* alignedData now uses std::align_alloc * imagedata is now allocated on the heap * m3 server fix for trimbits and badchannels that are shifted by 1 * formatting * binary in * added check for proper memory allocation * commenting out the example in receiver data call back changing size as it affects users using debugging mode to print out headers * fixed warnings * commenting out the example in receiver data call back changing size as it affects users using debugging mode to print out headers * got rid of cast to uint64 * got rid of Reorder function * added sanity check to only enable for chipttestboard and xilinx * removed Gotthard stuff * update the comment about how to modify data on a data call back from the receiver * autogenerated commands and make format * changed font size in GUI * clang-format with clang-format version 17 * updated update_image_size in xilinx * version number automated for python build * mistakenly set version back to 0.0.0 * updated github workflow scripts to support automatic version numbering with environment variable * managed to load VERSION file in yaml file - simplifies things * saving changes in git workflow failed * got typo in github workflow * updatet regex pattern to support postfix * normalized version to PEP 440 specification in update_version.py * bug did not support version 0.0.0 * upgrading to c++17 from c++11 and patch command has to be found before applying patch on libzmq (#1195) * Dev/allow localhost for virtual tests (#1190) * remove the check for localhost being used in rx_hostname for python test for simulators, run rx_arping test only if hostname is not 'localhost' * fix tests for fpath: cannot set back to empty anymore (empty is default) * default rx_hostname arg = localhost, and default settings path =../../settingsdir * changed virtual tests script for better printout on exceptions * fix for catching generaltests exceptions and exiting instead of continuing * fix minor * fixed shared memeory tests to include current env and fixed prints for errors --------- Co-authored-by: Erik Fröjdh <erik.frojdh@gmail.com> * added regex pattern matching to version in toml file * Dev/gitea docker (#1194) * gitea workflows for RH8 and RH9 * using our docker images * version now supports . before postfix * rough draft of test acquire of all detectors for frames caught and file size. ctb not included yet * moved dbitoffset, dbitreorder and dbitlist to GeneralData * added error message on receiver side, throw error * removed log as error already printed * added tests to check file size and frames caught with an acquire (virtual) for every detector * minor printout removed * typo fixed * removed minor printout * incorrect counter mask tested * fix 10g adc enable mask, switched with 1g * fixed hardcoded values of nchip nchan etc from detPArameters * fixed ctb tests, need to fix in develoepr (if digital modfe not enabled, should not take into accoutn dbitlist or dbitoffset or dbitreorder * only reorder bits if some sort of digital readout mode enabled * trying to fix acquire for xilinx * fix for xilinx ctb virtual * alloweing all tests * typo * fix for slsreceiver killed but complaining for virtual tests with script * fixed bug found by @AliceMazzoleni99 that for ctb server is still shown in pgrep -f if xilinx server running, so now the pid is killed and looking for any DetectorServer_virtual instead. also reset color coding after Log * check if process running for kill -9 slsReceiver fail * removed -9 to kill with cleanup * frame synchonrizer fixes: typo of iterator for loop and zmg_msg_t list cleaned up before sending multi part zmq; test written for the frame synchronizer, test_simulator.py rewritten for more robustness and refactoring commonality between both scripts * better error messageS * minor * typo * moving the erasure of the fnum to after sending the zmg packets and also deleteing all old frames when end of acquisition * fix bug in blackfin read access to firmware registers * updates api version based on version file & converted shell script files to python * updated all makefiles * refactoring code and compiling binary * formatting * rewrote end() for StaticVector * rearranged receiver topics, differentiated btween receiver variants and added info about slsFrameSynchronizer * typo * minor aesthetics * minor * added extra fs link and fixed execute_program warning * and now with link * updating pmods * adresses review comments * dummy commit for versionAPI * formatted and updated versionAPI.h * added expat to host section * updated documentation for pip installation as well * Dev/add numpy (#1227) * added numpy dependency * added build specifications for python version and platform * updates files/variants for pmods for 9.2.0 (#1233) * tests for bool in ToString/StringTo (#1230) - Added tests for ToString/StringTo<bool> - Added overload for ToString of bool (previously went through int) * added docs for SLSDETNAME (#1228) * added docs for SLSDETNAME * clarification on hostname * added examples on module index * fixes * fixed typo * Dev/update test framesynchronizer (#1221) * raise an exception if the pull socket python script had errors at startup (for eg if pyzmq was not installed) * minor changes that got lost in the merge of automate_version_part 2 PR --------- Co-authored-by: Erik Fröjdh <erik.frojdh@gmail.com> * added workflow for python wheels * wip * formatting * wip * wip to parse vector of rois at command line * wip * first level test * can get individual rois, but not connected to command yet * rois shoudl work. left to implement tests for individual rois, create multiple datasets (1 for each roi) in the virutal data file. currently virutal dataset with roi is not implemented and a warning is given instead. wonder why since the inviduviaual roi files are clipped * all tests pased * minor * fixed rx_roi for multi modules jungfrau , tests for eiger, multi modules jungfrau in x and 2 interfaces * works for eiger as well * switched to vector instead of std::array<ROI, 2>>, which prints extra [-1, -1] when theres only 1 udp interface * wip * fix for empty roi vectors (which shouldnt be) as you cant know if its all or not in roi * wip: to map roi to virutal * fix for eiger, added python test for testig roi in different module and detector type configurations * wip, fails with master and virtual * works for complete roi * wip, works for a single roi * works for all rois * wip to fix tests * 1d fixes * rois test work on 1d as well * check master file creation as well in rx_roi tests * get rx_roi from metadata from rxr, cant reconstruct. fixed clear roi should give 1 roi min * gui shows roi now * format * updated python bindings * updated master file versions * cmd generation and formatting * minor fixes in command line and help * minor * doesnt happen anymore * comment * minor * redundant getRxROI in Detector class for multi level and module level * refactor cmd parsing (detid can be parsed directly) * refactor cmd line * refactor command line parsing of roi * modified comments about ctb and xilinx not using roi * refactoring * refactorign * refactoring wip * wip refactoring * formattin * to avoid confusion, moved default initialized, single sized declared vector of roi to be created at setDetectorType * pybind only 1 function for getRxROI * command line help * specified number of receiver error message * minor comment * refactored to take out repetitive code, need to adjust for slsMulti and slsFrameSync * wip * works, need to add tests * made Commadnlineoptions into a class * wip test * fixed tests * cleaning up properly , semaphore leaks, child process/thread throwing handled * getuid issue on github workflow * constexpr and checking if options object type is same * unnecessary capture * remove testing code, minor * fixed help, -t for multi should not be supported as it never had it * Formatting * hdf5 definitions in test when not compiled with hdf5 * typo * moved optstring and long options to the constructor * raising a SIGINT when the child thread has an exception so that the parent thread can exit all the threads and clean up gracefully * minor test typo * check status of child exiting and use that to send sigint to all the child processes from the parent * fixed validation in network_utils, added a tests to throw for port 65535 in test mode (option on for sls_use_tests), multi:parent process checks child process exit status to send sigint to others * moving set signal handler to network utils * readoutspeed in rx master file and other master file inconsistencies (#1245) readout speed added to json and h5 master files. Also fixed master file inconsistencies Sserver binaries - update server binaries because readoutspeed needs to be sent to receiver with rx_hostname command API - added const to Detector class set/getburstmode Python - updated python bindings (burstmode const and roi arguments) Cmd generation - added pragma once in Caller.in.h as Caller is included in test files m3: num channels due to #counters < 3 * workaround for m3 for messed up num channels (client always assumes all counters enabled and adds them to num channels), fix for hdf5 g2: exptime master file inconsistency - exptime didnt match because of round of when setting burst mode (sets to a different clk divider) - so updating actual time for all timers (exptime, period, subexptime etc, ) in Module class, get timer values from detector when setting it and then send to receiver to write in master file ctb image size incorrect: - write actual size into master file and not the reserved size (digital reduces depending on dbit list and dbit offset) - added a calculate ctb image size free function in generalData.h that is used there as well as for the tests. master file inconsistencies - refactored master attributes writing using templates - names changed to keep it consistent between json and hdf5 master file (Version, Pixels, Exposure Times, GateDelays, Acquisition Period, etc.) - datatypes changed to keep it simple where possible: imageSize, dynamicRange, tengiga, quad, readnrows, analog, analogsamples, digital, digitalsamples, dbitreorder, dbitoffset, transceivermask, transeiver, transceiversamples, countermask, gates =>int - replacing "toString" with arrays, objects etc for eg for scan, rois, etc. - json header always written (empty dataset or empty brackets) - hdf5 needs const char* so have to convert strings to it, but taking care that strings exist prior to push_back - master attributes (redundant string literals->error prone tests for master file - suppressed deprecated functions in rapidjson warnings just for the tests - added slsREceiverSoftware/src to allow access to receiver_defs.h to test binary/hdf5 version - refactored acquire tests by moving all the acquire tests from individual detector type files to a single one=test-Caller-acquire.cpp - set some default settings (loadBasicSettings) for a basic acquire at load config part for the test_simulator python scripts. so minimum number of settings for detector to be set for any acquire tests. - added tests to test master files for json and hdf5= test-Caller-master-attributes.cpp - added option to add '-m' markers for tests using test_simulator python script * doc: added inst on how to set persistentn NIC changes after reboot for each ethernet interface such as rx 4096, rx-usecs, adaptive-rx and gro etc. * added permanent ethtool settings also for fedora or modern rhel * troubleshooting doc: permanent changes for 10g pc tuning (#1247) * doc: added inst on how to set persistentn NIC changes after reboot for each ethernet interface such as rx 4096, rx-usecs, adaptive-rx and gro etc. * added permanent ethtool settings also for fedora or modern rhel * ifcfg scripts still work on rhel8, just not preferred * added dataformat for jungfrau * eiger basic mod * eiger doc done * added moench * done * free shm exposed in python as free function and detector function * minimum change * added quad and updated about 1gbe/10gbe * more info * remove arguments info * replacing commands with links * minor * detail explanation of eiger * fixed imagesize ctb issue (out values not transferred, setting any dbit values was not recalculatign image size in generaldata) * fixed ctb dbit clock changing period in tests as it was setting run clock instead * python accessing freed shared memory object (#1253) * added a 'isValid' member in shared memory (also updated shm version) with default true, any access to shared memory() checks also for validity. any free will set this to false and then unmap shm. Any access to shm will then check validity in python. * fixed tests for shm * added tests in python as well --------- Co-authored-by: Alice <alice.mazzoleni@psi.ch> * updated error message * made markers argument in ParseArguments a boolean instead of an int * removed relative path compared to where executable run in test script for settingsdir * fix roi test * updating versions (#1258) * updating package version, client version, server versions. Renaming server versions, using hardlinks in serverBin. Removing ctb servers in serverBin. (#1259) * fixed no interpolation mode for moench (#1262) Co-authored-by: Anna Bergamaschi <anna.bergamaschi@psi.ch> * fixed multi receiver and frames sync help throw of bad variant access (#1265) * 1000/doc c standard (#1267) * updated c++11 to c++17 * more about c++11 and updating readme * updated documentation for receiver arguments and also making receiver constructor explicit * minor fix for rxr err message * fixed doc about gcc version * 1000/release notes (#1269) * updated firmware and server version in release notes * release notes wip * updated notes(prs done) * updated release notes. wip * Release notes * minor * minor fix * 1000/doc architecture commands (#1271) * sw architecture and setup commands * 1000/shm free obsolete (#1273) * freeing obsolete shm withoua a 'isValid' should access raw pointers. Need to move this all into the shm class * fixed obsolete shm free issue * minor * ensuring the test works platform independent for size of int * removed verify, update, fixed getUser to be a free function, generated commands, python bindings yet to do * python bindings * fixed tests * minor * minor * format * userdetails refinedg * fixed caller test * updated client api version (#1277) * one doesnt need to open shared memory to call removesharedmemory, and calling hasMemoryvalid without opening will cause segfault (not used now, but could in the future) * fix test on shm * minor * added image source files from draw.io to create the images (#1280) * 1000/fix_actual_tests (#1282) - fix acquire fail in tests (adcreg test) - roi tests fail after overlapping invalid test and acquire after - print udp dest mac in server properly - fixed udp dst list get (server was not sending entry proper size to match proper struct size in client) - updated server binaries and updated hard links in serverBin - added documentation regarding gui: zmqport and zmqip in terms of gui, rx_zmqstream - removed print - probably ended there for debuggung --------- Co-authored-by: Alice <alice.mazzoleni@psi.ch> * 1000/fix_m3_tests (#1286) * testing clkdiv one must ensure the exptime delay etc all are reset to the exact values for tests * change dac max values for vth values for m3 in client side (set module * 1000/doc_cmake (#1289) * more detail documentation in installation * more detail documentation in installation * added links to api examples * reverted back that vthreshold dacs in m3 have min and max as 200 and 2400 (#1294) * update release notes and date (#1298) --------- Co-authored-by: Mazzoleni Alice Francesca <mazzol_a@pc17378.psi.ch> Co-authored-by: Erik Fröjdh <erik.frojdh@gmail.com> Co-authored-by: AliceMazzoleni99 <alice.mazzoleni@psi.ch> Co-authored-by: Martin Mueller <martin.mueller@psi.ch> Co-authored-by: froejdh_e <erik.frojdh@psi.ch> Co-authored-by: Anna Bergamaschi <anna.bergamaschi@psi.ch>
6.4 KiB
Pattern
This is a test and development feature implemented only for Ctb, Xilinx_Ctb and Mythen3.
A pattern is a sequence of 64-bit words which is executed using a clock on the FPGA. Each of the 64 bits is connected to a pin or internal signal of the FPGA. The purpose of a pattern is to provide a way to change these 64 signals with precise timing. Commands run by the detector server could manipulate the same signals as the pattern, but they cannot enforce a change in a specific clock cycle.
Usage
A pattern is written to memory on the FPGA using the patword command.
patword 0x0000 0x000000000000000A
patword 0x0001 0x000000000000000B
patword 0x0002 0x000000000000000C
patword 0x0003 0x000000000000000D
patword 0x0004 0x000000000000000E
The example above writes a five-word pattern into FPGA memory. The first argument is the memory address, the second argument is the content to be written into this address. Before executing a pattern one has to set the address limits of the pattern:
patlimits 0x0000 0x0004
This instructs the firmware to execute the commands from address 0 to 4 (including 0 and 4). The execution can be started from the pyctbgui or with the commands
start [Ctb, Xilinx_Ctb]
patternstart [Mythen3]
The maximal number of patword addresses is 8192. However, it is possible to extend the length of the pattern sequence using loops and wait commands. Loops can be configured with the following commands:
patloop 0 0x0001 0x0003
patnloop 0 7
The first argument of both commands is the ID of the loop. Ctb and Xilinx_Ctb can have 6 loops (ID 0-5), Mythen3 can have 4 loop definitions. The commands above configure the loop with ID 0 to run 7 times and jump from the patword with address 3 to the patword with address 1. Important: If patnloop is set to 1, the addresses 0x1-0x3 will execute exactly once; if it is set to 0, the pattern addresses will be skipped.
The same idea is used to introduce wait times. The example below causes the patword at address 0x0002 to be active for 9 clock cycles before the execution continues.
patwait 0 0x0002
patwaittime 0 9
Waits can be placed inside a loop and loops can be nested.
patioctrl
The function of each bit in the sequence of 64-bit words depends on the connected detector and firmware version. Some of the 64 bits might connect directly to pads of a chip. The patioctrl command is used to configure the direction of some of these signals (not all of them !! See tables below). Signals where the corresponding bit in the argument of patioctrl is set to 1 will be driven from the FPGA.
patsetbit and patmask
The functions patsetbit and patmask can be used to ignore a specific bit of the pattern. Example:
patmask 0x0101
patsetbit 0x0001
Patmask configures bit 0 and 8 of the pattern to be set to their value in patsetbit. These bits will be ignored during pattern execution and will always be 0 (bit 8) and 1 (bit 0).
The mappings of bit positions in the pattern word to signals/pads of the FPGA are listed below for the three detector types where patterns are used. In the case of the two CTB's, connections of the signals to actual pads of a chip depend on the layout of the used detector adapter board. Therefore, each type of detector adapter board adds an additional mapping layer.
CTB Pattern Bit Mapping
| 63 | 62 | 61-57 | 56 |
|
|
|
|
|
|
|
EXTIO | DO, stream source | DIO |
DIO: Driving the 32 FPGA pins corresponding to the lowest 32 bits of the patioctrl command. If bits in patioctrl are 0, the same bit positions in DIO will switch to input pins and connect to dbit sampling. Additionally, some of these 32 bits have an automatic override by detector-specific statemachines which is active whenever one of these statemachines is running (currently bits 7,8,11,14 and 20).
DO: Directly connected to 16 FPGA pins. Output only. Not influenced by patioctrl. Also connected to bit 47-32 in all Ctb dbit samples. All of them can be used as dbit sample trigger. In addition, every bit of DO can be selected as trigger for sending out a udp packet with samples to the receiver.
EXTIO: Similar to DIO, but not used as input to the fpga. With the corresponding patioctrl bits set to 0 these pins will switch to a high impedance mode and be ignored by the firmware.
T: trigger output
D: enable signal for digital sampling
A: adc enable
Xilinx_CTB Pattern Bit Mapping
| 63-32 |
|
|
DIO |
DIO: Driving the 32 FPGA pins corresponding to the lowest 32 bits of the patioctrl command. If bits in patioctrl are 0, the same bit positions in DIO will switch to input pins and connect to dbit sampling. Additionally, some of these 32 bits have an automatic override by detector-specific statemachines which is active whenever these sm's are running (currently bits 7,8,11,14 and 20).
Mythen3 Pattern Bit Mapping
| 63-33 |
|
31-25 |
|
23 |
|
21 | 20 | 19 |
|
signARD |
|
CHSclk |
|
|
STATLOAD | STO | SIN |
| 18 | 17 | 16-14 | 13 | 12 | 11 | 10 | 9-0 |
| SR_MODE | clk | EN | PULSE | RD | CHSIN | ANAMode | TBLOAD |
For Mythen3 the pattern word only connects to output pins of the FPGA when the pattern is running. Afterwards the signals will switch back to other logic in the FPGA. Both CTB's hold the last executed pattern word until a new pattern is started.