patches for DPDK stable branches
 help / color / mirror / Atom feed
* [PATCH] acl: fix autotest failures on ppc64le with gcc 11
@ 2023-03-22 17:29 David Christensen
  2023-03-28 16:29 ` Thomas Monjalon
  0 siblings, 1 reply; 2+ messages in thread
From: David Christensen @ 2023-03-22 17:29 UTC (permalink / raw)
  To: dev; +Cc: David Christensen, stable, Thinh Tran

Original acl implementation of Altivec optimized code included an
explicit -O2 optimization level for a particular inlined function.
When DPDK is built with the default -O3 optimization level on gcc
11.x or later, the resulting code may generate a segmentation
fault as observed in acl_autotest.

Since there is no explicit reason given for the local function
optimization level in the original commit, and testing with gcc
versions 8.x through 12.x results in working code at all -O
optimization settings, the local optimization is removed.

Bugzilla ID: 1197
Cc: stable@dpdk.org

Signed-off-by: David Christensen <drc@linux.vnet.ibm.com>
Tested-by: Thinh Tran <thinhtr@linux.vnet.ibm.com>
---
 lib/acl/acl_run_altivec.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/lib/acl/acl_run_altivec.h b/lib/acl/acl_run_altivec.h
index 4dfe7a14b4..4556e1503b 100644
--- a/lib/acl/acl_run_altivec.h
+++ b/lib/acl/acl_run_altivec.h
@@ -102,7 +102,7 @@ acl_match_check_x4(int slot, const struct rte_acl_ctx *ctx, struct parms *parms,
 /*
  * Process 4 transitions (in 2 XMM registers) in parallel
  */
-static inline __attribute__((optimize("O2"))) xmm_t
+static __rte_always_inline xmm_t
 transition4(xmm_t next_input, const uint64_t *trans,
 	xmm_t *indices1, xmm_t *indices2)
 {
-- 
2.31.1


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2023-03-28 16:29 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-03-22 17:29 [PATCH] acl: fix autotest failures on ppc64le with gcc 11 David Christensen
2023-03-28 16:29 ` Thomas Monjalon

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).