Merge pull request #2288 from jamescowens/PR2286

build: add upstream compiler warnings
gridcoin-community · Aug 18, 2021 · d57ca01 · d57ca01
2 parents 9b378fe + 2656083
commit d57ca01
Show file tree

Hide file tree

Showing 7 changed files with 110 additions and 70 deletions.
diff --git a/Makefile.am b/Makefile.am
@@ -2,6 +2,10 @@
 # Distributed under the MIT software license, see the accompanying
 # file COPYING or https://opensource.org/licenses/mit-license.php.
 
+# Pattern rule to print variables, e.g. make print-top_srcdir
+print-%: FORCE
+	@echo '$*'='$($*)'
+
 ACLOCAL_AMFLAGS = -I build-aux/m4
 SUBDIRS = src
 .PHONY: deploy FORCE

diff --git a/configure.ac b/configure.ac
@@ -180,8 +180,8 @@ AC_ARG_ENABLE([glibc-back-compat],
   [use_glibc_compat=no])
 
 AC_ARG_ENABLE([asm],
-  [AS_HELP_STRING([--enable-asm],
-  [Enable assembly routines (default is yes)])],
+  [AS_HELP_STRING([--disable-asm],
+  [disable assembly routines (enabled by default)])],
   [use_asm=$enableval],
   [use_asm=yes])
 
@@ -196,34 +196,6 @@ case $BUILD_TARGET in
     ;;
 esac
 
-if test x$enable_assembly != xno -a x$have_x86_64 = xtrue
-then
-  AC_MSG_CHECKING(whether we can compile AVX code)
-  AC_COMPILE_IFELSE([AC_LANG_PROGRAM(,[asm ("vmovdqa %ymm0, %ymm1");])],
-    AC_DEFINE(ENABLE_AVX, 1, [Define to 1 if AVX assembly is available.])
-    AC_MSG_RESULT(yes)
-    AC_MSG_CHECKING(whether we can compile XOP code)
-    AC_COMPILE_IFELSE([AC_LANG_PROGRAM(,[asm ("vprotd \$7, %xmm0, %xmm1");])],
-      AC_DEFINE(ENABLE_XOP, 1, [Define to 1 if XOP assembly is available.])
-      AC_MSG_RESULT(yes)
-    ,
-      AC_MSG_RESULT(no)
-      AC_MSG_WARN([The assembler does not support the XOP instruction set.])
-    )
-    AC_MSG_CHECKING(whether we can compile AVX2 code)
-    AC_COMPILE_IFELSE([AC_LANG_PROGRAM(,[asm ("vpaddd %ymm0, %ymm1, %ymm2");])],
-      AC_DEFINE(enable_avx2=yes; ENABLE_AVX2, 1, [Define to 1 if AVX2 assembly is available.])
-      AC_MSG_RESULT(yes)
-    ,
-      AC_MSG_RESULT(no)
-      AC_MSG_WARN([The assembler does not support the AVX2 instruction set.])
-    )
-  ,
-    AC_MSG_RESULT(no)
-    AC_MSG_WARN([The assembler does not support the AVX instruction set.])
-  )
-fi
-
 # Enable debug
 AC_ARG_ENABLE([debug],
     [AS_HELP_STRING([--enable-debug],
@@ -257,17 +229,42 @@ if test "x$enable_werror" = "xyes"; then
   if test "x$CXXFLAG_WERROR" = "x"; then
     AC_MSG_ERROR("enable-werror set but -Werror is not usable")
   fi
+  AX_CHECK_COMPILE_FLAG([-Werror=gnu],[ERROR_CXXFLAGS="$ERROR_CXXFLAGS -Werror=gnu"],,[[$CXXFLAG_WERROR]])
   AX_CHECK_COMPILE_FLAG([-Werror=vla],[ERROR_CXXFLAGS="$ERROR_CXXFLAGS -Werror=vla"],,[[$CXXFLAG_WERROR]])
-  AX_CHECK_COMPILE_FLAG([-Werror=thread-safety-analysis],[ERROR_CXXFLAGS="$ERROR_CXXFLAGS -Werror=thread-safety-analysis"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Werror=shadow-field],[ERROR_CXXFLAGS="$ERROR_CXXFLAGS -Werror=shadow-field"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Werror=switch],[ERROR_CXXFLAGS="$ERROR_CXXFLAGS -Werror=switch"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Werror=thread-safety],[ERROR_CXXFLAGS="$ERROR_CXXFLAGS -Werror=thread-safety"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Werror=range-loop-analysis],[ERROR_CXXFLAGS="$ERROR_CXXFLAGS -Werror=range-loop-analysis"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Werror=unused-variable],[ERROR_CXXFLAGS="$ERROR_CXXFLAGS -Werror=unused-variable"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Werror=date-time],[ERROR_CXXFLAGS="$ERROR_CXXFLAGS -Werror=date-time"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Werror=return-type],[ERROR_CXXFLAGS="$ERROR_CXXFLAGS -Werror=return-type"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Werror=conditional-uninitialized],[ERROR_CXXFLAGS="$ERROR_CXXFLAGS -Werror=conditional-uninitialized"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Werror=unreachable-code-loop-increment],[ERROR_CXXFLAGS="$ERROR_CXXFLAGS -Werror=unreachable-code-loop-increment"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Werror=mismatched-tags], [ERROR_CXXFLAGS="$ERROR_CXXFLAGS -Werror=mismatched-tags"], [], [$CXXFLAG_WERROR])
 fi
 
 if test "x$CXXFLAGS_overridden" = "xno"; then
   AX_CHECK_COMPILE_FLAG([-Wall],[CXXFLAGS="$CXXFLAGS -Wall"],,[[$CXXFLAG_WERROR]])
   AX_CHECK_COMPILE_FLAG([-Wextra],[CXXFLAGS="$CXXFLAGS -Wextra"],,[[$CXXFLAG_WERROR]])
-  AX_CHECK_COMPILE_FLAG([-Wformat],[CXXFLAGS="$CXXFLAGS -Wformat"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Wgnu],[CXXFLAGS="$CXXFLAGS -Wgnu"],,[[$CXXFLAG_WERROR]])
+  dnl some compilers will ignore -Wformat-security without -Wformat, so just combine the two here.
+  AX_CHECK_COMPILE_FLAG([-Wformat -Wformat-security],[CXXFLAGS="$CXXFLAGS -Wformat -Wformat-security"],,[[$CXXFLAG_WERROR]])
   AX_CHECK_COMPILE_FLAG([-Wvla],[CXXFLAGS="$CXXFLAGS -Wvla"],,[[$CXXFLAG_WERROR]])
-  AX_CHECK_COMPILE_FLAG([-Wformat-security],[CXXFLAGS="$CXXFLAGS -Wformat-security"],,[[$CXXFLAG_WERROR]])
-  AX_CHECK_COMPILE_FLAG([-Wthread-safety-analysis],[CXXFLAGS="$CXXFLAGS -Wthread-safety-analysis"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Wshadow-field],[CXXFLAGS="$CXXFLAGS -Wshadow-field"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Wswitch],[CXXFLAGS="$CXXFLAGS -Wswitch"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Wthread-safety],[CXXFLAGS="$CXXFLAGS -Wthread-safety"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Wrange-loop-analysis],[CXXFLAGS="$CXXFLAGS -Wrange-loop-analysis"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Wunused-variable],[CXXFLAGS="$CXXFLAGS -Wunused-variable"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Wunused-member-function],[CXXFLAGS="$CXXFLAGS -Wunused-member-function"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Wdate-time],[CXXFLAGS="$CXXFLAGS -Wdate-time"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Wconditional-uninitialized],[CXXFLAGS="$CXXFLAGS -Wconditional-uninitialized"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Wsign-compare],[CXXFLAGS="$CXXFLAGS -Wsign-compare"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Wduplicated-branches],[CXXFLAGS="$CXXFLAGS -Wduplicated-branches"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Wduplicated-cond],[CXXFLAGS="$CXXFLAGS -Wduplicated-cond"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Wlogical-op],[CXXFLAGS="$CXXFLAGS -Wlogical-op"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Woverloaded-virtual],[CXXFLAGS="$CXXFLAGS -Woverloaded-virtual"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Wunreachable-code-loop-increment],[CXXFLAGS="$CXXFLAGS -Wunreachable-code-loop-increment"],,[[$CXXFLAG_WERROR]])
+  AX_CHECK_COMPILE_FLAG([-Wimplicit-fallthrough], [CXXFLAGS="$CXXFLAGS -Wimplicit-fallthrough"], [], [$CXXFLAG_WERROR])
 
   ## Some compilers (gcc) ignore unknown -Wno-* options, but warn about all
   ## unknown options if any other warning is produced. Test the -Wfoo case, and
@@ -281,24 +278,26 @@ if test "x$CXXFLAGS_overridden" = "xno"; then
   AX_CHECK_COMPILE_FLAG([-Wdeprecated-copy],[CXXFLAGS="$CXXFLAGS -Wno-deprecated-copy"],,[[$CXXFLAG_WERROR]])
 fi
 
-enable_hwcrc32=no
+enable_sse42=no
 enable_sse41=no
 enable_avx2=no
 enable_shani=no
 
 if test "x$use_asm" = "xyes"; then
 
-# Check for optional instruction set support. Enabling these does _not_ imply that all code will
-# be compiled with them, rather that specific objects/libs may use them after checking for runtime
-# compatibility.
+dnl Check for optional instruction set support. Enabling these does _not_ imply that all code will
+dnl be compiled with them, rather that specific objects/libs may use them after checking for runtime
+dnl compatibility.
+
+dnl x86
 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"]],,[[$CXXFLAG_WERROR]])
 AX_CHECK_COMPILE_FLAG([-msse4 -msha],[[SHANI_CXXFLAGS="-msse4 -msha"]],,[[$CXXFLAG_WERROR]])
 
 TEMP_CXXFLAGS="$CXXFLAGS"
 CXXFLAGS="$CXXFLAGS $SSE42_CXXFLAGS"
-AC_MSG_CHECKING(for assembler crc32 support)
+AC_MSG_CHECKING(for SSE4.2 intrinsics)
 AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[
     #include <stdint.h>
     #if defined(_MSC_VER)
@@ -313,7 +312,7 @@ AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[
     l = _mm_crc32_u64(l, 0);
     return l;
   ]])],
- [ AC_MSG_RESULT(yes); enable_hwcrc32=yes],
+ [ AC_MSG_RESULT(yes); enable_sse42=yes],
  [ AC_MSG_RESULT(no)]
 )
 CXXFLAGS="$TEMP_CXXFLAGS"
@@ -323,11 +322,7 @@ CXXFLAGS="$CXXFLAGS $SSE41_CXXFLAGS"
 AC_MSG_CHECKING(for SSE4.1 intrinsics)
 AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[
     #include <stdint.h>
-    #if defined(_MSC_VER)
     #include <immintrin.h>
-    #elif defined(__GNUC__)
-    #include <x86intrin.h>
-    #endif
   ]],[[
     __m128i l = _mm_set1_epi32(0);
     return _mm_extract_epi32(l, 3);
@@ -337,6 +332,21 @@ AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[
 )
 CXXFLAGS="$TEMP_CXXFLAGS"
 
+TEMP_CXXFLAGS="$CXXFLAGS"
+CXXFLAGS="$CXXFLAGS $AVX2_CXXFLAGS"
+AC_MSG_CHECKING(for AVX2 intrinsics)
+AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[
+    #include <stdint.h>
+    #include <immintrin.h>
+  ]],[[
+    __m256i l = _mm256_set1_epi32(0);
+    return _mm256_extract_epi32(l, 7);
+  ]])],
+ [ AC_MSG_RESULT(yes); enable_avx2=yes; AC_DEFINE(ENABLE_AVX2, 1, [Define this symbol to build code that uses AVX2 intrinsics]) ],
+ [ AC_MSG_RESULT(no)]
+)
+CXXFLAGS="$TEMP_CXXFLAGS"
+
 TEMP_CXXFLAGS="$CXXFLAGS"
 CXXFLAGS="$CXXFLAGS $SHANI_CXXFLAGS"
 AC_MSG_CHECKING(for SHA-NI intrinsics)
@@ -354,6 +364,24 @@ AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[
 )
 CXXFLAGS="$TEMP_CXXFLAGS"
 
+# ARM
+AX_CHECK_COMPILE_FLAG([-march=armv8-a+crc+crypto],[[ARM_CRC_CXXFLAGS="-march=armv8-a+crc+crypto"]],,[[$CXXFLAG_WERROR]])
+
+TEMP_CXXFLAGS="$CXXFLAGS"
+CXXFLAGS="$CXXFLAGS $ARM_CRC_CXXFLAGS"
+AC_MSG_CHECKING(for ARM CRC32 intrinsics)
+AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[
+    #include <arm_acle.h>
+    #include <arm_neon.h>
+  ]],[[
+    __crc32cb(0, 0); __crc32ch(0, 0); __crc32cw(0, 0); __crc32cd(0, 0);
+    vmull_p64(0, 0);
+  ]])],
+ [ AC_MSG_RESULT(yes); enable_arm_crc=yes; ],
+ [ AC_MSG_RESULT(no)]
+)
+CXXFLAGS="$TEMP_CXXFLAGS"
+
 fi
 
 CPPFLAGS="$CPPFLAGS -DBOOST_SPIRIT_THREADSAFE -DHAVE_BUILD_INFO -D__STDC_FORMAT_MACROS"
@@ -1117,10 +1145,11 @@ AM_CONDITIONAL([USE_QRCODE], [test x$use_qr = xyes])
 AM_CONDITIONAL([USE_LCOV],[test x$use_lcov = xyes])
 AM_CONDITIONAL([GLIBC_BACK_COMPAT],[test x$use_glibc_compat = xyes])
 AM_CONDITIONAL([HARDEN],[test x$use_hardening = xyes])
-AM_CONDITIONAL([ENABLE_HWCRC32],[test x$enable_hwcrc32 = xyes])
+AM_CONDITIONAL([ENABLE_SSE42],[test x$enable_sse42 = xyes])
 AM_CONDITIONAL([ENABLE_SSE41],[test x$enable_sse41 = xyes])
 AM_CONDITIONAL([ENABLE_AVX2],[test x$enable_avx2 = xyes])
 AM_CONDITIONAL([ENABLE_SHANI],[test x$enable_shani = xyes])
+AM_CONDITIONAL([ENABLE_ARM_CRC],[test x$enable_arm_crc = xyes])
 AM_CONDITIONAL([USE_ASM],[test x$use_asm = xyes])
 AM_CONDITIONAL([ARCH_x86], [test x$have_x86 = xtrue])
 AM_CONDITIONAL([ARCH_x86_64], [test x$have_x86_64 = xtrue])

diff --git a/src/Makefile.am b/src/Makefile.am
@@ -2,6 +2,10 @@
 # Distributed under the MIT software license, see the accompanying
 # file COPYING or https://opensource.org/licenses/mit-license.php.
 
+# Pattern rule to print variables, e.g. make print-top_srcdir
+print-%: FORCE
+	@echo '$*'='$($*)'
+
 DIST_SUBDIRS = univalue
 
 AM_LDFLAGS = ${libcurl_LIBS} $(PTHREAD_CFLAGS) $(LIBTOOL_LDFLAGS) $(HARDENED_LDFLAGS)

diff --git a/src/Makefile.leveldb.include b/src/Makefile.leveldb.include
@@ -142,7 +142,7 @@ leveldb_libmemenv_a_SOURCES += leveldb/helpers/memenv/memenv.h
 
 leveldb_libleveldb_sse42_a_CPPFLAGS = $(leveldb_libleveldb_a_CPPFLAGS)
 leveldb_libleveldb_sse42_a_CXXFLAGS = $(leveldb_libleveldb_a_CXXFLAGS)
-if ENABLE_HWCRC32
+if ENABLE_SSE42
 leveldb_libleveldb_sse42_a_CPPFLAGS += -DLEVELDB_PLATFORM_POSIX_SSE
 leveldb_libleveldb_sse42_a_CXXFLAGS += $(SSE42_CXXFLAGS)
 endif

diff --git a/src/crypto/sha256_avx2.cpp b/src/crypto/sha256_avx2.cpp
@@ -1,9 +1,12 @@
+// Copyright (c) 2017-2019 The Bitcoin Core developers
+// Distributed under the MIT software license, see the accompanying
+// file COPYING or https://www.opensource.org/licenses/mit-license.php.
+
 #ifdef ENABLE_AVX2
 
 #include <stdint.h>
 #include <immintrin.h>
 
-#include <crypto/sha256.h>
 #include <crypto/common.h>
 
 namespace sha256d64_avx2 {

diff --git a/src/crypto/sha256_shani.cpp b/src/crypto/sha256_shani.cpp
@@ -1,6 +1,6 @@
-// Copyright (c) 2018 The Bitcoin Core developers
+// Copyright (c) 2018-2020 The Bitcoin Core developers
 // Distributed under the MIT software license, see the accompanying
-// file COPYING or https://opensource.org/licenses/mit-license.php.
+// file COPYING or https://www.opensource.org/licenses/mit-license.php.
 //
 // Based on https://github.com/noloader/SHA-Intrinsics/blob/master/sha256-x86.c,
 // Written and placed in public domain by Jeffrey Walton.
@@ -11,14 +11,11 @@
 #include <stdint.h>
 #include <immintrin.h>
 
-#include <crypto/common.h>
-
-
 namespace {
 
-const __m128i MASK = _mm_set_epi64x(0x0c0d0e0f08090a0bULL, 0x0405060700010203ULL);
-const __m128i INIT0 = _mm_set_epi64x(0x6a09e667bb67ae85ull, 0x510e527f9b05688cull);
-const __m128i INIT1 = _mm_set_epi64x(0x3c6ef372a54ff53aull, 0x1f83d9ab5be0cd19ull);
+alignas(__m128i) const uint8_t MASK[16] = {0x03, 0x02, 0x01, 0x00, 0x07, 0x06, 0x05, 0x04, 0x0b, 0x0a, 0x09, 0x08, 0x0f, 0x0e, 0x0d, 0x0c};
+alignas(__m128i) const uint8_t INIT0[16] = {0x8c, 0x68, 0x05, 0x9b, 0x7f, 0x52, 0x0e, 0x51, 0x85, 0xae, 0x67, 0xbb, 0x67, 0xe6, 0x09, 0x6a};
+alignas(__m128i) const uint8_t INIT1[16] = {0x19, 0xcd, 0xe0, 0x5b, 0xab, 0xd9, 0x83, 0x1f, 0x3a, 0xf5, 0x4f, 0xa5, 0x72, 0xf3, 0x6e, 0x3c};
 
 void inline  __attribute__((always_inline)) QuadRound(__m128i& state0, __m128i& state1, uint64_t k1, uint64_t k0)
 {
@@ -68,12 +65,12 @@ void inline __attribute__((always_inline)) Unshuffle(__m128i& s0, __m128i& s1)
 
 __m128i inline  __attribute__((always_inline)) Load(const unsigned char* in)
 {
-    return _mm_shuffle_epi8(_mm_loadu_si128((const __m128i*)in), MASK);
+    return _mm_shuffle_epi8(_mm_loadu_si128((const __m128i*)in), _mm_load_si128((const __m128i*)MASK));
 }
 
 void inline  __attribute__((always_inline)) Save(unsigned char* out, __m128i s)
 {
-    _mm_storeu_si128((__m128i*)out, _mm_shuffle_epi8(s, MASK));
+    _mm_storeu_si128((__m128i*)out, _mm_shuffle_epi8(s, _mm_load_si128((const __m128i*)MASK)));
 }
 }
 
@@ -150,8 +147,8 @@ void Transform_2way(unsigned char* out, const unsigned char* in)
     __m128i bm0, bm1, bm2, bm3, bs0, bs1, bso0, bso1;
 
     /* Transform 1 */
-    bs0 = as0 = INIT0;
-    bs1 = as1 = INIT1;
+    bs0 = as0 = _mm_load_si128((const __m128i*)INIT0);
+    bs1 = as1 = _mm_load_si128((const __m128i*)INIT1);
     am0 = Load(in);
     bm0 = Load(in + 64);
     QuadRound(as0, as1, am0, 0xe9b5dba5b5c0fbcfull, 0x71374491428a2f98ull);
@@ -220,10 +217,10 @@ void Transform_2way(unsigned char* out, const unsigned char* in)
     ShiftMessageC(bm1, bm2, bm3);
     QuadRound(as0, as1, am3, 0xc67178f2bef9A3f7ull, 0xa4506ceb90befffaull);
     QuadRound(bs0, bs1, bm3, 0xc67178f2bef9A3f7ull, 0xa4506ceb90befffaull);
-    as0 = _mm_add_epi32(as0, INIT0);
-    bs0 = _mm_add_epi32(bs0, INIT0);
-    as1 = _mm_add_epi32(as1, INIT1);
-    bs1 = _mm_add_epi32(bs1, INIT1);
+    as0 = _mm_add_epi32(as0, _mm_load_si128((const __m128i*)INIT0));
+    bs0 = _mm_add_epi32(bs0, _mm_load_si128((const __m128i*)INIT0));
+    as1 = _mm_add_epi32(as1, _mm_load_si128((const __m128i*)INIT1));
+    bs1 = _mm_add_epi32(bs1, _mm_load_si128((const __m128i*)INIT1));
 
     /* Transform 2 */
     aso0 = as0;
@@ -276,8 +273,8 @@ void Transform_2way(unsigned char* out, const unsigned char* in)
     bm1 = bs1;
 
     /* Transform 3 */
-    bs0 = as0 = INIT0;
-    bs1 = as1 = INIT1;
+    bs0 = as0 = _mm_load_si128((const __m128i*)INIT0);
+    bs1 = as1 = _mm_load_si128((const __m128i*)INIT1);
     QuadRound(as0, as1, am0, 0xe9b5dba5B5c0fbcfull, 0x71374491428a2f98ull);
     QuadRound(bs0, bs1, bm0, 0xe9b5dba5B5c0fbcfull, 0x71374491428a2f98ull);
     QuadRound(as0, as1, am1, 0xab1c5ed5923f82a4ull, 0x59f111f13956c25bull);
@@ -340,10 +337,10 @@ void Transform_2way(unsigned char* out, const unsigned char* in)
     ShiftMessageC(bm1, bm2, bm3);
     QuadRound(as0, as1, am3, 0xc67178f2bef9a3f7ull, 0xa4506ceb90befffaull);
     QuadRound(bs0, bs1, bm3, 0xc67178f2bef9a3f7ull, 0xa4506ceb90befffaull);
-    as0 = _mm_add_epi32(as0, INIT0);
-    bs0 = _mm_add_epi32(bs0, INIT0);
-    as1 = _mm_add_epi32(as1, INIT1);
-    bs1 = _mm_add_epi32(bs1, INIT1);
+    as0 = _mm_add_epi32(as0, _mm_load_si128((const __m128i*)INIT0));
+    bs0 = _mm_add_epi32(bs0, _mm_load_si128((const __m128i*)INIT0));
+    as1 = _mm_add_epi32(as1, _mm_load_si128((const __m128i*)INIT1));
+    bs1 = _mm_add_epi32(bs1, _mm_load_si128((const __m128i*)INIT1));
 
     /* Extract hash into out */
     Unshuffle(as0, as1);

diff --git a/src/crypto/sha256_sse41.cpp b/src/crypto/sha256_sse41.cpp
@@ -1,9 +1,12 @@
+// Copyright (c) 2018-2019 The Bitcoin Core developers
+// Distributed under the MIT software license, see the accompanying
+// file COPYING or https://www.opensource.org/licenses/mit-license.php.
+
 #ifdef ENABLE_SSE41
 
 #include <stdint.h>
 #include <immintrin.h>
 
-#include <crypto/sha256.h>
 #include <crypto/common.h>
 
 namespace sha256d64_sse41 {