v4l2grab: optimize conversion routines

The conversion routines came from the vimc driver, where they
also re-scale and flip images vertically and horizontally
and can convert into several formats including Bayer.

So, it was done in two steps. As we just want to convert
everything into RGB24, this can be simplified in order
to avoid double-buffering.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

1 files changed, 15 insertions, 38 deletions

diff --git a/contrib/test/v4l2grab.c b/contrib/test/v4l2grab.c
index 5a874c4f..70b29da3 100644
--- a/contrib/test/v4l2grab.c
+++ b/contrib/test/v4l2grab.c
@@ -191,32 +191,25 @@ static void copy_two_pixels(uint32_t fourcc,
 			    unsigned char *src[2], unsigned char **dst,
 			    int ypos)
 {
-	unsigned char _r[2], _g[2], _b[2], *r, *g, *b;
 	int i;
 
-	/* Step 1: read two consecutive pixels from src pointer */
-
-	r = _r;
-	g = _g;
-	b = _b;
-
 	switch (fourcc) {
 	case V4L2_PIX_FMT_RGB565: /* rrrrrggg gggbbbbb */
 		for (i = 0; i < 2; i++) {
 			uint16_t pix = (src[i][0] << 8) + src[i][1];
 
-			*r++ = (unsigned char)(((pix & 0xf800) >> 11) << 3) | 0x07;
-			*g++ = (unsigned char)((((pix & 0x07e0) >> 5)) << 2) | 0x03;
-			*b++ = (unsigned char)((pix & 0x1f) << 3) | 0x07;
+			*(*dst)++ = (unsigned char)(((pix & 0xf800) >> 11) << 3) | 0x07;
+			*(*dst)++ = (unsigned char)((((pix & 0x07e0) >> 5)) << 2) | 0x03;
+			*(*dst)++ = (unsigned char)((pix & 0x1f) << 3) | 0x07;
 		}
 		break;
 	case V4L2_PIX_FMT_RGB565X: /* gggbbbbb rrrrrggg */
 		for (i = 0; i < 2; i++) {
 			uint16_t pix = (src[i][1] << 8) + src[i][0];
 
-			*r++ = (unsigned char)(((pix & 0xf800) >> 11) << 3) | 0x07;
-			*g++ = (unsigned char)((((pix & 0x07e0) >> 5)) << 2) | 0x03;
-			*b++ = (unsigned char)((pix & 0x1f) << 3) | 0x07;
+			*(*dst)++ = (unsigned char)(((pix & 0xf800) >> 11) << 3) | 0x07;
+			*(*dst)++ = (unsigned char)((((pix & 0x07e0) >> 5)) << 2) | 0x03;
+			*(*dst)++ = (unsigned char)((pix & 0x1f) << 3) | 0x07;
 		}
 		break;
 
@@ -229,42 +222,26 @@ static void copy_two_pixels(uint32_t fourcc,
 		y_off = (fourcc == V4L2_PIX_FMT_YUYV || fourcc == V4L2_PIX_FMT_YVYU) ? 0 : 1;
 		u_off = (fourcc == V4L2_PIX_FMT_YUYV || fourcc == V4L2_PIX_FMT_UYVY) ? 0 : 1;
 
-		u = src[1 - y_off][u_off];
-		v = src[1 - y_off][1 - u_off];
+		u = src[1 - y_off][u_off] - 128;
+		v = src[1 - y_off][1 - u_off] - 128;
 
 		for (i = 0; i < 2; i++) {
-			int y = src[i][y_off];
+			int y = src[i][y_off] - 16;
 
-			int luma = y - 16;
-			int cr = u - 128;
-			int cb = v - 128;
-
-			*r++ = CLAMP((298 * luma + 409 * cb + 128) >> 8);
-			*g++ = CLAMP((298 * luma - 100 * cr - 208 * cb + 128) >> 8);
-			*b++ = CLAMP((298 * luma + 516 * cr + 128) >> 8);
+			*(*dst)++ = CLAMP((298 * y + 409 * v + 128) >> 8);
+			*(*dst)++ = CLAMP((298 * y - 100 * u - 208 * v + 128) >> 8);
+			*(*dst)++ = CLAMP((298 * y + 516 * u + 128) >> 8);
 		}
 		break;
 	default:
 	case V4L2_PIX_FMT_BGR24:
 		for (i = 0; i < 2; i++) {
-			*b++ = src[i][0];
-			*g++ = src[i][1];
-			*r++ = src[i][2];
+			*(*dst)++ = src[i][2];
+			*(*dst)++ = src[i][1];
+			*(*dst)++ = src[i][0];
 		}
 		break;
 	}
-
-	/* Step 2: store two consecutive points in RGB24 format */
-
-	r = _r;
-	g = _g;
-	b = _b;
-
-	for (i = 0; i < 2; i++) {
-		*(*dst)++ = *r++;
-		*(*dst)++ = *g++;
-		*(*dst)++ = *b++;
-	}
 }
 
 static unsigned int convert_to_rgb24(uint32_t fourcc, unsigned char *p_in,