From patchwork Sun Sep 29 12:02:27 2024 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Zhao Zhili X-Patchwork-Id: 51931 Delivered-To: ffmpegpatchwork2@gmail.com Received: by 2002:a59:d8ca:0:b0:48e:c0f8:d0de with SMTP id dy10csp1384956vqb; Sun, 29 Sep 2024 05:03:59 -0700 (PDT) X-Forwarded-Encrypted: i=2; AJvYcCUI129/h+EwtvDGvAuN19FEltsPOE5FTcflO+OWx55Xld0s3dd2inHkD5UARPZF232zvQ2VlmDVvJdwo6OIHKXe@gmail.com X-Google-Smtp-Source: AGHT+IGEYzywNCWBI3hTZl+vic4mNRboj7tpm7gwZ0hzeaqlULcke9UZxd2wwBxDWC+VyM27Ni4E X-Received: by 2002:a17:907:9604:b0:a8a:906d:b485 with SMTP id a640c23a62f3a-a93c490e060mr1151465766b.26.1727611439126; Sun, 29 Sep 2024 05:03:59 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1727611439; cv=none; d=google.com; s=arc-20240605; b=fCqpRWV2jr0rSWkkZ5yPwUHatpgkxLKY3+UmN7Uk/NOy4SMF27M1e2s2qytBMBmwHG ywK/+a0xTgtAnlPgkTZD7I7pkv4K4+IAe2Oc0tbkb4VlKg9FOTHsolu892NYsD98H7Gg hbwn5pKcPYii0Hk4MXmnj0mV+EtRQPUgqT2k8AOvulTx8taMMABd4tP2IY61kp0qj0wn I2hi0AHk5qp0RxFdOa54lg5rhVFEgSs3naVK9K4Sm5SV6e9BNb2F5EtB9JWHLFPU8nx9 0Izxcf/pRenJdI9cHDDqLc0Aw6M2p/UaEi7svR4EgLuZ1LkP2o5vB6RGH40rfSHg69kn ouDA== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20240605; h=sender:errors-to:content-transfer-encoding:cc:reply-to :list-subscribe:list-help:list-post:list-archive:list-unsubscribe :list-id:precedence:subject:mime-version:date:to:from:message-id :dkim-signature:delivered-to; bh=qJuJOqBQyMgE4YinpoqlKWTkxPcTPB8xiWqxu2kh9QM=; fh=HnHYuZ9XgUo86ZRXTLWWmQxhslYEI9B9taZ5X1DLFfc=; b=iqK/6Qvf6S6lSoaITXIFq3039fihC2dPgb5l8fQ4T3F5xkGAfsQLPBanPqIs2CvxeX S3hkPMCitDxyz1mm0DTLRrRJ/04tX/nG0QKhGwaORbujYkyF4RmtIYmEW4RYnko6/aOJ qLpgxrrK1h69eYupPrMjwLK2Wy7Yb84hSdrvK1Wc/CU0th7sfXB6SPvGw2EvB5QMJATq LK5H8CGu0HjUM4/JsB2dQK9ustuRCaM04TsxPj8uqV5f5WzyhEzL8ONtLIRbB3hdRROy yfxlQQv6vKZM0YPIMXC3Nm+lAPj6bELc0nkAke3N0OAX/DRIE1FMz2/dR9E6/45qR3Nf T6Gg==; dara=google.com ARC-Authentication-Results: i=1; mx.google.com; dkim=neutral (body hash did not verify) header.i=@foxmail.com header.s=s201512 header.b=PjN+Haxv; spf=pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) smtp.mailfrom=ffmpeg-devel-bounces@ffmpeg.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=foxmail.com Return-Path: Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org. [79.124.17.100]) by mx.google.com with ESMTP id a640c23a62f3a-a93c2a28653si442115166b.732.2024.09.29.05.03.58; Sun, 29 Sep 2024 05:03:59 -0700 (PDT) Received-SPF: pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) client-ip=79.124.17.100; Authentication-Results: mx.google.com; dkim=neutral (body hash did not verify) header.i=@foxmail.com header.s=s201512 header.b=PjN+Haxv; spf=pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) smtp.mailfrom=ffmpeg-devel-bounces@ffmpeg.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=foxmail.com Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 4FB8468DB29; Sun, 29 Sep 2024 15:03:56 +0300 (EEST) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from out203-205-221-233.mail.qq.com (out203-205-221-233.mail.qq.com [203.205.221.233]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id E1AC468DAA6 for ; Sun, 29 Sep 2024 15:03:48 +0300 (EEST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=foxmail.com; s=s201512; t=1727611425; bh=eUAUBZ8QAyhbpMFgf8/QkYUlXTty7YIgqWr7uWLg8y4=; h=From:To:Cc:Subject:Date; b=PjN+HaxvixEKWSsf51/HAJJPLrcnIB0/RTjCKqdyskElHSWEv71zocPnyodiuS8Pu /jh3cNcj0b00hcUCcFDMF/PdRl7lWY1kaTb+AJdAi8GxPnv9YkLSG36UnIxn889T8o i2qtGTcIbsDIbyp52jHsabst0598mSuPIgPB21so= Received: from ZHILIZHAO-MB1.tencent.com ([119.147.10.196]) by newxmesmtplogicsvrszb9-0.qq.com (NewEsmtp) with SMTP id A12BEE2; Sun, 29 Sep 2024 20:02:33 +0800 X-QQ-mid: xmsmtpt1727611353tfqw5sizi Message-ID: X-QQ-XMAILINFO: NB2Y38G+/V2YwckhGIh7C/ox4kFfdlOWkZZieIiVTtQru8vJcteK1NSq7qkoY4 mMvSunF0vwImT2sBuXAjVOFfbTtKG1F2eY0VD782kYuc+3ZDwwqPTozOOt4fAHdn8jfHy/+b4rST wQZVDIkPlHGTgM83KdjSqdXrPyYmpu4fvhBcLlH+yS9998sHFsalfvqurb3C5I11vtVHk+yHM9/M YSnOhQ+hHe/x51lYz70USnpvUVuMgjC3fnVCrzgh5rQ0Xx/o+n+uNIatjuV3lPWMj+y/bVFDaa1I PGT96PAsZLOtAHXcK5l0QD1TYbNfZa747+ZIWqzGOc8dqOxRef9zSxPLlHEepU4RhmeIam6SH6F9 ge88S0JeMFHcGFOqOqXHDwIeWbMVNwW/XzAXz2NiBn3Eq4TP4p2tR8qdF3ZJvpD67CMdWQoMX4Dk pd/Dd57rmwTzmLCpttGY42Akd5oviYLJm+wZvIwvNAhgwtu0brJfS5/b78eEVcv1JWYfynZkC4FN J8wFlzq+irjqC0OCuCb5e/FOke5uHCStvN+zUCc374ehOsr9Ba4N55DXI5SuDDNL8qBg6hUD9UKm 5oBGclBWTHwsnKrYohI81xUE1OLJqk0Q2+4F+j0gL1fivYD+PRwB271Wp6BqvnhLpblUKLVhPqTZ QJU399EsOBIMS8K6rUsplu5T9nuSM3ngOUu3ut2qkDmVB/LFuyYaECNAdrraYt5pY4MDa2aZFIYF VwOYAqpTyuB2Nk0DOicvPlrQxmWtwN4IeNQTMMSAZDnKuwX5QaPNZ2oGPeSg3VrD6fX0uws1na/r JkcM0XlQvtPL627ZwbgwZMGBpdwza7U/bWG1eZ3ic3AtDOZvYpJ7xcdGuRkTOpryW97RF0FZ2K9+ RPM9dC8TCK/fcdio/bsWNxrqfB8KfmJ4GfCk7gxoU+WgpaBQlL1GFjz7bLSFz5KVeIg4unGJ8Tzy jGw8qi8yJTmQm5ifStkPF929pdQPh3qHrNH1zhQbSIXFFjnPXIAA38MVnsPM1M7DtKe2nr/TVO5r jj3S7f4Q== X-QQ-XMRINFO: Mp0Kj//9VHAxr69bL5MkOOs= From: Zhao Zhili To: ffmpeg-devel@ffmpeg.org Date: Sun, 29 Sep 2024 20:02:27 +0800 X-OQ-MSGID: <20240929120229.41608-1-quinkblack@foxmail.com> X-Mailer: git-send-email 2.46.0 MIME-Version: 1.0 Subject: [FFmpeg-devel] [PATCH v4 1/3] aarch64/vvc: Add w_avg X-BeenThere: ffmpeg-devel@ffmpeg.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: FFmpeg development discussions and patches List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Reply-To: FFmpeg development discussions and patches Cc: Zhao Zhili Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" X-TUID: 2Mq87I4bRTqB From: Zhao Zhili w_avg_8_2x2_c: 0.0 ( 0.00x) w_avg_8_2x2_neon: 0.0 ( 0.00x) w_avg_8_4x4_c: 0.2 ( 1.00x) w_avg_8_4x4_neon: 0.0 ( 0.00x) w_avg_8_8x8_c: 1.2 ( 1.00x) w_avg_8_8x8_neon: 0.2 ( 5.00x) w_avg_8_16x16_c: 4.2 ( 1.00x) w_avg_8_16x16_neon: 0.8 ( 5.67x) w_avg_8_32x32_c: 16.2 ( 1.00x) w_avg_8_32x32_neon: 2.5 ( 6.50x) w_avg_8_64x64_c: 64.5 ( 1.00x) w_avg_8_64x64_neon: 9.0 ( 7.17x) w_avg_8_128x128_c: 269.5 ( 1.00x) w_avg_8_128x128_neon: 35.5 ( 7.59x) w_avg_10_2x2_c: 0.2 ( 1.00x) w_avg_10_2x2_neon: 0.2 ( 1.00x) w_avg_10_4x4_c: 0.2 ( 1.00x) w_avg_10_4x4_neon: 0.2 ( 1.00x) w_avg_10_8x8_c: 1.0 ( 1.00x) w_avg_10_8x8_neon: 0.2 ( 4.00x) w_avg_10_16x16_c: 4.2 ( 1.00x) w_avg_10_16x16_neon: 0.8 ( 5.67x) w_avg_10_32x32_c: 16.2 ( 1.00x) w_avg_10_32x32_neon: 2.5 ( 6.50x) w_avg_10_64x64_c: 66.2 ( 1.00x) w_avg_10_64x64_neon: 10.0 ( 6.62x) w_avg_10_128x128_c: 277.8 ( 1.00x) w_avg_10_128x128_neon: 39.8 ( 6.99x) w_avg_12_2x2_c: 0.0 ( 0.00x) w_avg_12_2x2_neon: 0.2 ( 0.00x) w_avg_12_4x4_c: 0.2 ( 1.00x) w_avg_12_4x4_neon: 0.0 ( 0.00x) w_avg_12_8x8_c: 1.2 ( 1.00x) w_avg_12_8x8_neon: 0.5 ( 2.50x) w_avg_12_16x16_c: 4.8 ( 1.00x) w_avg_12_16x16_neon: 0.8 ( 6.33x) w_avg_12_32x32_c: 17.0 ( 1.00x) w_avg_12_32x32_neon: 2.8 ( 6.18x) w_avg_12_64x64_c: 64.0 ( 1.00x) w_avg_12_64x64_neon: 10.0 ( 6.40x) w_avg_12_128x128_c: 269.2 ( 1.00x) w_avg_12_128x128_neon: 42.0 ( 6.41x) Signed-off-by: Zhao Zhili --- libavcodec/aarch64/vvc/dsp_init.c | 36 +++++++++++ libavcodec/aarch64/vvc/inter.S | 99 +++++++++++++++++++++++++------ 2 files changed, 118 insertions(+), 17 deletions(-) diff --git a/libavcodec/aarch64/vvc/dsp_init.c b/libavcodec/aarch64/vvc/dsp_init.c index ad767d17e2..ebe58a2ba5 100644 --- a/libavcodec/aarch64/vvc/dsp_init.c +++ b/libavcodec/aarch64/vvc/dsp_init.c @@ -52,6 +52,39 @@ void ff_vvc_avg_12_neon(uint8_t *dst, ptrdiff_t dst_stride, const int16_t *src0, const int16_t *src1, int width, int height); +void ff_vvc_w_avg_8_neon(uint8_t *_dst, ptrdiff_t _dst_stride, + const int16_t *src0, const int16_t *src1, + int width, int height, + uintptr_t w0_w1, uintptr_t offset_shift); +void ff_vvc_w_avg_10_neon(uint8_t *_dst, ptrdiff_t _dst_stride, + const int16_t *src0, const int16_t *src1, + int width, int height, + uintptr_t w0_w1, uintptr_t offset_shift); +void ff_vvc_w_avg_12_neon(uint8_t *_dst, ptrdiff_t _dst_stride, + const int16_t *src0, const int16_t *src1, + int width, int height, + uintptr_t w0_w1, uintptr_t offset_shift); +/* When passing arguments to functions, Apple platforms diverge from the ARM64 + * standard ABI for functions that require passing arguments on the stack. To + * simplify portability in the assembly function interface, use a different + * function signature that doesn't require passing arguments on the stack. + */ +#define W_AVG_FUN(bit_depth) \ +static void vvc_w_avg_ ## bit_depth(uint8_t *dst, ptrdiff_t dst_stride, \ + const int16_t *src0, const int16_t *src1, int width, int height, \ + int denom, int w0, int w1, int o0, int o1) \ +{ \ + int shift = denom + FFMAX(3, 15 - bit_depth); \ + int offset = ((o0 + o1) * (1 << (bit_depth - 8)) + 1) * (1 << (shift - 1)); \ + uintptr_t w0_w1 = ((uintptr_t)w0 << 32) | (uint32_t)w1; \ + uintptr_t offset_shift = ((uintptr_t)offset << 32) | (uint32_t)shift; \ + ff_vvc_w_avg_ ## bit_depth ## _neon(dst, dst_stride, src0, src1, width, height, w0_w1, offset_shift); \ +} + +W_AVG_FUN(8) +W_AVG_FUN(10) +W_AVG_FUN(12) + void ff_vvc_dsp_init_aarch64(VVCDSPContext *const c, const int bd) { int cpu_flags = av_get_cpu_flags(); @@ -123,6 +156,7 @@ void ff_vvc_dsp_init_aarch64(VVCDSPContext *const c, const int bd) c->inter.put_uni_w[0][6][0][0] = ff_vvc_put_pel_uni_w_pixels128_8_neon; c->inter.avg = ff_vvc_avg_8_neon; + c->inter.w_avg = vvc_w_avg_8; for (int i = 0; i < FF_ARRAY_ELEMS(c->sao.band_filter); i++) c->sao.band_filter[i] = ff_h26x_sao_band_filter_8x8_8_neon; @@ -163,11 +197,13 @@ void ff_vvc_dsp_init_aarch64(VVCDSPContext *const c, const int bd) } } else if (bd == 10) { c->inter.avg = ff_vvc_avg_10_neon; + c->inter.w_avg = vvc_w_avg_10; c->alf.filter[LUMA] = alf_filter_luma_10_neon; c->alf.filter[CHROMA] = alf_filter_chroma_10_neon; } else if (bd == 12) { c->inter.avg = ff_vvc_avg_12_neon; + c->inter.w_avg = vvc_w_avg_12; c->alf.filter[LUMA] = alf_filter_luma_12_neon; c->alf.filter[CHROMA] = alf_filter_chroma_12_neon; diff --git a/libavcodec/aarch64/vvc/inter.S b/libavcodec/aarch64/vvc/inter.S index 2f69274b86..c4c6ab1a72 100644 --- a/libavcodec/aarch64/vvc/inter.S +++ b/libavcodec/aarch64/vvc/inter.S @@ -22,9 +22,9 @@ #define VVC_MAX_PB_SIZE 128 -.macro vvc_avg, bit_depth +.macro vvc_avg type, bit_depth -.macro vvc_avg_\bit_depth\()_2_4, tap +.macro vvc_\type\()_\bit_depth\()_2_4 tap .if \tap == 2 ldr s0, [src0] ldr s2, [src1] @@ -32,9 +32,19 @@ ldr d0, [src0] ldr d2, [src1] .endif + +.ifc \type, avg saddl v4.4s, v0.4h, v2.4h add v4.4s, v4.4s, v16.4s sqshrn v4.4h, v4.4s, #(15 - \bit_depth) +.else + mov v4.16b, v16.16b + smlal v4.4s, v0.4h, v19.4h + smlal v4.4s, v2.4h, v20.4h + sqshl v4.4s, v4.4s, v22.4s + sqxtn v4.4h, v4.4s +.endif + .if \bit_depth == 8 sqxtun v4.8b, v4.8h .if \tap == 2 @@ -57,7 +67,7 @@ add dst, dst, dst_stride .endm -function ff_vvc_avg_\bit_depth\()_neon, export=1 +function ff_vvc_\type\()_\bit_depth\()_neon, export=1 dst .req x0 dst_stride .req x1 src0 .req x2 @@ -67,42 +77,64 @@ function ff_vvc_avg_\bit_depth\()_neon, export=1 mov x10, #(VVC_MAX_PB_SIZE * 2) cmp width, #8 -.if \bit_depth == 8 - movi v16.4s, #64 -.else -.if \bit_depth == 10 - mov w6, #1023 - movi v16.4s, #16 +.ifc \type, avg + movi v16.4s, #(1 << (14 - \bit_depth)) .else - mov w6, #4095 - movi v16.4s, #4 -.endif + lsr x11, x6, #32 // weight0 + mov w12, w6 // weight1 + lsr x13, x7, #32 // offset + mov w14, w7 // shift + + dup v19.8h, w11 + neg w14, w14 // so we can use sqshl + dup v20.8h, w12 + dup v16.4s, w13 + dup v22.4s, w14 +.endif // avg + + .if \bit_depth >= 10 + // clip pixel + mov w6, #((1 << \bit_depth) - 1) movi v18.8h, #0 dup v17.8h, w6 .endif + b.eq 8f b.hi 16f cmp width, #4 b.eq 4f 2: // width == 2 subs height, height, #1 - vvc_avg_\bit_depth\()_2_4 2 + vvc_\type\()_\bit_depth\()_2_4 2 b.ne 2b b 32f 4: // width == 4 subs height, height, #1 - vvc_avg_\bit_depth\()_2_4 4 + vvc_\type\()_\bit_depth\()_2_4 4 b.ne 4b b 32f 8: // width == 8 ld1 {v0.8h}, [src0], x10 ld1 {v2.8h}, [src1], x10 +.ifc \type, avg saddl v4.4s, v0.4h, v2.4h saddl2 v5.4s, v0.8h, v2.8h add v4.4s, v4.4s, v16.4s add v5.4s, v5.4s, v16.4s sqshrn v4.4h, v4.4s, #(15 - \bit_depth) sqshrn2 v4.8h, v5.4s, #(15 - \bit_depth) +.else + mov v4.16b, v16.16b + mov v5.16b, v16.16b + smlal v4.4s, v0.4h, v19.4h + smlal v4.4s, v2.4h, v20.4h + smlal2 v5.4s, v0.8h, v19.8h + smlal2 v5.4s, v2.8h, v20.8h + sqshl v4.4s, v4.4s, v22.4s + sqshl v5.4s, v5.4s, v22.4s + sqxtn v4.4h, v4.4s + sqxtn2 v4.8h, v5.4s +.endif subs height, height, #1 .if \bit_depth == 8 sqxtun v4.8b, v4.8h @@ -122,6 +154,7 @@ function ff_vvc_avg_\bit_depth\()_neon, export=1 17: ldp q0, q1, [x7], #32 ldp q2, q3, [x8], #32 +.ifc \type, avg saddl v4.4s, v0.4h, v2.4h saddl2 v5.4s, v0.8h, v2.8h saddl v6.4s, v1.4h, v3.4h @@ -134,6 +167,28 @@ function ff_vvc_avg_\bit_depth\()_neon, export=1 sqshrn2 v4.8h, v5.4s, #(15 - \bit_depth) sqshrn v6.4h, v6.4s, #(15 - \bit_depth) sqshrn2 v6.8h, v7.4s, #(15 - \bit_depth) +.else // avg + mov v4.16b, v16.16b + mov v5.16b, v16.16b + mov v6.16b, v16.16b + mov v7.16b, v16.16b + smlal v4.4s, v0.4h, v19.4h + smlal v4.4s, v2.4h, v20.4h + smlal2 v5.4s, v0.8h, v19.8h + smlal2 v5.4s, v2.8h, v20.8h + smlal v6.4s, v1.4h, v19.4h + smlal v6.4s, v3.4h, v20.4h + smlal2 v7.4s, v1.8h, v19.8h + smlal2 v7.4s, v3.8h, v20.8h + sqshl v4.4s, v4.4s, v22.4s + sqshl v5.4s, v5.4s, v22.4s + sqshl v6.4s, v6.4s, v22.4s + sqshl v7.4s, v7.4s, v22.4s + sqxtn v4.4h, v4.4s + sqxtn v6.4h, v6.4s + sqxtn2 v4.8h, v5.4s + sqxtn2 v6.8h, v7.4s +.endif // w_avg subs w6, w6, #16 .if \bit_depth == 8 sqxtun v4.8b, v4.8h @@ -155,9 +210,19 @@ function ff_vvc_avg_\bit_depth\()_neon, export=1 b.ne 16b 32: ret + +.unreq dst +.unreq dst_stride +.unreq src0 +.unreq src1 +.unreq width +.unreq height endfunc .endm -vvc_avg 8 -vvc_avg 10 -vvc_avg 12 +vvc_avg avg, 8 +vvc_avg avg, 10 +vvc_avg avg, 12 +vvc_avg w_avg, 8 +vvc_avg w_avg, 10 +vvc_avg w_avg, 12