From 4437d6e1f3107a20a8c7b66be8b4b972a82e3b28 Mon Sep 17 00:00:00 2001
From: Pieter Wuille <pieter.wuille@gmail.com>
Date: Wed, 27 Sep 2017 01:45:12 -0700
Subject: 8-way AVX2 implementation for double SHA256 on 64-byte inputs

---
 configure.ac | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

(limited to 'configure.ac')

diff --git a/configure.ac b/configure.ac
index 7396ac74e8..42d8bb76ea 100644
--- a/configure.ac
+++ b/configure.ac
@@ -305,6 +305,7 @@ fi
 # compatibility.
 AX_CHECK_COMPILE_FLAG([-msse4.2],[[SSE42_CXXFLAGS="-msse4.2"]],,[[$CXXFLAG_WERROR]])
 AX_CHECK_COMPILE_FLAG([-msse4.1],[[SSE41_CXXFLAGS="-msse4.1"]],,[[$CXXFLAG_WERROR]])
+AX_CHECK_COMPILE_FLAG([-mavx -mavx2],[[AVX2_CXXFLAGS="-mavx -mavx2"]],,[[$CXXFLAG_WERROR]])
 
 TEMP_CXXFLAGS="$CXXFLAGS"
 CXXFLAGS="$CXXFLAGS $SSE42_CXXFLAGS"
@@ -347,6 +348,25 @@ AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[
 )
 CXXFLAGS="$TEMP_CXXFLAGS"
 
+TEMP_CXXFLAGS="$CXXFLAGS"
+CXXFLAGS="$CXXFLAGS $AVX2_CXXFLAGS"
+AC_MSG_CHECKING(for AVX2 intrinsics)
+AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[
+    #include <stdint.h>
+    #if defined(_MSC_VER)
+    #include <immintrin.h>
+    #elif defined(__GNUC__) && defined(__AVX2__)
+    #include <x86intrin.h>
+    #endif
+  ]],[[
+    __m256i l = _mm256_set1_epi32(0);
+    return _mm256_extract_epi32(l, 7);
+  ]])],
+ [ AC_MSG_RESULT(yes); enable_avx2=yes; AC_DEFINE(ENABLE_AVX2, 1, [Define this symbol to build code that uses AVX2 intrinsics]) ],
+ [ AC_MSG_RESULT(no)]
+)
+CXXFLAGS="$TEMP_CXXFLAGS"
+
 CPPFLAGS="$CPPFLAGS -DHAVE_BUILD_INFO -D__STDC_FORMAT_MACROS"
 
 AC_ARG_WITH([utils],
@@ -1266,6 +1286,7 @@ AM_CONDITIONAL([GLIBC_BACK_COMPAT],[test x$use_glibc_compat = xyes])
 AM_CONDITIONAL([HARDEN],[test x$use_hardening = xyes])
 AM_CONDITIONAL([ENABLE_HWCRC32],[test x$enable_hwcrc32 = xyes])
 AM_CONDITIONAL([ENABLE_SSE41],[test x$enable_sse41 = xyes])
+AM_CONDITIONAL([ENABLE_AVX2],[test x$enable_avx2 = xyes])
 AM_CONDITIONAL([USE_ASM],[test x$use_asm = xyes])
 
 AC_DEFINE(CLIENT_VERSION_MAJOR, _CLIENT_VERSION_MAJOR, [Major version])
@@ -1305,6 +1326,7 @@ AC_SUBST(SANITIZER_CXXFLAGS)
 AC_SUBST(SANITIZER_LDFLAGS)
 AC_SUBST(SSE42_CXXFLAGS)
 AC_SUBST(SSE41_CXXFLAGS)
+AC_SUBST(AVX2_CXXFLAGS)
 AC_SUBST(LIBTOOL_APP_LDFLAGS)
 AC_SUBST(USE_UPNP)
 AC_SUBST(USE_QRCODE)
-- 
cgit v1.2.3