Closed TommyMurphyTM1234 closed 1 week ago
I don't know what triggered this?
This is like related to changes in upstream spike. If you look at the current changelog there is lot's of changes related to how RVV configuration is handled.
I don't know why the spike32.py uses a different ISA specification (including V) to other targets and why there isn't more consistency here?
I think this we should discuss this in https://github.com/riscv-software-src/riscv-tests/issues/568
I don't know why the [CI runs](https://github.com/riscv-collab/riscv-openocd/actions) don't seem to be affected (yet?)?
The current TOT in spike is 1b1a33376 . Which correspond to the latest pipeline we have in OpenOCD. It seems that you spike version is deprecated.
@TommyMurphyTM1234
This is like related to changes in upstream spike. If you look at the current changelog there is lot's of changes related to how RVV configuration is handled.
In addition to my point above you can also see here: https://github.com/riscv-software-src/riscv-tests/commit/e06a435c1e545def71e833031356372f0828f165 . So yes, this is changes in upstream spike
Thanks @aap-sc.
The current TOT in spike is 1b1a33376 . Which correspond to the latest pipeline we have in OpenOCD. It seems that you spike version is deprecated.
I just used the version that the riscv-gnu-toolchain
repo uses as a submodule. So maybe that needs to be bumped over there?
I just used the version that the riscv-gnu-toolchain repo uses as a submodule. So maybe that needs to be bumped over there?
Have no opinion on the matter since we don't use riscv-gnu-toolchain
:(
As mentioned above, riscv-gnu-toolchain
currently uses https://github.com/riscv-software-src/riscv-isa-sim/commit/20a2b6d05a562236a32ca3ce1642a2bdbbd907a0 for its Spike submodule. This is about three months old and predates the changes related to V extension handling.
I did a local build of Spike at the current top of tree commit - https://github.com/riscv-software-src/riscv-isa-sim/commit/f7d0dba6012d49cb75b2f175a31a85cf4b2db3b3 - and with this version I no longer have the aforementioned spike32.py
test failures.
So the issue is indeed that riscv-gnu-toolchain
uses a Spike that is not up to date with the requirements of the OpenOCD tests.
I will see if I can safely bump riscv-gnu-tolchain
's Spike submodule to use the latest Spike commit. I need to run the tests suites for the bare-metal/Newlib and Linux toolchains just to check that there is no adverse impact from doing this.
Addressed by this PR:
which is also discussed in this issue:
I am using the following:
riscv-gnu-toolchain
built from commit 763c862riscv-openocd
built from commit c6bb902629924eb66aae2a08c0ab8654261c9d71riscv-tests
at commit b0eb63aand when I run the Spike32 tests:
all tests fail:
The problem seems to be this:
But...
spike32.py
uses a different ISA specification (includingV
) to other targets and why there isn't more consistency here?