Index: source/libvpx/vp9/common/x86/vp9_asm_stubs.c |
=================================================================== |
--- source/libvpx/vp9/common/x86/vp9_asm_stubs.c (revision 232232) |
+++ source/libvpx/vp9/common/x86/vp9_asm_stubs.c (working copy) |
@@ -36,91 +36,29 @@ |
{ 8, 8, 8, 8, 120, 120, 120, 120 } |
}; |
+typedef void filter8_1dfunction ( |
+ const unsigned char *src_ptr, |
+ const unsigned int src_pitch, |
+ unsigned char *output_ptr, |
+ unsigned int out_pitch, |
+ unsigned int output_height, |
+ const short *filter |
+); |
+ |
#if HAVE_SSSE3 |
-void vp9_filter_block1d16_v8_ssse3(const unsigned char *src_ptr, |
- const unsigned int src_pitch, |
- unsigned char *output_ptr, |
- unsigned int out_pitch, |
- unsigned int output_height, |
- const short *filter); |
+filter8_1dfunction vp9_filter_block1d16_v8_ssse3; |
+filter8_1dfunction vp9_filter_block1d16_h8_ssse3; |
+filter8_1dfunction vp9_filter_block1d8_v8_ssse3; |
+filter8_1dfunction vp9_filter_block1d8_h8_ssse3; |
+filter8_1dfunction vp9_filter_block1d4_v8_ssse3; |
+filter8_1dfunction vp9_filter_block1d4_h8_ssse3; |
+filter8_1dfunction vp9_filter_block1d16_v8_avg_ssse3; |
+filter8_1dfunction vp9_filter_block1d16_h8_avg_ssse3; |
+filter8_1dfunction vp9_filter_block1d8_v8_avg_ssse3; |
+filter8_1dfunction vp9_filter_block1d8_h8_avg_ssse3; |
+filter8_1dfunction vp9_filter_block1d4_v8_avg_ssse3; |
+filter8_1dfunction vp9_filter_block1d4_h8_avg_ssse3; |
-void vp9_filter_block1d16_h8_ssse3(const unsigned char *src_ptr, |
- const unsigned int src_pitch, |
- unsigned char *output_ptr, |
- unsigned int out_pitch, |
- unsigned int output_height, |
- const short *filter); |
- |
-void vp9_filter_block1d8_v8_ssse3(const unsigned char *src_ptr, |
- const unsigned int src_pitch, |
- unsigned char *output_ptr, |
- unsigned int out_pitch, |
- unsigned int output_height, |
- const short *filter); |
- |
-void vp9_filter_block1d8_h8_ssse3(const unsigned char *src_ptr, |
- const unsigned int src_pitch, |
- unsigned char *output_ptr, |
- unsigned int out_pitch, |
- unsigned int output_height, |
- const short *filter); |
- |
-void vp9_filter_block1d4_v8_ssse3(const unsigned char *src_ptr, |
- const unsigned int src_pitch, |
- unsigned char *output_ptr, |
- unsigned int out_pitch, |
- unsigned int output_height, |
- const short *filter); |
- |
-void vp9_filter_block1d4_h8_ssse3(const unsigned char *src_ptr, |
- const unsigned int src_pitch, |
- unsigned char *output_ptr, |
- unsigned int out_pitch, |
- unsigned int output_height, |
- const short *filter); |
- |
-void vp9_filter_block1d16_v8_avg_ssse3(const unsigned char *src_ptr, |
- const unsigned int src_pitch, |
- unsigned char *output_ptr, |
- unsigned int out_pitch, |
- unsigned int output_height, |
- const short *filter); |
- |
-void vp9_filter_block1d16_h8_avg_ssse3(const unsigned char *src_ptr, |
- const unsigned int src_pitch, |
- unsigned char *output_ptr, |
- unsigned int out_pitch, |
- unsigned int output_height, |
- const short *filter); |
- |
-void vp9_filter_block1d8_v8_avg_ssse3(const unsigned char *src_ptr, |
- const unsigned int src_pitch, |
- unsigned char *output_ptr, |
- unsigned int out_pitch, |
- unsigned int output_height, |
- const short *filter); |
- |
-void vp9_filter_block1d8_h8_avg_ssse3(const unsigned char *src_ptr, |
- const unsigned int src_pitch, |
- unsigned char *output_ptr, |
- unsigned int out_pitch, |
- unsigned int output_height, |
- const short *filter); |
- |
-void vp9_filter_block1d4_v8_avg_ssse3(const unsigned char *src_ptr, |
- const unsigned int src_pitch, |
- unsigned char *output_ptr, |
- unsigned int out_pitch, |
- unsigned int output_height, |
- const short *filter); |
- |
-void vp9_filter_block1d4_h8_avg_ssse3(const unsigned char *src_ptr, |
- const unsigned int src_pitch, |
- unsigned char *output_ptr, |
- unsigned int out_pitch, |
- unsigned int output_height, |
- const short *filter); |
- |
void vp9_convolve8_horiz_ssse3(const uint8_t *src, ptrdiff_t src_stride, |
uint8_t *dst, ptrdiff_t dst_stride, |
const int16_t *filter_x, int x_step_q4, |
@@ -279,7 +217,7 @@ |
const int16_t *filter_x, int x_step_q4, |
const int16_t *filter_y, int y_step_q4, |
int w, int h) { |
- DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 64*71); |
+ DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 64 * 71); |
assert(w <= 64); |
assert(h <= 64); |
@@ -300,7 +238,7 @@ |
const int16_t *filter_x, int x_step_q4, |
const int16_t *filter_y, int y_step_q4, |
int w, int h) { |
- DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 64*71); |
+ DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 64 * 71); |
assert(w <= 64); |
assert(h <= 64); |
@@ -317,3 +255,214 @@ |
} |
} |
#endif |
+ |
+#if HAVE_SSE2 |
+filter8_1dfunction vp9_filter_block1d16_v8_sse2; |
+filter8_1dfunction vp9_filter_block1d16_h8_sse2; |
+filter8_1dfunction vp9_filter_block1d8_v8_sse2; |
+filter8_1dfunction vp9_filter_block1d8_h8_sse2; |
+filter8_1dfunction vp9_filter_block1d4_v8_sse2; |
+filter8_1dfunction vp9_filter_block1d4_h8_sse2; |
+filter8_1dfunction vp9_filter_block1d16_v8_avg_sse2; |
+filter8_1dfunction vp9_filter_block1d16_h8_avg_sse2; |
+filter8_1dfunction vp9_filter_block1d8_v8_avg_sse2; |
+filter8_1dfunction vp9_filter_block1d8_h8_avg_sse2; |
+filter8_1dfunction vp9_filter_block1d4_v8_avg_sse2; |
+filter8_1dfunction vp9_filter_block1d4_h8_avg_sse2; |
+ |
+void vp9_convolve8_horiz_sse2(const uint8_t *src, ptrdiff_t src_stride, |
+ uint8_t *dst, ptrdiff_t dst_stride, |
+ const int16_t *filter_x, int x_step_q4, |
+ const int16_t *filter_y, int y_step_q4, |
+ int w, int h) { |
+ /* Ensure the filter can be compressed to int16_t. */ |
+ if (x_step_q4 == 16 && filter_x[3] != 128) { |
+ while (w >= 16) { |
+ vp9_filter_block1d16_h8_sse2(src, src_stride, |
+ dst, dst_stride, |
+ h, filter_x); |
+ src += 16; |
+ dst += 16; |
+ w -= 16; |
+ } |
+ while (w >= 8) { |
+ vp9_filter_block1d8_h8_sse2(src, src_stride, |
+ dst, dst_stride, |
+ h, filter_x); |
+ src += 8; |
+ dst += 8; |
+ w -= 8; |
+ } |
+ while (w >= 4) { |
+ vp9_filter_block1d4_h8_sse2(src, src_stride, |
+ dst, dst_stride, |
+ h, filter_x); |
+ src += 4; |
+ dst += 4; |
+ w -= 4; |
+ } |
+ } |
+ if (w) { |
+ vp9_convolve8_horiz_c(src, src_stride, dst, dst_stride, |
+ filter_x, x_step_q4, filter_y, y_step_q4, |
+ w, h); |
+ } |
+} |
+ |
+void vp9_convolve8_vert_sse2(const uint8_t *src, ptrdiff_t src_stride, |
+ uint8_t *dst, ptrdiff_t dst_stride, |
+ const int16_t *filter_x, int x_step_q4, |
+ const int16_t *filter_y, int y_step_q4, |
+ int w, int h) { |
+ if (y_step_q4 == 16 && filter_y[3] != 128) { |
+ while (w >= 16) { |
+ vp9_filter_block1d16_v8_sse2(src - src_stride * 3, src_stride, |
+ dst, dst_stride, |
+ h, filter_y); |
+ src += 16; |
+ dst += 16; |
+ w -= 16; |
+ } |
+ while (w >= 8) { |
+ vp9_filter_block1d8_v8_sse2(src - src_stride * 3, src_stride, |
+ dst, dst_stride, |
+ h, filter_y); |
+ src += 8; |
+ dst += 8; |
+ w -= 8; |
+ } |
+ while (w >= 4) { |
+ vp9_filter_block1d4_v8_sse2(src - src_stride * 3, src_stride, |
+ dst, dst_stride, |
+ h, filter_y); |
+ src += 4; |
+ dst += 4; |
+ w -= 4; |
+ } |
+ } |
+ if (w) { |
+ vp9_convolve8_vert_c(src, src_stride, dst, dst_stride, |
+ filter_x, x_step_q4, filter_y, y_step_q4, |
+ w, h); |
+ } |
+} |
+ |
+void vp9_convolve8_avg_horiz_sse2(const uint8_t *src, ptrdiff_t src_stride, |
+ uint8_t *dst, ptrdiff_t dst_stride, |
+ const int16_t *filter_x, int x_step_q4, |
+ const int16_t *filter_y, int y_step_q4, |
+ int w, int h) { |
+ if (x_step_q4 == 16 && filter_x[3] != 128) { |
+ while (w >= 16) { |
+ vp9_filter_block1d16_h8_avg_sse2(src, src_stride, |
+ dst, dst_stride, |
+ h, filter_x); |
+ src += 16; |
+ dst += 16; |
+ w -= 16; |
+ } |
+ while (w >= 8) { |
+ vp9_filter_block1d8_h8_avg_sse2(src, src_stride, |
+ dst, dst_stride, |
+ h, filter_x); |
+ src += 8; |
+ dst += 8; |
+ w -= 8; |
+ } |
+ while (w >= 4) { |
+ vp9_filter_block1d4_h8_avg_sse2(src, src_stride, |
+ dst, dst_stride, |
+ h, filter_x); |
+ src += 4; |
+ dst += 4; |
+ w -= 4; |
+ } |
+ } |
+ if (w) { |
+ vp9_convolve8_avg_horiz_c(src, src_stride, dst, dst_stride, |
+ filter_x, x_step_q4, filter_y, y_step_q4, |
+ w, h); |
+ } |
+} |
+ |
+void vp9_convolve8_avg_vert_sse2(const uint8_t *src, ptrdiff_t src_stride, |
+ uint8_t *dst, ptrdiff_t dst_stride, |
+ const int16_t *filter_x, int x_step_q4, |
+ const int16_t *filter_y, int y_step_q4, |
+ int w, int h) { |
+ if (y_step_q4 == 16 && filter_y[3] != 128) { |
+ while (w >= 16) { |
+ vp9_filter_block1d16_v8_avg_sse2(src - src_stride * 3, src_stride, |
+ dst, dst_stride, |
+ h, filter_y); |
+ src += 16; |
+ dst += 16; |
+ w -= 16; |
+ } |
+ while (w >= 8) { |
+ vp9_filter_block1d8_v8_avg_sse2(src - src_stride * 3, src_stride, |
+ dst, dst_stride, |
+ h, filter_y); |
+ src += 8; |
+ dst += 8; |
+ w -= 8; |
+ } |
+ while (w >= 4) { |
+ vp9_filter_block1d4_v8_avg_sse2(src - src_stride * 3, src_stride, |
+ dst, dst_stride, |
+ h, filter_y); |
+ src += 4; |
+ dst += 4; |
+ w -= 4; |
+ } |
+ } |
+ if (w) { |
+ vp9_convolve8_avg_vert_c(src, src_stride, dst, dst_stride, |
+ filter_x, x_step_q4, filter_y, y_step_q4, |
+ w, h); |
+ } |
+} |
+ |
+void vp9_convolve8_sse2(const uint8_t *src, ptrdiff_t src_stride, |
+ uint8_t *dst, ptrdiff_t dst_stride, |
+ const int16_t *filter_x, int x_step_q4, |
+ const int16_t *filter_y, int y_step_q4, |
+ int w, int h) { |
+ DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 64 * 71); |
+ |
+ assert(w <= 64); |
+ assert(h <= 64); |
+ if (x_step_q4 == 16 && y_step_q4 == 16) { |
+ vp9_convolve8_horiz_sse2(src - 3 * src_stride, src_stride, fdata2, 64, |
+ filter_x, x_step_q4, filter_y, y_step_q4, |
+ w, h + 7); |
+ vp9_convolve8_vert_sse2(fdata2 + 3 * 64, 64, dst, dst_stride, |
+ filter_x, x_step_q4, filter_y, y_step_q4, w, h); |
+ } else { |
+ vp9_convolve8_c(src, src_stride, dst, dst_stride, |
+ filter_x, x_step_q4, filter_y, y_step_q4, w, h); |
+ } |
+} |
+ |
+void vp9_convolve8_avg_sse2(const uint8_t *src, ptrdiff_t src_stride, |
+ uint8_t *dst, ptrdiff_t dst_stride, |
+ const int16_t *filter_x, int x_step_q4, |
+ const int16_t *filter_y, int y_step_q4, |
+ int w, int h) { |
+ DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 64 * 71); |
+ |
+ assert(w <= 64); |
+ assert(h <= 64); |
+ if (x_step_q4 == 16 && y_step_q4 == 16) { |
+ vp9_convolve8_horiz_sse2(src - 3 * src_stride, src_stride, fdata2, 64, |
+ filter_x, x_step_q4, filter_y, y_step_q4, |
+ w, h + 7); |
+ vp9_convolve8_avg_vert_sse2(fdata2 + 3 * 64, 64, dst, dst_stride, |
+ filter_x, x_step_q4, filter_y, y_step_q4, |
+ w, h); |
+ } else { |
+ vp9_convolve8_avg_c(src, src_stride, dst, dst_stride, |
+ filter_x, x_step_q4, filter_y, y_step_q4, w, h); |
+ } |
+} |
+#endif |