[FFmpeg-cvslog] indeo4: implement haar 8x1 and 1x8 transforms

[Michael Niedermayer]

ffmpeg | branch: master | Michael Niedermayer <michaelni at gmx.at> | Mon Apr 22 02:24:08 2013 +0200| [de9fbf3dc41abb8eb7a70b8a6bbfce64cee21bc5] | committer: Michael Niedermayer

indeo4: implement haar 8x1 and 1x8 transforms

Fixes Ticket1984
Signed-off-by: Michael Niedermayer <michaelni at gmx.at>

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=de9fbf3dc41abb8eb7a70b8a6bbfce64cee21bc5
---

 libavcodec/indeo4.c  |    4 ++--
 libavcodec/ivi_dsp.c |   54 ++++++++++++++++++++++++++++++++++++++++++++++++++
 libavcodec/ivi_dsp.h |    4 ++++
 3 files changed, 60 insertions(+), 2 deletions(-)

diff --git a/libavcodec/indeo4.c b/libavcodec/indeo4.c
index a958b50..5b107fb 100644
--- a/libavcodec/indeo4.c
+++ b/libavcodec/indeo4.c
@@ -56,8 +56,8 @@ static const struct {
     int             is_2d_trans;
 } transforms[18] = {
     { ff_ivi_inverse_haar_8x8,  ff_ivi_dc_haar_2d,       1 },
-    { NULL, NULL, 0 }, /* inverse Haar 8x1 */
-    { NULL, NULL, 0 }, /* inverse Haar 1x8 */
+    { ff_ivi_inverse_haar_8x1,  ff_ivi_dc_haar_2d,       1 },
+    { ff_ivi_inverse_haar_1x8,  ff_ivi_dc_haar_2d,       1 },
     { ff_ivi_put_pixels_8x8,    ff_ivi_put_dc_pixel_8x8, 1 },
     { ff_ivi_inverse_slant_8x8, ff_ivi_dc_slant_2d,      1 },
     { ff_ivi_row_slant8,        ff_ivi_dc_row_slant,     1 },
diff --git a/libavcodec/ivi_dsp.c b/libavcodec/ivi_dsp.c
index 84e2436..9c35aa0 100644
--- a/libavcodec/ivi_dsp.c
+++ b/libavcodec/ivi_dsp.c
@@ -320,6 +320,60 @@ void ff_ivi_inverse_haar_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
 #undef  COMPENSATE
 }
 
+void ff_ivi_inverse_haar_1x8(const int32_t *in, int16_t *out, uint32_t pitch,
+                             const uint8_t *flags)
+{
+    int     i;
+    const int32_t *src;
+    int     t0, t1, t2, t3, t4, t5, t6, t7, t8;
+
+    /* apply the InvHaar8 to all columns */
+#define COMPENSATE(x) (x)
+    src = in;
+    for (i = 0; i < 8; i++) {
+        if (flags[i]) {
+            INV_HAAR8(src[ 0], src[ 8], src[16], src[24],
+                      src[32], src[40], src[48], src[56],
+                      out[ 0], out[pitch], out[2*pitch], out[3*pitch],
+                      out[4*pitch], out[5*pitch], out[6*pitch], out[7*pitch],
+                      t0, t1, t2, t3, t4, t5, t6, t7, t8);
+        } else
+            out[      0]= out[  pitch]= out[2*pitch]= out[3*pitch]=
+            out[4*pitch]= out[5*pitch]= out[6*pitch]= out[7*pitch]= 0;
+
+        src++;
+        out++;
+    }
+#undef  COMPENSATE
+}
+
+void ff_ivi_inverse_haar_8x1(const int32_t *in, int16_t *out, uint32_t pitch,
+                             const uint8_t *flags)
+{
+    int     i;
+    const int32_t *src;
+    int     t0, t1, t2, t3, t4, t5, t6, t7, t8;
+
+    /* apply the InvHaar8 to all rows */
+#define COMPENSATE(x) (x)
+    src = in;
+    for (i = 0; i < 8; i++) {
+        if (   !src[0] && !src[1] && !src[2] && !src[3]
+            && !src[4] && !src[5] && !src[6] && !src[7]) {
+            memset(out, 0, 8 * sizeof(out[0]));
+        } else {
+            INV_HAAR8(src[0], src[1], src[2], src[3],
+                      src[4], src[5], src[6], src[7],
+                      out[0], out[1], out[2], out[3],
+                      out[4], out[5], out[6], out[7],
+                      t0, t1, t2, t3, t4, t5, t6, t7, t8);
+        }
+        src += 8;
+        out += pitch;
+    }
+#undef  COMPENSATE
+}
+
 void ff_ivi_dc_haar_2d(const int32_t *in, int16_t *out, uint32_t pitch,
                        int blk_size)
 {
diff --git a/libavcodec/ivi_dsp.h b/libavcodec/ivi_dsp.h
index e95bb11..748b3f4 100644
--- a/libavcodec/ivi_dsp.h
+++ b/libavcodec/ivi_dsp.h
@@ -64,6 +64,10 @@ void ff_ivi_recompose_haar(const IVIPlaneDesc *plane, uint8_t *dst,
  */
 void ff_ivi_inverse_haar_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
                              const uint8_t *flags);
+void ff_ivi_inverse_haar_8x1(const int32_t *in, int16_t *out, uint32_t pitch,
+                             const uint8_t *flags);
+void ff_ivi_inverse_haar_1x8(const int32_t *in, int16_t *out, uint32_t pitch,
+                             const uint8_t *flags);
 
 /**
  *  DC-only two-dimensional inverse Haar transform for Indeo 4.