Support AVX for 32 bit platforms

_mm256_extract_epi64 is not available for 32 bit platforms,
but it can be replaced by "a very simple workaround".

Signed-off-by: Stefan Weil <sw@weilnetz.de>

Author: stweil

arch/dotproductavx.cpp

@@ -16,9 +16,8 @@
 // limitations under the License.
 ///////////////////////////////////////////////////////////////////////
 
-#if !defined(__AVX__) || defined(__i386__)
+#if !defined(__AVX__)
 // Implementation for non-avx archs.
-// Also used for 32 bit AVX archs because of missing _mm256_extract_epi64.
 
 #include "dotproductavx.h"
 #include <stdio.h>
@@ -92,13 +91,13 @@ double DotProductAVX(const double* u, const double* v, int n) {
   // fool the instrinsics into thinking we are extracting the bottom int64.
   auto cast_sum = _mm256_castpd_si256(sum);
   *(reinterpret_cast<inT64*>(&result)) =
-#ifndef _WIN32
-      _mm256_extract_epi64(cast_sum, 0)
-#else
-      // this is a very simple workaround that probably could be activated
-      // for all other platforms that do not have _mm256_extract_epi64
+#if defined(_WIN32) || defined(__i386__)
+      // This is a very simple workaround that is activated
+      // for all platforms that do not have _mm256_extract_epi64.
       // _mm256_extract_epi64(X, Y) == ((uint64_t*)&X)[Y]
       ((uint64_t*)&cast_sum)[0]
+#else
+      _mm256_extract_epi64(cast_sum, 0)
 #endif
       ;
   while (offset < n) {

configure.ac

@@ -119,13 +119,9 @@ esac
 AM_CONDITIONAL([AVX_OPT], false)
 AM_CONDITIONAL([SSE41_OPT], false)
 
-# The current implementation for AVX uses 64 bit code.
-AC_CHECK_SIZEOF([void *])
-if test "$ac_cv_sizeof_void_p" = "8"; then
-    AX_CHECK_COMPILE_FLAG([-mavx], [avx=true], [avx=false])
-    if $avx; then
-        AM_CONDITIONAL([AVX_OPT], true)
-    fi
+AX_CHECK_COMPILE_FLAG([-mavx], [avx=true], [avx=false])
+if $avx; then
+    AM_CONDITIONAL([AVX_OPT], true)
 fi
 
 AX_CHECK_COMPILE_FLAG([-msse4.1], [sse41=true], [sse41=false])