summaryrefslogtreecommitdiff
path: root/libavcodec/ivi_dsp.c
diff options
context:
space:
mode:
authorDiego Biurrun <diego@biurrun.de>2016-08-26 12:26:50 +0200
committerDiego Biurrun <diego@biurrun.de>2016-09-04 11:57:15 +0200
commit746c56b7730ce09397d3a8354acc131285e9d829 (patch)
treece950689016e319a0ff996ad5a77a01912e2c511 /libavcodec/ivi_dsp.c
parent4fb311c804098d78e5ce5f527f9a9c37536d3a08 (diff)
indeo: Change type of array pitch parameters to ptrdiff_t
ptrdiff_t is the correct type for array pitches and similar.
Diffstat (limited to 'libavcodec/ivi_dsp.c')
-rw-r--r--libavcodec/ivi_dsp.c52
1 files changed, 26 insertions, 26 deletions
diff --git a/libavcodec/ivi_dsp.c b/libavcodec/ivi_dsp.c
index 9b74f730e4..b8a476deb6 100644
--- a/libavcodec/ivi_dsp.c
+++ b/libavcodec/ivi_dsp.c
@@ -31,13 +31,13 @@
#include "ivi_dsp.h"
void ff_ivi_recompose53(const IVIPlaneDesc *plane, uint8_t *dst,
- const int dst_pitch)
+ const ptrdiff_t dst_pitch)
{
int x, y, indx;
int32_t p0, p1, p2, p3, tmp0, tmp1, tmp2;
int32_t b0_1, b0_2, b1_1, b1_2, b1_3, b2_1, b2_2, b2_3, b2_4, b2_5, b2_6;
int32_t b3_1, b3_2, b3_3, b3_4, b3_5, b3_6, b3_7, b3_8, b3_9;
- int32_t pitch, back_pitch;
+ ptrdiff_t pitch, back_pitch;
const short *b0_ptr, *b1_ptr, *b2_ptr, *b3_ptr;
const int num_bands = 4;
@@ -178,11 +178,11 @@ void ff_ivi_recompose53(const IVIPlaneDesc *plane, uint8_t *dst,
}
void ff_ivi_recompose_haar(const IVIPlaneDesc *plane, uint8_t *dst,
- const int dst_pitch)
+ const ptrdiff_t dst_pitch)
{
int x, y, indx, b0, b1, b2, b3, p0, p1, p2, p3;
const short *b0_ptr, *b1_ptr, *b2_ptr, *b3_ptr;
- int32_t pitch;
+ ptrdiff_t pitch;
/* all bands should have the same pitch */
pitch = plane->bands[0].pitch;
@@ -257,7 +257,7 @@ void ff_ivi_recompose_haar(const IVIPlaneDesc *plane, uint8_t *dst,
d3 = COMPENSATE(t2);\
d4 = COMPENSATE(t3); }
-void ff_ivi_inverse_haar_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
+void ff_ivi_inverse_haar_8x8(const int32_t *in, int16_t *out, ptrdiff_t pitch,
const uint8_t *flags)
{
int i, shift, sp1, sp2, sp3, sp4;
@@ -312,7 +312,7 @@ void ff_ivi_inverse_haar_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
#undef COMPENSATE
}
-void ff_ivi_row_haar8(const int32_t *in, int16_t *out, uint32_t pitch,
+void ff_ivi_row_haar8(const int32_t *in, int16_t *out, ptrdiff_t pitch,
const uint8_t *flags)
{
int i;
@@ -337,7 +337,7 @@ void ff_ivi_row_haar8(const int32_t *in, int16_t *out, uint32_t pitch,
#undef COMPENSATE
}
-void ff_ivi_col_haar8(const int32_t *in, int16_t *out, uint32_t pitch,
+void ff_ivi_col_haar8(const int32_t *in, int16_t *out, ptrdiff_t pitch,
const uint8_t *flags)
{
int i;
@@ -366,7 +366,7 @@ void ff_ivi_col_haar8(const int32_t *in, int16_t *out, uint32_t pitch,
#undef COMPENSATE
}
-void ff_ivi_inverse_haar_4x4(const int32_t *in, int16_t *out, uint32_t pitch,
+void ff_ivi_inverse_haar_4x4(const int32_t *in, int16_t *out, ptrdiff_t pitch,
const uint8_t *flags)
{
int i, shift, sp1, sp2;
@@ -413,7 +413,7 @@ void ff_ivi_inverse_haar_4x4(const int32_t *in, int16_t *out, uint32_t pitch,
#undef COMPENSATE
}
-void ff_ivi_row_haar4(const int32_t *in, int16_t *out, uint32_t pitch,
+void ff_ivi_row_haar4(const int32_t *in, int16_t *out, ptrdiff_t pitch,
const uint8_t *flags)
{
int i;
@@ -435,7 +435,7 @@ void ff_ivi_row_haar4(const int32_t *in, int16_t *out, uint32_t pitch,
#undef COMPENSATE
}
-void ff_ivi_col_haar4(const int32_t *in, int16_t *out, uint32_t pitch,
+void ff_ivi_col_haar4(const int32_t *in, int16_t *out, ptrdiff_t pitch,
const uint8_t *flags)
{
int i;
@@ -459,7 +459,7 @@ void ff_ivi_col_haar4(const int32_t *in, int16_t *out, uint32_t pitch,
#undef COMPENSATE
}
-void ff_ivi_dc_haar_2d(const int32_t *in, int16_t *out, uint32_t pitch,
+void ff_ivi_dc_haar_2d(const int32_t *in, int16_t *out, ptrdiff_t pitch,
int blk_size)
{
int x, y;
@@ -523,7 +523,7 @@ void ff_ivi_dc_haar_2d(const int32_t *in, int16_t *out, uint32_t pitch,
d3 = COMPENSATE(t3);\
d4 = COMPENSATE(t4);}
-void ff_ivi_inverse_slant_8x8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
+void ff_ivi_inverse_slant_8x8(const int32_t *in, int16_t *out, ptrdiff_t pitch, const uint8_t *flags)
{
int i;
const int32_t *src;
@@ -563,7 +563,7 @@ void ff_ivi_inverse_slant_8x8(const int32_t *in, int16_t *out, uint32_t pitch, c
#undef COMPENSATE
}
-void ff_ivi_inverse_slant_4x4(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
+void ff_ivi_inverse_slant_4x4(const int32_t *in, int16_t *out, ptrdiff_t pitch, const uint8_t *flags)
{
int i;
const int32_t *src;
@@ -603,7 +603,7 @@ void ff_ivi_inverse_slant_4x4(const int32_t *in, int16_t *out, uint32_t pitch, c
#undef COMPENSATE
}
-void ff_ivi_dc_slant_2d(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
+void ff_ivi_dc_slant_2d(const int32_t *in, int16_t *out, ptrdiff_t pitch, int blk_size)
{
int x, y;
int16_t dc_coeff;
@@ -616,7 +616,7 @@ void ff_ivi_dc_slant_2d(const int32_t *in, int16_t *out, uint32_t pitch, int blk
}
}
-void ff_ivi_row_slant8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
+void ff_ivi_row_slant8(const int32_t *in, int16_t *out, ptrdiff_t pitch, const uint8_t *flags)
{
int i;
int t0, t1, t2, t3, t4, t5, t6, t7, t8;
@@ -636,7 +636,7 @@ void ff_ivi_row_slant8(const int32_t *in, int16_t *out, uint32_t pitch, const ui
#undef COMPENSATE
}
-void ff_ivi_dc_row_slant(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
+void ff_ivi_dc_row_slant(const int32_t *in, int16_t *out, ptrdiff_t pitch, int blk_size)
{
int x, y;
int16_t dc_coeff;
@@ -654,7 +654,7 @@ void ff_ivi_dc_row_slant(const int32_t *in, int16_t *out, uint32_t pitch, int bl
}
}
-void ff_ivi_col_slant8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
+void ff_ivi_col_slant8(const int32_t *in, int16_t *out, ptrdiff_t pitch, const uint8_t *flags)
{
int i, row2, row4, row8;
int t0, t1, t2, t3, t4, t5, t6, t7, t8;
@@ -681,7 +681,7 @@ void ff_ivi_col_slant8(const int32_t *in, int16_t *out, uint32_t pitch, const ui
#undef COMPENSATE
}
-void ff_ivi_dc_col_slant(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
+void ff_ivi_dc_col_slant(const int32_t *in, int16_t *out, ptrdiff_t pitch, int blk_size)
{
int x, y;
int16_t dc_coeff;
@@ -695,7 +695,7 @@ void ff_ivi_dc_col_slant(const int32_t *in, int16_t *out, uint32_t pitch, int bl
}
}
-void ff_ivi_row_slant4(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
+void ff_ivi_row_slant4(const int32_t *in, int16_t *out, ptrdiff_t pitch, const uint8_t *flags)
{
int i;
int t0, t1, t2, t3, t4;
@@ -715,7 +715,7 @@ void ff_ivi_row_slant4(const int32_t *in, int16_t *out, uint32_t pitch, const ui
#undef COMPENSATE
}
-void ff_ivi_col_slant4(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
+void ff_ivi_col_slant4(const int32_t *in, int16_t *out, ptrdiff_t pitch, const uint8_t *flags)
{
int i, row2;
int t0, t1, t2, t3, t4;
@@ -738,7 +738,7 @@ void ff_ivi_col_slant4(const int32_t *in, int16_t *out, uint32_t pitch, const ui
#undef COMPENSATE
}
-void ff_ivi_put_pixels_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
+void ff_ivi_put_pixels_8x8(const int32_t *in, int16_t *out, ptrdiff_t pitch,
const uint8_t *flags)
{
int x, y;
@@ -748,7 +748,7 @@ void ff_ivi_put_pixels_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
out[x] = in[x];
}
-void ff_ivi_put_dc_pixel_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
+void ff_ivi_put_dc_pixel_8x8(const int32_t *in, int16_t *out, ptrdiff_t pitch,
int blk_size)
{
int y;
@@ -763,9 +763,9 @@ void ff_ivi_put_dc_pixel_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
#define IVI_MC_TEMPLATE(size, suffix, OP) \
static void ivi_mc_ ## size ##x## size ## suffix(int16_t *buf, \
- uint32_t dpitch, \
+ ptrdiff_t dpitch, \
const int16_t *ref_buf, \
- uint32_t pitch, int mc_type) \
+ ptrdiff_t pitch, int mc_type) \
{ \
int i, j; \
const int16_t *wptr; \
@@ -799,7 +799,7 @@ static void ivi_mc_ ## size ##x## size ## suffix(int16_t *buf, \
} \
\
void ff_ivi_mc_ ## size ##x## size ## suffix(int16_t *buf, const int16_t *ref_buf, \
- uint32_t pitch, int mc_type) \
+ ptrdiff_t pitch, int mc_type) \
{ \
ivi_mc_ ## size ##x## size ## suffix(buf, pitch, ref_buf, pitch, mc_type); \
} \
@@ -808,7 +808,7 @@ void ff_ivi_mc_ ## size ##x## size ## suffix(int16_t *buf, const int16_t *ref_bu
void ff_ivi_mc_avg_ ## size ##x## size ## suffix(int16_t *buf, \
const int16_t *ref_buf, \
const int16_t *ref_buf2, \
- uint32_t pitch, \
+ ptrdiff_t pitch, \
int mc_type, int mc_type2) \
{ \
int16_t tmp[size * size]; \