Having the default avoids having builds which don't work in some endpoints (controller built for newer CPU than the endpoint).
Supporting a micro-architecture means that
- we need to accurately identify what the micro-architecture of the endpoint is. We probably need to rely on gcc for this, as that is what's deciding what optimizations to use.
- We need to include the micro-architecture in the md5sum for the container image tag
- We might need to support running workshop on the endpoint, especially if the userenv is different (and have a different compiler version) and if the cpu architecture is completely different (or we have to cross-compile). This could also vastly speed up container builds if we have to build for many endpoints.
Having the default avoids having builds which don't work in some endpoints (controller built for newer CPU than the endpoint).
Supporting a micro-architecture means that