File 0012-add-support-to-convert-to-YUY2-YUYV-color-format.patch of Package gst-plugins-base

From 57cbb98e546309ce6b1a3e8c9cb2586cf731572d Mon Sep 17 00:00:00 2001
From: Rob Clark <rob@ti.com>
Date: Tue, 1 Dec 2009 22:42:43 -0600
Subject: [PATCH 12/38] add support to convert to YUY2/YUYV color format

---
 gst/stride/armv7.s   |   63 ++++++++++++++++++++++++++++++++++++++++++
 gst/stride/convert.c |   74 +++++++++++++++++++++++++++++++++++++++++++++++---
 2 files changed, 133 insertions(+), 4 deletions(-)

diff --git a/gst/stride/armv7.s b/gst/stride/armv7.s
index ed636f7..2697a14 100644
--- a/gst/stride/armv7.s
+++ b/gst/stride/armv7.s
@@ -69,6 +69,69 @@ stride_copy_zip2_3:
        bx lr
 @}
 
+
+       .align
+       .global stride_copy_zip3a
+       .type   stride_copy_zip3a, %function
+@void
+@stride_copy_zip3a (guchar *new_buf,
+@    guchar *orig_buf1, guchar *orig_buf2, guchar *orig_buf3, gint sz)
+@{
+@@@@ note: r0-r3, q0-3, and q8-q15 do not need to be preserved
+stride_copy_zip3a:
+       pld [r1, #64]
+       pld [r2, #64]
+       pld [r3, #64]
+       ldr ip, [sp]     @ the sz arg
+@ interleave remaining >= 32 bytes:
+       cmp ip, #32
+       blt stride_copy_zip3a_2
+stride_copy_zip3a_1:
+       vld1.8 {q8},      [r1]!  @ Y
+       vld1.8 {q10},     [r1]!  @ Y
+       vld1.8 {q9},      [r2]!  @ U
+       vld1.8 {q11},     [r3]!  @ V
+
+       pld [r1, #64]
+       pld [r2, #64]
+       pld [r3, #64]
+
+       vzip.8 q9, q11           @ interleave U&V
+       vzip.8 q8, q9            @ interleave Y1UV1
+       vzip.8 q10, q11          @ interleave Y2UV2
+
+       vst1.8 {q8,q9},   [r0]!
+       vst1.8 {q10,q11}, [r0]!
+
+       sub ip, ip, #32
+
+       cmp ip, #32
+       bge stride_copy_zip3a_1
+@ interleave remaining >= 16 bytes:
+stride_copy_zip3a_2:
+       cmp ip, #16
+       blt stride_copy_zip3a_3
+
+       vld1.8 {d16},     [r1]!  @ Y
+       vld1.8 {d18},     [r1]!  @ Y
+       vld1.8 {d17},     [r2]!  @ U
+       vld1.8 {d19},     [r3]!  @ V
+
+       vzip.8 d17, d19          @ interleave U&V
+       vzip.8 d16, d17          @ interleave Y1UV1
+       vzip.8 d18, d19          @ interleave Y2UV2
+
+       vst1.8 {d16,d17}, [r0]!
+       vst1.8 {d18,d19}, [r0]!
+
+       sub ip, ip, #16
+@ copy remaining >= 8 bytes:
+stride_copy_zip3a_3:
+@XXX
+       bx lr
+@}
+
+
        .align
        .global stride_copy
        .type   stride_copy, %function
diff --git a/gst/stride/convert.c b/gst/stride/convert.c
index a15063b..0f59e78 100644
--- a/gst/stride/convert.c
+++ b/gst/stride/convert.c
@@ -58,6 +58,19 @@ stride_copy_zip2 (guchar *new_buf, guchar *orig_buf1, guchar *orig_buf2, gint sz
 }
 
 WEAK void
+stride_copy_zip3a (guchar *new_buf,
+    guchar *orig_buf1, guchar *orig_buf2, guchar *orig_buf3, gint sz)
+{
+  while (sz > 1) {
+    *new_buf++ = *orig_buf1++;
+    *new_buf++ = *orig_buf2++;
+    *new_buf++ = *orig_buf1++;
+    *new_buf++ = *orig_buf3++;
+    sz -= 2;
+  }
+}
+
+WEAK void
 stride_copy (guchar *new_buf, guchar *orig_buf, gint sz)
 {
   memcpy (new_buf, orig_buf, sz);
@@ -99,6 +112,36 @@ stridemove_zip2 (guchar *new_buf, guchar *orig_buf1, guchar *orig_buf2, gint new
 }
 
 /**
+ * move to strided buffer, interleaving three planes, where the first plane
+ * (orig_buf1) has 2x as many samples.. Ie. ABACABAC..
+ */
+static void
+stridemove_zip3a (guchar *new_buf,
+    guchar *orig_buf1, guchar *orig_buf2, guchar *orig_buf3,
+    guint new_width, gint orig_width, gint height)
+{
+  gint copy_width = (new_width < orig_width) ? new_width : orig_width;
+
+  while (height > 0) {
+
+    /* even row */
+    stride_copy_zip3a (new_buf, orig_buf1, orig_buf2, orig_buf3, copy_width);
+    new_buf += new_width;
+    orig_buf1 += orig_width;
+
+    /* odd row, recycles same U & V */
+    stride_copy_zip3a (new_buf, orig_buf1, orig_buf2, orig_buf3, copy_width);
+    new_buf += new_width;
+    orig_buf1 += orig_width;
+
+    orig_buf2 += orig_width/2;
+    orig_buf3 += orig_width/2;
+
+    height -= 2;
+  }
+}
+
+/**
  * Convert from one stride to another... like memmove, but can convert stride in
  * the process.  This function is not aware of pixels, only of bytes.  So widths
  * are given in bytes, not pixels.  The new_buf and orig_buf can point to the
@@ -250,14 +293,36 @@ stridify_i420_nv12 (GstStrideTransform *self, guchar *strided, guchar *unstrided
   /* XXX widths/heights/strides that are not multiple of four??: */
   stridemove_zip2 (
       strided + (height*stride),
-      unstrided + (height*width),
-      unstrided + (int)(height*width*1.25),
-      stride, width/2, height/2);                           /* interleave U&V */
-  stridemove (strided, unstrided, stride, width, height);   /* move Y */
+      unstrided + (height*width),                           /* U */
+      unstrided + (int)(height*width*1.25),                 /* V */
+      stride, width/2, height/2);
+  stridemove (strided, unstrided, stride, width, height);   /* Y */
+
+  return GST_FLOW_OK;
+}
+
+/** convert I420 unstrided to YUY2 strided */
+static GstFlowReturn
+stridify_i420_yuy2 (GstStrideTransform *self, guchar *strided, guchar *unstrided)
+{
+  gint width  = self->width;
+  gint height = self->height;
+  gint stride = self->out_rowstride;
+
+  g_return_val_if_fail (stride >= width, GST_FLOW_ERROR);
+
+  /* XXX widths/heights/strides that are not multiple of four??: */
+  stridemove_zip3a (
+      strided,
+      unstrided,                                            /* Y */
+      unstrided + (height*width),                           /* U */
+      unstrided + (int)(height*width*1.25),                 /* V */
+      stride, width, height);
 
   return GST_FLOW_OK;
 }
 
+
 /* last entry has GST_VIDEO_FORMAT_UNKNOWN for in/out formats */
 Conversion stride_conversions[] = {
   { { GST_VIDEO_FORMAT_NV12, GST_VIDEO_FORMAT_NV12 }, stridify_420sp_420sp, unstridify_420sp_420sp },
@@ -266,6 +331,7 @@ Conversion stride_conversions[] = {
   { { GST_VIDEO_FORMAT_YUY2, GST_VIDEO_FORMAT_YUY2 }, stridify_422i_422i,   unstridify_422i_422i },
   { { GST_VIDEO_FORMAT_UYVY, GST_VIDEO_FORMAT_UYVY }, stridify_422i_422i,   unstridify_422i_422i },
   { { GST_VIDEO_FORMAT_I420, GST_VIDEO_FORMAT_NV12 }, stridify_i420_nv12,   NULL },
+  { { GST_VIDEO_FORMAT_I420, GST_VIDEO_FORMAT_YUY2 }, stridify_i420_yuy2,   NULL },
   /* add new entries before here */
   { { GST_VIDEO_FORMAT_UNKNOWN } }
 };
-- 
1.7.5.4