[FFmpeg-devel] [PATCH] swresample/arm: avoid conditional branch to PLT in THUMB-2.

[Rahul Chaudhry]

On Wed, Apr 18, 2018 at 3:46 PM, Michael Niedermayer
<michael at niedermayer.cc> wrote:
> please make sure this works on apple based arm (unless you know it works)
> (ive tested qemu linux based)
>
> Also please add a commit message

If by 'apple based arm' you mean llvm/clang assembler, then yes, I've verified
that the assembly works with armv7a-cros-linux-gnueabi-clang (version 7.0.0).

Updated patch with new commit message is attached.

Thanks,
Rahul
-------------- next part --------------
From 2e3318acf266b519e98b680102a07196d6ddbf93 Mon Sep 17 00:00:00 2001
From: Rahul Chaudhry <rahulchaudhry at chromium.org>
Date: Wed, 18 Apr 2018 16:29:39 -0700
Subject: [PATCH] swresample/arm: remove unintentional relocation.

Branch to global symbol results in reference to PLT, and when compiling
for THUMB-2 - in a R_ARM_THM_JUMP19 relocation. Some linkers don't
support this relocation (ld.gold), while others can end up truncating
the relocation to fit (ld.bfd).

Convert this branch through PLT into a direct branch that the assembler
can resolve locally.

See https://github.com/android-ndk/ndk/issues/337 for background.

The current workaround is to disable neon during gstreamer build,
which is not optimal and can be reverted after this patch:
https://github.com/freedesktop/gstreamer-cerbero/commit/41556c415739fbc3a72c7eaee7e70a565b719b2f
---
 libswresample/arm/audio_convert_neon.S | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git libswresample/arm/audio_convert_neon.S libswresample/arm/audio_convert_neon.S
index 1f88316ddec838dfe791b08cbe72533207994741..7729514701d3a02d04146f24cd9b12b11548ac64 100644
--- libswresample/arm/audio_convert_neon.S
+++ libswresample/arm/audio_convert_neon.S
@@ -22,6 +22,7 @@
 #include "libavutil/arm/asm.S"
 
 function swri_oldapi_conv_flt_to_s16_neon, export=1
+_swri_oldapi_conv_flt_to_s16_neon:
         subs            r2,  r2,  #8
         vld1.32         {q0},     [r1,:128]!
         vcvt.s32.f32    q8,  q0,  #31
@@ -66,6 +67,7 @@ function swri_oldapi_conv_flt_to_s16_neon, export=1
 endfunc
 
 function swri_oldapi_conv_fltp_to_s16_2ch_neon, export=1
+_swri_oldapi_conv_fltp_to_s16_2ch_neon:
         ldm             r1,  {r1, r3}
         subs            r2,  r2,  #8
         vld1.32         {q0},     [r1,:128]!
@@ -133,8 +135,8 @@ function swri_oldapi_conv_fltp_to_s16_nch_neon, export=1
         cmp             r3,  #2
         itt             lt
         ldrlt           r1,  [r1]
-        blt             X(swri_oldapi_conv_flt_to_s16_neon)
-        beq             X(swri_oldapi_conv_fltp_to_s16_2ch_neon)
+        blt             _swri_oldapi_conv_flt_to_s16_neon
+        beq             _swri_oldapi_conv_fltp_to_s16_2ch_neon
 
         push            {r4-r8, lr}
         cmp             r3,  #4
-- 
2.13.5