On 08/11/2025 18:54, Kito Cheng wrote:
I incline to set default abi by march if mabi is not given, rather than by mcpu.
one point is I don’t want to introduce an incompatible behavior with clang, and
the rule of
inference abi from march is already in clang.
makes sense (didn't know of that behaviour), and besides, I looked at the other
cpu's arch specs to figure out the abi!
Nathan Sidwell <[email protected] <mailto:[email protected]>>於 2025年11月9日 週日，
01:43寫道：
We have multiple cpus with different abis (combinations of 32 vs 64, float
vs
no-float, etc). It's annoying having to specify both -mcpu=thing and the
correct -mabi=thing. Just as with -mcpu selecting a default ISA set, this
adds
smarts for it to select a default ABI.
You can still specify -mabi if you want to go off-road, but this gives us
one
source of truth. It's possible I've made a mistake with selecting the
default
abi for the current set of cpus.
Also while there, cleaned up riscv_expend_arch_from_cpu. I found it over
complex, and as it's potentially getting data from a user-provided specs
file,
best not to use asserts for errors, just fail safely.
nathan
--
Nathan Sidwell
--
Nathan Sidwell
