From patchwork Wed Feb 14 17:01:12 2024 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Hans de Goede X-Patchwork-Id: 19491 Return-Path: X-Original-To: parsemail@patchwork.libcamera.org Delivered-To: parsemail@patchwork.libcamera.org Received: from lancelot.ideasonboard.com (lancelot.ideasonboard.com [92.243.16.209]) by patchwork.libcamera.org (Postfix) with ESMTPS id C10FDC3259 for ; Wed, 14 Feb 2024 17:01:50 +0000 (UTC) Received: from lancelot.ideasonboard.com (localhost [IPv6:::1]) by lancelot.ideasonboard.com (Postfix) with ESMTP id 74AF16281F; Wed, 14 Feb 2024 18:01:50 +0100 (CET) Authentication-Results: lancelot.ideasonboard.com; dkim=fail reason="signature verification failed" (1024-bit key; unprotected) header.d=redhat.com header.i=@redhat.com header.b="b7YfpJY8"; dkim-atps=neutral Received: from us-smtp-delivery-124.mimecast.com (us-smtp-delivery-124.mimecast.com [170.10.133.124]) by lancelot.ideasonboard.com (Postfix) with ESMTPS id 3363762811 for ; Wed, 14 Feb 2024 18:01:47 +0100 (CET) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=mimecast20190719; t=1707930106; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references; bh=cKd27moigvkhk+ZECPl58Mt5IX2BOhq3gHck4LotUF4=; b=b7YfpJY8yFk3aGIgk6Q1lkoonnVSX1Lp9R6tezD2l/4W6zNWs366WOywRZk9T9FjIn6pCv TICiWGu/SDjN5SKXXVvfc/ZxMGzLZ0VzYLFD3SFkYLovSI7m9zMSwJ58SqAcFY+KSqL9cX 4Hnpn6/IuDlDGvXnJAFpcdMS4uxkGQs= Received: from mimecast-mx02.redhat.com (mimecast-mx02.redhat.com [66.187.233.88]) by relay.mimecast.com with ESMTP with STARTTLS (version=TLSv1.3, cipher=TLS_AES_256_GCM_SHA384) id us-mta-154-2YXov3EIM0GC3OIa9yWvmw-1; Wed, 14 Feb 2024 12:01:43 -0500 X-MC-Unique: 2YXov3EIM0GC3OIa9yWvmw-1 Received: from smtp.corp.redhat.com (int-mx06.intmail.prod.int.rdu2.redhat.com [10.11.54.6]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange X25519 server-signature RSA-PSS (2048 bits) server-digest SHA256) (No client certificate requested) by mimecast-mx02.redhat.com (Postfix) with ESMTPS id 33F108493EB; Wed, 14 Feb 2024 17:01:43 +0000 (UTC) Received: from localhost.localdomain (unknown [10.39.192.30]) by smtp.corp.redhat.com (Postfix) with ESMTP id E64E02166B4F; Wed, 14 Feb 2024 17:01:41 +0000 (UTC) From: Hans de Goede To: libcamera-devel@lists.libcamera.org Subject: [PATCH v3 08/16] libcamera: software_isp: Add DebayerCpu class Date: Wed, 14 Feb 2024 18:01:12 +0100 Message-ID: <20240214170122.60754-9-hdegoede@redhat.com> In-Reply-To: <20240214170122.60754-1-hdegoede@redhat.com> References: <20240214170122.60754-1-hdegoede@redhat.com> MIME-Version: 1.0 X-Scanned-By: MIMEDefang 3.4.1 on 10.11.54.6 X-Mimecast-Spam-Score: 0 X-Mimecast-Originator: redhat.com X-BeenThere: libcamera-devel@lists.libcamera.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: Maxime Ripard , Pavel Machek , Bryan O'Donoghue , Dennis Bonke Errors-To: libcamera-devel-bounces@lists.libcamera.org Sender: "libcamera-devel" Add CPU based debayering implementation. This initial implementation only supports debayering packed 10 bits per pixel bayer data in the 4 standard bayer orders. Doxygen documentation by Dennis Bonke. Tested-by: Bryan O'Donoghue # sc8280xp Lenovo x13s Tested-by: Pavel Machek Reviewed-by: Pavel Machek Co-developed-by: Dennis Bonke Signed-off-by: Dennis Bonke Co-developed-by: Andrey Konovalov Signed-off-by: Andrey Konovalov Co-developed-by: Pavel Machek Signed-off-by: Pavel Machek Signed-off-by: Hans de Goede --- Changes in v3: - Move debayer_cpu.h to src/libcamera/software_isp/ - Move documentation to .cpp file - Document how/why an array of src pointers is passed to the debayer functions --- src/libcamera/software_isp/debayer_cpu.cpp | 619 +++++++++++++++++++++ src/libcamera/software_isp/debayer_cpu.h | 143 +++++ src/libcamera/software_isp/meson.build | 1 + 3 files changed, 763 insertions(+) create mode 100644 src/libcamera/software_isp/debayer_cpu.cpp create mode 100644 src/libcamera/software_isp/debayer_cpu.h diff --git a/src/libcamera/software_isp/debayer_cpu.cpp b/src/libcamera/software_isp/debayer_cpu.cpp new file mode 100644 index 00000000..53e90776 --- /dev/null +++ b/src/libcamera/software_isp/debayer_cpu.cpp @@ -0,0 +1,619 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ +/* + * Copyright (C) 2023, Linaro Ltd + * Copyright (C) 2023, Red Hat Inc. + * + * Authors: + * Hans de Goede + * + * debayer_cpu.cpp - CPU based debayering class + */ + +#include "debayer_cpu.h" + +#include +#include +#include + +#include + +#include "libcamera/internal/bayer_format.h" +#include "libcamera/internal/framebuffer.h" +#include "libcamera/internal/mapped_framebuffer.h" + +namespace libcamera { + +/** + * \class DebayerCpu + * \brief Class for debayering on the CPU + * + * Implementation for CPU based debayering + */ + +/** + * \brief Constructs a DebayerCpu object. + * \param[in] stats Pointer to the stats object to use. + */ +DebayerCpu::DebayerCpu(std::unique_ptr stats) + : stats_(std::move(stats)), gamma_correction_(1.0) +{ +#ifdef __x86_64__ + enableInputMemcpy_ = false; +#else + enableInputMemcpy_ = true; +#endif + /* Initialize gamma to 1.0 curve */ + for (unsigned int i = 0; i < kGammaLookupSize; i++) + gamma_[i] = i / 4; + + for (unsigned int i = 0; i < kMaxLineBuffers; i++) + lineBuffers_[i] = nullptr; +} + +DebayerCpu::~DebayerCpu() +{ + for (unsigned int i = 0; i < kMaxLineBuffers; i++) + free(lineBuffers_[i]); +} + +// RGR +// GBG +// RGR +#define BGGR_BGR888(p, n, div) \ + *dst++ = blue_[curr[x] / (div)]; \ + *dst++ = green_[(prev[x] + curr[x - p] + curr[x + n] + next[x]) / (4 * (div))]; \ + *dst++ = red_[(prev[x - p] + prev[x + n] + next[x - p] + next[x + n]) / (4 * (div))]; \ + x++; + +// GBG +// RGR +// GBG +#define GRBG_BGR888(p, n, div) \ + *dst++ = blue_[(prev[x] + next[x]) / (2 * (div))]; \ + *dst++ = green_[curr[x] / (div)]; \ + *dst++ = red_[(curr[x - p] + curr[x + n]) / (2 * (div))]; \ + x++; + +// GRG +// BGB +// GRG +#define GBRG_BGR888(p, n, div) \ + *dst++ = blue_[(curr[x - p] + curr[x + n]) / (2 * (div))]; \ + *dst++ = green_[curr[x] / (div)]; \ + *dst++ = red_[(prev[x] + next[x]) / (2 * (div))]; \ + x++; + +// BGB +// GRG +// BGB +#define RGGB_BGR888(p, n, div) \ + *dst++ = blue_[(prev[x - p] + prev[x + n] + next[x - p] + next[x + n]) / (4 * (div))]; \ + *dst++ = green_[(prev[x] + curr[x - p] + curr[x + n] + next[x]) / (4 * (div))]; \ + *dst++ = red_[curr[x] / (div)]; \ + x++; + +void DebayerCpu::debayer10P_BGBG_BGR888(uint8_t *dst, const uint8_t *src[]) +{ + const int width_in_bytes = window_.width * 5 / 4; + const uint8_t *prev = (const uint8_t *)src[0]; + const uint8_t *curr = (const uint8_t *)src[1]; + const uint8_t *next = (const uint8_t *)src[2]; + + /* + * For the first pixel getting a pixel from the previous column uses + * x - 2 to skip the 5th byte with least-significant bits for 4 pixels. + * Same for last pixel (uses x + 2) and looking at the next column. + * x++ in the for-loop skips the 5th byte with 4 x 2 lsb-s for 10bit packed. + */ + for (int x = 0; x < width_in_bytes; x++) { + /* First pixel */ + BGGR_BGR888(2, 1, 1) + /* Second pixel BGGR -> GBRG */ + GBRG_BGR888(1, 1, 1) + /* Same thing for third and fourth pixels */ + BGGR_BGR888(1, 1, 1) + GBRG_BGR888(1, 2, 1) + } +} + +void DebayerCpu::debayer10P_GRGR_BGR888(uint8_t *dst, const uint8_t *src[]) +{ + const int width_in_bytes = window_.width * 5 / 4; + const uint8_t *prev = (const uint8_t *)src[0]; + const uint8_t *curr = (const uint8_t *)src[1]; + const uint8_t *next = (const uint8_t *)src[2]; + + for (int x = 0; x < width_in_bytes; x++) { + /* First pixel */ + GRBG_BGR888(2, 1, 1) + /* Second pixel GRBG -> RGGB */ + RGGB_BGR888(1, 1, 1) + /* Same thing for third and fourth pixels */ + GRBG_BGR888(1, 1, 1) + RGGB_BGR888(1, 2, 1) + } +} + +void DebayerCpu::debayer10P_GBGB_BGR888(uint8_t *dst, const uint8_t *src[]) +{ + const int width_in_bytes = window_.width * 5 / 4; + const uint8_t *prev = (const uint8_t *)src[0]; + const uint8_t *curr = (const uint8_t *)src[1]; + const uint8_t *next = (const uint8_t *)src[2]; + + for (int x = 0; x < width_in_bytes; x++) { + /* Even pixel */ + GBRG_BGR888(2, 1, 1) + /* Odd pixel GBGR -> BGGR */ + BGGR_BGR888(1, 1, 1) + /* Same thing for next 2 pixels */ + GBRG_BGR888(1, 1, 1) + BGGR_BGR888(1, 2, 1) + } +} + +void DebayerCpu::debayer10P_RGRG_BGR888(uint8_t *dst, const uint8_t *src[]) +{ + const int width_in_bytes = window_.width * 5 / 4; + const uint8_t *prev = (const uint8_t *)src[0]; + const uint8_t *curr = (const uint8_t *)src[1]; + const uint8_t *next = (const uint8_t *)src[2]; + + for (int x = 0; x < width_in_bytes; x++) { + /* Even pixel */ + RGGB_BGR888(2, 1, 1) + /* Odd pixel RGGB -> GRBG*/ + GRBG_BGR888(1, 1, 1) + /* Same thing for next 2 pixels */ + RGGB_BGR888(1, 1, 1) + GRBG_BGR888(1, 2, 1) + } +} + +static bool isStandardBayerOrder(BayerFormat::Order order) +{ + return order == BayerFormat::BGGR || order == BayerFormat::GBRG || + order == BayerFormat::GRBG || order == BayerFormat::RGGB; +} + +/* + * Setup the Debayer object according to the passed in parameters. + * Return 0 on success, a negative errno value on failure + * (unsupported parameters). + */ +int DebayerCpu::getInputConfig(PixelFormat inputFormat, DebayerInputConfig &config) +{ + BayerFormat bayerFormat = + BayerFormat::fromPixelFormat(inputFormat); + + if (bayerFormat.bitDepth == 10 && + bayerFormat.packing == BayerFormat::Packing::CSI2 && + isStandardBayerOrder(bayerFormat.order)) { + config.bpp = 10; + config.patternSize.width = 4; /* 5 bytes per *4* pixels */ + config.patternSize.height = 2; + config.outputFormats = std::vector({ formats::RGB888 }); + return 0; + } + + LOG(Debayer, Info) + << "Unsupported input format " << inputFormat.toString(); + return -EINVAL; +} + +int DebayerCpu::getOutputConfig(PixelFormat outputFormat, DebayerOutputConfig &config) +{ + if (outputFormat == formats::RGB888) { + config.bpp = 24; + return 0; + } + + LOG(Debayer, Info) + << "Unsupported output format " << outputFormat.toString(); + return -EINVAL; +} + +/* TODO: this ignores outputFormat since there is only 1 supported outputFormat for now */ +int DebayerCpu::setDebayerFunctions(PixelFormat inputFormat, [[maybe_unused]] PixelFormat outputFormat) +{ + BayerFormat bayerFormat = + BayerFormat::fromPixelFormat(inputFormat); + + if (bayerFormat.bitDepth == 10 && + bayerFormat.packing == BayerFormat::Packing::CSI2) { + switch (bayerFormat.order) { + case BayerFormat::BGGR: + debayer0_ = &DebayerCpu::debayer10P_BGBG_BGR888; + debayer1_ = &DebayerCpu::debayer10P_GRGR_BGR888; + return 0; + case BayerFormat::GBRG: + debayer0_ = &DebayerCpu::debayer10P_GBGB_BGR888; + debayer1_ = &DebayerCpu::debayer10P_RGRG_BGR888; + return 0; + case BayerFormat::GRBG: + debayer0_ = &DebayerCpu::debayer10P_GRGR_BGR888; + debayer1_ = &DebayerCpu::debayer10P_BGBG_BGR888; + return 0; + case BayerFormat::RGGB: + debayer0_ = &DebayerCpu::debayer10P_RGRG_BGR888; + debayer1_ = &DebayerCpu::debayer10P_GBGB_BGR888; + return 0; + default: + break; + } + } + + LOG(Debayer, Error) << "Unsupported input output format combination"; + return -EINVAL; +} + +int DebayerCpu::configure(const StreamConfiguration &inputCfg, + const std::vector> &outputCfgs) +{ + if (getInputConfig(inputCfg.pixelFormat, inputConfig_) != 0) + return -EINVAL; + + if (stats_->configure(inputCfg) != 0) + return -EINVAL; + + const Size &stats_pattern_size = stats_->patternSize(); + if (inputConfig_.patternSize.width != stats_pattern_size.width || + inputConfig_.patternSize.height != stats_pattern_size.height) { + LOG(Debayer, Error) + << "mismatching stats and debayer pattern sizes for " + << inputCfg.pixelFormat.toString(); + return -EINVAL; + } + + inputConfig_.stride = inputCfg.stride; + + if (outputCfgs.size() != 1) { + LOG(Debayer, Error) + << "Unsupported number of output streams: " + << outputCfgs.size(); + return -EINVAL; + } + + const StreamConfiguration &outputCfg = outputCfgs[0]; + SizeRange outSizeRange = sizes(inputCfg.pixelFormat, inputCfg.size); + std::tie(outputConfig_.stride, outputConfig_.frameSize) = + strideAndFrameSize(outputCfg.pixelFormat, outputCfg.size); + + if (!outSizeRange.contains(outputCfg.size) || outputConfig_.stride != outputCfg.stride) { + LOG(Debayer, Error) + << "Invalid output size/stride: " + << "\n " << outputCfg.size << " (" << outSizeRange << ")" + << "\n " << outputCfg.stride << " (" << outputConfig_.stride << ")"; + return -EINVAL; + } + + if (setDebayerFunctions(inputCfg.pixelFormat, outputCfg.pixelFormat) != 0) + return -EINVAL; + + window_.x = ((inputCfg.size.width - outputCfg.size.width) / 2) & + ~(inputConfig_.patternSize.width - 1); + window_.y = ((inputCfg.size.height - outputCfg.size.height) / 2) & + ~(inputConfig_.patternSize.height - 1); + window_.width = outputCfg.size.width; + window_.height = outputCfg.size.height; + + /* Don't pass x,y since process() already adjusts src before passing it */ + stats_->setWindow(Rectangle(window_.size())); + + /* pad with patternSize.Width on both left and right side */ + lineBufferPadding_ = inputConfig_.patternSize.width * inputConfig_.bpp / 8; + lineBufferLength_ = window_.width * inputConfig_.bpp / 8 + + 2 * lineBufferPadding_; + for (unsigned int i = 0; + i < (inputConfig_.patternSize.height + 1) && enableInputMemcpy_; + i++) { + free(lineBuffers_[i]); + lineBuffers_[i] = (uint8_t *)malloc(lineBufferLength_); + if (!lineBuffers_[i]) + return -ENOMEM; + } + + measuredFrames_ = 0; + frameProcessTime_ = 0; + + return 0; +} + +/* + * Get width and height at which the bayer-pattern repeats. + * Return pattern-size or an empty Size for an unsupported inputFormat. + */ +Size DebayerCpu::patternSize(PixelFormat inputFormat) +{ + DebayerCpu::DebayerInputConfig config; + + if (getInputConfig(inputFormat, config) != 0) + return {}; + + return config.patternSize; +} + +std::vector DebayerCpu::formats(PixelFormat inputFormat) +{ + DebayerCpu::DebayerInputConfig config; + + if (getInputConfig(inputFormat, config) != 0) + return std::vector(); + + return config.outputFormats; +} + +std::tuple +DebayerCpu::strideAndFrameSize(const PixelFormat &outputFormat, const Size &size) +{ + DebayerCpu::DebayerOutputConfig config; + + if (getOutputConfig(outputFormat, config) != 0) + return std::make_tuple(0, 0); + + /* round up to multiple of 8 for 64 bits alignment */ + unsigned int stride = (size.width * config.bpp / 8 + 7) & ~7; + + return std::make_tuple(stride, stride * size.height); +} + +void DebayerCpu::setupInputMemcpy(const uint8_t *linePointers[]) +{ + const unsigned int patternHeight = inputConfig_.patternSize.height; + + if (!enableInputMemcpy_) + return; + + for (unsigned int i = 0; i < patternHeight; i++) { + memcpy(lineBuffers_[i], linePointers[i + 1] - lineBufferPadding_, + lineBufferLength_); + linePointers[i + 1] = lineBuffers_[i] + lineBufferPadding_; + } + + /* Point lineBufferIndex_ to first unused lineBuffer */ + lineBufferIndex_ = patternHeight; +} + +void DebayerCpu::shiftLinePointers(const uint8_t *linePointers[], const uint8_t *src) +{ + const unsigned int patternHeight = inputConfig_.patternSize.height; + + for (unsigned int i = 0; i < patternHeight; i++) + linePointers[i] = linePointers[i + 1]; + + linePointers[patternHeight] = src + + (patternHeight / 2) * (int)inputConfig_.stride; +} + +void DebayerCpu::memcpyNextLine(const uint8_t *linePointers[]) +{ + const unsigned int patternHeight = inputConfig_.patternSize.height; + + if (!enableInputMemcpy_) + return; + + memcpy(lineBuffers_[lineBufferIndex_], linePointers[patternHeight] - lineBufferPadding_, + lineBufferLength_); + linePointers[patternHeight] = lineBuffers_[lineBufferIndex_] + lineBufferPadding_; + + lineBufferIndex_ = (lineBufferIndex_ + 1) % (patternHeight + 1); +} + +void DebayerCpu::process2(const uint8_t *src, uint8_t *dst) +{ + unsigned int y_end = window_.y + window_.height; + /* Holds [0] previous- [1] current- [2] next-line */ + const uint8_t *linePointers[3]; + + /* Adjust src to top left corner of the window */ + src += window_.y * inputConfig_.stride + window_.x * inputConfig_.bpp / 8; + + /* [x] becomes [x - 1] after initial shiftLinePointers() call */ + if (window_.y) { + linePointers[1] = src - inputConfig_.stride; /* previous-line */ + linePointers[2] = src; + } else { + /* window_.y == 0, use the next line as prev line */ + linePointers[1] = src + inputConfig_.stride; + linePointers[2] = src; + /* Last 2 lines also need special handling */ + y_end -= 2; + } + + setupInputMemcpy(linePointers); + + for (unsigned int y = window_.y; y < y_end; y += 2) { + shiftLinePointers(linePointers, src); + memcpyNextLine(linePointers); + stats_->processLine0(y, linePointers); + (this->*debayer0_)(dst, linePointers); + src += inputConfig_.stride; + dst += outputConfig_.stride; + + shiftLinePointers(linePointers, src); + memcpyNextLine(linePointers); + (this->*debayer1_)(dst, linePointers); + src += inputConfig_.stride; + dst += outputConfig_.stride; + } + + if (window_.y == 0) { + shiftLinePointers(linePointers, src); + memcpyNextLine(linePointers); + stats_->processLine0(y_end, linePointers); + (this->*debayer0_)(dst, linePointers); + src += inputConfig_.stride; + dst += outputConfig_.stride; + + shiftLinePointers(linePointers, src); + /* next line may point outside of src, use prev. */ + linePointers[2] = linePointers[0]; + (this->*debayer1_)(dst, linePointers); + src += inputConfig_.stride; + dst += outputConfig_.stride; + } +} + +void DebayerCpu::process4(const uint8_t *src, uint8_t *dst) +{ + const unsigned int y_end = window_.y + window_.height; + /* + * This holds pointers to [0] 2-lines-up [1] 1-line-up [2] current-line + * [3] 1-line-down [4] 2-lines-down. + */ + const uint8_t *linePointers[5]; + + /* Adjust src to top left corner of the window */ + src += window_.y * inputConfig_.stride + window_.x * inputConfig_.bpp / 8; + + /* [x] becomes [x - 1] after initial shiftLinePointers() call */ + linePointers[1] = src - 2 * inputConfig_.stride; + linePointers[2] = src - inputConfig_.stride; + linePointers[3] = src; + linePointers[4] = src + inputConfig_.stride; + + setupInputMemcpy(linePointers); + + for (unsigned int y = window_.y; y < y_end; y += 4) { + shiftLinePointers(linePointers, src); + memcpyNextLine(linePointers); + stats_->processLine0(y, linePointers); + (this->*debayer0_)(dst, linePointers); + src += inputConfig_.stride; + dst += outputConfig_.stride; + + shiftLinePointers(linePointers, src); + memcpyNextLine(linePointers); + (this->*debayer1_)(dst, linePointers); + src += inputConfig_.stride; + dst += outputConfig_.stride; + + shiftLinePointers(linePointers, src); + memcpyNextLine(linePointers); + stats_->processLine2(y, linePointers); + (this->*debayer2_)(dst, linePointers); + src += inputConfig_.stride; + dst += outputConfig_.stride; + + shiftLinePointers(linePointers, src); + memcpyNextLine(linePointers); + (this->*debayer3_)(dst, linePointers); + src += inputConfig_.stride; + dst += outputConfig_.stride; + } +} + +static inline int64_t timeDiff(timespec &after, timespec &before) +{ + return (after.tv_sec - before.tv_sec) * 1000000000LL + + (int64_t)after.tv_nsec - (int64_t)before.tv_nsec; +} + +void DebayerCpu::process(FrameBuffer *input, FrameBuffer *output, DebayerParams params) +{ + timespec frameStartTime; + + if (measuredFrames_ < DebayerCpu::kLastFrameToMeasure) { + frameStartTime = {}; + clock_gettime(CLOCK_MONOTONIC_RAW, &frameStartTime); + } + + /* Apply DebayerParams */ + if (params.gamma != gamma_correction_) { + for (unsigned int i = 0; i < kGammaLookupSize; i++) + gamma_[i] = UINT8_MAX * powf(i / (kGammaLookupSize - 1.0), params.gamma); + + gamma_correction_ = params.gamma; + } + + for (unsigned int i = 0; i < kRGBLookupSize; i++) { + constexpr unsigned int div = + kRGBLookupSize * DebayerParams::kGain10 / kGammaLookupSize; + unsigned int idx; + + /* Apply gamma after gain! */ + idx = std::min({ i * params.gainR / div, (kGammaLookupSize - 1) }); + red_[i] = gamma_[idx]; + + idx = std::min({ i * params.gainG / div, (kGammaLookupSize - 1) }); + green_[i] = gamma_[idx]; + + idx = std::min({ i * params.gainB / div, (kGammaLookupSize - 1) }); + blue_[i] = gamma_[idx]; + } + + /* Copy metadata from the input buffer */ + FrameMetadata &metadata = output->_d()->metadata(); + metadata.status = input->metadata().status; + metadata.sequence = input->metadata().sequence; + metadata.timestamp = input->metadata().timestamp; + + MappedFrameBuffer in(input, MappedFrameBuffer::MapFlag::Read); + MappedFrameBuffer out(output, MappedFrameBuffer::MapFlag::Write); + if (!in.isValid() || !out.isValid()) { + LOG(Debayer, Error) << "mmap-ing buffer(s) failed"; + metadata.status = FrameMetadata::FrameError; + return; + } + + stats_->startFrame(); + + if (inputConfig_.patternSize.height == 2) + process2(in.planes()[0].data(), out.planes()[0].data()); + else + process4(in.planes()[0].data(), out.planes()[0].data()); + + metadata.planes()[0].bytesused = out.planes()[0].size(); + + /* Measure before emitting signals */ + if (measuredFrames_ < DebayerCpu::kLastFrameToMeasure && + ++measuredFrames_ > DebayerCpu::kFramesToSkip) { + timespec frameEndTime = {}; + clock_gettime(CLOCK_MONOTONIC_RAW, &frameEndTime); + frameProcessTime_ += timeDiff(frameEndTime, frameStartTime); + if (measuredFrames_ == DebayerCpu::kLastFrameToMeasure) { + const unsigned int measuredFrames = DebayerCpu::kLastFrameToMeasure - + DebayerCpu::kFramesToSkip; + LOG(Debayer, Info) + << "Processed " << measuredFrames + << " frames in " << frameProcessTime_ / 1000 << "us, " + << frameProcessTime_ / (1000 * measuredFrames) + << " us/frame"; + } + } + + stats_->finishFrame(); + outputBufferReady.emit(output); + inputBufferReady.emit(input); +} + +SizeRange DebayerCpu::sizes(PixelFormat inputFormat, const Size &inputSize) +{ + Size pattern_size = patternSize(inputFormat); + unsigned int border_height = pattern_size.height; + + if (pattern_size.isNull()) + return {}; + + /* No need for top/bottom border with a pattern height of 2 */ + if (pattern_size.height == 2) + border_height = 0; + + /* + * For debayer interpolation a border is kept around the entire image + * and the minimum output size is pattern-height x pattern-width. + */ + if (inputSize.width < (3 * pattern_size.width) || + inputSize.height < (2 * border_height + pattern_size.height)) { + LOG(Debayer, Warning) + << "Input format size too small: " << inputSize.toString(); + return {}; + } + + return SizeRange(Size(pattern_size.width, pattern_size.height), + Size((inputSize.width - 2 * pattern_size.width) & ~(pattern_size.width - 1), + (inputSize.height - 2 * border_height) & ~(pattern_size.height - 1)), + pattern_size.width, pattern_size.height); +} + +} /* namespace libcamera */ diff --git a/src/libcamera/software_isp/debayer_cpu.h b/src/libcamera/software_isp/debayer_cpu.h new file mode 100644 index 00000000..e0c4b9a8 --- /dev/null +++ b/src/libcamera/software_isp/debayer_cpu.h @@ -0,0 +1,143 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ +/* + * Copyright (C) 2023, Linaro Ltd + * Copyright (C) 2023, Red Hat Inc. + * + * Authors: + * Hans de Goede + * + * debayer_cpu.h - CPU based debayering header + */ + +#pragma once + +#include +#include +#include + +#include + +#include "debayer.h" +#include "swstats_cpu.h" + +namespace libcamera { + +class DebayerCpu : public Debayer, public Object +{ +public: + DebayerCpu(std::unique_ptr stats); + ~DebayerCpu(); + + int configure(const StreamConfiguration &inputCfg, + const std::vector> &outputCfgs); + Size patternSize(PixelFormat inputFormat); + std::vector formats(PixelFormat input); + std::tuple + strideAndFrameSize(const PixelFormat &outputFormat, const Size &size); + void process(FrameBuffer *input, FrameBuffer *output, DebayerParams params); + SizeRange sizes(PixelFormat inputFormat, const Size &inputSize); + + /** + * \brief Get the file descriptor for the statistics. + * + * \return the file descriptor pointing to the statistics. + */ + const SharedFD &getStatsFD() { return stats_->getStatsFD(); } + + /** + * \brief Get the output frame size. + * + * \return The output frame size. + */ + unsigned int frameSize() { return outputConfig_.frameSize; } + +private: + /** + * \brief Called to debayer 1 line of Bayer input data to output format + * \param[out] dst Pointer to the start of the output line to write + * \param[in] src The input data + * + * Input data is an array of (patternSize_.height + 1) src + * pointers each pointing to a line in the Bayer source. The middle + * element of the array will point to the actual line being processed. + * Earlier element(s) will point to the previous line(s) and later + * element(s) to the next line(s). + * + * These functions take an array of src pointers, rather then + * a single src pointer + a stride for the source, so that when the src + * is slow uncached memory it can be copied to faster memory before + * debayering. Debayering a standard 2x2 Bayer pattern requires access + * to the previous and next src lines for interpolating the missing + * colors. To allow copying the src lines only once 3 buffers each + * holding a single line are used, re-using the oldest buffer for + * the next line and the pointers are swizzled so that: + * src[0] = previous-line, src[1] = currrent-line, src[2] = next-line. + * This way the 3 pointers passed to the debayer functions form + * a sliding window over the src avoiding the need to copy each + * line more then once. + * + * Similarly for bayer patterns which repeat every 4 lines, 5 src + * pointers are passed holding: src[0] = 2-lines-up, src[1] = 1-line-up + * src[2] = current-line, src[3] = 1-line-down, src[4] = 2-lines-down. + */ + typedef void (DebayerCpu::*debayerFn)(uint8_t *dst, const uint8_t *src[]); + + /* CSI-2 packed 10-bit raw bayer format (all the 4 orders) */ + void debayer10P_BGBG_BGR888(uint8_t *dst, const uint8_t *src[]); + void debayer10P_GRGR_BGR888(uint8_t *dst, const uint8_t *src[]); + void debayer10P_GBGB_BGR888(uint8_t *dst, const uint8_t *src[]); + void debayer10P_RGRG_BGR888(uint8_t *dst, const uint8_t *src[]); + + struct DebayerInputConfig { + Size patternSize; + unsigned int bpp; /* Memory used per pixel, not precision */ + unsigned int stride; + std::vector outputFormats; + }; + + struct DebayerOutputConfig { + unsigned int bpp; /* Memory used per pixel, not precision */ + unsigned int stride; + unsigned int frameSize; + }; + + int getInputConfig(PixelFormat inputFormat, DebayerInputConfig &config); + int getOutputConfig(PixelFormat outputFormat, DebayerOutputConfig &config); + int setDebayerFunctions(PixelFormat inputFormat, PixelFormat outputFormat); + void setupInputMemcpy(const uint8_t *linePointers[]); + void shiftLinePointers(const uint8_t *linePointers[], const uint8_t *src); + void memcpyNextLine(const uint8_t *linePointers[]); + void process2(const uint8_t *src, uint8_t *dst); + void process4(const uint8_t *src, uint8_t *dst); + + static constexpr unsigned int kGammaLookupSize = 1024; + static constexpr unsigned int kRGBLookupSize = 256; + /* Max. supported Bayer pattern height is 4, debayering this requires 5 lines */ + static constexpr unsigned int kMaxLineBuffers = 5; + + std::array gamma_; + std::array red_; + std::array green_; + std::array blue_; + debayerFn debayer0_; + debayerFn debayer1_; + debayerFn debayer2_; + debayerFn debayer3_; + Rectangle window_; + DebayerInputConfig inputConfig_; + DebayerOutputConfig outputConfig_; + std::unique_ptr stats_; + uint8_t *lineBuffers_[kMaxLineBuffers]; + unsigned int lineBufferLength_; + unsigned int lineBufferPadding_; + unsigned int lineBufferIndex_; + bool enableInputMemcpy_; + float gamma_correction_; + unsigned int measuredFrames_; + int64_t frameProcessTime_; + /* Skip 30 frames for things to stabilize then measure 30 frames */ + static constexpr unsigned int kFramesToSkip = 30; + static constexpr unsigned int kLastFrameToMeasure = 60; +}; + +} /* namespace libcamera */ diff --git a/src/libcamera/software_isp/meson.build b/src/libcamera/software_isp/meson.build index 92fc90f3..dd5e70c1 100644 --- a/src/libcamera/software_isp/meson.build +++ b/src/libcamera/software_isp/meson.build @@ -2,5 +2,6 @@ libcamera_sources += files([ 'debayer.cpp', + 'debayer_cpu.cpp', 'swstats_cpu.cpp', ])