From: "Logan.Lyu" <Logan.Lyu@myais.com.cn>
To: "Martin Storsjö" <martin@martin.st>,
"FFmpeg development discussions and patches"
<ffmpeg-devel@ffmpeg.org>
Subject: Re: [FFmpeg-devel] [PATCH 5/5] lavc/aarch64: new optimization for 8-bit hevc_epel_uni_w_hv
Date: Sun, 18 Jun 2023 16:25:19 +0800
Message-ID: <08a60626-b5f4-935b-5a56-2e001b59b8d7@myais.com.cn> (raw)
In-Reply-To: <c07cf4d-6cd8-36c4-124c-11da8aed149@martin.st>
[-- Attachment #1: Type: text/plain, Size: 13840 bytes --]
Hi, Martin,
I modified it according to your comments. Please review again.
And here are the checkasm benchmark results of the related functions:
put_hevc_epel_uni_w_hv4_8_c: 254.6
put_hevc_epel_uni_w_hv4_8_i8mm: 102.9
put_hevc_epel_uni_w_hv6_8_c: 411.6
put_hevc_epel_uni_w_hv6_8_i8mm: 221.6
put_hevc_epel_uni_w_hv8_8_c: 669.4
put_hevc_epel_uni_w_hv8_8_i8mm: 214.9
put_hevc_epel_uni_w_hv12_8_c: 1412.6
put_hevc_epel_uni_w_hv12_8_i8mm: 481.4
put_hevc_epel_uni_w_hv16_8_c: 2425.4
put_hevc_epel_uni_w_hv16_8_i8mm: 647.4
put_hevc_epel_uni_w_hv24_8_c: 5384.1
put_hevc_epel_uni_w_hv24_8_i8mm: 1450.6
put_hevc_epel_uni_w_hv32_8_c: 9470.9
put_hevc_epel_uni_w_hv32_8_i8mm: 2497.1
put_hevc_epel_uni_w_hv48_8_c: 20930.1
put_hevc_epel_uni_w_hv48_8_i8mm: 5635.9
put_hevc_epel_uni_w_hv64_8_c: 36682.9
put_hevc_epel_uni_w_hv64_8_i8mm: 9712.6
在 2023/6/12 16:19, Martin Storsjö 写道:
> On Sun, 4 Jun 2023, Logan.Lyu@myais.com.cn wrote:
>
>> From: Logan Lyu <Logan.Lyu@myais.com.cn>
>>
>> Signed-off-by: Logan Lyu <Logan.Lyu@myais.com.cn>
>> ---
>> libavcodec/aarch64/hevcdsp_epel_neon.S | 703 ++++++++++++++++++++++
>> libavcodec/aarch64/hevcdsp_init_aarch64.c | 7 +
>> 2 files changed, 710 insertions(+)
>>
>> diff --git a/libavcodec/aarch64/hevcdsp_epel_neon.S
>> b/libavcodec/aarch64/hevcdsp_epel_neon.S
>> index 32f052a7b1..24a74d2c7d 100644
>> --- a/libavcodec/aarch64/hevcdsp_epel_neon.S
>> +++ b/libavcodec/aarch64/hevcdsp_epel_neon.S
>> @@ -718,6 +718,709 @@ function
>> ff_hevc_put_hevc_epel_uni_w_h64_8_neon_i8mm, export=1
>> ret
>> endfunc
>>
>> +.macro epel_uni_w_hv_start
>> + mov x15, x5 //denom
>> + mov x16, x6 //wx
>> + mov x17, x7 //ox
>> + add w15, w15, #6 //shift = denom+6
>> +
>> +
>> + ldp x5, x6, [sp]
>> + ldp x7, xzr, [sp, #16]
>
> Why ldp into xzr, that seems pointless?
>
>> +
>> + sub sp, sp, #128
>> + stp q12, q13, [sp]
>
> This could be "stp q12, q13, [sp, #-128]!"
>
>> + stp q14, q15, [sp, #32]
>> + stp q8, q9, [sp, #64]
>> + stp q10, q11, [sp, #96]
>> +
>> + dup v13.8h, w16 //wx
>> + dup v14.4s, w17 //ox
>> +
>> + mov w17, #1
>> + lsl w17, w17, w15
>> + lsr w17, w17, #1
>> + dup v15.4s, w17
>> +
>> + neg w15, w15 // -shift
>> + dup v12.4s, w15 //shift
>> +.endm
>> +
>> +.macro epel_uni_w_hv_end
>> + smull v28.4s, v4.4h, v13.4h
>> + smull2 v29.4s, v4.8h, v13.8h
>> + add v28.4s, v28.4s, v15.4s
>> + add v29.4s, v29.4s, v15.4s
>> + sshl v28.4s, v28.4s, v12.4s
>> + sshl v29.4s, v29.4s, v12.4s
>> + add v28.4s, v28.4s, v14.4s
>> + add v29.4s, v29.4s, v14.4s
>> + sqxtn v4.4h, v28.4s
>> + sqxtn2 v4.8h, v29.4s
>> +.endm
>> +
>> +.macro epel_uni_w_hv_end2
>> + smull v28.4s, v4.4h, v13.4h
>> + smull2 v29.4s, v4.8h, v13.8h
>> + smull v30.4s, v5.4h, v13.4h
>> + smull2 v31.4s, v5.8h, v13.8h
>> + add v28.4s, v28.4s, v15.4s
>> + add v29.4s, v29.4s, v15.4s
>> + add v30.4s, v30.4s, v15.4s
>> + add v31.4s, v31.4s, v15.4s
>> +
>> + sshl v28.4s, v28.4s, v12.4s
>> + sshl v29.4s, v29.4s, v12.4s
>> + sshl v30.4s, v30.4s, v12.4s
>> + sshl v31.4s, v31.4s, v12.4s
>> +
>> + add v28.4s, v28.4s, v14.4s
>> + add v29.4s, v29.4s, v14.4s
>> + add v30.4s, v30.4s, v14.4s
>> + add v31.4s, v31.4s, v14.4s
>> +
>> + sqxtn v4.4h, v28.4s
>> + sqxtn2 v4.8h, v29.4s
>> + sqxtn v5.4h, v30.4s
>> + sqxtn2 v5.8h, v31.4s
>> +.endm
>> +
>> +.macro epel_uni_w_hv_end3
>> + smull v1.4s, v4.4h, v13.4h
>> + smull2 v2.4s, v4.8h, v13.8h
>> + smull v28.4s, v5.4h, v13.4h
>> + smull2 v29.4s, v5.8h, v13.8h
>> + smull v30.4s, v6.4h, v13.4h
>> + smull2 v31.4s, v6.8h, v13.8h
>> + add v1.4s, v1.4s, v15.4s
>> + add v2.4s, v2.4s, v15.4s
>> + add v28.4s, v28.4s, v15.4s
>> + add v29.4s, v29.4s, v15.4s
>> + add v30.4s, v30.4s, v15.4s
>> + add v31.4s, v31.4s, v15.4s
>> +
>> + sshl v1.4s, v1.4s, v12.4s
>> + sshl v2.4s, v2.4s, v12.4s
>> + sshl v28.4s, v28.4s, v12.4s
>> + sshl v29.4s, v29.4s, v12.4s
>> + sshl v30.4s, v30.4s, v12.4s
>> + sshl v31.4s, v31.4s, v12.4s
>> + add v1.4s, v1.4s, v14.4s
>> + add v2.4s, v2.4s, v14.4s
>> + add v28.4s, v28.4s, v14.4s
>> + add v29.4s, v29.4s, v14.4s
>> + add v30.4s, v30.4s, v14.4s
>> + add v31.4s, v31.4s, v14.4s
>> +
>> + sqxtn v4.4h, v1.4s
>> + sqxtn2 v4.8h, v2.4s
>> + sqxtn v5.4h, v28.4s
>> + sqxtn2 v5.8h, v29.4s
>> + sqxtn v6.4h, v30.4s
>> + sqxtn2 v6.8h, v31.4s
>> +.endm
>> +
>> +.macro calc_epelh dst, src0, src1, src2, src3
>> + smull \dst\().4s, \src0\().4h, v0.h[0]
>> + smlal \dst\().4s, \src1\().4h, v0.h[1]
>> + smlal \dst\().4s, \src2\().4h, v0.h[2]
>> + smlal \dst\().4s, \src3\().4h, v0.h[3]
>> + sqshrn \dst\().4h, \dst\().4s, #6
>> +.endm
>> +
>> +.macro calc_epelh2 dst, tmp, src0, src1, src2, src3
>> + smull2 \tmp\().4s, \src0\().8h, v0.h[0]
>> + smlal2 \tmp\().4s, \src1\().8h, v0.h[1]
>> + smlal2 \tmp\().4s, \src2\().8h, v0.h[2]
>> + smlal2 \tmp\().4s, \src3\().8h, v0.h[3]
>> + sqshrn2 \dst\().8h, \tmp\().4s, #6
>> +.endm
>> +
>> +.macro load_epel_filterh freg, xreg
>> + movrel \xreg, epel_filters
>> + add \xreg, \xreg, \freg, lsl #2
>> + ld1 {v0.8b}, [\xreg]
>> + sxtl v0.8h, v0.8b
>> +.endm
>> +
>> +function ff_hevc_put_hevc_epel_uni_w_hv4_8_neon_i8mm, export=1
>> + epel_uni_w_hv_start
>> + and x4, x4, 0xffffffff
>
> What does this "and" do here? Is it a case where the argument is
> "int", while the upper bits of the register is undefined? In those
> cases, you're best off by just using "w4", possibly "w4, uxtw" (or
> sxtw) instead of manually doing such an "and" here.
>
>> +
>> + add x10, x4, #3
>> + lsl x10, x10, #7
>> + sub sp, sp, x10 // tmp_array
>> + stp x0, x1, [sp, #-16]!
>> + stp x4, x6, [sp, #-16]!
>> + stp xzr, x30, [sp, #-16]!
>
> Don't do consecutive decrements like this, but do one "stp ..., [sp,
> #-48]!" followed by "stp ..., [sp, #16]" etc.
>
>> + add x0, sp, #48
>> + sub x1, x2, x3
>> + mov x2, x3
>> + add x3, x4, #3
>> + mov x4, x5
>> + bl X(ff_hevc_put_hevc_epel_h4_8_neon_i8mm)
>> + ldp xzr, x30, [sp], #16
>> + ldp x4, x6, [sp], #16
>> + ldp x0, x1, [sp], #16
>> + load_epel_filterh x6, x5
>> + mov x10, #(MAX_PB_SIZE * 2)
>> + ld1 {v16.4h}, [sp], x10
>> + ld1 {v17.4h}, [sp], x10
>> + ld1 {v18.4h}, [sp], x10
>> +1: ld1 {v19.4h}, [sp], x10
>> + calc_epelh v4, v16, v17, v18, v19
>> + epel_uni_w_hv_end
>> + sqxtun v4.8b, v4.8h
>> + str s4, [x0]
>> + add x0, x0, x1
>> + subs x4, x4, #1
>> + b.eq 2f
>> +
>> + ld1 {v16.4h}, [sp], x10
>> + calc_epelh v4, v17, v18, v19, v16
>> + epel_uni_w_hv_end
>> + sqxtun v4.8b, v4.8h
>> + str s4, [x0]
>> + add x0, x0, x1
>> + subs x4, x4, #1
>> + b.eq 2f
>> +
>> + ld1 {v17.4h}, [sp], x10
>> + calc_epelh v4, v18, v19, v16, v17
>> + epel_uni_w_hv_end
>> + sqxtun v4.8b, v4.8h
>> + str s4, [x0]
>> + add x0, x0, x1
>> + subs x4, x4, #1
>> + b.eq 2f
>> +
>> + ld1 {v18.4h}, [sp], x10
>> + calc_epelh v4, v19, v16, v17, v18
>> + epel_uni_w_hv_end
>> + sqxtun v4.8b, v4.8h
>> + str s4, [x0]
>> + add x0, x0, x1
>> + subs x4, x4, #1
>> + b.ne 1b
>> +2:
>> + ldp q12, q13, [sp]
>> + ldp q14, q15, [sp, #32]
>> + ldp q8, q9, [sp, #64]
>> + ldp q10, q11, [sp, #96]
>> + add sp, sp, #128
>
> Fold the stack increment into ldp, like "ldp q12, q13, [sp], #128".
>
> The same thing applies to all other functions in this patch too.
>
>> diff --git a/libavcodec/aarch64/hevcdsp_init_aarch64.c
>> b/libavcodec/aarch64/hevcdsp_init_aarch64.c
>> index 348497bbbe..fbbc4e6071 100644
>> --- a/libavcodec/aarch64/hevcdsp_init_aarch64.c
>> +++ b/libavcodec/aarch64/hevcdsp_init_aarch64.c
>> @@ -189,6 +189,11 @@ NEON8_FNPROTO(qpel_uni_w_h, (uint8_t *_dst,
>> ptrdiff_t _dststride,
>> int height, int denom, int wx, int ox,
>> intptr_t mx, intptr_t my, int width), _i8mm);
>>
>> +NEON8_FNPROTO(epel_uni_w_hv, (uint8_t *_dst, ptrdiff_t _dststride,
>> + const uint8_t *_src, ptrdiff_t _srcstride,
>> + int height, int denom, int wx, int ox,
>> + intptr_t mx, intptr_t my, int width), _i8mm);
>> +
>> NEON8_FNPROTO_PARTIAL_5(qpel_uni_w_hv, (uint8_t *_dst, ptrdiff_t
>> _dststride,
>> const uint8_t *_src, ptrdiff_t _srcstride,
>> int height, int denom, int wx, int ox,
>> @@ -286,11 +291,13 @@ av_cold void
>> ff_hevc_dsp_init_aarch64(HEVCDSPContext *c, const int bit_depth)
>> NEON8_FNASSIGN(c->put_hevc_epel_uni_w, 1, 0, epel_uni_w_v,);
>> NEON8_FNASSIGN_PARTIAL_4(c->put_hevc_qpel_uni_w, 1, 0,
>> qpel_uni_w_v,);
>>
>> +
>> if (have_i8mm(cpu_flags)) {
>
> Stray whitespace change.
>
> // Martin
>
[-- Attachment #2: 0005-lavc-aarch64-new-optimization-for-8-bit-hevc_epel_uni_w_hv.patch --]
[-- Type: text/plain, Size: 28754 bytes --]
From 47b7f7af634add7680b56a216fff7dbe1f08cd11 Mon Sep 17 00:00:00 2001
From: Logan Lyu <Logan.Lyu@myais.com.cn>
Date: Sun, 28 May 2023 10:35:43 +0800
Subject: [PATCH 5/5] lavc/aarch64: new optimization for 8-bit
hevc_epel_uni_w_hv
Signed-off-by: Logan Lyu <Logan.Lyu@myais.com.cn>
---
libavcodec/aarch64/hevcdsp_epel_neon.S | 694 ++++++++++++++++++++++
libavcodec/aarch64/hevcdsp_init_aarch64.c | 6 +
2 files changed, 700 insertions(+)
diff --git a/libavcodec/aarch64/hevcdsp_epel_neon.S b/libavcodec/aarch64/hevcdsp_epel_neon.S
index 8b6f396a0b..355679af29 100644
--- a/libavcodec/aarch64/hevcdsp_epel_neon.S
+++ b/libavcodec/aarch64/hevcdsp_epel_neon.S
@@ -717,6 +717,700 @@ function ff_hevc_put_hevc_epel_uni_w_h64_8_neon_i8mm, export=1
ret
endfunc
+.macro epel_uni_w_hv_start
+ mov x15, x5 //denom
+ mov x16, x6 //wx
+ mov x17, x7 //ox
+ add w15, w15, #6 //shift = denom+6
+
+
+ ldp x5, x6, [sp]
+ ldr x7, [sp, #16]
+
+ stp q12, q13, [sp, #-128]!
+ stp q14, q15, [sp, #32]
+ stp q8, q9, [sp, #64]
+ stp q10, q11, [sp, #96]
+
+ dup v13.8h, w16 //wx
+ dup v14.4s, w17 //ox
+
+ mov w17, #1
+ lsl w17, w17, w15
+ lsr w17, w17, #1
+ dup v15.4s, w17
+
+ neg w15, w15 // -shift
+ dup v12.4s, w15 //shift
+.endm
+
+.macro epel_uni_w_hv_end
+ smull v28.4s, v4.4h, v13.4h
+ smull2 v29.4s, v4.8h, v13.8h
+ add v28.4s, v28.4s, v15.4s
+ add v29.4s, v29.4s, v15.4s
+ sshl v28.4s, v28.4s, v12.4s
+ sshl v29.4s, v29.4s, v12.4s
+ add v28.4s, v28.4s, v14.4s
+ add v29.4s, v29.4s, v14.4s
+ sqxtn v4.4h, v28.4s
+ sqxtn2 v4.8h, v29.4s
+.endm
+
+.macro epel_uni_w_hv_end2
+ smull v28.4s, v4.4h, v13.4h
+ smull2 v29.4s, v4.8h, v13.8h
+ smull v30.4s, v5.4h, v13.4h
+ smull2 v31.4s, v5.8h, v13.8h
+ add v28.4s, v28.4s, v15.4s
+ add v29.4s, v29.4s, v15.4s
+ add v30.4s, v30.4s, v15.4s
+ add v31.4s, v31.4s, v15.4s
+
+ sshl v28.4s, v28.4s, v12.4s
+ sshl v29.4s, v29.4s, v12.4s
+ sshl v30.4s, v30.4s, v12.4s
+ sshl v31.4s, v31.4s, v12.4s
+
+ add v28.4s, v28.4s, v14.4s
+ add v29.4s, v29.4s, v14.4s
+ add v30.4s, v30.4s, v14.4s
+ add v31.4s, v31.4s, v14.4s
+
+ sqxtn v4.4h, v28.4s
+ sqxtn2 v4.8h, v29.4s
+ sqxtn v5.4h, v30.4s
+ sqxtn2 v5.8h, v31.4s
+.endm
+
+.macro epel_uni_w_hv_end3
+ smull v1.4s, v4.4h, v13.4h
+ smull2 v2.4s, v4.8h, v13.8h
+ smull v28.4s, v5.4h, v13.4h
+ smull2 v29.4s, v5.8h, v13.8h
+ smull v30.4s, v6.4h, v13.4h
+ smull2 v31.4s, v6.8h, v13.8h
+ add v1.4s, v1.4s, v15.4s
+ add v2.4s, v2.4s, v15.4s
+ add v28.4s, v28.4s, v15.4s
+ add v29.4s, v29.4s, v15.4s
+ add v30.4s, v30.4s, v15.4s
+ add v31.4s, v31.4s, v15.4s
+
+ sshl v1.4s, v1.4s, v12.4s
+ sshl v2.4s, v2.4s, v12.4s
+ sshl v28.4s, v28.4s, v12.4s
+ sshl v29.4s, v29.4s, v12.4s
+ sshl v30.4s, v30.4s, v12.4s
+ sshl v31.4s, v31.4s, v12.4s
+ add v1.4s, v1.4s, v14.4s
+ add v2.4s, v2.4s, v14.4s
+ add v28.4s, v28.4s, v14.4s
+ add v29.4s, v29.4s, v14.4s
+ add v30.4s, v30.4s, v14.4s
+ add v31.4s, v31.4s, v14.4s
+
+ sqxtn v4.4h, v1.4s
+ sqxtn2 v4.8h, v2.4s
+ sqxtn v5.4h, v28.4s
+ sqxtn2 v5.8h, v29.4s
+ sqxtn v6.4h, v30.4s
+ sqxtn2 v6.8h, v31.4s
+.endm
+
+.macro calc_epelh dst, src0, src1, src2, src3
+ smull \dst\().4s, \src0\().4h, v0.h[0]
+ smlal \dst\().4s, \src1\().4h, v0.h[1]
+ smlal \dst\().4s, \src2\().4h, v0.h[2]
+ smlal \dst\().4s, \src3\().4h, v0.h[3]
+ sqshrn \dst\().4h, \dst\().4s, #6
+.endm
+
+.macro calc_epelh2 dst, tmp, src0, src1, src2, src3
+ smull2 \tmp\().4s, \src0\().8h, v0.h[0]
+ smlal2 \tmp\().4s, \src1\().8h, v0.h[1]
+ smlal2 \tmp\().4s, \src2\().8h, v0.h[2]
+ smlal2 \tmp\().4s, \src3\().8h, v0.h[3]
+ sqshrn2 \dst\().8h, \tmp\().4s, #6
+.endm
+
+.macro load_epel_filterh freg, xreg
+ movrel \xreg, epel_filters
+ add \xreg, \xreg, \freg, lsl #2
+ ld1 {v0.8b}, [\xreg]
+ sxtl v0.8h, v0.8b
+.endm
+
+function ff_hevc_put_hevc_epel_uni_w_hv4_8_neon_i8mm, export=1
+ epel_uni_w_hv_start
+ sxtw x4, w4
+
+ add x10, x4, #3
+ lsl x10, x10, #7
+ sub sp, sp, x10 // tmp_array
+ stp xzr, x30, [sp, #-48]!
+ stp x4, x6, [sp, #16]
+ stp x0, x1, [sp, #32]
+ add x0, sp, #48
+ sub x1, x2, x3
+ mov x2, x3
+ add x3, x4, #3
+ mov x4, x5
+ bl X(ff_hevc_put_hevc_epel_h4_8_neon_i8mm)
+ ldp x4, x6, [sp, #16]
+ ldp x0, x1, [sp, #32]
+ ldp xzr, x30, [sp], #48
+ load_epel_filterh x6, x5
+ mov x10, #(MAX_PB_SIZE * 2)
+ ld1 {v16.4h}, [sp], x10
+ ld1 {v17.4h}, [sp], x10
+ ld1 {v18.4h}, [sp], x10
+1: ld1 {v19.4h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v16, v17, v18, v19
+ epel_uni_w_hv_end
+ sqxtun v4.8b, v4.8h
+ str s4, [x0]
+ add x0, x0, x1
+ b.eq 2f
+
+ ld1 {v16.4h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v17, v18, v19, v16
+ epel_uni_w_hv_end
+ sqxtun v4.8b, v4.8h
+ str s4, [x0]
+ add x0, x0, x1
+ b.eq 2f
+
+ ld1 {v17.4h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v18, v19, v16, v17
+ epel_uni_w_hv_end
+ sqxtun v4.8b, v4.8h
+ str s4, [x0]
+ add x0, x0, x1
+ b.eq 2f
+
+ ld1 {v18.4h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v19, v16, v17, v18
+ epel_uni_w_hv_end
+ sqxtun v4.8b, v4.8h
+ str s4, [x0]
+ add x0, x0, x1
+ b.ne 1b
+2:
+ ldp q14, q15, [sp, #32]
+ ldp q8, q9, [sp, #64]
+ ldp q10, q11, [sp, #96]
+ ldp q12, q13, [sp], #128
+ ret
+endfunc
+
+function ff_hevc_put_hevc_epel_uni_w_hv6_8_neon_i8mm, export=1
+ epel_uni_w_hv_start
+ sxtw x4, w4
+
+ add x10, x4, #3
+ lsl x10, x10, #7
+ sub sp, sp, x10 // tmp_array
+ stp xzr, x30, [sp, #-48]!
+ stp x4, x6, [sp, #16]
+ stp x0, x1, [sp, #32]
+ add x0, sp, #48
+ sub x1, x2, x3
+ mov x2, x3
+ add x3, x4, #3
+ mov x4, x5
+ bl X(ff_hevc_put_hevc_epel_h6_8_neon_i8mm)
+ ldp x4, x6, [sp, #16]
+ ldp x0, x1, [sp, #32]
+ ldp xzr, x30, [sp], #48
+ load_epel_filterh x6, x5
+ sub x1, x1, #4
+ mov x10, #(MAX_PB_SIZE * 2)
+ ld1 {v16.8h}, [sp], x10
+ ld1 {v17.8h}, [sp], x10
+ ld1 {v18.8h}, [sp], x10
+1: ld1 {v19.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v16, v17, v18, v19
+ calc_epelh2 v4, v5, v16, v17, v18, v19
+ epel_uni_w_hv_end
+ sqxtun v4.8b, v4.8h
+ st1 {v4.s}[0], [x0], #4
+ st1 {v4.h}[2], [x0], x1
+ b.eq 2f
+
+ ld1 {v16.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v17, v18, v19, v16
+ calc_epelh2 v4, v5, v17, v18, v19, v16
+ epel_uni_w_hv_end
+ sqxtun v4.8b, v4.8h
+ st1 {v4.s}[0], [x0], #4
+ st1 {v4.h}[2], [x0], x1
+ b.eq 2f
+
+ ld1 {v17.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v18, v19, v16, v17
+ calc_epelh2 v4, v5, v18, v19, v16, v17
+ epel_uni_w_hv_end
+ sqxtun v4.8b, v4.8h
+ st1 {v4.s}[0], [x0], #4
+ st1 {v4.h}[2], [x0], x1
+ b.eq 2f
+
+ ld1 {v18.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v19, v16, v17, v18
+ calc_epelh2 v4, v5, v19, v16, v17, v18
+ epel_uni_w_hv_end
+ sqxtun v4.8b, v4.8h
+ st1 {v4.s}[0], [x0], #4
+ st1 {v4.h}[2], [x0], x1
+ b.ne 1b
+2:
+ ldp q14, q15, [sp, #32]
+ ldp q8, q9, [sp, #64]
+ ldp q10, q11, [sp, #96]
+ ldp q12, q13, [sp], #128
+ ret
+endfunc
+
+function ff_hevc_put_hevc_epel_uni_w_hv8_8_neon_i8mm, export=1
+ epel_uni_w_hv_start
+ sxtw x4, w4
+
+ add x10, x4, #3
+ lsl x10, x10, #7
+ sub sp, sp, x10 // tmp_array
+ stp xzr, x30, [sp, #-48]!
+ stp x4, x6, [sp, #16]
+ stp x0, x1, [sp, #32]
+ add x0, sp, #48
+ sub x1, x2, x3
+ mov x2, x3
+ add x3, x4, #3
+ mov x4, x5
+ bl X(ff_hevc_put_hevc_epel_h8_8_neon_i8mm)
+ ldp x4, x6, [sp, #16]
+ ldp x0, x1, [sp, #32]
+ ldp xzr, x30, [sp], #48
+ load_epel_filterh x6, x5
+ mov x10, #(MAX_PB_SIZE * 2)
+ ld1 {v16.8h}, [sp], x10
+ ld1 {v17.8h}, [sp], x10
+ ld1 {v18.8h}, [sp], x10
+1: ld1 {v19.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v16, v17, v18, v19
+ calc_epelh2 v4, v5, v16, v17, v18, v19
+ epel_uni_w_hv_end
+ sqxtun v4.8b, v4.8h
+ st1 {v4.8b}, [x0], x1
+ b.eq 2f
+
+ ld1 {v16.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v17, v18, v19, v16
+ calc_epelh2 v4, v5, v17, v18, v19, v16
+ epel_uni_w_hv_end
+ sqxtun v4.8b, v4.8h
+ st1 {v4.8b}, [x0], x1
+ b.eq 2f
+
+ ld1 {v17.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v18, v19, v16, v17
+ calc_epelh2 v4, v5, v18, v19, v16, v17
+ epel_uni_w_hv_end
+ sqxtun v4.8b, v4.8h
+ st1 {v4.8b}, [x0], x1
+ b.eq 2f
+
+ ld1 {v18.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v19, v16, v17, v18
+ calc_epelh2 v4, v5, v19, v16, v17, v18
+ epel_uni_w_hv_end
+ sqxtun v4.8b, v4.8h
+ st1 {v4.8b}, [x0], x1
+ b.ne 1b
+2:
+ ldp q14, q15, [sp, #32]
+ ldp q8, q9, [sp, #64]
+ ldp q10, q11, [sp, #96]
+ ldp q12, q13, [sp], #128
+ ret
+endfunc
+
+function ff_hevc_put_hevc_epel_uni_w_hv12_8_neon_i8mm, export=1
+ epel_uni_w_hv_start
+ sxtw x4, w4
+
+ add x10, x4, #3
+ lsl x10, x10, #7
+ sub sp, sp, x10 // tmp_array
+ stp xzr, x30, [sp, #-48]!
+ stp x4, x6, [sp, #16]
+ stp x0, x1, [sp, #32]
+ add x0, sp, #48
+ sub x1, x2, x3
+ mov x2, x3
+ add x3, x4, #3
+ mov x4, x5
+ bl X(ff_hevc_put_hevc_epel_h12_8_neon_i8mm)
+ ldp x4, x6, [sp, #16]
+ ldp x0, x1, [sp, #32]
+ ldp xzr, x30, [sp], #48
+ load_epel_filterh x6, x5
+ sub x1, x1, #8
+ mov x10, #(MAX_PB_SIZE * 2)
+ ld1 {v16.8h, v17.8h}, [sp], x10
+ ld1 {v18.8h, v19.8h}, [sp], x10
+ ld1 {v20.8h, v21.8h}, [sp], x10
+1: ld1 {v22.8h, v23.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v16, v18, v20, v22
+ calc_epelh2 v4, v5, v16, v18, v20, v22
+ calc_epelh v5, v17, v19, v21, v23
+ epel_uni_w_hv_end2
+ sqxtun v4.8b, v4.8h
+ sqxtun2 v4.16b, v5.8h
+ st1 {v4.8b}, [x0], #8
+ st1 {v4.s}[2], [x0], x1
+ b.eq 2f
+
+ ld1 {v16.8h, v17.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v18, v20, v22, v16
+ calc_epelh2 v4, v5, v18, v20, v22, v16
+ calc_epelh v5, v19, v21, v23, v17
+ epel_uni_w_hv_end2
+ sqxtun v4.8b, v4.8h
+ sqxtun2 v4.16b, v5.8h
+ st1 {v4.8b}, [x0], #8
+ st1 {v4.s}[2], [x0], x1
+ b.eq 2f
+
+ ld1 {v18.8h, v19.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v20, v22, v16, v18
+ calc_epelh2 v4, v5, v20, v22, v16, v18
+ calc_epelh v5, v21, v23, v17, v19
+ epel_uni_w_hv_end2
+ sqxtun v4.8b, v4.8h
+ sqxtun2 v4.16b, v5.8h
+ st1 {v4.8b}, [x0], #8
+ st1 {v4.s}[2], [x0], x1
+ b.eq 2f
+
+ ld1 {v20.8h, v21.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v22, v16, v18, v20
+ calc_epelh2 v4, v5, v22, v16, v18, v20
+ calc_epelh v5, v23, v17, v19, v21
+ epel_uni_w_hv_end2
+ sqxtun v4.8b, v4.8h
+ sqxtun2 v4.16b, v5.8h
+ st1 {v4.8b}, [x0], #8
+ st1 {v4.s}[2], [x0], x1
+ b.ne 1b
+2:
+ ldp q14, q15, [sp, #32]
+ ldp q8, q9, [sp, #64]
+ ldp q10, q11, [sp, #96]
+ ldp q12, q13, [sp], #128
+ ret
+endfunc
+
+function ff_hevc_put_hevc_epel_uni_w_hv16_8_neon_i8mm, export=1
+ epel_uni_w_hv_start
+ sxtw x4, w4
+
+ add x10, x4, #3
+ lsl x10, x10, #7
+ sub sp, sp, x10 // tmp_array
+ stp xzr, x30, [sp, #-48]!
+ stp x4, x6, [sp, #16]
+ stp x0, x1, [sp, #32]
+ add x0, sp, #48
+ sub x1, x2, x3
+ mov x2, x3
+ add x3, x4, #3
+ mov x4, x5
+ bl X(ff_hevc_put_hevc_epel_h16_8_neon_i8mm)
+ ldp x4, x6, [sp, #16]
+ ldp x0, x1, [sp, #32]
+ ldp xzr, x30, [sp], #48
+ load_epel_filterh x6, x5
+ mov x10, #(MAX_PB_SIZE * 2)
+ ld1 {v16.8h, v17.8h}, [sp], x10
+ ld1 {v18.8h, v19.8h}, [sp], x10
+ ld1 {v20.8h, v21.8h}, [sp], x10
+1: ld1 {v22.8h, v23.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v16, v18, v20, v22
+ calc_epelh2 v4, v5, v16, v18, v20, v22
+ calc_epelh v5, v17, v19, v21, v23
+ calc_epelh2 v5, v6, v17, v19, v21, v23
+ epel_uni_w_hv_end2
+ sqxtun v4.8b, v4.8h
+ sqxtun2 v4.16b, v5.8h
+ st1 {v4.16b}, [x0], x1
+ b.eq 2f
+
+ ld1 {v16.8h, v17.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v18, v20, v22, v16
+ calc_epelh2 v4, v5, v18, v20, v22, v16
+ calc_epelh v5, v19, v21, v23, v17
+ calc_epelh2 v5, v6, v19, v21, v23, v17
+ epel_uni_w_hv_end2
+ sqxtun v4.8b, v4.8h
+ sqxtun2 v4.16b, v5.8h
+ st1 {v4.16b}, [x0], x1
+ b.eq 2f
+
+ ld1 {v18.8h, v19.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v20, v22, v16, v18
+ calc_epelh2 v4, v5, v20, v22, v16, v18
+ calc_epelh v5, v21, v23, v17, v19
+ calc_epelh2 v5, v6, v21, v23, v17, v19
+ epel_uni_w_hv_end2
+ sqxtun v4.8b, v4.8h
+ sqxtun2 v4.16b, v5.8h
+ st1 {v4.16b}, [x0], x1
+ b.eq 2f
+
+ ld1 {v20.8h, v21.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v22, v16, v18, v20
+ calc_epelh2 v4, v5, v22, v16, v18, v20
+ calc_epelh v5, v23, v17, v19, v21
+ calc_epelh2 v5, v6, v23, v17, v19, v21
+ epel_uni_w_hv_end2
+ sqxtun v4.8b, v4.8h
+ sqxtun2 v4.16b, v5.8h
+ st1 {v4.16b}, [x0], x1
+ b.ne 1b
+2:
+ ldp q14, q15, [sp, #32]
+ ldp q8, q9, [sp, #64]
+ ldp q10, q11, [sp, #96]
+ ldp q12, q13, [sp], #128
+ ret
+endfunc
+
+function ff_hevc_put_hevc_epel_uni_w_hv24_8_neon_i8mm, export=1
+ epel_uni_w_hv_start
+ sxtw x4, w4
+
+ add x10, x4, #3
+ lsl x10, x10, #7
+ sub sp, sp, x10 // tmp_array
+ stp xzr, x30, [sp, #-48]!
+ stp x4, x6, [sp, #16]
+ stp x0, x1, [sp, #32]
+ add x0, sp, #48
+ sub x1, x2, x3
+ mov x2, x3
+ add x3, x4, #3
+ mov x4, x5
+ bl X(ff_hevc_put_hevc_epel_h24_8_neon_i8mm)
+ ldp x4, x6, [sp, #16]
+ ldp x0, x1, [sp, #32]
+ ldp xzr, x30, [sp], #48
+ load_epel_filterh x6, x5
+ mov x10, #(MAX_PB_SIZE * 2)
+ ld1 {v16.8h, v17.8h, v18.8h}, [sp], x10
+ ld1 {v19.8h, v20.8h, v21.8h}, [sp], x10
+ ld1 {v22.8h, v23.8h, v24.8h}, [sp], x10
+1: ld1 {v25.8h, v26.8h, v27.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v16, v19, v22, v25
+ calc_epelh2 v4, v5, v16, v19, v22, v25
+ calc_epelh v5, v17, v20, v23, v26
+ calc_epelh2 v5, v6, v17, v20, v23, v26
+ calc_epelh v6, v18, v21, v24, v27
+ calc_epelh2 v6, v7, v18, v21, v24, v27
+
+ epel_uni_w_hv_end3
+ sqxtun v4.8b, v4.8h
+ sqxtun v5.8b, v5.8h
+ sqxtun v6.8b, v6.8h
+ st1 {v4.8b, v5.8b, v6.8b}, [x0], x1
+ b.eq 2f
+
+ ld1 {v16.8h, v17.8h, v18.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v19, v22, v25, v16
+ calc_epelh2 v4, v5, v19, v22, v25, v16
+ calc_epelh v5, v20, v23, v26, v17
+ calc_epelh2 v5, v6, v20, v23, v26, v17
+ calc_epelh v6, v21, v24, v27, v18
+ calc_epelh2 v6, v7, v21, v24, v27, v18
+ epel_uni_w_hv_end3
+
+ sqxtun v4.8b, v4.8h
+ sqxtun v5.8b, v5.8h
+ sqxtun v6.8b, v6.8h
+ st1 {v4.8b, v5.8b, v6.8b}, [x0], x1
+ b.eq 2f
+
+ ld1 {v19.8h, v20.8h, v21.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v22, v25, v16, v19
+ calc_epelh2 v4, v5, v22, v25, v16, v19
+ calc_epelh v5, v23, v26, v17, v20
+ calc_epelh2 v5, v6, v23, v26, v17, v20
+ calc_epelh v6, v24, v27, v18, v21
+ calc_epelh2 v6, v7, v24, v27, v18, v21
+ epel_uni_w_hv_end3
+
+ sqxtun v4.8b, v4.8h
+ sqxtun v5.8b, v5.8h
+ sqxtun v6.8b, v6.8h
+ st1 {v4.8b, v5.8b, v6.8b}, [x0], x1
+ b.eq 2f
+
+ ld1 {v22.8h, v23.8h, v24.8h}, [sp], x10
+ subs x4, x4, #1
+ calc_epelh v4, v25, v16, v19, v22
+ calc_epelh2 v4, v5, v25, v16, v19, v22
+ calc_epelh v5, v26, v17, v20, v23
+ calc_epelh2 v5, v6, v26, v17, v20, v23
+ calc_epelh v6, v27, v18, v21, v24
+ calc_epelh2 v6, v7, v27, v18, v21, v24
+ epel_uni_w_hv_end3
+
+ sqxtun v4.8b, v4.8h
+ sqxtun v5.8b, v5.8h
+ sqxtun v6.8b, v6.8h
+ st1 {v4.8b, v5.8b, v6.8b}, [x0], x1
+ b.ne 1b
+2:
+ ldp q14, q15, [sp, #32]
+ ldp q8, q9, [sp, #64]
+ ldp q10, q11, [sp, #96]
+ ldp q12, q13, [sp], #128
+ ret
+endfunc
+
+function ff_hevc_put_hevc_epel_uni_w_hv32_8_neon_i8mm, export=1
+ ldp x15, x16, [sp]
+ stp x0, x30, [sp, #-16]!
+ stp x1, x2, [sp, #-16]!
+ stp x3, x4, [sp, #-16]!
+ stp x5, x6, [sp, #-16]!
+ mov x17, #16
+ stp x17, x7, [sp, #-16]!
+ stp x15, x16, [sp, #-16]!
+ bl X(ff_hevc_put_hevc_epel_uni_w_hv16_8_neon_i8mm)
+ ldp x15, x16, [sp], #16
+ ldp x17, x7, [sp], #16
+ ldp x5, x6, [sp], #16
+ ldp x3, x4, [sp], #16
+ ldp x1, x2, [sp], #16
+ ldr x0, [sp]
+ add x0, x0, #16
+ add x2, x2, #16
+ mov x17, #16
+ stp x17, xzr, [sp, #-16]!
+ stp x15, x16, [sp, #-16]!
+ bl X(ff_hevc_put_hevc_epel_uni_w_hv16_8_neon_i8mm)
+ ldp x15, x16, [sp], #16
+ ldp x17, xzr, [sp], #16
+ ldp xzr, x30, [sp], #16
+ ret
+endfunc
+
+function ff_hevc_put_hevc_epel_uni_w_hv48_8_neon_i8mm, export=1
+ ldp x15, x16, [sp]
+ stp x0, x30, [sp, #-16]!
+ stp x1, x2, [sp, #-16]!
+ stp x3, x4, [sp, #-16]!
+ stp x5, x6, [sp, #-16]!
+ mov x17, #24
+ stp x17, x7, [sp, #-16]!
+ stp x15, x16, [sp, #-16]!
+ bl X(ff_hevc_put_hevc_epel_uni_w_hv24_8_neon_i8mm)
+ ldp x15, x16, [sp], #16
+ ldp x17, x7, [sp], #16
+ ldp x5, x6, [sp], #16
+ ldp x3, x4, [sp], #16
+ ldp x1, x2, [sp], #16
+ ldr x0, [sp]
+ add x0, x0, #24
+ add x2, x2, #24
+ mov x17, #24
+ stp x17, xzr, [sp, #-16]!
+ stp x15, x16, [sp, #-16]!
+ bl X(ff_hevc_put_hevc_epel_uni_w_hv24_8_neon_i8mm)
+ ldp x15, x16, [sp], #16
+ ldp x17, xzr, [sp], #16
+ ldp xzr, x30, [sp], #16
+ ret
+endfunc
+
+function ff_hevc_put_hevc_epel_uni_w_hv64_8_neon_i8mm, export=1
+ ldp x15, x16, [sp]
+ stp x0, x30, [sp, #-16]!
+ stp x1, x2, [sp, #-16]!
+ stp x3, x4, [sp, #-16]!
+ stp x5, x6, [sp, #-16]!
+ mov x17, #16
+ stp x17, x7, [sp, #-16]!
+ stp x15, x16, [sp, #-16]!
+ bl X(ff_hevc_put_hevc_epel_uni_w_hv16_8_neon_i8mm)
+ ldp x15, x16, [sp], #16
+ ldp x17, x7, [sp], #16
+ ldp x5, x6, [sp]
+ ldp x3, x4, [sp, #16]
+ ldp x1, x2, [sp, #32]
+ ldr x0, [sp, #48]
+ add x0, x0, #16
+ add x2, x2, #16
+ mov x17, #16
+ stp x17, x7, [sp, #-16]!
+ stp x15, x16, [sp, #-16]!
+
+ bl X(ff_hevc_put_hevc_epel_uni_w_hv16_8_neon_i8mm)
+ ldp x15, x16, [sp], #16
+ ldp x17, x7, [sp], #16
+ ldp x5, x6, [sp]
+ ldp x3, x4, [sp, #16]
+ ldp x1, x2, [sp, #32]
+ ldr x0, [sp, #48]
+ add x0, x0, #32
+ add x2, x2, #32
+ mov x17, #16
+ stp x17, x7, [sp, #-16]!
+ stp x15, x16, [sp, #-16]!
+ bl X(ff_hevc_put_hevc_epel_uni_w_hv16_8_neon_i8mm)
+ ldp x15, x16, [sp], #16
+ ldp x17, x7, [sp], #16
+ ldp x5, x6, [sp], #16
+ ldp x3, x4, [sp], #16
+ ldp x1, x2, [sp], #16
+ ldr x0, [sp]
+ add x0, x0, #48
+ add x2, x2, #48
+ mov x17, #16
+ stp x17, xzr, [sp, #-16]!
+ stp x15, x16, [sp, #-16]!
+ bl X(ff_hevc_put_hevc_epel_uni_w_hv16_8_neon_i8mm)
+ ldp x15, x16, [sp], #16
+ ldp x17, xzr, [sp], #16
+ ldp xzr, x30, [sp], #16
+ ret
+endfunc
+
+
#endif
diff --git a/libavcodec/aarch64/hevcdsp_init_aarch64.c b/libavcodec/aarch64/hevcdsp_init_aarch64.c
index b448d755b9..e125b0cfb2 100644
--- a/libavcodec/aarch64/hevcdsp_init_aarch64.c
+++ b/libavcodec/aarch64/hevcdsp_init_aarch64.c
@@ -189,6 +189,11 @@ NEON8_FNPROTO(qpel_uni_w_h, (uint8_t *_dst, ptrdiff_t _dststride,
int height, int denom, int wx, int ox,
intptr_t mx, intptr_t my, int width), _i8mm);
+NEON8_FNPROTO(epel_uni_w_hv, (uint8_t *_dst, ptrdiff_t _dststride,
+ const uint8_t *_src, ptrdiff_t _srcstride,
+ int height, int denom, int wx, int ox,
+ intptr_t mx, intptr_t my, int width), _i8mm);
+
NEON8_FNPROTO_PARTIAL_5(qpel_uni_w_hv, (uint8_t *_dst, ptrdiff_t _dststride,
const uint8_t *_src, ptrdiff_t _srcstride,
int height, int denom, int wx, int ox,
@@ -291,6 +296,7 @@ av_cold void ff_hevc_dsp_init_aarch64(HEVCDSPContext *c, const int bit_depth)
NEON8_FNASSIGN(c->put_hevc_epel_uni_w, 0, 1, epel_uni_w_h ,_i8mm);
NEON8_FNASSIGN(c->put_hevc_qpel, 0, 1, qpel_h, _i8mm);
NEON8_FNASSIGN(c->put_hevc_qpel_uni_w, 0, 1, qpel_uni_w_h, _i8mm);
+ NEON8_FNASSIGN(c->put_hevc_epel_uni_w, 1, 1, epel_uni_w_hv, _i8mm);
NEON8_FNASSIGN_PARTIAL_5(c->put_hevc_qpel_uni_w, 1, 1, qpel_uni_w_hv, _i8mm);
}
--
2.38.0.windows.1
[-- Attachment #3: Type: text/plain, Size: 251 bytes --]
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
next prev parent reply other threads:[~2023-06-18 8:25 UTC|newest]
Thread overview: 22+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-06-04 4:17 [FFmpeg-devel] [PATCH 1/5] lavc/aarch64: new optimization for 8-bit hevc_pel_uni_pixels Logan.Lyu
2023-06-04 4:17 ` [FFmpeg-devel] [PATCH 2/5] lavc/aarch64: new optimization for 8-bit hevc_epel_uni_w_h Logan.Lyu
2023-06-12 7:59 ` Martin Storsjö
2023-06-18 8:21 ` Logan.Lyu
2023-06-04 4:17 ` [FFmpeg-devel] [PATCH 3/5] lavc/aarch64: new optimization for 8-bit hevc_epel_uni_w_v Logan.Lyu
2023-06-12 8:09 ` Martin Storsjö
2023-06-12 9:08 ` Martin Storsjö
2023-06-18 8:22 ` Logan.Lyu
2023-07-01 21:21 ` Martin Storsjö
2023-06-04 4:17 ` [FFmpeg-devel] [PATCH 4/5] lavc/aarch64: new optimization for 8-bit hevc_epel_h Logan.Lyu
2023-06-12 8:12 ` Martin Storsjö
2023-06-18 8:23 ` Logan.Lyu
2023-06-18 8:26 ` Logan.Lyu
2023-06-04 4:17 ` [FFmpeg-devel] [PATCH 5/5] lavc/aarch64: new optimization for 8-bit hevc_epel_uni_w_hv Logan.Lyu
2023-06-12 8:19 ` Martin Storsjö
2023-06-18 8:25 ` Logan.Lyu [this message]
2023-07-01 21:28 ` Martin Storsjö
2023-07-13 14:54 ` Logan.Lyu
2023-07-14 9:28 ` Martin Storsjö
2023-06-12 7:47 ` [FFmpeg-devel] [PATCH 1/5] lavc/aarch64: new optimization for 8-bit hevc_pel_uni_pixels Martin Storsjö
2023-06-18 8:29 ` Logan.Lyu
2023-07-01 21:16 ` Martin Storsjö
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=08a60626-b5f4-935b-5a56-2e001b59b8d7@myais.com.cn \
--to=logan.lyu@myais.com.cn \
--cc=ffmpeg-devel@ffmpeg.org \
--cc=martin@martin.st \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
This inbox may be cloned and mirrored by anyone:
git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git
# If you have public-inbox 1.1+ installed, you may
# initialize and index your mirror using the following commands:
public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \
ffmpegdev@gitmailbox.com
public-inbox-index ffmpegdev
Example config snippet for mirrors.
AGPL code for this site: git clone https://public-inbox.org/public-inbox.git