[FFmpeg-devel] [PATCH 1/2] alpha/idctdsp: move disabled cruft out of dsputil
James Almer
jamrial at gmail.com
Fri Jul 18 04:45:34 CEST 2014
Signed-off-by: James Almer <jamrial at gmail.com>
---
libavcodec/alpha/dsputil_alpha.c | 72 ----------------------------------------
libavcodec/alpha/idctdsp_alpha.c | 72 ++++++++++++++++++++++++++++++++++++++++
2 files changed, 72 insertions(+), 72 deletions(-)
diff --git a/libavcodec/alpha/dsputil_alpha.c b/libavcodec/alpha/dsputil_alpha.c
index d99a74b..e00497d 100644
--- a/libavcodec/alpha/dsputil_alpha.c
+++ b/libavcodec/alpha/dsputil_alpha.c
@@ -24,78 +24,6 @@
#include "dsputil_alpha.h"
#include "asm.h"
-#if 0
-/* These functions were the base for the optimized assembler routines,
- and remain here for documentation purposes. */
-static void put_pixels_clamped_mvi(const int16_t *block, uint8_t *pixels,
- ptrdiff_t line_size)
-{
- int i = 8;
- uint64_t clampmask = zap(-1, 0xaa); /* 0x00ff00ff00ff00ff */
-
- do {
- uint64_t shorts0, shorts1;
-
- shorts0 = ldq(block);
- shorts0 = maxsw4(shorts0, 0);
- shorts0 = minsw4(shorts0, clampmask);
- stl(pkwb(shorts0), pixels);
-
- shorts1 = ldq(block + 4);
- shorts1 = maxsw4(shorts1, 0);
- shorts1 = minsw4(shorts1, clampmask);
- stl(pkwb(shorts1), pixels + 4);
-
- pixels += line_size;
- block += 8;
- } while (--i);
-}
-
-void add_pixels_clamped_mvi(const int16_t *block, uint8_t *pixels,
- ptrdiff_t line_size)
-{
- int h = 8;
- /* Keep this function a leaf function by generating the constants
- manually (mainly for the hack value ;-). */
- uint64_t clampmask = zap(-1, 0xaa); /* 0x00ff00ff00ff00ff */
- uint64_t signmask = zap(-1, 0x33);
- signmask ^= signmask >> 1; /* 0x8000800080008000 */
-
- do {
- uint64_t shorts0, pix0, signs0;
- uint64_t shorts1, pix1, signs1;
-
- shorts0 = ldq(block);
- shorts1 = ldq(block + 4);
-
- pix0 = unpkbw(ldl(pixels));
- /* Signed subword add (MMX paddw). */
- signs0 = shorts0 & signmask;
- shorts0 &= ~signmask;
- shorts0 += pix0;
- shorts0 ^= signs0;
- /* Clamp. */
- shorts0 = maxsw4(shorts0, 0);
- shorts0 = minsw4(shorts0, clampmask);
-
- /* Next 4. */
- pix1 = unpkbw(ldl(pixels + 4));
- signs1 = shorts1 & signmask;
- shorts1 &= ~signmask;
- shorts1 += pix1;
- shorts1 ^= signs1;
- shorts1 = maxsw4(shorts1, 0);
- shorts1 = minsw4(shorts1, clampmask);
-
- stl(pkwb(shorts0), pixels);
- stl(pkwb(shorts1), pixels + 4);
-
- pixels += line_size;
- block += 8;
- } while (--h);
-}
-#endif
-
av_cold void ff_dsputil_init_alpha(DSPContext *c, AVCodecContext *avctx)
{
/* amask clears all bits that correspond to present features. */
diff --git a/libavcodec/alpha/idctdsp_alpha.c b/libavcodec/alpha/idctdsp_alpha.c
index 73f4821..1050697 100644
--- a/libavcodec/alpha/idctdsp_alpha.c
+++ b/libavcodec/alpha/idctdsp_alpha.c
@@ -33,6 +33,78 @@ void (*put_pixels_clamped_axp_p)(const int16_t *block, uint8_t *pixels,
void (*add_pixels_clamped_axp_p)(const int16_t *block, uint8_t *pixels,
int line_size);
+#if 0
+/* These functions were the base for the optimized assembler routines,
+ and remain here for documentation purposes. */
+static void put_pixels_clamped_mvi(const int16_t *block, uint8_t *pixels,
+ ptrdiff_t line_size)
+{
+ int i = 8;
+ uint64_t clampmask = zap(-1, 0xaa); /* 0x00ff00ff00ff00ff */
+
+ do {
+ uint64_t shorts0, shorts1;
+
+ shorts0 = ldq(block);
+ shorts0 = maxsw4(shorts0, 0);
+ shorts0 = minsw4(shorts0, clampmask);
+ stl(pkwb(shorts0), pixels);
+
+ shorts1 = ldq(block + 4);
+ shorts1 = maxsw4(shorts1, 0);
+ shorts1 = minsw4(shorts1, clampmask);
+ stl(pkwb(shorts1), pixels + 4);
+
+ pixels += line_size;
+ block += 8;
+ } while (--i);
+}
+
+void add_pixels_clamped_mvi(const int16_t *block, uint8_t *pixels,
+ ptrdiff_t line_size)
+{
+ int h = 8;
+ /* Keep this function a leaf function by generating the constants
+ manually (mainly for the hack value ;-). */
+ uint64_t clampmask = zap(-1, 0xaa); /* 0x00ff00ff00ff00ff */
+ uint64_t signmask = zap(-1, 0x33);
+ signmask ^= signmask >> 1; /* 0x8000800080008000 */
+
+ do {
+ uint64_t shorts0, pix0, signs0;
+ uint64_t shorts1, pix1, signs1;
+
+ shorts0 = ldq(block);
+ shorts1 = ldq(block + 4);
+
+ pix0 = unpkbw(ldl(pixels));
+ /* Signed subword add (MMX paddw). */
+ signs0 = shorts0 & signmask;
+ shorts0 &= ~signmask;
+ shorts0 += pix0;
+ shorts0 ^= signs0;
+ /* Clamp. */
+ shorts0 = maxsw4(shorts0, 0);
+ shorts0 = minsw4(shorts0, clampmask);
+
+ /* Next 4. */
+ pix1 = unpkbw(ldl(pixels + 4));
+ signs1 = shorts1 & signmask;
+ shorts1 &= ~signmask;
+ shorts1 += pix1;
+ shorts1 ^= signs1;
+ shorts1 = maxsw4(shorts1, 0);
+ shorts1 = minsw4(shorts1, clampmask);
+
+ stl(pkwb(shorts0), pixels);
+ stl(pkwb(shorts1), pixels + 4);
+
+ pixels += line_size;
+ block += 8;
+ } while (--h);
+}
+#endif
+
av_cold void ff_idctdsp_init_alpha(IDCTDSPContext *c, AVCodecContext *avctx,
unsigned high_bit_depth)
{
--
1.8.5.5
More information about the ffmpeg-devel
mailing list