Hi Torbjörn， >(1) Why do you set up paths for zen (as a fallback)?
>Doing that seems wrong unless all these 3 CPUs support every zen >instruction. Do they? >Also, passing k8 to the compiler adn choosing zen asm code makes very >little sense to me. If zen makes sense for asm code, why does it not >make sence for compiler-generated code? >A more conservative strategy is to pick file-by-file from the desired >x86_64 subdir. E.g., if mpn_mul_basecase from x86_64/foo runs great for >the kx6000 CPU, do >MULFUNC_PROLOGUE(mpn_mul_basecase) >include_mpn(`x86_64/foo/mul_basecase.asm') >in x86_64/kx6000/mul_basecase.asm. >(At some point, providing specific asm code for these CPUs might be >desirable, of course.) I try to use paths for zen and k8 compiler parameter. This runs great in test. I now realize that this is a wrong setup. I'm going to modify it with the strategy you recommend. >(2) Testing specific stepping. >You check for stepping 0xE specifically. I assume the "nano" CPU has >lower stepping values than 0xE, and Zhaoxin variants currently have >exactly 0xE. It then feels more future-proof to check >= 0xE, as that >would allow Zhaoxin to do stepping iterations without GMP silently >falling back to "nano" for such steppings. Right. Zhaoxin will do iterations by more model values. The "nano" CPU and the "zxc" CPU have the same model (model=15), so I distinguish between them by stepping. Thanks Dylan _______________________________________________ gmp-devel mailing list gmp-devel@gmplib.org https://gmplib.org/mailman/listinfo/gmp-devel
