diff options
author | Luca Barbato <lu_zero@gentoo.org> | 2017-04-26 15:43:19 +0000 |
---|---|---|
committer | Luca Barbato <lu_zero@gentoo.org> | 2017-04-29 15:47:25 +0200 |
commit | 63860ba7b8aa0133049dddd9b9dd39b033636b79 (patch) | |
tree | 25890c1da6a8becf25d11b2a93b2d4aeb443be35 /vpx_dsp | |
parent | ef5918098d5c7f8ffda960274e3f8e38f02cb487 (diff) | |
download | libvpx-63860ba7b8aa0133049dddd9b9dd39b033636b79.tar libvpx-63860ba7b8aa0133049dddd9b9dd39b033636b79.tar.gz libvpx-63860ba7b8aa0133049dddd9b9dd39b033636b79.tar.bz2 libvpx-63860ba7b8aa0133049dddd9b9dd39b033636b79.zip |
ppc: Add convolve_copy
Change-Id: Ie26d6dbe090e711d84bac01ba7da270db983f405
Diffstat (limited to 'vpx_dsp')
-rw-r--r-- | vpx_dsp/ppc/vpx_convolve_vsx.c | 85 | ||||
-rw-r--r-- | vpx_dsp/vpx_dsp.mk | 2 | ||||
-rw-r--r-- | vpx_dsp/vpx_dsp_rtcd_defs.pl | 2 |
3 files changed, 88 insertions, 1 deletions
diff --git a/vpx_dsp/ppc/vpx_convolve_vsx.c b/vpx_dsp/ppc/vpx_convolve_vsx.c new file mode 100644 index 000000000..32cdbb4d0 --- /dev/null +++ b/vpx_dsp/ppc/vpx_convolve_vsx.c @@ -0,0 +1,85 @@ +/* + * Copyright (c) 2017 The WebM project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ +#include <string.h> +#include "./vpx_dsp_rtcd.h" +#include "vpx_dsp/ppc/types_vsx.h" + +// TODO(lu_zero): unroll +static inline void copy_w16(const uint8_t *src, ptrdiff_t src_stride, + uint8_t *dst, ptrdiff_t dst_stride, int32_t h) { + int i; + + for (i = h; i--;) { + vec_vsx_st(vec_vsx_ld(0, src), 0, dst); + src += src_stride; + dst += dst_stride; + } +} + +static inline void copy_w32(const uint8_t *src, ptrdiff_t src_stride, + uint8_t *dst, ptrdiff_t dst_stride, int32_t h) { + int i; + + for (i = h; i--;) { + vec_vsx_st(vec_vsx_ld(0, src), 0, dst); + vec_vsx_st(vec_vsx_ld(16, src), 16, dst); + src += src_stride; + dst += dst_stride; + } +} + +static inline void copy_w64(const uint8_t *src, ptrdiff_t src_stride, + uint8_t *dst, ptrdiff_t dst_stride, int32_t h) { + int i; + + for (i = h; i--;) { + vec_vsx_st(vec_vsx_ld(0, src), 0, dst); + vec_vsx_st(vec_vsx_ld(16, src), 16, dst); + vec_vsx_st(vec_vsx_ld(32, src), 32, dst); + vec_vsx_st(vec_vsx_ld(48, src), 48, dst); + src += src_stride; + dst += dst_stride; + } +} + +void vpx_convolve_copy_vsx(const uint8_t *src, ptrdiff_t src_stride, + uint8_t *dst, ptrdiff_t dst_stride, + const int16_t *filter_x, int32_t filter_x_stride, + const int16_t *filter_y, int32_t filter_y_stride, + int32_t w, int32_t h) { + (void)filter_x; + (void)filter_y; + (void)filter_x_stride; + (void)filter_y_stride; + + switch (w) { + case 16: { + copy_w16(src, src_stride, dst, dst_stride, h); + break; + } + case 32: { + copy_w32(src, src_stride, dst, dst_stride, h); + break; + } + case 64: { + copy_w64(src, src_stride, dst, dst_stride, h); + break; + } + default: { + int i; + for (i = h; i--;) { + memcpy(dst, src, w); + src += src_stride; + dst += dst_stride; + } + break; + } + } +} diff --git a/vpx_dsp/vpx_dsp.mk b/vpx_dsp/vpx_dsp.mk index 64df59cfe..d1c991782 100644 --- a/vpx_dsp/vpx_dsp.mk +++ b/vpx_dsp/vpx_dsp.mk @@ -144,6 +144,8 @@ DSP_SRCS-$(HAVE_DSPR2) += mips/convolve8_dspr2.c DSP_SRCS-$(HAVE_DSPR2) += mips/convolve8_horiz_dspr2.c DSP_SRCS-$(HAVE_DSPR2) += mips/convolve8_vert_dspr2.c +DSP_SRCS-$(HAVE_VSX) += ppc/vpx_convolve_vsx.c + # loop filters DSP_SRCS-yes += loopfilter.c diff --git a/vpx_dsp/vpx_dsp_rtcd_defs.pl b/vpx_dsp/vpx_dsp_rtcd_defs.pl index 667352a64..0a9327ba6 100644 --- a/vpx_dsp/vpx_dsp_rtcd_defs.pl +++ b/vpx_dsp/vpx_dsp_rtcd_defs.pl @@ -332,7 +332,7 @@ if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") { # Sub Pixel Filters # add_proto qw/void vpx_convolve_copy/, "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"; -specialize qw/vpx_convolve_copy neon dspr2 msa sse2/; +specialize qw/vpx_convolve_copy neon dspr2 msa sse2 vsx/; add_proto qw/void vpx_convolve_avg/, "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"; specialize qw/vpx_convolve_avg neon dspr2 msa sse2/; |