[meta-intel] [meta-dpdk][PATCH] dpdk: workaround sse mismatch error

Hongxu Jia hongxu.jia at windriver.com
Wed Oct 10 08:18:06 PDT 2018


Since commit `7f6bffb dpdk: support build with external & multilib toolchains' applied,
it added var-HOST_CC_ARCH in which has option `-march=**' to enable specific SSE for
specific arch that causing SSE mismatch error while compiling
[snip]
| recipe-sysroot-native/usr/lib/i586-wrs-linux/gcc/i586-wrs-linux/8.2.0/include/
xmmintrin.h:1232:1: error: inlining failed in call to always_inline '_mm_sfence':
target specific option mismatch
|  _mm_sfence (void)
|  ^~~~~~~~~~
| In file included from git/lib/librte_eal/linuxapp/eal/eal_thread.c:16:
| git/i686-native-linuxapp-gcc/include/rte_atomic.h:26:19: note: called from here
|  #define rte_wmb() _mm_sfence()
|                    ^~~~~~~~~~~~
| git/lib/librte_eal/linuxapp/eal/eal_thread.c:155:3: note: in expansion of macro 'rte_wmb'
|    rte_wmb();
[snip]

Choose `-march=native', which will select what is available on the processor. Although
it disable previously `-march=**' in var-HOST_CC_ARCH, but others options in var-HOST_CC_ARCH
(such as security flags `-fstack-protector-strong -D_FORTIFY_SOURCE=2') still works.

Refers:
https://stackoverflow.com/questions/10686638/whats-the-differrence-among-cflgs-sse-options-of-msse-msse2-mssse3-msse4
https://stackoverflow.com/questions/35772562/inlining-failed-in-call-to-always-inline-m128i-mm-cvtepu8-epi32-m128i-t
https://bugzilla.redhat.com/show_bug.cgi?id=1092991
https://bugzilla.gnome.org/show_bug.cgi?id=670690
https://stackoverflow.com/questions/51003218/gcc-target-for-avx2-disabling-sse-instruction-set

Signed-off-by: Hongxu Jia <hongxu.jia at windriver.com>
---
 recipes-extended/dpdk/dpdk.inc | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/recipes-extended/dpdk/dpdk.inc b/recipes-extended/dpdk/dpdk.inc
index 194986b..b893e16 100644
--- a/recipes-extended/dpdk/dpdk.inc
+++ b/recipes-extended/dpdk/dpdk.inc
@@ -102,18 +102,18 @@ do_compile () {
 
 	cd ${S}/${RTE_TARGET}
 	oe_runmake EXTRA_LDFLAGS="-L${STAGING_LIBDIR} --hash-style=gnu" \
-		   EXTRA_CFLAGS="${HOST_CC_ARCH} ${TOOLCHAIN_OPTIONS} -I${STAGING_INCDIR}" \
+		   EXTRA_CFLAGS="${HOST_CC_ARCH} -march=native ${TOOLCHAIN_OPTIONS} -I${STAGING_INCDIR}" \
 		   CROSS="${TARGET_PREFIX}" \
 		   prefix=""  LDFLAGS="${TUNE_LDARGS}"  WERROR_FLAGS="-w" V=1
 
 	cd ${S}/examples/
 	oe_runmake EXTRA_LDFLAGS="-L${STAGING_LIBDIR} --hash-style=gnu -fuse-ld=bfd" \
-		   EXTRA_CFLAGS="${HOST_CC_ARCH} ${TOOLCHAIN_OPTIONS} -I${STAGING_INCDIR}" \
+		   EXTRA_CFLAGS="${HOST_CC_ARCH} -march=native ${TOOLCHAIN_OPTIONS} -I${STAGING_INCDIR}" \
 		   CROSS="${TARGET_PREFIX}" O="${S}/examples/$@/"
 
 	cd ${S}/test/
 	oe_runmake EXTRA_LDFLAGS="-L${STAGING_LIBDIR} --hash-style=gnu -fuse-ld=bfd" \
-		   EXTRA_CFLAGS="${HOST_CC_ARCH} ${TOOLCHAIN_OPTIONS} -I${STAGING_INCDIR}" \
+		   EXTRA_CFLAGS="${HOST_CC_ARCH} -march=native  ${TOOLCHAIN_OPTIONS} -I${STAGING_INCDIR}" \
 		   CROSS="${TARGET_PREFIX}" O="${S}/test/$@/"
 }
 
-- 
2.7.4



More information about the meta-intel mailing list