddb483
add examples for OSDI noise
2023-11-04 19:36:30 +01:00
f66e0bf2ac
implement osdi noise support
2023-11-04 19:36:30 +01:00
04c68d5f30
Cleanup child processes on Windows VisualC and Mingw. On Linux those processes quit when ngspice quits.
2023-10-28 19:45:24 +02:00
864ef7925c
Add notes on the structure and organization of an external d_process program.
2023-10-28 19:43:50 +02:00
4219efb5d5
Make it explicit that null output ports are not allowed. The output port vector bounds are always >= 1.
2023-10-28 19:43:46 +02:00
5c6b9f03b5
Fix the zero count.
2023-10-28 19:43:41 +02:00
1f5f7ae439
Update d_process examples.
2023-10-28 19:43:36 +02:00
c83d11c9c0
Call cm_cexit() to prevent the simulator running when d_process initialization errors occur.
2023-10-28 19:43:23 +02:00
21761207bc
Change license to 3-clause BSD (see
...
https://sourceforge.net/p/ngspice/ngspice/merge-requests/6/?limit=25#fd4f/73e6 )
2023-10-28 14:55:13 +02:00
09f070f582
Error handling improvements in cfunc.mod. Ensure that d_process.h wiil always respond to version and interface checks sent from sendheader. This is needed so that the pipe reads in sendheader do not cause Windows to hang when the interface version and in/out counts do not match. This hang was the cause of errors not being reported and the Windows gui hanging. Startup and header checks are now detected in cm_d_process, and the simulator will run but with runtime errors since a d_process model cannot be completely instantiated after initial errors. It would be good to find a means of gracefully halting the simulation run.
2023-10-28 11:00:33 +02:00
8c72a76844
Another attempt at clean up of error handling and reporting. It's OK on limux and cygwin, but Windows VisualC is being stubborn. I would prefer not to have to call exit(1) just for Windows, since then nothing is reported and the gui disappears in a puff of smoke. Even without the exit calls nothing gets reported with the Windows gui which will hang when there are errors detected by d_process.
2023-10-28 11:00:28 +02:00
4eae039779
Add modifications note.
2023-10-28 11:00:23 +02:00
4530cde8e2
Use Xspice cm_message_send to report errors rathen than printing to stderr and calling exit. When a d_process model has errors found in start(), sendheader(), and dprocess_exchangedata() these are reported, but if the model is run a SIGINT is raised. There must be a better way of stopping the simulator.
2023-10-28 11:00:18 +02:00
182764a894
Add examples/xspice/d_process.
2023-10-28 11:00:12 +02:00
df28fabc42
Restore the deleted unused params to be backword compatible. Handle the case where the d_process instance has a null in vector, so that N_din is zero. This allows the model to be just a pattern generator. Add include to cmproto.h to avoid a forward ref. to an enum type.
2023-10-28 11:00:03 +02:00
307ba1985b
Remove unused reset_state param; do not allow null in and out, so there is at least 1 input and 1 output.
2023-10-28 10:59:52 +02:00
7820940be3
Remove the unused reset_delay param.
2023-10-28 10:59:42 +02:00
e762047fc2
Add array bounds check, fix typos.
2023-10-28 10:59:34 +02:00
022a340ddb
Fix VisualC and Mingw compiler warnings.
2023-10-28 10:59:11 +02:00
cccb4a1559
Add callback to free process memory.
2023-10-28 10:58:39 +02:00
33780055ee
Add MODIFICATIONS note, cleanup white space.
2023-10-28 10:58:28 +02:00
5c4352d02f
Add __MINGW64__ defines for builds on mingw with msys.
2023-10-28 10:58:15 +02:00
3bb6e86015
Add Windows Visual Studio support for the Isotel d_process xspice digital model. The Isotel mixedsim-master/examples/embedded/motorforce C code needs to have minor changes for _MSC_VER defines, and are not included with this commit. The changes relate to using #pragma pack(push, 1)...#pragma pack(pop) around some structs, and setting stdin/stdout in binary mode. It is doubtful that the cfunc.mod will compile on mingw or cygwin since I don't know if they support pipe/fork/exec/spawn calls. The code builds and runs on Windows, Linux, and MacOS. In the future, Giles Atkinson is looking at allowing Xspice models to load shared libraries (.dll, .so) which will avoid creating and communicating with an external program.
2023-10-28 10:58:06 +02:00
046be0cdc8
Use the PSP103 model pspnqs103va (including nqs option)
...
as this is the standard in the IHP Open PDK.
2023-10-27 23:25:42 +02:00
efdf332061
Prevent a crash if hash model table was not generated,
...
e.g. due to missing valid models.
2023-10-27 23:13:41 +02:00
Holger Vogt
dwarning
Giles Atkinson
Brian Taylor
Pascal Kuthe
Pascal Kuthe