Peter Maydell writes: > On 5 April 2016 at 13:57, Lluís Vilanova <vilan...@ac.upc.edu> wrote: >> Emilio G Cota writes: >> >>> I'm assuming windows compilers don't support this attribute. >>> Signed-off-by: Emilio G. Cota <c...@braap.org> >>> --- >>> include/qemu/compiler.h | 10 ++++++++++ >>> 1 file changed, 10 insertions(+) >> >>> diff --git a/include/qemu/compiler.h b/include/qemu/compiler.h >>> index 8f1cc7b..fb946f1 100644 >>> --- a/include/qemu/compiler.h >>> +++ b/include/qemu/compiler.h >>> @@ -41,6 +41,16 @@ >>> # define QEMU_PACKED __attribute__((packed)) >>> #endif >> >>> +#define QEMU_CACHELINE (64) >> [...] >> >> You should make this a value taken from configure. In linux you could use >> something like: >> >> getconf LEVEL1_DCACHE_LINESIZE >> >> Or an equivalent program using sysconf(3) if getconf is not present.
> The cache line size at compile time is not necessarily the cache > line size at runtime... True, but exposing it as a configure flag allows tunning for the target arch during cross-compilation. It would be bettwe to use the compiler to tell you about the target cache line size, but I'm not aware of a way to do it. Got it! gcc -march=native --help=params -v 2>&1 | grep "param l1-cache-line-size" | sed -e 's/.* --param l1-cache-line-size=\([0-9]\+\) .*/\1/' Cheers, Lluis
