[martin-frbg]

Needs testing as the implementation of cpuid_count may break some systems. Not sure if there is an easier solution to the problem of passing subleaf values to cpuid page 4. (This is intended to fix #1232 at some point)

[wangshankun]

Yes, it's work

./getarch 1
#define HASWELL
#define L1_CODE_SIZE 32768
#define L1_CODE_ASSOCIATIVE 8
#define L1_CODE_LINESIZE 64
#define L1_DATA_SIZE 32768
#define L1_DATA_ASSOCIATIVE 8
#define L1_DATA_LINESIZE 64
#define L2_SIZE 262144
#define L2_ASSOCIATIVE 8
#define L2_LINESIZE 64
#define ITB_SIZE 2097152
#define ITB_ASSOCIATIVE 0
#define ITB_ENTRIES 8
#define DTB_SIZE 4096
#define DTB_ASSOCIATIVE 4
#define DTB_DEFAULT_ENTRIES 64
#define HAVE_CMOV
#define HAVE_MMX
#define HAVE_SSE
#define HAVE_SSE2
#define HAVE_SSE3
#define HAVE_SSSE3
#define HAVE_SSE4_1
#define HAVE_SSE4_2
#define HAVE_AVX
#define HAVE_FMA3
#define HAVE_CFLUSH
#define NUM_SHAREDCACHE 1
#define NUM_CORES 1
#define CORE_HASWELL
#define CHAR_CORENAME "HASWELL"