]> git.karo-electronics.de Git - linux-beck.git/commitdiff
[media] coda: add support for partial interleaved YCbCr 4:2:0 (NV12) format
authorPhilipp Zabel <p.zabel@pengutronix.de>
Mon, 29 Sep 2014 12:53:47 +0000 (09:53 -0300)
committerMauro Carvalho Chehab <mchehab@osg.samsung.com>
Tue, 28 Oct 2014 16:31:01 +0000 (14:31 -0200)
This patch adds support for the two-plane NV12 format with one luma plane
and one interleaved chroma plane.

Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Kamil Debski <k.debski@samsung.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
drivers/media/platform/coda/coda-bit.c
drivers/media/platform/coda/coda-common.c

index 3839e3507628e17dc08361d22a5cb24eed3f0ec1..fde777573f7becec53d8ee8f49e1a919a81ae1e8 100644 (file)
@@ -729,6 +729,9 @@ static int coda_start_encoding(struct coda_ctx *ctx)
                break;
        }
 
+       ctx->frame_mem_ctrl &= ~CODA_FRAME_CHROMA_INTERLEAVE;
+       if (q_data_src->fourcc == V4L2_PIX_FMT_NV12)
+               ctx->frame_mem_ctrl |= CODA_FRAME_CHROMA_INTERLEAVE;
        coda_write(dev, ctx->frame_mem_ctrl, CODA_REG_BIT_FRAME_MEM_CTRL);
 
        if (dev->devtype->product == CODA_DX6) {
@@ -1128,7 +1131,6 @@ static int coda_prepare_encode(struct coda_ctx *ctx)
        coda_write(dev, rot_mode, CODA_CMD_ENC_PIC_ROT_MODE);
        coda_write(dev, quant_param, CODA_CMD_ENC_PIC_QS);
 
-
        if (dev->devtype->product == CODA_960) {
                coda_write(dev, 4/*FIXME: 0*/, CODA9_CMD_ENC_PIC_SRC_INDEX);
                coda_write(dev, q_data_src->width, CODA9_CMD_ENC_PIC_SRC_STRIDE);
@@ -1273,7 +1275,7 @@ static int __coda_start_decoding(struct coda_ctx *ctx)
        u32 bitstream_buf, bitstream_size;
        struct coda_dev *dev = ctx->dev;
        int width, height;
-       u32 src_fourcc;
+       u32 src_fourcc, dst_fourcc;
        u32 val;
        int ret;
 
@@ -1283,6 +1285,7 @@ static int __coda_start_decoding(struct coda_ctx *ctx)
        bitstream_buf = ctx->bitstream.paddr;
        bitstream_size = ctx->bitstream.size;
        src_fourcc = q_data_src->fourcc;
+       dst_fourcc = q_data_dst->fourcc;
 
        /* Allocate per-instance buffers */
        ret = coda_alloc_context_buffers(ctx, q_data_src);
@@ -1294,6 +1297,9 @@ static int __coda_start_decoding(struct coda_ctx *ctx)
        /* Update coda bitstream read and write pointers from kfifo */
        coda_kfifo_sync_to_device_full(ctx);
 
+       ctx->frame_mem_ctrl &= ~CODA_FRAME_CHROMA_INTERLEAVE;
+       if (dst_fourcc == V4L2_PIX_FMT_NV12)
+               ctx->frame_mem_ctrl |= CODA_FRAME_CHROMA_INTERLEAVE;
        coda_write(dev, ctx->frame_mem_ctrl, CODA_REG_BIT_FRAME_MEM_CTRL);
 
        ctx->display_idx = -1;
@@ -1424,13 +1430,23 @@ static int __coda_start_decoding(struct coda_ctx *ctx)
        }
 
        if (dev->devtype->product == CODA_960) {
-               coda_write(dev, -1, CODA9_CMD_SET_FRAME_DELAY);
+               int cbb_size, crb_size;
 
+               coda_write(dev, -1, CODA9_CMD_SET_FRAME_DELAY);
+               /* Luma 2x0 page, 2x6 cache, chroma 2x0 page, 2x4 cache size */
                coda_write(dev, 0x20262024, CODA9_CMD_SET_FRAME_CACHE_SIZE);
+
+               if (dst_fourcc == V4L2_PIX_FMT_NV12) {
+                       cbb_size = 0;
+                       crb_size = 16;
+               } else {
+                       cbb_size = 8;
+                       crb_size = 8;
+               }
                coda_write(dev, 2 << CODA9_CACHE_PAGEMERGE_OFFSET |
                                32 << CODA9_CACHE_LUMA_BUFFER_SIZE_OFFSET |
-                               8 << CODA9_CACHE_CB_BUFFER_SIZE_OFFSET |
-                               8 << CODA9_CACHE_CR_BUFFER_SIZE_OFFSET,
+                               cbb_size << CODA9_CACHE_CB_BUFFER_SIZE_OFFSET |
+                               crb_size << CODA9_CACHE_CR_BUFFER_SIZE_OFFSET,
                                CODA9_CMD_SET_FRAME_CACHE_CONFIG);
        }
 
index feb270f2860b336ff14b304e7c5b17cae143b8df..02d47fabae651280de3bac79b5def063905ed4b3 100644 (file)
@@ -95,6 +95,7 @@ void coda_write_base(struct coda_ctx *ctx, struct coda_q_data *q_data,
                base_cb = base_cr + q_data->bytesperline * q_data->height / 4;
                break;
        case V4L2_PIX_FMT_YUV420:
+       case V4L2_PIX_FMT_NV12:
        default:
                base_cb = base_y + q_data->bytesperline * q_data->height;
                base_cr = base_cb + q_data->bytesperline * q_data->height / 4;
@@ -118,6 +119,10 @@ static const struct coda_fmt coda_formats[] = {
                .name = "YUV 4:2:0 Planar, YCrCb",
                .fourcc = V4L2_PIX_FMT_YVU420,
        },
+       {
+               .name = "YUV 4:2:0 Partial interleaved Y/CbCr",
+               .fourcc = V4L2_PIX_FMT_NV12,
+       },
        {
                .name = "H264 Encoded Stream",
                .fourcc = V4L2_PIX_FMT_H264,
@@ -162,6 +167,7 @@ static bool coda_format_is_yuv(u32 fourcc)
        switch (fourcc) {
        case V4L2_PIX_FMT_YUV420:
        case V4L2_PIX_FMT_YVU420:
+       case V4L2_PIX_FMT_NV12:
                return true;
        default:
                return false;
@@ -366,6 +372,7 @@ static int coda_try_fmt(struct coda_ctx *ctx, const struct coda_codec *codec,
        switch (f->fmt.pix.pixelformat) {
        case V4L2_PIX_FMT_YUV420:
        case V4L2_PIX_FMT_YVU420:
+       case V4L2_PIX_FMT_NV12:
        case V4L2_PIX_FMT_H264:
        case V4L2_PIX_FMT_MPEG4:
        case V4L2_PIX_FMT_JPEG:
@@ -380,6 +387,7 @@ static int coda_try_fmt(struct coda_ctx *ctx, const struct coda_codec *codec,
        switch (f->fmt.pix.pixelformat) {
        case V4L2_PIX_FMT_YUV420:
        case V4L2_PIX_FMT_YVU420:
+       case V4L2_PIX_FMT_NV12:
                /* Frame stride must be multiple of 8, but 16 for h.264 */
                f->fmt.pix.bytesperline = round_up(f->fmt.pix.width, 16);
                f->fmt.pix.sizeimage = f->fmt.pix.bytesperline *