Commit 95a2d705 authored by James Almer's avatar James Almer Committed by Michael Niedermayer

alpha/idctdsp: move disabled cruft out of dsputil

Signed-off-by: 's avatarJames Almer <jamrial@gmail.com>
Signed-off-by: 's avatarMichael Niedermayer <michaelni@gmx.at>
parent d5ba19e6
......@@ -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. */
......
......@@ -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)
{
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment