Removing LLD breaks kernel #23

New Issue

pca006132 · 2020-06-29T16:53:29+08:00

pca006132 commented

2020-06-29 16:53:29 +08:00

Removing the dependency for lld would break the kernel for some functions and exception.
From what I've tested, it breaks when delay is called or an exception is raised.
Adding lld_9 back to shell.nix as dependency would fix it.

Minimal example with the latest master:

from artiq.experiment import *


class BlinkForever(EnvExperiment):
    def build(self):
        self.setattr_device("core")
        self.setattr_device("led0")

    @kernel
    def run(self):
        self.core.reset()
        print("OK!")
        delay(100*ms)

With lld_9 in shell.nix:

OK!
(terminated)

Log:

[     9.137520s] DEBUG(runtime::comms): received connection
[     9.152439s] DEBUG(runtime::kernel): Core1 started
[     9.157213s] DEBUG(runtime::kernel): FPU enabled on Core1
[     9.162800s] DEBUG(dyld): ELF target: 12336 bytes, align to 1000, allocated at 00161000
[     9.170824s] DEBUG(dyld): Relocating 0 rela, 0 rel, 64 pltrel
[     9.176592s] DEBUG(runtime::kernel): kernel loaded
[     9.182102s] DEBUG(runtime::kernel): kernel starting
[     9.187052s] DEBUG(runtime::rpc): send<2>(String)->None
[     9.193567s] DEBUG(runtime::rpc): recv ...->None
[     9.198178s] DEBUG(runtime::rpc): send<1>(None)->None
[     9.203216s] DEBUG(runtime::kernel): kernel finished
[     9.208976s] DEBUG(runtime::comms): peer closed connection

Without lld_9:

OK!
(not terminated)

Log:

[     7.110199s] DEBUG(runtime::comms): received connection
[     7.124930s] DEBUG(runtime::kernel): Core1 started
[     7.129706s] DEBUG(runtime::kernel): FPU enabled on Core1
[     7.135351s] DEBUG(dyld): ELF target: 16388 bytes, align to 1000, allocated at 00161000
[     7.143430s] DEBUG(dyld): Relocating 0 rela, 0 rel, 64 pltrel
[     7.149196s] DEBUG(runtime::kernel): kernel loaded
[     7.154518s] DEBUG(runtime::kernel): kernel starting
[     7.159466s] DEBUG(runtime::rpc): send<2>(String)->None

Removing the dependency for lld would break the kernel for some functions and exception. From what I've tested, it breaks when `delay` is called or an exception is raised. Adding `lld_9` back to `shell.nix` as dependency would fix it. Minimal example with the latest master: ```python from artiq.experiment import * class BlinkForever(EnvExperiment): def build(self): self.setattr_device("core") self.setattr_device("led0") @kernel def run(self): self.core.reset() print("OK!") delay(100*ms) ``` With `lld_9` in `shell.nix`: ``` OK! (terminated) ``` Log: ``` [ 9.137520s] DEBUG(runtime::comms): received connection [ 9.152439s] DEBUG(runtime::kernel): Core1 started [ 9.157213s] DEBUG(runtime::kernel): FPU enabled on Core1 [ 9.162800s] DEBUG(dyld): ELF target: 12336 bytes, align to 1000, allocated at 00161000 [ 9.170824s] DEBUG(dyld): Relocating 0 rela, 0 rel, 64 pltrel [ 9.176592s] DEBUG(runtime::kernel): kernel loaded [ 9.182102s] DEBUG(runtime::kernel): kernel starting [ 9.187052s] DEBUG(runtime::rpc): send<2>(String)->None [ 9.193567s] DEBUG(runtime::rpc): recv ...->None [ 9.198178s] DEBUG(runtime::rpc): send<1>(None)->None [ 9.203216s] DEBUG(runtime::kernel): kernel finished [ 9.208976s] DEBUG(runtime::comms): peer closed connection ``` ----- Without `lld_9`: ``` OK! (not terminated) ``` Log: ``` [ 7.110199s] DEBUG(runtime::comms): received connection [ 7.124930s] DEBUG(runtime::kernel): Core1 started [ 7.129706s] DEBUG(runtime::kernel): FPU enabled on Core1 [ 7.135351s] DEBUG(dyld): ELF target: 16388 bytes, align to 1000, allocated at 00161000 [ 7.143430s] DEBUG(dyld): Relocating 0 rela, 0 rel, 64 pltrel [ 7.149196s] DEBUG(runtime::kernel): kernel loaded [ 7.154518s] DEBUG(runtime::kernel): kernel starting [ 7.159466s] DEBUG(runtime::rpc): send<2>(String)->None ```

pca006132 commented

2020-06-30 10:35:19 +08:00

Poster

OK this is due to incorrect version of the artiq compiler, updating the nix channel solves the issue.

pca006132 closed this issue

2020-06-30 10:35:22 +08:00

Sign in to join this conversation.