Xavier Leroy

Results 360 comments of Xavier Leroy

Thanks for the detailed analysis. Allocation of multiple blocks (as opposed to OCaml 4's "allocate one big block and carve sub-blocks inside" strategy) is probably unavoidable in OCaml 5, but...

Yes, let's fix this for 5.3. But the problem has been there since 5.0, right?

I need to be convinced this is not a regression compared to just using winpthreads, or a more modern implementation of the POSIX API for Win32. There are benefits to...

Somewhat related (but not directly): #12243 .

I hope ocamltest's output format can be made not only more suitable for CI tools, but also more readable for humans. What about having only one line of output for...

When I implemented hash table randomization back in 2012, I wanted to make it the default (and actually the only) mode. I was quickly dissuaded by two serious industrial users...

Well spotted, thank you for the report. I think the main offenders here are the `tbz` and `tbnz` instructions, since they are limited to +/- 32K. The other conditional branches...

@monniaux : thanks for the Chamois plug, but branch relaxation for PowerPC was implemented 14 years ago in 14b44ae202de640f2a7a6e973e5fb912b736c002, so nothing is new under the sun. @bagnalla: thanks for the...

Note that the runtime library functions in `runtime/x86_64` are not annotated with `endbr64` instructions, as these functions are intended to be called directly, but not through a function pointer.

Also: IBT support is for 64-bit x86 only. Intel processors support IBT in 32-bit mode as well, but I don't think it's worth supporting in CompCert.