From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org [79.124.17.100]) by master.gitmailbox.com (Postfix) with ESMTP id 9AADD48A20 for ; Mon, 22 Jul 2024 18:12:29 +0000 (UTC) Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 590B068D6FD; Mon, 22 Jul 2024 21:12:16 +0300 (EEST) Received: from ursule.remlab.net (vps-a2bccee9.vps.ovh.net [51.75.19.47]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id CC20968D695 for ; Mon, 22 Jul 2024 21:12:02 +0300 (EEST) Received: from basile.remlab.net (localhost [IPv6:::1]) by ursule.remlab.net (Postfix) with ESMTP id 58405C01F0 for ; Mon, 22 Jul 2024 21:12:02 +0300 (EEST) From: =?UTF-8?q?R=C3=A9mi=20Denis-Courmont?= To: ffmpeg-devel@ffmpeg.org Date: Mon, 22 Jul 2024 21:12:00 +0300 Message-ID: <20240722181201.24563-3-remi@remlab.net> X-Mailer: git-send-email 2.45.2 In-Reply-To: <20240722181201.24563-1-remi@remlab.net> References: <20240722181201.24563-1-remi@remlab.net> MIME-Version: 1.0 Subject: [FFmpeg-devel] [PATCH 3/4] lavc/vp9dsp: copy 8 pixels at once 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 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" Archived-At: List-Archive: List-Post: In the 8-bit case, we can actually read/write 8 aligned pixel values per load/store, which unsurprisingly tends to be faster on 64-bit systems (and makes no differences on 32-bit systems). This requires ifdef'ing though. --- libavcodec/vp9dsp_template.c | 32 ++++++++++++++++++++++++++++++++ 1 file changed, 32 insertions(+) diff --git a/libavcodec/vp9dsp_template.c b/libavcodec/vp9dsp_template.c index da3cc28e5e..9e5b25142d 100644 --- a/libavcodec/vp9dsp_template.c +++ b/libavcodec/vp9dsp_template.c @@ -49,14 +49,22 @@ static void vert_8x8_c(uint8_t *restrict _dst, ptrdiff_t stride, { pixel *dst = (pixel *) _dst; const pixel *top = (const pixel *) _top; +#if BIT_DEPTH == 8 + uint64_t p8 = AV_RN64A(top); +#else pixel4 p4a = AV_RN4PA(top + 0); pixel4 p4b = AV_RN4PA(top + 4); +#endif int y; stride /= sizeof(pixel); for (y = 0; y < 8; y++) { +#if BIT_DEPTH == 8 + AV_WN64A(dst, p8); +#else AV_WN4PA(dst + 0, p4a); AV_WN4PA(dst + 4, p4b); +#endif dst += stride; } } @@ -66,18 +74,28 @@ static void vert_16x16_c(uint8_t *restrict _dst, ptrdiff_t stride, { pixel *dst = (pixel *) _dst; const pixel *top = (const pixel *) _top; +#if BIT_DEPTH == 8 + uint64_t p8a = AV_RN64A(top); + uint64_t p8b = AV_RN64A(top + 8); +#else pixel4 p4a = AV_RN4PA(top + 0); pixel4 p4b = AV_RN4PA(top + 4); pixel4 p4c = AV_RN4PA(top + 8); pixel4 p4d = AV_RN4PA(top + 12); +#endif int y; stride /= sizeof(pixel); for (y = 0; y < 16; y++) { +#if BIT_DEPTH == 8 + AV_WN64A(dst + 0, p8a); + AV_WN64A(dst + 8, p8b); +#else AV_WN4PA(dst + 0, p4a); AV_WN4PA(dst + 4, p4b); AV_WN4PA(dst + 8, p4c); AV_WN4PA(dst + 12, p4d); +#endif dst += stride; } } @@ -87,6 +105,12 @@ static void vert_32x32_c(uint8_t *restrict _dst, ptrdiff_t stride, { pixel *dst = (pixel *) _dst; const pixel *top = (const pixel *) _top; +#if BIT_DEPTH == 8 + uint64_t p8a = AV_RN64A(top); + uint64_t p8b = AV_RN64A(top + 8); + uint64_t p8c = AV_RN64A(top + 16); + uint64_t p8d = AV_RN64A(top + 24); +#else pixel4 p4a = AV_RN4PA(top + 0); pixel4 p4b = AV_RN4PA(top + 4); pixel4 p4c = AV_RN4PA(top + 8); @@ -95,10 +119,17 @@ static void vert_32x32_c(uint8_t *restrict _dst, ptrdiff_t stride, pixel4 p4f = AV_RN4PA(top + 20); pixel4 p4g = AV_RN4PA(top + 24); pixel4 p4h = AV_RN4PA(top + 28); +#endif int y; stride /= sizeof(pixel); for (y = 0; y < 32; y++) { +#if BIT_DEPTH == 8 + AV_WN64A(dst + 0, p8a); + AV_WN64A(dst + 8, p8b); + AV_WN64A(dst + 16, p8c); + AV_WN64A(dst + 24, p8d); +#else AV_WN4PA(dst + 0, p4a); AV_WN4PA(dst + 4, p4b); AV_WN4PA(dst + 8, p4c); @@ -107,6 +138,7 @@ static void vert_32x32_c(uint8_t *restrict _dst, ptrdiff_t stride, AV_WN4PA(dst + 20, p4f); AV_WN4PA(dst + 24, p4g); AV_WN4PA(dst + 28, p4h); +#endif dst += stride; } } -- 2.45.2 _______________________________________________ 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".