HDAWG: Multiple issues during .elf file upload #751
Description
There are multiple issues present with the upload of .elf files. In the default case that auto-upload is enabled, those issues only manifest in error messages, but the upload itself works. With auto-upload disabled no error is thrown but .elf files are also not uploaded. However, this is compensated by another bug which makes it difficult to actually deactivate auto-upload.
Auto-Upload enabled
The following lines:
HardwareSetup.remove_program("playground_program")
HardwareSetup.register_program("playground_program", pulse_program, run_callback=awg.channel_tuples[0].run_current_program)
HardwareSetup.run_program("playground_program")Cause the following error:
...
File successfully uploaded
Uploading ELF file to device DEV8212, core 3
File successfully uploaded
Output exceeds the [size limit](command:workbench.action.openSettings?[). Open the full output data [in a text editor](command:workbench.action.openLargeOutput?c0865e13-c6e5-4064-8c7e-461877ce63d8)
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
~AppData\Local\Temp\ipykernel_21684\1993656821.py in <module>
9 my_awg.awg_module.set('compiler/upload', 1)
10
---> 11 HardwareSetup.run_program("playground_program")
c:\users\beer\repos\qupulse\qupulse\hardware\setup.py in run_program(self, name)
221 def run_program(self, name) -> None:
222 """Calls arm program and starts it using the run callback"""
--> 223 self.arm_program(name)
224 self._registered_programs[name].run_callback()
225
c:\users\beer\repos\qupulse\qupulse\hardware\setup.py in arm_program(self, name)
212 for awg in self.known_awgs:
213 if awg in awgs_to_upload_to:
--> 214 awg.arm(name)
215 else:
216 # The other AWGs should ignore the trigger
c:\Users\Beer\venvs\qupulse\lib\site-packages\autologging.py in autologging_traced_instancemethod_delegator(self_, *args, **keywords)
1039 return proxy(method, args, keywords)
1040 else:
-> 1041 return method(*args, **keywords)
...
-> 1007 _, status = self._upload_job
1008 if isinstance(status, int):
1009 assert status in (-1, 0, 1)
TypeError: cannot unpack non-iterable NoneType objectThis is caused by self._upload_job in zihdawg.py still being at it's default value, being None. This can relatively easily be fixed by modifying ELFManager._update_upload_job_status while still keeping validation of successful upload in the follow way:
...
def _update_upload_job_status(self):
elf_upload = self.awg_module.elf_upload
elf_file = self.awg_module.elf_file
old_status = None
if self._upload_job is None:
assert not elf_upload
else:
elf_file, old_status = self._upload_job
assert self.awg_module.elf_file == elf_file
if isinstance(old_status, float) or old_status is None:
status_int, progress = self.awg_module.elf_status
...While this is not changed the problem can be circumvented on the user side by first arming the program, which works but throws the same error, which can now be caught, and then running the program:
try:
HardwareSetup.arm_program("playground_program")
except TypeError:
print("Catch Type error due to bug when using auto upload")
HardwareSetup.run_program("playground_program")I am unsure while this error does not occur for other users. If my solution is satisfactory I can create a pull request to deploy the fix.
Auto-Upload disabled
More problematic is what happens when auto-upload is disabled, which can be seen by this:
HardwareSetup.remove_program("playground_program")
HardwareSetup.register_program("playground_program", pulse_program, run_callback=awg.channel_tuples[0].run_current_program)
for awg_i in HardwareSetup.known_awgs:
my_awg = awg_i
my_awg.awg_module.set('compiler/upload', 0)
HardwareSetup.run_program("playground_program")not causing an error on program run. This is a problem however because it turns out that the the program is not actually uploaded. This can easily be checked by either observing no output signal from the awg or by checking that sequence on devXXXX in LabOne stays empty. This is caused by the following:
Upon this:
def _upload(self, elf_file) -> Generator[str, str, None]:
if self.awg_module.compiler_upload:
pass
else:
self._start_elf_upload(elf_file)
...Manual upload is supposed to be triggered if compiler/upload is 0 by zihdawg._start_elf_upload(elf_file). In there it is checked wehter the program is already uploaded in order to not perform unnecessary uploads by the following:
def _start_elf_upload(self, elf_file):
...
current_elf = self.awg_module.elf_file
if current_elf != elf_file:
... (start upload)The problem is that current_elf != elf_file is always False as elf/file does not return the uploaded file but rather the compiled one as current_elf. This can be seen in the docs:
As elf_file is set when instructing the ZIdriver to compile the file, these values are always equal (see top-left corner):
It does not appear possible to me to get the actually uploaded .elf file to check for successful upload. Note that this must be possible in some way as apparently LabOne can do this as the sequence on devXXXX-tab suggests.
Possible solutions:
- Disable manual-uploads and fail if aut-upload is disabled with an error
- Always upload whenever the file that is compiled changes or is set for the first time
The reason while this did not occur earlier:
It appears to be impossible to disable Auto Upload from the LabOne interface in 20.07, the Auto Upload button simply does nothing. Unsetting it resets to be set on page refresh (which is different to other settings) and changing it's state has no influence on compiler/upload, it is always 1 if not set from the API. Setting it to 0 from the API does also have no influence on LabOne while the result is apparently registered and applied in the device itself. Therefore Auto-Upload is always enabled for almost all users.