run idle kernel on flash #325

srenblad · 2024-09-12T17:51:32+08:00

srenblad commented

2024-09-12 17:51:32 +08:00

Run idle kernel on flash

Implements https://github.com/m-labs/artiq/pull/2569 for artiq-zynq as closely as possible.
Related issue: Discussed in https://github.com/m-labs/artiq/issues/2041.

At the moment the only difference in idle kernel behavior between Zynq and RISC-V is:

In RISC-V, idle kernels are restarted when they finish or encounter any error except KernelNotFound.
In Zynq, idle kernels only restart on artiq_coremgmt config calls, after an experiment is run or on startup.

I don't hold a strong preference either way. Thoughts @mwojcik, @sb10q?

Testing

All tests performed with the Kasli-SoC v1.1 demo.json variant.

idle kernel restarted with artiq_coremgmt config write idle_kernel ...
idle kernel stops running after artiq_coremgmt config remove idle_kernel
stored idle kernel starts on startup
idle kernel starts after experiment is run
sequential artiq_coremgmt config write/remove idle_kernel ... calls are properly handled

## Run idle kernel on flash Implements https://github.com/m-labs/artiq/pull/2569 for artiq-zynq as closely as possible. Related issue: Discussed in https://github.com/m-labs/artiq/issues/2041. At the moment the only difference in idle kernel behavior between Zynq and RISC-V is: + In RISC-V, idle kernels are restarted when they finish or encounter any error except `KernelNotFound`. + In Zynq, idle kernels only restart on `artiq_coremgmt config` calls, after an experiment is run or on startup. I don't hold a strong preference either way. Thoughts @mwojcik, @sb10q? ## Testing All tests performed with the Kasli-SoC v1.1 demo.json variant. - [x] idle kernel restarted with `artiq_coremgmt config write idle_kernel ...` - [x] idle kernel stops running after `artiq_coremgmt config remove idle_kernel` - [x] stored idle kernel starts on startup - [x] idle kernel starts after experiment is run - [x] sequential `artiq_coremgmt config write/remove idle_kernel ...` calls are properly handled

srenblad added 1 commit 2024-09-12 17:51:32 +08:00

491e426222 run idle kernel on flash

mwojcik commented

2024-09-12 21:53:38 +08:00

I believe "idle" would mean the device would be running it, when otherwise it would be doing nothing; RISC-V behaviour seems more fitting that the kernel would fully fill the idle state. Even weirder if any irrelevant artiq_coremgmt config call would cause the idle kernel to restart;

I believe "idle" would mean the device would be running it, when otherwise it would be doing nothing; RISC-V behaviour seems more fitting that the kernel would fully fill the idle state. Even weirder if any irrelevant ``artiq_coremgmt config`` call would cause the idle kernel to restart;

👍 1

srenblad commented

2024-09-13 09:18:50 +08:00

Even weirder if any irrelevant artiq_coremgmt config call would cause the idle kernel to restart;

Just to clarify, only config calls that edit the idle_kernel cause a restart.

> Even weirder if any irrelevant artiq_coremgmt config call would cause the idle kernel to restart; Just to clarify, only config calls that edit the idle_kernel cause a restart.

👍 1

srenblad changed title from ~~WIP: run idle kernel on flash~~ to run idle kernel on flash

2024-09-13 09:26:49 +08:00

srenblad added 1 commit 2024-09-13 09:46:45 +08:00

e7614d2e8e rerun idle kernel on finish

srenblad commented

2024-09-13 09:47:23 +08:00

Updated with idle kernel rerunning on finish.

sb10q merged commit e7614d2e8e into master

2024-09-14 11:38:03 +08:00

srenblad referenced this pull request

2024-09-19 10:10:33 +08:00