[igt-dev] [PATCH i-g-t 2/3] lib/igt_fb: Add support for Y410/Y416 formats, v4.
Ville Syrjälä
ville.syrjala at linux.intel.com
Wed Mar 27 17:20:18 UTC 2019
On Wed, Mar 27, 2019 at 10:29:29AM +0100, Maarten Lankhorst wrote:
> Y410 is packed with compressed a channel and only 32 bpp, like
> 10 bits RGB formats. Y416 is a packed 16 bits per component format.
>
> Changes since v1:
> - Rebase on top of upstream YUV changes.
> Changes since v2:
> - Discard alpha channel, not used upstream.
> Changes since v3:
> - Handle the XVYU formats and Y41x formats correctly.
>
> Signed-off-by: Maarten Lankhorst <maarten.lankhorst at linux.intel.com>
> ---
> lib/igt_color_encoding.c | 6 +
> lib/igt_fb.c | 263 +++++++++++++++++++++++++++++++++++----
> 2 files changed, 244 insertions(+), 25 deletions(-)
>
> diff --git a/lib/igt_color_encoding.c b/lib/igt_color_encoding.c
> index 9f9dc1439ee4..7de6d5abd690 100644
> --- a/lib/igt_color_encoding.c
> +++ b/lib/igt_color_encoding.c
> @@ -154,6 +154,12 @@ static const struct color_encoding_format {
> { DRM_FORMAT_Y210, 65472.f, 4096.f, 60160.f, 4096.f, 32768.f, 61440.f },
> { DRM_FORMAT_Y212, 65520.f, 4096.f, 60160.f, 4096.f, 32768.f, 61440.f },
> { DRM_FORMAT_Y216, 65535.f, 4096.f, 60160.f, 4096.f, 32768.f, 61440.f },
> + { DRM_FORMAT_Y410, 1023.f, 64.f, 940.f, 64.f, 512.f, 960.f },
> + { DRM_FORMAT_Y412, 65520.f, 4096.f, 60160.f, 4096.f, 32768.f, 61440.f },
> + { DRM_FORMAT_Y416, 65535.f, 4096.f, 60160.f, 4096.f, 32768.f, 61440.f },
> + { DRM_FORMAT_XVYU2101010, 1023.f, 64.f, 940.f, 64.f, 512.f, 960.f },
> + { DRM_FORMAT_XVYU12_16161616, 65520.f, 4096.f, 60160.f, 4096.f, 32768.f, 61440.f },
> + { DRM_FORMAT_XVYU16161616, 65535.f, 4096.f, 60160.f, 4096.f, 32768.f, 61440.f },
> };
>
> static const struct color_encoding_format *lookup_fourcc(uint32_t fourcc)
> diff --git a/lib/igt_fb.c b/lib/igt_fb.c
> index 1e0bcb9ea2cd..2d55687a7aa3 100644
> --- a/lib/igt_fb.c
> +++ b/lib/igt_fb.c
> @@ -221,6 +221,36 @@ static const struct format_desc_struct {
> .num_planes = 3, .plane_bpp = { 8, 8, 8, },
> .hsub = 2, .vsub = 1,
> },
> + { .name = "Y410", .depth = -1, .drm_id = DRM_FORMAT_Y410,
> + .cairo_id = CAIRO_FORMAT_RGBA128F,
> + .num_planes = 1, .plane_bpp = { 32, },
> + .hsub = 1, .vsub = 1,
> + },
> + { .name = "Y412", .depth = -1, .drm_id = DRM_FORMAT_Y412,
> + .cairo_id = CAIRO_FORMAT_RGBA128F,
> + .num_planes = 1, .plane_bpp = { 64, },
> + .hsub = 1, .vsub = 1,
> + },
> + { .name = "Y416", .depth = -1, .drm_id = DRM_FORMAT_Y416,
> + .cairo_id = CAIRO_FORMAT_RGBA128F,
> + .num_planes = 1, .plane_bpp = { 64, },
> + .hsub = 1, .vsub = 1,
> + },
> + { .name = "XV30", .depth = -1, .drm_id = DRM_FORMAT_XVYU2101010,
> + .cairo_id = CAIRO_FORMAT_RGB96F,
> + .num_planes = 1, .plane_bpp = { 32, },
> + .hsub = 1, .vsub = 1,
> + },
> + { .name = "XV36", .depth = -1, .drm_id = DRM_FORMAT_XVYU12_16161616,
> + .cairo_id = CAIRO_FORMAT_RGB96F,
> + .num_planes = 1, .plane_bpp = { 64, },
> + .hsub = 1, .vsub = 1,
> + },
> + { .name = "XV48", .depth = -1, .drm_id = DRM_FORMAT_XVYU16161616,
> + .cairo_id = CAIRO_FORMAT_RGB96F,
> + .num_planes = 1, .plane_bpp = { 64, },
> + .hsub = 1, .vsub = 1,
> + },
> { .name = "P010", .depth = -1, .drm_id = DRM_FORMAT_P010,
> .cairo_id = CAIRO_FORMAT_RGB96F,
> .num_planes = 2, .plane_bpp = { 16, 32 },
> @@ -684,6 +714,33 @@ static void clear_yuv_buffer(struct igt_fb *fb)
> fb->strides[0] * fb->plane_height[0] / sizeof(wchar_t));
> break;
>
> + case DRM_FORMAT_XVYU2101010:
> + case DRM_FORMAT_Y410:
> + wmemset(ptr + fb->offsets[0],
> + full_range ? 0xe0000200 : 0xe0010200,
For everything else we leave alpha zeroed.
> + fb->strides[0] * fb->plane_height[0] / sizeof(wchar_t));
> + break;
> +
> + case DRM_FORMAT_XVYU12_16161616:
> + case DRM_FORMAT_XVYU16161616:
> + case DRM_FORMAT_Y412:
> + case DRM_FORMAT_Y416: {
> + struct ayuv16 { uint16_t u, y, v, a; };
> + const struct ayuv16 pixel = {
> + .a = 0xffff,
> + .v = full_range ? 0 : 0x1000,
> + .y = 0x8000,
> + .u = full_range ? 0 : 0x1000
> + };
> +
> + for (int i = 0; i < fb->plane_height[0]; i++) {
> + struct ayuv16 *cur = ptr + fb->offsets[0] + fb->strides[0] * i;
> +
> + for (int j = 0; j < fb->plane_width[0]; j++)
> + *cur++ = pixel;
> + }
Maybe we want a memset64()?
> + break;
> + }
> }
>
> igt_fb_unmap_buffer(fb, ptr);
> @@ -1897,10 +1954,11 @@ static void convert_src_put(const struct fb_convert *cvt,
> }
>
> struct yuv_parameters {
> - unsigned y_inc;
> + unsigned ay_inc;
> unsigned uv_inc;
> - unsigned y_stride;
> + unsigned ay_stride;
> unsigned uv_stride;
> + unsigned a_offset;
> unsigned y_offset;
> unsigned u_offset;
> unsigned v_offset;
> @@ -1918,7 +1976,7 @@ static void get_yuv_parameters(struct igt_fb *fb, struct yuv_parameters *params)
> case DRM_FORMAT_P010:
> case DRM_FORMAT_P012:
> case DRM_FORMAT_P016:
> - params->y_inc = 1;
> + params->ay_inc = 1;
> params->uv_inc = 2;
> break;
>
> @@ -1926,7 +1984,7 @@ static void get_yuv_parameters(struct igt_fb *fb, struct yuv_parameters *params)
> case DRM_FORMAT_YUV422:
> case DRM_FORMAT_YVU420:
> case DRM_FORMAT_YVU422:
> - params->y_inc = 1;
> + params->ay_inc = 1;
> params->uv_inc = 1;
> break;
>
> @@ -1937,12 +1995,16 @@ static void get_yuv_parameters(struct igt_fb *fb, struct yuv_parameters *params)
> case DRM_FORMAT_Y210:
> case DRM_FORMAT_Y212:
> case DRM_FORMAT_Y216:
> - params->y_inc = 2;
> + params->ay_inc = 2;
> params->uv_inc = 4;
> break;
>
> + case DRM_FORMAT_XVYU12_16161616:
> + case DRM_FORMAT_XVYU16161616:
> + case DRM_FORMAT_Y412:
> + case DRM_FORMAT_Y416:
> case DRM_FORMAT_XYUV8888:
> - params->y_inc = 4;
> + params->ay_inc = 4;
> params->uv_inc = 4;
> break;
> }
> @@ -1959,7 +2021,7 @@ static void get_yuv_parameters(struct igt_fb *fb, struct yuv_parameters *params)
> case DRM_FORMAT_P010:
> case DRM_FORMAT_P012:
> case DRM_FORMAT_P016:
> - params->y_stride = fb->strides[0];
> + params->ay_stride = fb->strides[0];
> params->uv_stride = fb->strides[1];
> break;
>
> @@ -1971,7 +2033,11 @@ static void get_yuv_parameters(struct igt_fb *fb, struct yuv_parameters *params)
> case DRM_FORMAT_Y212:
> case DRM_FORMAT_Y216:
> case DRM_FORMAT_XYUV8888:
> - params->y_stride = fb->strides[0];
> + case DRM_FORMAT_XVYU12_16161616:
> + case DRM_FORMAT_XVYU16161616:
> + case DRM_FORMAT_Y412:
> + case DRM_FORMAT_Y416:
> + params->ay_stride = fb->strides[0];
> params->uv_stride = fb->strides[0];
> break;
> }
> @@ -2045,6 +2111,16 @@ static void get_yuv_parameters(struct igt_fb *fb, struct yuv_parameters *params)
> params->v_offset = fb->offsets[0] + 6;
> break;
>
> + case DRM_FORMAT_XVYU12_16161616:
> + case DRM_FORMAT_XVYU16161616:
> + case DRM_FORMAT_Y412:
> + case DRM_FORMAT_Y416:
> + params->a_offset = fb->offsets[0] + 6;
> + params->y_offset = fb->offsets[0] + 2;
> + params->u_offset = fb->offsets[0];
> + params->v_offset = fb->offsets[0] + 4;
> + break;
> +
> case DRM_FORMAT_XYUV8888:
> params->y_offset = fb->offsets[0] + 1;
> params->u_offset = fb->offsets[0] + 2;
> @@ -2096,7 +2172,7 @@ static void convert_yuv_to_rgb24(struct fb_convert *cvt)
> write_rgb(rgb_tmp, &rgb);
>
> rgb_tmp += bpp;
> - y_tmp += params.y_inc;
> + y_tmp += params.ay_inc;
>
> if ((src_fmt->hsub == 1) || (j % src_fmt->hsub)) {
> u_tmp += params.uv_inc;
> @@ -2105,7 +2181,7 @@ static void convert_yuv_to_rgb24(struct fb_convert *cvt)
> }
>
> rgb24 += rgb24_stride;
> - y += params.y_stride;
> + y += params.ay_stride;
>
> if ((src_fmt->vsub == 1) || (i % src_fmt->vsub)) {
> u += params.uv_stride;
> @@ -2156,7 +2232,7 @@ static void convert_rgb24_to_yuv(struct fb_convert *cvt)
> rgb_tmp += bpp;
>
> *y_tmp = yuv.d[0];
> - y_tmp += params.y_inc;
> + y_tmp += params.ay_inc;
>
> if ((i % dst_fmt->vsub) || (j % dst_fmt->hsub))
> continue;
> @@ -2193,7 +2269,7 @@ static void convert_rgb24_to_yuv(struct fb_convert *cvt)
> }
>
> rgb24 += rgb24_stride;
> - y += params.y_stride;
> + y += params.ay_stride;
>
> if ((i % dst_fmt->vsub) == (dst_fmt->vsub - 1)) {
> u += params.uv_stride;
> @@ -2217,13 +2293,13 @@ static void write_rgbf(float *rgb24, const struct igt_vec4 *rgb)
> rgb24[2] = rgb->d[2];
> }
>
> -static void convert_yuv16_to_float(struct fb_convert *cvt)
> +static void convert_yuv16_to_float(struct fb_convert *cvt, bool alpha)
> {
> const struct format_desc_struct *src_fmt =
> lookup_drm_format(cvt->src.fb->drm_format);
> int i, j;
> - uint8_t fpp = 3;
> - uint16_t *y, *u, *v;
> + uint8_t fpp = alpha ? 4 : 3;
> + uint16_t *a, *y, *u, *v;
> float *ptr = cvt->dst.ptr;
> unsigned int float_stride = cvt->dst.fb->strides[0] / sizeof(*ptr);
> struct igt_mat4 m = igt_ycbcr_to_rgb_matrix(cvt->src.fb->drm_format,
> @@ -2242,11 +2318,13 @@ static void convert_yuv16_to_float(struct fb_convert *cvt)
> !(params.u_offset % sizeof(*buf)) &&
> !(params.v_offset % sizeof(*buf)));
>
> + a = buf + params.a_offset / sizeof(*buf);
> y = buf + params.y_offset / sizeof(*buf);
> u = buf + params.u_offset / sizeof(*buf);
> v = buf + params.v_offset / sizeof(*buf);
>
> for (i = 0; i < cvt->dst.fb->height; i++) {
> + const uint16_t *a_tmp = a;
> const uint16_t *y_tmp = y;
> const uint16_t *u_tmp = u;
> const uint16_t *v_tmp = v;
> @@ -2263,8 +2341,13 @@ static void convert_yuv16_to_float(struct fb_convert *cvt)
> rgb = igt_matrix_transform(&m, &yuv);
> write_rgbf(rgb_tmp, &rgb);
>
> + if (alpha) {
> + rgb_tmp[3] = ((float)*a_tmp) / 65535.f;
> + a_tmp += params.ay_inc;
> + }
> +
> rgb_tmp += fpp;
> - y_tmp += params.y_inc;
> + y_tmp += params.ay_inc;
>
> if ((src_fmt->hsub == 1) || (j % src_fmt->hsub)) {
> u_tmp += params.uv_inc;
> @@ -2273,7 +2356,9 @@ static void convert_yuv16_to_float(struct fb_convert *cvt)
> }
>
> ptr += float_stride;
> - y += params.y_stride / sizeof(*y);
> +
> + a += params.ay_stride / sizeof(*a);
> + y += params.ay_stride / sizeof(*y);
>
> if ((src_fmt->vsub == 1) || (i % src_fmt->vsub)) {
> u += params.uv_stride / sizeof(*u);
> @@ -2284,14 +2369,14 @@ static void convert_yuv16_to_float(struct fb_convert *cvt)
> convert_src_put(cvt, buf);
> }
>
> -static void convert_float_to_yuv16(struct fb_convert *cvt)
> +static void convert_float_to_yuv16(struct fb_convert *cvt, bool alpha)
> {
> const struct format_desc_struct *dst_fmt =
> lookup_drm_format(cvt->dst.fb->drm_format);
> int i, j;
> - uint16_t *y, *u, *v;
> + uint16_t *a, *y, *u, *v;
> const float *ptr = cvt->src.ptr;
> - uint8_t fpp = 3;
> + uint8_t fpp = alpha ? 4 : 3;
> unsigned float_stride = cvt->src.fb->strides[0] / sizeof(*ptr);
> struct igt_mat4 m = igt_rgb_to_ycbcr_matrix(cvt->src.fb->drm_format,
> cvt->dst.fb->drm_format,
> @@ -2303,16 +2388,19 @@ static void convert_float_to_yuv16(struct fb_convert *cvt)
> igt_format_is_yuv(cvt->dst.fb->drm_format));
>
> get_yuv_parameters(cvt->dst.fb, ¶ms);
> - igt_assert(!(params.y_offset % sizeof(*y)) &&
> + igt_assert(!(params.a_offset % sizeof(*a)) &&
> + !(params.y_offset % sizeof(*y)) &&
> !(params.u_offset % sizeof(*u)) &&
> !(params.v_offset % sizeof(*v)));
>
> + a = cvt->dst.ptr + params.a_offset;
> y = cvt->dst.ptr + params.y_offset;
> u = cvt->dst.ptr + params.u_offset;
> v = cvt->dst.ptr + params.v_offset;
>
> for (i = 0; i < cvt->dst.fb->height; i++) {
> const float *rgb_tmp = ptr;
> + uint16_t *a_tmp = a;
> uint16_t *y_tmp = y;
> uint16_t *u_tmp = u;
> uint16_t *v_tmp = v;
> @@ -2325,10 +2413,15 @@ static void convert_float_to_yuv16(struct fb_convert *cvt)
> read_rgbf(&rgb, rgb_tmp);
> yuv = igt_matrix_transform(&m, &rgb);
>
> + if (alpha) {
> + *a_tmp = rgb_tmp[3] * 65535.f + .5f;
> + a_tmp += params.ay_inc;
> + }
> +
> rgb_tmp += fpp;
>
> *y_tmp = yuv.d[0];
> - y_tmp += params.y_inc;
> + y_tmp += params.ay_inc;
>
> if ((i % dst_fmt->vsub) || (j % dst_fmt->hsub))
> continue;
> @@ -2365,7 +2458,8 @@ static void convert_float_to_yuv16(struct fb_convert *cvt)
> }
>
> ptr += float_stride;
> - y += params.y_stride / sizeof(*y);
> + a += params.ay_stride / sizeof(*a);
> + y += params.ay_stride / sizeof(*y);
>
> if ((i % dst_fmt->vsub) == (dst_fmt->vsub - 1)) {
> u += params.uv_stride / sizeof(*u);
> @@ -2374,6 +2468,95 @@ static void convert_float_to_yuv16(struct fb_convert *cvt)
> }
> }
>
> +static void convert_Y410_to_float(struct fb_convert *cvt, bool alpha)
> +{
> + int i, j;
> + const uint32_t *uyv;
> + uint32_t *buf;
> + float *ptr = cvt->dst.ptr;
> + unsigned int float_stride = cvt->dst.fb->strides[0] / sizeof(*ptr);
> + unsigned int uyv_stride = cvt->src.fb->strides[0] / sizeof(*uyv);
> + struct igt_mat4 m = igt_ycbcr_to_rgb_matrix(cvt->src.fb->drm_format,
> + cvt->dst.fb->drm_format,
> + cvt->src.fb->color_encoding,
> + cvt->src.fb->color_range);
> + unsigned bpp = alpha ? 4 : 3;
> +
> + igt_assert((cvt->src.fb->drm_format == DRM_FORMAT_Y410 ||
> + cvt->src.fb->drm_format == DRM_FORMAT_XVYU2101010) &&
> + cvt->dst.fb->drm_format == IGT_FORMAT_FLOAT);
> +
> + uyv = buf = convert_src_get(cvt);
> +
> + for (i = 0; i < cvt->dst.fb->height; i++) {
> + for (j = 0; j < cvt->dst.fb->width; j++) {
> + /* Convert 2x1 pixel blocks */
> + struct igt_vec4 yuv;
> + struct igt_vec4 rgb;
> +
> + yuv.d[0] = uyv[j] & 0x3ff;
> + yuv.d[1] = (uyv[j] >> 10) & 0x3ff;
> + yuv.d[2] = (uyv[j] >> 20) & 0x3ff;
These seem to be in the wrong order. At least I think we wanted
[0]=y, [1]=cb, [2]=cr
> + yuv.d[3] = 1.f;
> +
> + rgb = igt_matrix_transform(&m, &yuv);
> +
> + write_rgbf(&ptr[j * bpp], &rgb);
> + if (alpha)
> + ptr[j * bpp + 3] = (float)(uyv[j] >> 30) / 3.f;
> + }
> +
> + ptr += float_stride;
> + uyv += uyv_stride;
> + }
> +
> + convert_src_put(cvt, buf);
> +}
> +
> +static void convert_float_to_Y410(struct fb_convert *cvt, bool alpha)
> +{
> + int i, j;
> + uint32_t *uyv = cvt->dst.ptr;
> + const float *ptr = cvt->src.ptr;
> + unsigned float_stride = cvt->src.fb->strides[0] / sizeof(*ptr);
> + unsigned uyv_stride = cvt->dst.fb->strides[0] / sizeof(*uyv);
> + struct igt_mat4 m = igt_rgb_to_ycbcr_matrix(cvt->src.fb->drm_format,
> + cvt->dst.fb->drm_format,
> + cvt->dst.fb->color_encoding,
> + cvt->dst.fb->color_range);
> + unsigned bpp = alpha ? 4 : 3;
> +
> + igt_assert(cvt->src.fb->drm_format == IGT_FORMAT_FLOAT &&
> + (cvt->dst.fb->drm_format == DRM_FORMAT_Y410 ||
> + cvt->dst.fb->drm_format == DRM_FORMAT_XVYU2101010));
> +
> + for (i = 0; i < cvt->dst.fb->height; i++) {
> + for (j = 0; j < cvt->dst.fb->width; j++) {
> + struct igt_vec4 rgb;
> + struct igt_vec4 yuv;
> + uint8_t a = 0;
> + uint16_t y, cb, cr;
> +
> + read_rgbf(&rgb, &ptr[j * bpp]);
> + if (alpha)
> + a = ptr[j * bpp + 3] * 3.f + .5f;
> +
> + yuv = igt_matrix_transform(&m, &rgb);
> + cr = yuv.d[0];
> + y = yuv.d[1];
> + cb = yuv.d[2];
Same here.
> +
> + uyv[j] = ((cr & 0x3ff) << 0) |
> + ((y & 0x3ff) << 10) |
> + ((cb & 0x3ff) << 20) |
> + (a << 30);
cb and cr swapped here.
> + }
> +
> + ptr += float_stride;
> + uyv += uyv_stride;
> + }
> +}
> +
> static void convert_pixman(struct fb_convert *cvt)
> {
> pixman_format_code_t src_pixman = drm_format_to_pixman(cvt->src.fb->drm_format);
> @@ -2463,7 +2646,19 @@ static void fb_convert(struct fb_convert *cvt)
> case DRM_FORMAT_Y210:
> case DRM_FORMAT_Y212:
> case DRM_FORMAT_Y216:
> - convert_yuv16_to_float(cvt);
> + case DRM_FORMAT_XVYU12_16161616:
> + case DRM_FORMAT_XVYU16161616:
> + convert_yuv16_to_float(cvt, false);
> + return;
> + case DRM_FORMAT_Y410:
> + convert_Y410_to_float(cvt, true);
> + return;
> + case DRM_FORMAT_XVYU2101010:
> + convert_Y410_to_float(cvt, false);
> + return;
> + case DRM_FORMAT_Y412:
> + case DRM_FORMAT_Y416:
> + convert_yuv16_to_float(cvt, true);
> return;
> }
> } else if (cvt->src.fb->drm_format == IGT_FORMAT_FLOAT) {
> @@ -2474,7 +2669,19 @@ static void fb_convert(struct fb_convert *cvt)
> case DRM_FORMAT_Y210:
> case DRM_FORMAT_Y212:
> case DRM_FORMAT_Y216:
> - convert_float_to_yuv16(cvt);
> + case DRM_FORMAT_XVYU12_16161616:
> + case DRM_FORMAT_XVYU16161616:
> + convert_float_to_yuv16(cvt, false);
> + return;
> + case DRM_FORMAT_Y410:
> + convert_float_to_Y410(cvt, true);
> + return;
> + case DRM_FORMAT_XVYU2101010:
> + convert_float_to_Y410(cvt, false);
> + return;
> + case DRM_FORMAT_Y412:
> + case DRM_FORMAT_Y416:
> + convert_float_to_yuv16(cvt, true);
> return;
> }
> }
> @@ -2941,6 +3148,12 @@ bool igt_format_is_yuv(uint32_t drm_format)
> case DRM_FORMAT_Y210:
> case DRM_FORMAT_Y212:
> case DRM_FORMAT_Y216:
> + case DRM_FORMAT_XVYU2101010:
> + case DRM_FORMAT_XVYU12_16161616:
> + case DRM_FORMAT_XVYU16161616:
> + case DRM_FORMAT_Y410:
> + case DRM_FORMAT_Y412:
> + case DRM_FORMAT_Y416:
> case DRM_FORMAT_YUYV:
> case DRM_FORMAT_YVYU:
> case DRM_FORMAT_UYVY:
> --
> 2.20.1
>
> _______________________________________________
> igt-dev mailing list
> igt-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/igt-dev
--
Ville Syrjälä
Intel
More information about the igt-dev
mailing list