volk: added xgetbv stuff from volk_work to maint

This ensures that the compiler has support for xgetbv.
This also fixes MSVC by checking for _xgetbv.
Also, restored copy of cpuid.h, this should not be modified.

1 files changed, 21 insertions, 0 deletions

diff --git a/volk/gen/make_cpuid_c.py b/volk/gen/make_cpuid_c.py
index c6bb5059e..674b4fb2e 100644
--- a/volk/gen/make_cpuid_c.py
+++ b/volk/gen/make_cpuid_c.py
@@ -41,11 +41,32 @@ struct VOLK_CPU volk_cpu;
 #include <gcc_x86_cpuid.h>
 #define cpuid_x86(op, r) __get_cpuid(op, (unsigned int *)r+0, (unsigned int *)r+1, (unsigned int *)r+2, (unsigned int *)r+3)
 
+    /* Return Intel AVX extended CPU capabilities register.
+     * This function will bomb on non-AVX-capable machines, so
+     * check for AVX capability before executing.
+     */
+    #if defined(__GNUC_PREREQ) && __GNUC_PREREQ(4, 4)
+    static inline unsigned long long _xgetbv(unsigned int index){
+        unsigned int eax, edx;
+        __asm__ __volatile__("xgetbv" : "=a"(eax), "=d"(edx) : "c"(index));
+        return ((unsigned long long)edx << 32) | eax;
+    }
+    #define __xgetbv() _xgetbv(0)
+    #else
+    #define __xgetbv() 0
+    #endif
+
 //implement get cpuid for MSVC compilers using __cpuid intrinsic
 #elif defined(_MSC_VER)
 #include <intrin.h>
 #define cpuid_x86(op, r) __cpuid(r, op)
 
+    #if defined(_XCR_XFEATURE_ENABLED_MASK)
+    #define __xgetbv() _xgetbv(_XCR_XFEATURE_ENABLED_MASK)
+    #else
+    #define __xgetbv() 0
+    #endif
+
 #else
 #error "A get cpuid for volk is not available on this compiler..."
 #endif