On 10/19/23 11:22, Laszlo Ersek wrote:
> On 10/19/23 08:48, Dhaval Sharma wrote:

>> (11) I agree that we should use symbolic names rather than
>> magic constants, but raw encodings of machine instructions don't belong
>> into a
>>      C header file. [Dhaval] This bytecode was introduced thinking what
>> if all compilers do not support it. but given the default compiler in
>> edk2 GCC 12 supports it
>>      we can eliminate this byte encoding completely to make it easy and
>> simple to consume for others.
>
> To be honest, I can't determine the minimum expected gcc version for
> edk2. "BaseTools/Conf/tools_def.template" states a minimum version for
> NASM, for example, but I can't find a similar gcc requirement there.
>
> gcc-12 does work for me personally, because my riscv cross-compiler is
> "riscv64-linux-gnu-gcc (GCC) 12.1.1 20220507 (Red Hat Cross 12.1.1-1)".
>
> If the CI environment that builds these patches also provides gcc-12+,
> then I figure you should be set.

Wait, for the assembly language source files, what matters is the
binutils version, not the gcc version. Mine is "GNU assembler version
2.38-3.el9" (from "binutils-riscv64-linux-gnu-2.38-3.el9.x86_64").

Is that sufficient for the instuctions in question?

(More generally -- what version does our CI env expect / provide?)

Thanks
Laszlo