diff options
author | Martin Storsjö <martin@martin.st> | 2017-01-09 00:04:19 +0200 |
---|---|---|
committer | Martin Storsjö <martin@martin.st> | 2017-03-19 22:53:32 +0200 |
commit | 21c89f3a26bb1331381b90e653277585447cfbb3 (patch) | |
tree | e29745bb6f8296b5a7952721a48822ab645003b2 /libavcodec/arm | |
parent | 70317b25aa35c0907720e4d2b7686408588c07aa (diff) | |
download | ffmpeg-21c89f3a26bb1331381b90e653277585447cfbb3.tar.gz |
arm/aarch64: vp9: Fix vertical alignment
Align the second/third operands as they usually are.
Due to the wildly varying sizes of the written out operands
in aarch64 assembly, the column alignment is usually not as clear
as in arm assembly.
This is cherrypicked from libav commit
7995ebfad12002033c73feed422a1cfc62081e8f.
Signed-off-by: Martin Storsjö <martin@martin.st>
Diffstat (limited to 'libavcodec/arm')
-rw-r--r-- | libavcodec/arm/vp9itxfm_neon.S | 14 | ||||
-rw-r--r-- | libavcodec/arm/vp9lpf_neon.S | 2 |
2 files changed, 8 insertions, 8 deletions
diff --git a/libavcodec/arm/vp9itxfm_neon.S b/libavcodec/arm/vp9itxfm_neon.S index 6d4d765c28..6c09922cae 100644 --- a/libavcodec/arm/vp9itxfm_neon.S +++ b/libavcodec/arm/vp9itxfm_neon.S @@ -530,7 +530,7 @@ function idct16x16_dc_add_neon movrel r12, idct_coeffs vld1.16 {d0}, [r12,:64] - vmov.i16 q2, #0 + vmov.i16 q2, #0 vld1.16 {d16[]}, [r2,:16] vmull.s16 q8, d16, d0[0] @@ -793,7 +793,7 @@ function \txfm\()16_1d_4x16_pass1_neon push {lr} mov r12, #32 - vmov.s16 q2, #0 + vmov.s16 q2, #0 .irp i, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 vld1.16 {d\i}, [r2,:64] vst1.16 {d4}, [r2,:64], r12 @@ -1142,7 +1142,7 @@ function idct32x32_dc_add_neon movrel r12, idct_coeffs vld1.16 {d0}, [r12,:64] - vmov.i16 q2, #0 + vmov.i16 q2, #0 vld1.16 {d16[]}, [r2,:16] vmull.s16 q8, d16, d0[0] @@ -1330,7 +1330,7 @@ function idct32_1d_4x32_pass1\suffix\()_neon @ Double stride of the input, since we only read every other line mov r12, #128 - vmov.s16 d4, #0 + vmov.s16 d4, #0 @ d16 = IN(0), d17 = IN(2) ... d31 = IN(30) .ifb \suffix @@ -1394,7 +1394,7 @@ function idct32_1d_4x32_pass1\suffix\()_neon .endif add r2, r2, #64 - vmov.s16 d8, #0 + vmov.s16 d8, #0 @ d16 = IN(1), d17 = IN(3) ... d31 = IN(31) .ifb \suffix .irp i, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 @@ -1533,9 +1533,9 @@ function idct32_1d_4x32_pass2\suffix\()_neon .endif vld1.32 {d12[]}, [r0,:32], r1 vld1.32 {d12[1]}, [r0,:32], r1 - vrshr.s16 q4, q4, #6 + vrshr.s16 q4, q4, #6 vld1.32 {d13[]}, [r0,:32], r1 - vrshr.s16 q5, q5, #6 + vrshr.s16 q5, q5, #6 vld1.32 {d13[1]}, [r0,:32], r1 sub r0, r0, r1, lsl #2 vaddw.u8 q4, q4, d12 diff --git a/libavcodec/arm/vp9lpf_neon.S b/libavcodec/arm/vp9lpf_neon.S index 8d44d58f32..4b3608064a 100644 --- a/libavcodec/arm/vp9lpf_neon.S +++ b/libavcodec/arm/vp9lpf_neon.S @@ -828,7 +828,7 @@ function ff_vp9_loop_filter_v_16_16_neon, export=1 endfunc function vp9_loop_filter_h_16_neon - sub r12, r0, #8 + sub r12, r0, #8 vld1.8 {d16}, [r12,:64], r1 vld1.8 {d24}, [r0, :64], r1 vld1.8 {d17}, [r12,:64], r1 |