[x265] [PATCH 3/6] pixel-util-sve2.S: Fix branch target in pixel_sub_ps_64x64_sve2

George Steed george.steed at arm.com
Mon Jan 6 17:17:02 UTC 2025


When branching to loops to handle longer SVE vectors, the branch target
was previously written incorrectly here as "vl_gt_16" causing an
infinite loop. Fix this by adjusting the branch target to correctly
refer to the "vl_gt_48" case instead.

Co-authored-by: Hari Limaye <hari.limaye at arm.com>
---
 source/common/aarch64/pixel-util-sve2.S | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/source/common/aarch64/pixel-util-sve2.S b/source/common/aarch64/pixel-util-sve2.S
index 00aa2f984..b2b4d24c1 100644
--- a/source/common/aarch64/pixel-util-sve2.S
+++ b/source/common/aarch64/pixel-util-sve2.S
@@ -408,8 +408,8 @@ function PFX(pixel_sub_ps_64x64_sve2)
     ret
 .vl_gt_16_pixel_sub_ps_64x64:
     rdvl            x9, #1
-    cmp             x9, #16
-    bgt             .vl_gt_16_pixel_sub_ps_64x64
+    cmp             x9, #32
+    bgt             .vl_gt_48_pixel_sub_ps_64x64
     ptrue           p0.b, vl32
     mov             w12, #16
 .vl_gt_16_loop_sub_ps_64_sve2:
-- 
2.34.1

-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0003-pixel-util-sve2.S-Fix-branch-target-in-pixel_sub_ps_.patch
Type: text/x-diff
Size: 1452 bytes
Desc: not available
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20250106/c46c51d7/attachment.patch>


More information about the x265-devel mailing list