Re: question about linker

On 06/12/2024 17:47, David Brown wrote:
 > On 05/12/2024 18:42, Bart wrote:
 >> A big one is when C is used an intermediate language for a compiler.
 >
 > Why would that be relevant?  If you compile language X by first transpiling to C, then compiling C, then the time to compile language X is the sum of these.  A fast or slow C compiler is no more and no less relevant here than if you were just compiling C.
You don't give do you?!
If your task to get from A to B was split into two, you'd be happy to do the first part by a fast car, then complete the rest of it on a horse and cart, for no reason at all?
After all the journey time is simply the sum of two halves!
Obviously you'd want both halves to be done as quickly as possible.
Compiler front-ends can be very fast, so you don't want to hit a brick wall on the backend. This is a common complaint of the LLVM backend which can take up 90% of compile-time.
While people may want faster execution than is possibly with an interpreter, they don't necessarily need the fastest possible native code. ANY native code will be sufficient.
With a C intermediate, using tcc for the final stage can easily give 10 times faster throughput than using even gcc-O0.
If one day they decide they need fast production version, THEN they pass it through gcc-O3.
What IS your problem with that?
Once the C-generation has been debugged, then there should be nothing wrong with the C code that requires detailed analysis.
Your hatred of tcc (I won't even mention my stuff) is bordering on the irrational.