[x265] [PATCH] Add aarch64 support - Part 2

[Suyimeng]

From: x265-devel [mailto:x265-devel-bounces at videolan.org] On Behalf Of chen
Sent: Wednesday, March 18, 2020 9:07 AM
To: Development for x265 <x265-devel at videolan.org>
Subject: Re: [x265] [PATCH] Add aarch64 support - Part 2





On Tue, Mar 17, 2020 at 2:59 PM Suyimeng <yimeng.su at huawei.com<mailto:yimeng.su at huawei.com>> wrote:

From: x265-devel [mailto:x265-devel-bounces at videolan.org<mailto:x265-devel-bounces at videolan.org>] On Behalf Of Gopi Satykrishna Akisetty
Sent: Tuesday, March 17, 2020 4:53 PM
To: Development for x265 <x265-devel at videolan.org<mailto:x265-devel at videolan.org>>
Subject: Re: [x265] [PATCH] Add aarch64 support - Part 2

diff --git a/source/common/pixel.cpp b/source/common/pixel.cpp
index 99b84449c..e4f890cd5 100644
--- a/source/common/pixel.cpp
+++ b/source/common/pixel.cpp
@@ -5,6 +5,7 @@
  *          Mandar Gurav <mandar at multicorewareinc.com<mailto:mandar at multicorewareinc.com>>
  *          Mahesh Pittala <mahesh at multicorewareinc.com<mailto:mahesh at multicorewareinc.com>>
  *          Min Chen <min.chen at multicorewareinc.com<mailto:min.chen at multicorewareinc.com>>
+ *          Hongbin Liu<liuhongbin1 at huawei.com<mailto:liuhongbin1 at huawei.com>>
  *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
@@ -265,6 +266,10 @@ int satd4(const pixel* pix1, intptr_t stride_pix1, const pixel* pix2, intptr_t s
 {
     int satd = 0;

+#if ENABLE_ASSEMBLY && X265_ARCH_ARM64
+    pixelcmp_t satd_4x4 = x265_pixel_satd_4x4_neon;
+#endif
is there any specific reason why the above code is added?? is this a kind of a temporary fix for the output mismatch between c and asm code?
No, c and asm output is matched. Currently we only complete partial satd primatives. This is a workaround that improve all satd primitives with asm code. Maybe there is a bad code style.
If I understand correctly, you are trying to use a combination of c and asm code for all other kernel sizes that you have not completed asm implementation yet?
Yes, you are right.
ok. If this code block is going to be removed in the future patches, where you will be implementing the asm for remaining satd  kernels, then this patch is good to be pushed.

before push the patches, I want to double check how about response for my review on Part-1?
I am not sure I missed these email, or my post still in pending.

I didn't see any questions on part-1. Maybe the mail is oversize and is blocked. You could let me know in any way.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20200318/5d5d5c27/attachment-0001.html>