SCons: Properly set SSE2 as baseline on x86_32

[akien-mga]

Setting it only for release templates on Windows and macOS was inconsistent, and Jolt requires it as a minimum.

Drop the -mxsave flag from the raycast module, this doesn't seem to be used explicitly by Embree, and unnecessarily makes our config and baseline muddy.

I'm aware of #59595, but this one requires more discussion as requiring SSE4.2 does cut off a significant amount of CPUs, and we need to evaluate whether we're willing to make that compromise.

SSE2 on the other hand should be a no brainer, it was already the default for MSVC, and for all other platforms indirectly by way of it being set for the raycast module, which would mean that our binaries can't run on i386 machines without SSE2 already.

[hpvb]

At the risk of bikeshedding, I think we might as well just set -mssse3 that is supported by every CPU that has at least 2 cores. The newest cpu to support it is AMD's "bobcat" architecture which is from early 2011.

Every single "atom" branded CPU has it too. On the Intel side this makes the oldest required CPU a "Core 2 duo" or a 64bit capable P4.

Now I don't want this to bikeshed forever, I am just approving this. But I would also just approve an -mssse3 if presented!

[akien-mga]

At the risk of bikeshedding, I think we might as well just set -mssse3 that is supported by every CPU that has at least 2 cores. The newest cpu to support it is AMD's "bobcat" architecture which is from early 2011.

Every single "atom" branded CPU has it too. On the Intel side this makes the oldest required CPU a "Core 2 duo" or a 64bit capable P4.

I would agree too. I propose we do this in two steps so we get an intermediate dev snapshot inbetween for testing purposes, if we need to debug reports of some users having Godot suddenly crash on their old PC.

This PR just affects x86_32 in a way that should practically be a no-op.

Then we can make another PR bumping to ssse3 that will affect both x86_32 and x86_64.
Or would we bump to ssse3 only for x86_64?

[hpvb]

Then we can make another PR bumping to ssse3 that will affect both x86_32 and x86_64. Or would we bump to ssse3 only for x86_64?

I'd do both x86_32 and x86_64. There's some reason to want to run a 32bit linux distro in particular on a machine that old. Moving from SSE2 to SSSE3 in practice doesn't really limit the amount of CPUs that can run Godot much either way.

[bruvzg]

For the reference - current min. supported macOS version (10.13) require SSE4.1 to work.

[Repiteo]

Thanks!