[PATCH] libcamera: debayer_cpu: Add 32bits/aligned output formats
Milan Zamazal
mzamazal at redhat.com
Tue Jun 11 16:29:10 CEST 2024
Hi Robert,
thank you for the patch.
Robert Mader <robert.mader at collabora.com> writes:
> In order to be more compatible with modern hardware and APIs. This
> notably allows GL implementations to directly import the buffers more
> often and seems to be required for Wayland.
>
> Further more, as we already enforce a 8 byte stride, these formats work
> better for clients that don't support padding - such as libwebrtc at the
> time of writing.
>
> Tested on the Librem5 and PinePhone.
>
> Signed-off-by: Robert Mader <robert.mader at collabora.com>
> ---
> src/libcamera/software_isp/debayer_cpu.cpp | 244 +++++++++++++++++++--
> src/libcamera/software_isp/debayer_cpu.h | 10 +
> 2 files changed, 238 insertions(+), 16 deletions(-)
>
> diff --git a/src/libcamera/software_isp/debayer_cpu.cpp b/src/libcamera/software_isp/debayer_cpu.cpp
> index c038eed4..73c66a88 100644
> --- a/src/libcamera/software_isp/debayer_cpu.cpp
> +++ b/src/libcamera/software_isp/debayer_cpu.cpp
> @@ -76,6 +76,13 @@ DebayerCpu::~DebayerCpu()
> *dst++ = red_[(prev[x - p] + prev[x + n] + next[x - p] + next[x + n]) / (4 * (div))]; \
> x++;
>
> +#define BGGR_XBGR8888(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))]; \
> + *dst++ = 255; \
> + x++;
> +
The level of code duplication here starts to exceed reasonable limits.
Maybe adding an argument to the macro deciding whether to append the
last `dst' assignment or not would have a significant performance
impact. But even then I guess there must be a way to let the compiler
generate the duplicate code (an inline function template?) rather than
doing it manually. What do the C++ experts around think?
If nothing better is possible then `*dst++ = 255' can be applied in the
callers rather than defining the alternative macro versions.
> /*
> * GBG
> * RGR
> @@ -87,6 +94,13 @@ DebayerCpu::~DebayerCpu()
> *dst++ = red_[(curr[x - p] + curr[x + n]) / (2 * (div))]; \
> x++;
>
> +#define GRBG_XBGR8888(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))]; \
> + *dst++ = 255; \
> + x++;
> +
> /*
> * GRG
> * BGB
> @@ -98,6 +112,13 @@ DebayerCpu::~DebayerCpu()
> *dst++ = red_[(prev[x] + next[x]) / (2 * (div))]; \
> x++;
>
> +#define GBRG_XBGR8888(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))]; \
> + *dst++ = 255; \
> + x++;
> +
> /*
> * BGB
> * GRG
> @@ -109,6 +130,13 @@ DebayerCpu::~DebayerCpu()
> *dst++ = red_[curr[x] / (div)]; \
> x++;
>
> +#define RGGB_XBGR8888(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)]; \
> + *dst++ = 255; \
> + x++;
> +
> void DebayerCpu::debayer8_BGBG_BGR888(uint8_t *dst, const uint8_t *src[])
> {
> DECLARE_SRC_POINTERS(uint8_t)
> @@ -119,6 +147,16 @@ void DebayerCpu::debayer8_BGBG_BGR888(uint8_t *dst, const uint8_t *src[])
> }
> }
>
> +void DebayerCpu::debayer8_BGBG_XBGR8888(uint8_t *dst, const uint8_t *src[])
> +{
> + DECLARE_SRC_POINTERS(uint8_t)
> +
> + for (int x = 0; x < (int)window_.width;) {
> + BGGR_XBGR8888(1, 1, 1)
> + GBRG_XBGR8888(1, 1, 1)
> + }
> +}
> +
... and then X and non-X versions of these methods could be merged too.
Using a template for the purpose should have no performance impact, I
suppose.
> void DebayerCpu::debayer8_GRGR_BGR888(uint8_t *dst, const uint8_t *src[])
> {
> DECLARE_SRC_POINTERS(uint8_t)
> @@ -129,6 +167,16 @@ void DebayerCpu::debayer8_GRGR_BGR888(uint8_t *dst, const uint8_t *src[])
> }
> }
>
> +void DebayerCpu::debayer8_GRGR_XBGR8888(uint8_t *dst, const uint8_t *src[])
> +{
> + DECLARE_SRC_POINTERS(uint8_t)
> +
> + for (int x = 0; x < (int)window_.width;) {
> + GRBG_XBGR8888(1, 1, 1)
> + RGGB_XBGR8888(1, 1, 1)
> + }
> +}
> +
> void DebayerCpu::debayer10_BGBG_BGR888(uint8_t *dst, const uint8_t *src[])
> {
> DECLARE_SRC_POINTERS(uint16_t)
> @@ -140,6 +188,17 @@ void DebayerCpu::debayer10_BGBG_BGR888(uint8_t *dst, const uint8_t *src[])
> }
> }
>
> +void DebayerCpu::debayer10_BGBG_XBGR8888(uint8_t *dst, const uint8_t *src[])
> +{
> + DECLARE_SRC_POINTERS(uint16_t)
> +
> + for (int x = 0; x < (int)window_.width;) {
> + /* divide values by 4 for 10 -> 8 bpp value */
> + BGGR_XBGR8888(1, 1, 4)
> + GBRG_XBGR8888(1, 1, 4)
> + }
> +}
> +
> void DebayerCpu::debayer10_GRGR_BGR888(uint8_t *dst, const uint8_t *src[])
> {
> DECLARE_SRC_POINTERS(uint16_t)
> @@ -151,6 +210,17 @@ void DebayerCpu::debayer10_GRGR_BGR888(uint8_t *dst, const uint8_t *src[])
> }
> }
>
> +void DebayerCpu::debayer10_GRGR_XBGR8888(uint8_t *dst, const uint8_t *src[])
> +{
> + DECLARE_SRC_POINTERS(uint16_t)
> +
> + for (int x = 0; x < (int)window_.width;) {
> + /* divide values by 4 for 10 -> 8 bpp value */
> + GRBG_XBGR8888(1, 1, 4)
> + RGGB_XBGR8888(1, 1, 4)
> + }
> +}
> +
> void DebayerCpu::debayer12_BGBG_BGR888(uint8_t *dst, const uint8_t *src[])
> {
> DECLARE_SRC_POINTERS(uint16_t)
> @@ -162,6 +232,17 @@ void DebayerCpu::debayer12_BGBG_BGR888(uint8_t *dst, const uint8_t *src[])
> }
> }
>
> +void DebayerCpu::debayer12_BGBG_XBGR8888(uint8_t *dst, const uint8_t *src[])
> +{
> + DECLARE_SRC_POINTERS(uint16_t)
> +
> + for (int x = 0; x < (int)window_.width;) {
> + /* divide values by 16 for 12 -> 8 bpp value */
> + BGGR_XBGR8888(1, 1, 16)
> + GBRG_XBGR8888(1, 1, 16)
> + }
> +}
> +
> void DebayerCpu::debayer12_GRGR_BGR888(uint8_t *dst, const uint8_t *src[])
> {
> DECLARE_SRC_POINTERS(uint16_t)
> @@ -173,6 +254,17 @@ void DebayerCpu::debayer12_GRGR_BGR888(uint8_t *dst, const uint8_t *src[])
> }
> }
>
> +void DebayerCpu::debayer12_GRGR_XBGR8888(uint8_t *dst, const uint8_t *src[])
> +{
> + DECLARE_SRC_POINTERS(uint16_t)
> +
> + for (int x = 0; x < (int)window_.width;) {
> + /* divide values by 16 for 12 -> 8 bpp value */
> + GRBG_XBGR8888(1, 1, 16)
> + RGGB_XBGR8888(1, 1, 16)
> + }
> +}
> +
> void DebayerCpu::debayer10P_BGBG_BGR888(uint8_t *dst, const uint8_t *src[])
> {
> const int widthInBytes = window_.width * 5 / 4;
> @@ -198,6 +290,31 @@ void DebayerCpu::debayer10P_BGBG_BGR888(uint8_t *dst, const uint8_t *src[])
> }
> }
>
> +void DebayerCpu::debayer10P_BGBG_XBGR8888(uint8_t *dst, const uint8_t *src[])
> +{
> + const int widthInBytes = window_.width * 5 / 4;
> + const uint8_t *prev = src[0];
> + const uint8_t *curr = src[1];
> + const uint8_t *next = 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.
> + */
> + for (int x = 0; x < widthInBytes;) {
> + /* First pixel */
> + BGGR_XBGR8888(2, 1, 1)
> + /* Second pixel BGGR -> GBRG */
> + GBRG_XBGR8888(1, 1, 1)
> + /* Same thing for third and fourth pixels */
> + BGGR_XBGR8888(1, 1, 1)
> + GBRG_XBGR8888(1, 2, 1)
> + /* Skip 5th src byte with 4 x 2 least-significant-bits */
> + x++;
> + }
> +}
> +
> void DebayerCpu::debayer10P_GRGR_BGR888(uint8_t *dst, const uint8_t *src[])
> {
> const int widthInBytes = window_.width * 5 / 4;
> @@ -218,6 +335,26 @@ void DebayerCpu::debayer10P_GRGR_BGR888(uint8_t *dst, const uint8_t *src[])
> }
> }
>
> +void DebayerCpu::debayer10P_GRGR_XBGR8888(uint8_t *dst, const uint8_t *src[])
> +{
> + const int widthInBytes = window_.width * 5 / 4;
> + const uint8_t *prev = src[0];
> + const uint8_t *curr = src[1];
> + const uint8_t *next = src[2];
> +
> + for (int x = 0; x < widthInBytes;) {
> + /* First pixel */
> + GRBG_XBGR8888(2, 1, 1)
> + /* Second pixel GRBG -> RGGB */
> + RGGB_XBGR8888(1, 1, 1)
> + /* Same thing for third and fourth pixels */
> + GRBG_XBGR8888(1, 1, 1)
> + RGGB_XBGR8888(1, 2, 1)
> + /* Skip 5th src byte with 4 x 2 least-significant-bits */
> + x++;
> + }
> +}
> +
> void DebayerCpu::debayer10P_GBGB_BGR888(uint8_t *dst, const uint8_t *src[])
> {
> const int widthInBytes = window_.width * 5 / 4;
> @@ -238,6 +375,26 @@ void DebayerCpu::debayer10P_GBGB_BGR888(uint8_t *dst, const uint8_t *src[])
> }
> }
>
> +void DebayerCpu::debayer10P_GBGB_XBGR8888(uint8_t *dst, const uint8_t *src[])
> +{
> + const int widthInBytes = window_.width * 5 / 4;
> + const uint8_t *prev = src[0];
> + const uint8_t *curr = src[1];
> + const uint8_t *next = src[2];
> +
> + for (int x = 0; x < widthInBytes;) {
> + /* Even pixel */
> + GBRG_XBGR8888(2, 1, 1)
> + /* Odd pixel GBGR -> BGGR */
> + BGGR_XBGR8888(1, 1, 1)
> + /* Same thing for next 2 pixels */
> + GBRG_XBGR8888(1, 1, 1)
> + BGGR_XBGR8888(1, 2, 1)
> + /* Skip 5th src byte with 4 x 2 least-significant-bits */
> + x++;
> + }
> +}
> +
> void DebayerCpu::debayer10P_RGRG_BGR888(uint8_t *dst, const uint8_t *src[])
> {
> const int widthInBytes = window_.width * 5 / 4;
> @@ -258,6 +415,26 @@ void DebayerCpu::debayer10P_RGRG_BGR888(uint8_t *dst, const uint8_t *src[])
> }
> }
>
> +void DebayerCpu::debayer10P_RGRG_XBGR8888(uint8_t *dst, const uint8_t *src[])
> +{
> + const int widthInBytes = window_.width * 5 / 4;
> + const uint8_t *prev = src[0];
> + const uint8_t *curr = src[1];
> + const uint8_t *next = src[2];
> +
> + for (int x = 0; x < widthInBytes;) {
> + /* Even pixel */
> + RGGB_XBGR8888(2, 1, 1)
> + /* Odd pixel RGGB -> GRBG */
> + GRBG_XBGR8888(1, 1, 1)
> + /* Same thing for next 2 pixels */
> + RGGB_XBGR8888(1, 1, 1)
> + GRBG_XBGR8888(1, 2, 1)
> + /* Skip 5th src byte with 4 x 2 least-significant-bits */
> + x++;
> + }
> +}
> +
> static bool isStandardBayerOrder(BayerFormat::Order order)
> {
> return order == BayerFormat::BGGR || order == BayerFormat::GBRG ||
> @@ -280,7 +457,14 @@ int DebayerCpu::getInputConfig(PixelFormat inputFormat, DebayerInputConfig &conf
> config.bpp = (bayerFormat.bitDepth + 7) & ~7;
> config.patternSize.width = 2;
> config.patternSize.height = 2;
> - config.outputFormats = std::vector<PixelFormat>({ formats::RGB888, formats::BGR888 });
> + config.outputFormats = std::vector<PixelFormat>({
> + formats::RGB888,
> + formats::XRGB8888,
> + formats::ARGB8888,
> + formats::BGR888,
> + formats::XBGR8888,
> + formats::ABGR8888
> + });
> return 0;
> }
>
> @@ -290,7 +474,14 @@ int DebayerCpu::getInputConfig(PixelFormat inputFormat, DebayerInputConfig &conf
> config.bpp = 10;
> config.patternSize.width = 4; /* 5 bytes per *4* pixels */
> config.patternSize.height = 2;
> - config.outputFormats = std::vector<PixelFormat>({ formats::RGB888, formats::BGR888 });
> + config.outputFormats = std::vector<PixelFormat>({
> + formats::RGB888,
> + formats::XRGB8888,
> + formats::ARGB8888,
> + formats::BGR888,
> + formats::XBGR8888,
> + formats::ABGR8888
> + });
> return 0;
> }
>
> @@ -306,6 +497,12 @@ int DebayerCpu::getOutputConfig(PixelFormat outputFormat, DebayerOutputConfig &c
> return 0;
> }
>
> + if (outputFormat == formats::XRGB8888 || outputFormat == formats::ARGB8888 ||
> + outputFormat == formats::XBGR8888 || outputFormat == formats::ABGR8888) {
> + config.bpp = 32;
> + return 0;
> + }
> +
> LOG(Debayer, Info)
> << "Unsupported output format " << outputFormat.toString();
> return -EINVAL;
> @@ -341,6 +538,7 @@ int DebayerCpu::setDebayerFunctions(PixelFormat inputFormat, PixelFormat outputF
> {
> BayerFormat bayerFormat =
> BayerFormat::fromPixelFormat(inputFormat);
> + bool is_aligned = false;
camelCase should be used for variable names.
> xShift_ = 0;
> swapRedBlueGains_ = false;
> @@ -351,8 +549,16 @@ int DebayerCpu::setDebayerFunctions(PixelFormat inputFormat, PixelFormat outputF
> };
>
> switch (outputFormat) {
> + case formats::XRGB8888:
> + case formats::ARGB8888:
> + is_aligned = true;
> + [[fallthrough]];
> case formats::RGB888:
> break;
> + case formats::XBGR8888:
> + case formats::ABGR8888:
> + is_aligned = true;
> + [[fallthrough]];
> case formats::BGR888:
> /* Swap R and B in bayer order to generate BGR888 instead of RGB888 */
> swapRedBlueGains_ = true;
> @@ -383,16 +589,19 @@ int DebayerCpu::setDebayerFunctions(PixelFormat inputFormat, PixelFormat outputF
> isStandardBayerOrder(bayerFormat.order)) {
> switch (bayerFormat.bitDepth) {
> case 8:
> - debayer0_ = &DebayerCpu::debayer8_BGBG_BGR888;
> - debayer1_ = &DebayerCpu::debayer8_GRGR_BGR888;
> + LOG(Debayer, Warning) << "8bit no packing";
Is this a debugging leftover or do you really mean to log something
here? If the latter, it should probably be Debug rather than Warning
and the message should be improved. The same applies to the other LOGs
below.
> + debayer0_ = is_aligned ? &DebayerCpu::debayer8_BGBG_XBGR8888 : &DebayerCpu::debayer8_BGBG_BGR888;
If the two methods were unified and extended with isAligned argument,
I'm not sure what would be the best way to pass the argument to them.
lambda, a template parameter or anything else?
> + debayer1_ = is_aligned ? &DebayerCpu::debayer8_GRGR_XBGR8888 : &DebayerCpu::debayer8_GRGR_BGR888;
> break;
> case 10:
> - debayer0_ = &DebayerCpu::debayer10_BGBG_BGR888;
> - debayer1_ = &DebayerCpu::debayer10_GRGR_BGR888;
> + LOG(Debayer, Warning) << "10bit no packing";
> + debayer0_ = is_aligned ? &DebayerCpu::debayer10_BGBG_XBGR8888 : &DebayerCpu::debayer10_BGBG_BGR888;
> + debayer1_ = is_aligned ? &DebayerCpu::debayer10_GRGR_XBGR8888 : &DebayerCpu::debayer10_GRGR_BGR888;
> break;
> case 12:
> - debayer0_ = &DebayerCpu::debayer12_BGBG_BGR888;
> - debayer1_ = &DebayerCpu::debayer12_GRGR_BGR888;
> + LOG(Debayer, Warning) << "12bit no packing";
> + debayer0_ = is_aligned ? &DebayerCpu::debayer12_BGBG_XBGR8888 : &DebayerCpu::debayer12_BGBG_BGR888;
> + debayer1_ = is_aligned ? &DebayerCpu::debayer12_GRGR_XBGR8888 : &DebayerCpu::debayer12_GRGR_BGR888;
> break;
> }
> setupStandardBayerOrder(bayerFormat.order);
> @@ -401,22 +610,23 @@ int DebayerCpu::setDebayerFunctions(PixelFormat inputFormat, PixelFormat outputF
>
> if (bayerFormat.bitDepth == 10 &&
> bayerFormat.packing == BayerFormat::Packing::CSI2) {
> + LOG(Debayer, Warning) << "10bit csi2";
> switch (bayerFormat.order) {
> case BayerFormat::BGGR:
> - debayer0_ = &DebayerCpu::debayer10P_BGBG_BGR888;
> - debayer1_ = &DebayerCpu::debayer10P_GRGR_BGR888;
> + debayer0_ = is_aligned ? &DebayerCpu::debayer10P_BGBG_XBGR8888 : &DebayerCpu::debayer10P_BGBG_BGR888;
> + debayer1_ = is_aligned ? &DebayerCpu::debayer10P_GRGR_XBGR8888 : &DebayerCpu::debayer10P_GRGR_BGR888;
> return 0;
> case BayerFormat::GBRG:
> - debayer0_ = &DebayerCpu::debayer10P_GBGB_BGR888;
> - debayer1_ = &DebayerCpu::debayer10P_RGRG_BGR888;
> + debayer0_ = is_aligned ? &DebayerCpu::debayer10P_GBGB_XBGR8888 : &DebayerCpu::debayer10P_GBGB_BGR888;
> + debayer1_ = is_aligned ? &DebayerCpu::debayer10P_RGRG_XBGR8888 : &DebayerCpu::debayer10P_RGRG_BGR888;
> return 0;
> case BayerFormat::GRBG:
> - debayer0_ = &DebayerCpu::debayer10P_GRGR_BGR888;
> - debayer1_ = &DebayerCpu::debayer10P_BGBG_BGR888;
> + debayer0_ = is_aligned ? &DebayerCpu::debayer10P_GRGR_XBGR8888 : &DebayerCpu::debayer10P_GRGR_BGR888;
> + debayer1_ = is_aligned ? &DebayerCpu::debayer10P_BGBG_XBGR8888 : &DebayerCpu::debayer10P_BGBG_BGR888;
> return 0;
> case BayerFormat::RGGB:
> - debayer0_ = &DebayerCpu::debayer10P_RGRG_BGR888;
> - debayer1_ = &DebayerCpu::debayer10P_GBGB_BGR888;
> + debayer0_ = is_aligned ? &DebayerCpu::debayer10P_RGRG_XBGR8888 : &DebayerCpu::debayer10P_RGRG_BGR888;
> + debayer1_ = is_aligned ? &DebayerCpu::debayer10P_GBGB_XBGR8888 : &DebayerCpu::debayer10P_GBGB_BGR888;
> return 0;
> default:
> break;
> @@ -533,6 +743,8 @@ DebayerCpu::strideAndFrameSize(const PixelFormat &outputFormat, const Size &size
> /* round up to multiple of 8 for 64 bits alignment */
> unsigned int stride = (size.width * config.bpp / 8 + 7) & ~7;
>
> + LOG(Debayer, Warning) << outputFormat.toString() << " " << size.width << " " << size.height << " " << config.bpp << " " << stride << " " << stride * size.height;
> +
> return std::make_tuple(stride, stride * size.height);
> }
>
> diff --git a/src/libcamera/software_isp/debayer_cpu.h b/src/libcamera/software_isp/debayer_cpu.h
> index be7dcdca..c30f44aa 100644
> --- a/src/libcamera/software_isp/debayer_cpu.h
> +++ b/src/libcamera/software_isp/debayer_cpu.h
> @@ -86,18 +86,28 @@ private:
>
> /* 8-bit raw bayer format */
> void debayer8_BGBG_BGR888(uint8_t *dst, const uint8_t *src[]);
> + void debayer8_BGBG_XBGR8888(uint8_t *dst, const uint8_t *src[]);
> void debayer8_GRGR_BGR888(uint8_t *dst, const uint8_t *src[]);
> + void debayer8_GRGR_XBGR8888(uint8_t *dst, const uint8_t *src[]);
> /* unpacked 10-bit raw bayer format */
> void debayer10_BGBG_BGR888(uint8_t *dst, const uint8_t *src[]);
> + void debayer10_BGBG_XBGR8888(uint8_t *dst, const uint8_t *src[]);
> void debayer10_GRGR_BGR888(uint8_t *dst, const uint8_t *src[]);
> + void debayer10_GRGR_XBGR8888(uint8_t *dst, const uint8_t *src[]);
> /* unpacked 12-bit raw bayer format */
> void debayer12_BGBG_BGR888(uint8_t *dst, const uint8_t *src[]);
> + void debayer12_BGBG_XBGR8888(uint8_t *dst, const uint8_t *src[]);
> void debayer12_GRGR_BGR888(uint8_t *dst, const uint8_t *src[]);
> + void debayer12_GRGR_XBGR8888(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_BGBG_XBGR8888(uint8_t *dst, const uint8_t *src[]);
> void debayer10P_GRGR_BGR888(uint8_t *dst, const uint8_t *src[]);
> + void debayer10P_GRGR_XBGR8888(uint8_t *dst, const uint8_t *src[]);
> void debayer10P_GBGB_BGR888(uint8_t *dst, const uint8_t *src[]);
> + void debayer10P_GBGB_XBGR8888(uint8_t *dst, const uint8_t *src[]);
> void debayer10P_RGRG_BGR888(uint8_t *dst, const uint8_t *src[]);
> + void debayer10P_RGRG_XBGR8888(uint8_t *dst, const uint8_t *src[]);
>
> struct DebayerInputConfig {
> Size patternSize;
More information about the libcamera-devel
mailing list