aboutsummaryrefslogtreecommitdiff
path: root/files/source/scale.cc
diff options
context:
space:
mode:
Diffstat (limited to 'files/source/scale.cc')
-rw-r--r--files/source/scale.cc106
1 files changed, 82 insertions, 24 deletions
diff --git a/files/source/scale.cc b/files/source/scale.cc
index e1335f1e..80b030dc 100644
--- a/files/source/scale.cc
+++ b/files/source/scale.cc
@@ -198,6 +198,51 @@ static void ScalePlaneDown2_16(int src_width,
}
}
+void ScalePlaneDown2_16To8(int src_width,
+ int src_height,
+ int dst_width,
+ int dst_height,
+ int src_stride,
+ int dst_stride,
+ const uint16_t* src_ptr,
+ uint8_t* dst_ptr,
+ int scale,
+ enum FilterMode filtering) {
+ int y;
+ void (*ScaleRowDown2)(const uint16_t* src_ptr, ptrdiff_t src_stride,
+ uint8_t* dst_ptr, int dst_width, int scale) =
+ (src_width & 1)
+ ? (filtering == kFilterNone
+ ? ScaleRowDown2_16To8_Odd_C
+ : (filtering == kFilterLinear ? ScaleRowDown2Linear_16To8_Odd_C
+ : ScaleRowDown2Box_16To8_Odd_C))
+ : (filtering == kFilterNone
+ ? ScaleRowDown2_16To8_C
+ : (filtering == kFilterLinear ? ScaleRowDown2Linear_16To8_C
+ : ScaleRowDown2Box_16To8_C));
+ int row_stride = src_stride * 2;
+ (void)dst_height;
+ if (!filtering) {
+ src_ptr += src_stride; // Point to odd rows.
+ src_stride = 0;
+ }
+
+ if (filtering == kFilterLinear) {
+ src_stride = 0;
+ }
+ for (y = 0; y < src_height / 2; ++y) {
+ ScaleRowDown2(src_ptr, src_stride, dst_ptr, dst_width, scale);
+ src_ptr += row_stride;
+ dst_ptr += dst_stride;
+ }
+ if (src_height & 1) {
+ if (!filtering) {
+ src_ptr -= src_stride; // Point to last row.
+ }
+ ScaleRowDown2(src_ptr, 0, dst_ptr, dst_width, scale);
+ }
+}
+
// Scale plane, 1/4
// This is an optimized version for scaling down a plane to 1/4 of
// its original size.
@@ -775,9 +820,11 @@ static void ScaleAddCols2_C(int dst_width,
int ix = x >> 16;
x += dx;
boxwidth = MIN1((x >> 16) - ix);
- *dst_ptr++ =
- SumPixels(boxwidth, src_ptr + ix) * scaletbl[boxwidth - minboxwidth] >>
- 16;
+ int scaletbl_index = boxwidth - minboxwidth;
+ assert((scaletbl_index == 0) || (scaletbl_index == 1));
+ *dst_ptr++ = (uint8_t)(SumPixels(boxwidth, src_ptr + ix) *
+ scaletbl[scaletbl_index] >>
+ 16);
}
}
@@ -797,9 +844,10 @@ static void ScaleAddCols2_16_C(int dst_width,
int ix = x >> 16;
x += dx;
boxwidth = MIN1((x >> 16) - ix);
- *dst_ptr++ = SumPixels_16(boxwidth, src_ptr + ix) *
- scaletbl[boxwidth - minboxwidth] >>
- 16;
+ int scaletbl_index = boxwidth - minboxwidth;
+ assert((scaletbl_index == 0) || (scaletbl_index == 1));
+ *dst_ptr++ =
+ SumPixels_16(boxwidth, src_ptr + ix) * scaletbl[scaletbl_index] >> 16;
}
}
@@ -814,7 +862,7 @@ static void ScaleAddCols0_C(int dst_width,
(void)dx;
src_ptr += (x >> 16);
for (i = 0; i < dst_width; ++i) {
- *dst_ptr++ = src_ptr[i] * scaleval >> 16;
+ *dst_ptr++ = (uint8_t)(src_ptr[i] * scaleval >> 16);
}
}
@@ -829,7 +877,7 @@ static void ScaleAddCols1_C(int dst_width,
int i;
x >>= 16;
for (i = 0; i < dst_width; ++i) {
- *dst_ptr++ = SumPixels(boxwidth, src_ptr + x) * scaleval >> 16;
+ *dst_ptr++ = (uint8_t)(SumPixels(boxwidth, src_ptr + x) * scaleval >> 16);
x += boxwidth;
}
}
@@ -1020,10 +1068,10 @@ void ScalePlaneBilinearDown(int src_width,
const int max_y = (src_height - 1) << 16;
int j;
- void (*ScaleFilterCols)(uint8_t * dst_ptr, const uint8_t* src_ptr,
+ void (*ScaleFilterCols)(uint8_t* dst_ptr, const uint8_t* src_ptr,
int dst_width, int x, int dx) =
(src_width >= 32768) ? ScaleFilterCols64_C : ScaleFilterCols_C;
- void (*InterpolateRow)(uint8_t * dst_ptr, const uint8_t* src_ptr,
+ void (*InterpolateRow)(uint8_t* dst_ptr, const uint8_t* src_ptr,
ptrdiff_t src_stride, int dst_width,
int source_y_fraction) = InterpolateRow_C;
ScaleSlope(src_width, src_height, dst_width, dst_height, filtering, &x, &y,
@@ -1070,6 +1118,11 @@ void ScalePlaneBilinearDown(int src_width,
}
}
#endif
+#if defined(HAS_INTERPOLATEROW_RVV)
+ if (TestCpuFlag(kCpuHasRVV)) {
+ InterpolateRow = InterpolateRow_RVV;
+ }
+#endif
#if defined(HAS_SCALEFILTERCOLS_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3) && src_width < 32768) {
@@ -1143,10 +1196,10 @@ void ScalePlaneBilinearDown_16(int src_width,
const int max_y = (src_height - 1) << 16;
int j;
- void (*ScaleFilterCols)(uint16_t * dst_ptr, const uint16_t* src_ptr,
+ void (*ScaleFilterCols)(uint16_t* dst_ptr, const uint16_t* src_ptr,
int dst_width, int x, int dx) =
(src_width >= 32768) ? ScaleFilterCols64_16_C : ScaleFilterCols_16_C;
- void (*InterpolateRow)(uint16_t * dst_ptr, const uint16_t* src_ptr,
+ void (*InterpolateRow)(uint16_t* dst_ptr, const uint16_t* src_ptr,
ptrdiff_t src_stride, int dst_width,
int source_y_fraction) = InterpolateRow_16_C;
ScaleSlope(src_width, src_height, dst_width, dst_height, filtering, &x, &y,
@@ -1231,10 +1284,10 @@ void ScalePlaneBilinearUp(int src_width,
int dx = 0;
int dy = 0;
const int max_y = (src_height - 1) << 16;
- void (*InterpolateRow)(uint8_t * dst_ptr, const uint8_t* src_ptr,
+ void (*InterpolateRow)(uint8_t* dst_ptr, const uint8_t* src_ptr,
ptrdiff_t src_stride, int dst_width,
int source_y_fraction) = InterpolateRow_C;
- void (*ScaleFilterCols)(uint8_t * dst_ptr, const uint8_t* src_ptr,
+ void (*ScaleFilterCols)(uint8_t* dst_ptr, const uint8_t* src_ptr,
int dst_width, int x, int dx) =
filtering ? ScaleFilterCols_C : ScaleCols_C;
ScaleSlope(src_width, src_height, dst_width, dst_height, filtering, &x, &y,
@@ -1265,6 +1318,11 @@ void ScalePlaneBilinearUp(int src_width,
}
}
#endif
+#if defined(HAS_INTERPOLATEROW_RVV)
+ if (TestCpuFlag(kCpuHasRVV)) {
+ InterpolateRow = InterpolateRow_RVV;
+ }
+#endif
if (filtering && src_width >= 32768) {
ScaleFilterCols = ScaleFilterCols64_C;
@@ -1315,11 +1373,11 @@ void ScalePlaneBilinearUp(int src_width,
const uint8_t* src = src_ptr + yi * (int64_t)src_stride;
// Allocate 2 row buffers.
- const int kRowSize = (dst_width + 31) & ~31;
- align_buffer_64(row, kRowSize * 2);
+ const int row_size = (dst_width + 31) & ~31;
+ align_buffer_64(row, row_size * 2);
uint8_t* rowptr = row;
- int rowstride = kRowSize;
+ int rowstride = row_size;
int lasty = yi;
ScaleFilterCols(rowptr, src, dst_width, x, dx);
@@ -1699,10 +1757,10 @@ void ScalePlaneBilinearUp_16(int src_width,
int dx = 0;
int dy = 0;
const int max_y = (src_height - 1) << 16;
- void (*InterpolateRow)(uint16_t * dst_ptr, const uint16_t* src_ptr,
+ void (*InterpolateRow)(uint16_t* dst_ptr, const uint16_t* src_ptr,
ptrdiff_t src_stride, int dst_width,
int source_y_fraction) = InterpolateRow_16_C;
- void (*ScaleFilterCols)(uint16_t * dst_ptr, const uint16_t* src_ptr,
+ void (*ScaleFilterCols)(uint16_t* dst_ptr, const uint16_t* src_ptr,
int dst_width, int x, int dx) =
filtering ? ScaleFilterCols_16_C : ScaleCols_16_C;
ScaleSlope(src_width, src_height, dst_width, dst_height, filtering, &x, &y,
@@ -1766,11 +1824,11 @@ void ScalePlaneBilinearUp_16(int src_width,
const uint16_t* src = src_ptr + yi * (int64_t)src_stride;
// Allocate 2 row buffers.
- const int kRowSize = (dst_width + 31) & ~31;
- align_buffer_64(row, kRowSize * 4);
+ const int row_size = (dst_width + 31) & ~31;
+ align_buffer_64(row, row_size * 4);
uint16_t* rowptr = (uint16_t*)row;
- int rowstride = kRowSize;
+ int rowstride = row_size;
int lasty = yi;
ScaleFilterCols(rowptr, src, dst_width, x, dx);
@@ -1827,7 +1885,7 @@ static void ScalePlaneSimple(int src_width,
const uint8_t* src_ptr,
uint8_t* dst_ptr) {
int i;
- void (*ScaleCols)(uint8_t * dst_ptr, const uint8_t* src_ptr, int dst_width,
+ void (*ScaleCols)(uint8_t* dst_ptr, const uint8_t* src_ptr, int dst_width,
int x, int dx) = ScaleCols_C;
// Initial source x/y coordinate and step values as 16.16 fixed point.
int x = 0;
@@ -1864,7 +1922,7 @@ static void ScalePlaneSimple_16(int src_width,
const uint16_t* src_ptr,
uint16_t* dst_ptr) {
int i;
- void (*ScaleCols)(uint16_t * dst_ptr, const uint16_t* src_ptr, int dst_width,
+ void (*ScaleCols)(uint16_t* dst_ptr, const uint16_t* src_ptr, int dst_width,
int x, int dx) = ScaleCols_16_C;
// Initial source x/y coordinate and step values as 16.16 fixed point.
int x = 0;
generated by cgit v1.2.3 (git 2.25.1) at 2024-05-28 16:18:06 +0000