xserver: Branch 'master'
Søren Sandmann Pedersen
sandmann at kemper.freedesktop.org
Tue Jun 5 14:55:57 PDT 2007
fb/fbmmx.c | 2728 ++------------------------------------------------------
fb/fbmmx.h | 220 ----
render/mipict.c | 2
3 files changed, 108 insertions(+), 2842 deletions(-)
New commits:
diff-tree 49ed31c0b323dd8c5887a803c199875e6f2330d8 (from dfbe32b5b828cc4e3da36a0e2e6ad641164eaa5e)
Author: Søren Sandmann Pedersen <sandmann at redhat.com>
Date: Tue Jun 5 17:44:21 2007 -0400
Remove most of the fast-path MMX operations from fbmmx. fbCopyAreammx
and fbSolidFillmmx are still needed by other code.
diff --git a/fb/fbmmx.c b/fb/fbmmx.c
index 53459a0..2e240c1 100644
--- a/fb/fbmmx.c
+++ b/fb/fbmmx.c
@@ -62,2663 +62,149 @@
#define CHECKPOINT()
#endif
-/* Notes about writing mmx code
- *
- * give memory operands as the second operand. If you give it as the
- * first, gcc will first load it into a register, then use that
- * register
- *
- * ie. use
- *
- * _mm_mullo_pi16 (x, mmx_constant);
- *
- * not
- *
- * _mm_mullo_pi16 (mmx_constant, x);
- *
- * Also try to minimize dependencies. i.e. when you need a value, try
- * to calculate it from a value that was calculated as early as
- * possible.
- */
-/* --------------- MMX primitivess ------------------------------------ */
-
-typedef unsigned long long ullong;
-
-#ifdef __GNUC__
-typedef ullong mmxdatafield;
-#endif
-#ifdef _MSC_VER
-typedef unsigned __int64 ullong;
-typedef __m64 mmxdatafield;
-#endif
-
-typedef struct
-{
- mmxdatafield mmx_4x00ff;
- mmxdatafield mmx_4x0080;
- mmxdatafield mmx_565_rgb;
- mmxdatafield mmx_565_unpack_multiplier;
- mmxdatafield mmx_565_r;
- mmxdatafield mmx_565_g;
- mmxdatafield mmx_565_b;
- mmxdatafield mmx_mask_0;
- mmxdatafield mmx_mask_1;
- mmxdatafield mmx_mask_2;
- mmxdatafield mmx_mask_3;
- mmxdatafield mmx_full_alpha;
- mmxdatafield mmx_ffff0000ffff0000;
- mmxdatafield mmx_0000ffff00000000;
- mmxdatafield mmx_000000000000ffff;
-} MMXData;
-
-static const MMXData c =
-{
-#ifdef __GNUC__
- .mmx_4x00ff = 0x00ff00ff00ff00ffULL,
- .mmx_4x0080 = 0x0080008000800080ULL,
- .mmx_565_rgb = 0x000001f0003f001fULL,
- .mmx_565_unpack_multiplier = 0x0000008404100840ULL,
- .mmx_565_r = 0x000000f800000000ULL,
- .mmx_565_g = 0x0000000000fc0000ULL,
- .mmx_565_b = 0x00000000000000f8ULL,
- .mmx_mask_0 = 0xffffffffffff0000ULL,
- .mmx_mask_1 = 0xffffffff0000ffffULL,
- .mmx_mask_2 = 0xffff0000ffffffffULL,
- .mmx_mask_3 = 0x0000ffffffffffffULL,
- .mmx_full_alpha = 0x00ff000000000000ULL,
- .mmx_ffff0000ffff0000 = 0xffff0000ffff0000ULL,
- .mmx_0000ffff00000000 = 0x0000ffff00000000ULL,
- .mmx_000000000000ffff = 0x000000000000ffffULL,
-#endif
-#ifdef _MSC_VER
- { 0x00ff00ff00ff00ffUI64 },
- { 0x0080008000800080UI64 },
- { 0x000001f0003f001fUI64 },
- { 0x0000008404100840UI64 },
- { 0x000000f800000000UI64 },
- { 0x0000000000fc0000UI64 },
- { 0x00000000000000f8UI64 },
- { 0xffffffffffff0000UI64 },
- { 0xffffffff0000ffffUI64 },
- { 0xffff0000ffffffffUI64 },
- { 0x0000ffffffffffffUI64 },
- { 0x00ff000000000000UI64 },
- { 0xffff0000ffff0000UI64 },
- { 0x0000ffff00000000UI64 },
- { 0x000000000000ffffUI64 },
-#endif
-};
-
-#ifdef _MSC_VER
-#undef inline
-#define inline __forceinline
-#endif
-
-#ifdef __GNUC__
-#define MC(x) ((__m64) c.mmx_##x)
-#endif
-#ifdef _MSC_VER
-#define MC(x) c.mmx_##x
-#endif
-
-static inline __m64
-shift (__m64 v, int s)
-{
- if (s > 0)
- return _mm_slli_si64 (v, s);
- else if (s < 0)
- return _mm_srli_si64 (v, -s);
- else
- return v;
-}
-
-static inline __m64
-negate (__m64 mask)
-{
- return _mm_xor_si64 (mask, MC(4x00ff));
-}
-
-static inline __m64
-pix_multiply (__m64 a, __m64 b)
-{
- __m64 res;
-
- res = _mm_mullo_pi16 (a, b);
- res = _mm_adds_pu16 (res, MC(4x0080));
- res = _mm_adds_pu16 (res, _mm_srli_pi16 (res, 8));
- res = _mm_srli_pi16 (res, 8);
-
- return res;
-}
-
-static inline __m64
-pix_add (__m64 a, __m64 b)
-{
- return _mm_adds_pu8 (a, b);
-}
-
-#ifdef USE_SSE
-
-static inline __m64
-expand_alpha (__m64 pixel)
-{
- return _mm_shuffle_pi16 (pixel, _MM_SHUFFLE(3, 3, 3, 3));
-}
-
-static inline __m64
-expand_alpha_rev (__m64 pixel)
-{
- return _mm_shuffle_pi16 (pixel, _MM_SHUFFLE(0, 0, 0, 0));
-}
-
-static inline __m64
-invert_colors (__m64 pixel)
-{
- return _mm_shuffle_pi16 (pixel, _MM_SHUFFLE(3, 0, 1, 2));
-}
-
-#else
-
-static inline __m64
-expand_alpha (__m64 pixel)
-{
- __m64 t1, t2;
-
- t1 = shift (pixel, -48);
- t2 = shift (t1, 16);
- t1 = _mm_or_si64 (t1, t2);
- t2 = shift (t1, 32);
- t1 = _mm_or_si64 (t1, t2);
-
- return t1;
-}
-
-static inline __m64
-expand_alpha_rev (__m64 pixel)
-{
- __m64 t1, t2;
-
- /* move alpha to low 16 bits and zero the rest */
- t1 = shift (pixel, 48);
- t1 = shift (t1, -48);
-
- t2 = shift (t1, 16);
- t1 = _mm_or_si64 (t1, t2);
- t2 = shift (t1, 32);
- t1 = _mm_or_si64 (t1, t2);
-
- return t1;
-}
-
-static inline __m64
-invert_colors (__m64 pixel)
-{
- __m64 x, y, z;
-
- x = y = z = pixel;
-
- x = _mm_and_si64 (x, MC(ffff0000ffff0000));
- y = _mm_and_si64 (y, MC(000000000000ffff));
- z = _mm_and_si64 (z, MC(0000ffff00000000));
-
- y = shift (y, 32);
- z = shift (z, -32);
-
- x = _mm_or_si64 (x, y);
- x = _mm_or_si64 (x, z);
-
- return x;
-}
-
-#endif
-
-static inline __m64
-over (__m64 src, __m64 srca, __m64 dest)
-{
- return _mm_adds_pu8 (src, pix_multiply(dest, negate(srca)));
-}
-
-static inline __m64
-over_rev_non_pre (__m64 src, __m64 dest)
-{
- __m64 srca = expand_alpha (src);
- __m64 srcfaaa = _mm_or_si64 (srca, MC(full_alpha));
-
- return over(pix_multiply(invert_colors(src), srcfaaa), srca, dest);
-}
-
-static inline __m64
-in (__m64 src,
- __m64 mask)
-{
- return pix_multiply (src, mask);
-}
-
-static inline __m64
-in_over_full_src_alpha (__m64 src, __m64 mask, __m64 dest)
-{
- src = _mm_or_si64 (src, MC(full_alpha));
-
- return over(in (src, mask), mask, dest);
-}
-
-#ifndef _MSC_VER
-static inline __m64
-in_over (__m64 src,
- __m64 srca,
- __m64 mask,
- __m64 dest)
-{
- return over(in(src, mask), pix_multiply(srca, mask), dest);
-}
-#else
-#define in_over(src, srca, mask, dest) over(in(src, mask), pix_multiply(srca, mask), dest)
-#endif
-
-static inline __m64
-load8888 (CARD32 v)
-{
- return _mm_unpacklo_pi8 (_mm_cvtsi32_si64 (v), _mm_setzero_si64());
-}
-
-static inline __m64
-pack8888 (__m64 lo, __m64 hi)
-{
- return _mm_packs_pu16 (lo, hi);
-}
-
-static inline CARD32
-store8888 (__m64 v)
-{
- return _mm_cvtsi64_si32(pack8888(v, _mm_setzero_si64()));
-}
-
-/* Expand 16 bits positioned at @pos (0-3) of a mmx register into
- *
- * 00RR00GG00BB
- *
- * --- Expanding 565 in the low word ---
- *
- * m = (m << (32 - 3)) | (m << (16 - 5)) | m;
- * m = m & (01f0003f001f);
- * m = m * (008404100840);
- * m = m >> 8;
- *
- * Note the trick here - the top word is shifted by another nibble to
- * avoid it bumping into the middle word
- */
-static inline __m64
-expand565 (__m64 pixel, int pos)
-{
- __m64 p = pixel;
- __m64 t1, t2;
-
- /* move pixel to low 16 bit and zero the rest */
- p = shift (shift (p, (3 - pos) * 16), -48);
-
- t1 = shift (p, 36 - 11);
- t2 = shift (p, 16 - 5);
-
- p = _mm_or_si64 (t1, p);
- p = _mm_or_si64 (t2, p);
- p = _mm_and_si64 (p, MC(565_rgb));
-
- pixel = _mm_mullo_pi16 (p, MC(565_unpack_multiplier));
- return _mm_srli_pi16 (pixel, 8);
-}
-
-static inline __m64
-expand8888 (__m64 in, int pos)
-{
- if (pos == 0)
- return _mm_unpacklo_pi8 (in, _mm_setzero_si64());
- else
- return _mm_unpackhi_pi8 (in, _mm_setzero_si64());
-}
-
-static inline __m64
-pack565 (__m64 pixel, __m64 target, int pos)
-{
- __m64 p = pixel;
- __m64 t = target;
- __m64 r, g, b;
-
- r = _mm_and_si64 (p, MC(565_r));
- g = _mm_and_si64 (p, MC(565_g));
- b = _mm_and_si64 (p, MC(565_b));
-
- r = shift (r, - (32 - 8) + pos * 16);
- g = shift (g, - (16 - 3) + pos * 16);
- b = shift (b, - (0 + 3) + pos * 16);
-
- if (pos == 0)
- t = _mm_and_si64 (t, MC(mask_0));
- else if (pos == 1)
- t = _mm_and_si64 (t, MC(mask_1));
- else if (pos == 2)
- t = _mm_and_si64 (t, MC(mask_2));
- else if (pos == 3)
- t = _mm_and_si64 (t, MC(mask_3));
-
- p = _mm_or_si64 (r, t);
- p = _mm_or_si64 (g, p);
-
- return _mm_or_si64 (b, p);
-}
-
-#ifndef _MSC_VER
-static inline __m64
-pix_add_mul (__m64 x, __m64 a, __m64 y, __m64 b)
-{
- x = _mm_mullo_pi16 (x, a);
- y = _mm_mullo_pi16 (y, b);
- x = _mm_adds_pu16 (x, MC(4x0080));
- x = _mm_adds_pu16 (x, y);
- x = _mm_adds_pu16 (x, _mm_srli_pi16 (x, 8));
- x = _mm_srli_pi16 (x, 8);
-
- return x;
-}
-#else
-#define pix_add_mul(x, a, y, b) \
-( x = _mm_mullo_pi16 (x, a), \
- y = _mm_mullo_pi16 (y, b), \
- x = _mm_adds_pu16 (x, MC(4x0080)), \
- x = _mm_adds_pu16 (x, y), \
- x = _mm_adds_pu16 (x, _mm_srli_pi16 (x, 8)), \
- _mm_srli_pi16 (x, 8) )
-#endif
-
-/* --------------- MMX code patch for fbcompose.c --------------------- */
-
-static FASTCALL void
-mmxCombineMaskU (CARD32 *src, const CARD32 *mask, int width)
-{
- const CARD32 *end = mask + width;
- while (mask < end) {
- __m64 a = load8888(*mask);
- __m64 s = load8888(*src);
- a = expand_alpha(a);
- s = pix_multiply(s, a);
- *src = store8888(s);
- ++src;
- ++mask;
- }
- _mm_empty();
-}
-
-
-static FASTCALL void
-mmxCombineOverU (CARD32 *dest, const CARD32 *src, int width)
-{
- const CARD32 *end = dest + width;
-
- while (dest < end) {
- __m64 s, sa;
- s = load8888(*src);
- sa = expand_alpha(s);
- *dest = store8888(over(s, sa, load8888(*dest)));
- ++dest;
- ++src;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineOverReverseU (CARD32 *dest, const CARD32 *src, int width)
-{
- const CARD32 *end = dest + width;
-
- while (dest < end) {
- __m64 d, da;
- d = load8888(*dest);
- da = expand_alpha(d);
- *dest = store8888(over (d, da, load8888(*src)));
- ++dest;
- ++src;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineInU (CARD32 *dest, const CARD32 *src, int width)
-{
- const CARD32 *end = dest + width;
-
- while (dest < end) {
- __m64 x, a;
- x = load8888(*src);
- a = load8888(*dest);
- a = expand_alpha(a);
- x = pix_multiply(x, a);
- *dest = store8888(x);
- ++dest;
- ++src;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineInReverseU (CARD32 *dest, const CARD32 *src, int width)
-{
- const CARD32 *end = dest + width;
-
- while (dest < end) {
- __m64 x, a;
- x = load8888(*dest);
- a = load8888(*src);
- a = expand_alpha(a);
- x = pix_multiply(x, a);
- *dest = store8888(x);
- ++dest;
- ++src;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineOutU (CARD32 *dest, const CARD32 *src, int width)
-{
- const CARD32 *end = dest + width;
-
- while (dest < end) {
- __m64 x, a;
- x = load8888(*src);
- a = load8888(*dest);
- a = expand_alpha(a);
- a = negate(a);
- x = pix_multiply(x, a);
- *dest = store8888(x);
- ++dest;
- ++src;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineOutReverseU (CARD32 *dest, const CARD32 *src, int width)
-{
- const CARD32 *end = dest + width;
-
- while (dest < end) {
- __m64 x, a;
- x = load8888(*dest);
- a = load8888(*src);
- a = expand_alpha(a);
- a = negate(a);
- x = pix_multiply(x, a);
- *dest = store8888(x);
- ++dest;
- ++src;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineAtopU (CARD32 *dest, const CARD32 *src, int width)
-{
- const CARD32 *end = dest + width;
-
- while (dest < end) {
- __m64 s, da, d, sia;
- s = load8888(*src);
- d = load8888(*dest);
- sia = expand_alpha(s);
- sia = negate(sia);
- da = expand_alpha(d);
- s = pix_add_mul (s, da, d, sia);
- *dest = store8888(s);
- ++dest;
- ++src;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineAtopReverseU (CARD32 *dest, const CARD32 *src, int width)
-{
- const CARD32 *end;
-
- end = dest + width;
-
- while (dest < end) {
- __m64 s, dia, d, sa;
- s = load8888(*src);
- d = load8888(*dest);
- sa = expand_alpha(s);
- dia = expand_alpha(d);
- dia = negate(dia);
- s = pix_add_mul (s, dia, d, sa);
- *dest = store8888(s);
- ++dest;
- ++src;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineXorU (CARD32 *dest, const CARD32 *src, int width)
-{
- const CARD32 *end = dest + width;
-
- while (dest < end) {
- __m64 s, dia, d, sia;
- s = load8888(*src);
- d = load8888(*dest);
- sia = expand_alpha(s);
- dia = expand_alpha(d);
- sia = negate(sia);
- dia = negate(dia);
- s = pix_add_mul (s, dia, d, sia);
- *dest = store8888(s);
- ++dest;
- ++src;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineAddU (CARD32 *dest, const CARD32 *src, int width)
-{
- const CARD32 *end = dest + width;
- while (dest < end) {
- __m64 s, d;
- s = load8888(*src);
- d = load8888(*dest);
- s = pix_add(s, d);
- *dest = store8888(s);
- ++dest;
- ++src;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineSaturateU (CARD32 *dest, const CARD32 *src, int width)
-{
- const CARD32 *end = dest + width;
- while (dest < end) {
- CARD32 s = *src;
- CARD32 d = *dest;
- __m64 ms = load8888(s);
- __m64 md = load8888(d);
- CARD32 sa = s >> 24;
- CARD32 da = ~d >> 24;
-
- if (sa > da) {
- __m64 msa = load8888(FbIntDiv(da, sa) << 24);
- msa = expand_alpha(msa);
- ms = pix_multiply(ms, msa);
- }
- md = pix_add(md, ms);
- *dest = store8888(md);
- ++src;
- ++dest;
- }
- _mm_empty();
-}
-
-
-static FASTCALL void
-mmxCombineSrcC (CARD32 *dest, CARD32 *src, CARD32 *mask, int width)
-{
- const CARD32 *end = src + width;
- while (src < end) {
- __m64 a = load8888(*mask);
- __m64 s = load8888(*src);
- s = pix_multiply(s, a);
- *dest = store8888(s);
- ++src;
- ++mask;
- ++dest;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineOverC (CARD32 *dest, CARD32 *src, CARD32 *mask, int width)
-{
- const CARD32 *end = src + width;
- while (src < end) {
- __m64 a = load8888(*mask);
- __m64 s = load8888(*src);
- __m64 d = load8888(*dest);
- __m64 sa = expand_alpha(s);
-
- *dest = store8888(in_over (s, sa, a, d));
-
- ++src;
- ++dest;
- ++mask;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineOverReverseC (CARD32 *dest, CARD32 *src, CARD32 *mask, int width)
-{
- const CARD32 *end = src + width;
- while (src < end) {
- __m64 a = load8888(*mask);
- __m64 s = load8888(*src);
- __m64 d = load8888(*dest);
- __m64 da = expand_alpha(d);
-
- *dest = store8888(over (d, da, in (s, a)));
-
- ++src;
- ++dest;
- ++mask;
- }
- _mm_empty();
-}
-
-
-static FASTCALL void
-mmxCombineInC (CARD32 *dest, CARD32 *src, CARD32 *mask, int width)
-{
- const CARD32 *end = src + width;
- while (src < end) {
- __m64 a = load8888(*mask);
- __m64 s = load8888(*src);
- __m64 d = load8888(*dest);
- __m64 da = expand_alpha(d);
- s = pix_multiply(s, a);
- s = pix_multiply(s, da);
- *dest = store8888(s);
- ++src;
- ++dest;
- ++mask;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineInReverseC (CARD32 *dest, CARD32 *src, CARD32 *mask, int width)
-{
- const CARD32 *end = src + width;
- while (src < end) {
- __m64 a = load8888(*mask);
- __m64 s = load8888(*src);
- __m64 d = load8888(*dest);
- __m64 sa = expand_alpha(s);
- a = pix_multiply(a, sa);
- d = pix_multiply(d, a);
- *dest = store8888(d);
- ++src;
- ++dest;
- ++mask;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineOutC (CARD32 *dest, CARD32 *src, CARD32 *mask, int width)
-{
- const CARD32 *end = src + width;
- while (src < end) {
- __m64 a = load8888(*mask);
- __m64 s = load8888(*src);
- __m64 d = load8888(*dest);
- __m64 da = expand_alpha(d);
- da = negate(da);
- s = pix_multiply(s, a);
- s = pix_multiply(s, da);
- *dest = store8888(s);
- ++src;
- ++dest;
- ++mask;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineOutReverseC (CARD32 *dest, CARD32 *src, CARD32 *mask, int width)
-{
- const CARD32 *end = src + width;
- while (src < end) {
- __m64 a = load8888(*mask);
- __m64 s = load8888(*src);
- __m64 d = load8888(*dest);
- __m64 sa = expand_alpha(s);
- a = pix_multiply(a, sa);
- a = negate(a);
- d = pix_multiply(d, a);
- *dest = store8888(d);
- ++src;
- ++dest;
- ++mask;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineAtopC (CARD32 *dest, CARD32 *src, CARD32 *mask, int width)
-{
- const CARD32 *end = src + width;
- while (src < end) {
- __m64 a = load8888(*mask);
- __m64 s = load8888(*src);
- __m64 d = load8888(*dest);
- __m64 da = expand_alpha(d);
- __m64 sa = expand_alpha(s);
- s = pix_multiply(s, a);
- a = pix_multiply(a, sa);
- a = negate(a);
- d = pix_add_mul (d, a, s, da);
- *dest = store8888(d);
- ++src;
- ++dest;
- ++mask;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineAtopReverseC (CARD32 *dest, CARD32 *src, CARD32 *mask, int width)
-{
- const CARD32 *end = src + width;
- while (src < end) {
- __m64 a = load8888(*mask);
- __m64 s = load8888(*src);
- __m64 d = load8888(*dest);
- __m64 da = expand_alpha(d);
- __m64 sa = expand_alpha(s);
- s = pix_multiply(s, a);
- a = pix_multiply(a, sa);
- da = negate(da);
- d = pix_add_mul (d, a, s, da);
- *dest = store8888(d);
- ++src;
- ++dest;
- ++mask;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineXorC (CARD32 *dest, CARD32 *src, CARD32 *mask, int width)
-{
- const CARD32 *end = src + width;
- while (src < end) {
- __m64 a = load8888(*mask);
- __m64 s = load8888(*src);
- __m64 d = load8888(*dest);
- __m64 da = expand_alpha(d);
- __m64 sa = expand_alpha(s);
- s = pix_multiply(s, a);
- a = pix_multiply(a, sa);
- da = negate(da);
- a = negate(a);
- d = pix_add_mul (d, a, s, da);
- *dest = store8888(d);
- ++src;
- ++dest;
- ++mask;
- }
- _mm_empty();
-}
-
-static FASTCALL void
-mmxCombineAddC (CARD32 *dest, CARD32 *src, CARD32 *mask, int width)
-{
- const CARD32 *end = src + width;
- while (src < end) {
- __m64 a = load8888(*mask);
- __m64 s = load8888(*src);
- __m64 d = load8888(*dest);
- s = pix_multiply(s, a);
- d = pix_add(s, d);
- *dest = store8888(d);
- ++src;
- ++dest;
- ++mask;
- }
- _mm_empty();
-}
-
-#if 0
-/* FIXME: this should be reinstated after adding fbmmx to pixman */
-extern FbComposeFunctions composeFunctions;
-#endif
-
-void fbComposeSetupMMX(void)
-{
-#if 0
-/* FIXME: this should be reinstated after adding fbmmx to pixman */
- /* check if we have MMX support and initialize accordingly */
- if (fbHaveMMX()) {
- composeFunctions.combineU[PictOpOver] = mmxCombineOverU;
- composeFunctions.combineU[PictOpOverReverse] = mmxCombineOverReverseU;
- composeFunctions.combineU[PictOpIn] = mmxCombineInU;
- composeFunctions.combineU[PictOpInReverse] = mmxCombineInReverseU;
- composeFunctions.combineU[PictOpOut] = mmxCombineOutU;
- composeFunctions.combineU[PictOpOutReverse] = mmxCombineOutReverseU;
- composeFunctions.combineU[PictOpAtop] = mmxCombineAtopU;
- composeFunctions.combineU[PictOpAtopReverse] = mmxCombineAtopReverseU;
- composeFunctions.combineU[PictOpXor] = mmxCombineXorU;
- composeFunctions.combineU[PictOpAdd] = mmxCombineAddU;
- composeFunctions.combineU[PictOpSaturate] = mmxCombineSaturateU;
-
- composeFunctions.combineC[PictOpSrc] = mmxCombineSrcC;
- composeFunctions.combineC[PictOpOver] = mmxCombineOverC;
- composeFunctions.combineC[PictOpOverReverse] = mmxCombineOverReverseC;
- composeFunctions.combineC[PictOpIn] = mmxCombineInC;
- composeFunctions.combineC[PictOpInReverse] = mmxCombineInReverseC;
- composeFunctions.combineC[PictOpOut] = mmxCombineOutC;
- composeFunctions.combineC[PictOpOutReverse] = mmxCombineOutReverseC;
- composeFunctions.combineC[PictOpAtop] = mmxCombineAtopC;
- composeFunctions.combineC[PictOpAtopReverse] = mmxCombineAtopReverseC;
- composeFunctions.combineC[PictOpXor] = mmxCombineXorC;
- composeFunctions.combineC[PictOpAdd] = mmxCombineAddC;
-
- composeFunctions.combineMaskU = mmxCombineMaskU;
- }
-#endif
-}
-
-
-/* ------------------ MMX code paths called from fbpict.c ----------------------- */
-
-void
-fbCompositeSolid_nx8888mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD32 src;
- CARD32 *dstLine, *dst;
- CARD16 w;
- FbStride dstStride;
- __m64 vsrc, vsrca;
-
- CHECKPOINT();
-
- fbComposeGetSolid(pSrc, src, pDst->format);
-
- if (src >> 24 == 0)
- return;
-
- fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1);
-
- vsrc = load8888 (src);
- vsrca = expand_alpha (vsrc);
-
- while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- w = width;
-
- CHECKPOINT();
-
- while (w && (unsigned long)dst & 7)
- {
- *dst = store8888(over(vsrc, vsrca, load8888(*dst)));
-
- w--;
- dst++;
- }
-
- while (w >= 2)
- {
- __m64 vdest;
- __m64 dest0, dest1;
-
- vdest = *(__m64 *)dst;
-
- dest0 = over(vsrc, vsrca, expand8888(vdest, 0));
- dest1 = over(vsrc, vsrca, expand8888(vdest, 1));
-
- *(__m64 *)dst = pack8888(dest0, dest1);
-
- dst += 2;
- w -= 2;
- }
-
- CHECKPOINT();
-
- while (w)
- {
- *dst = store8888(over(vsrc, vsrca, load8888(*dst)));
-
- w--;
- dst++;
- }
- }
-
- _mm_empty();
-}
-
-void
-fbCompositeSolid_nx0565mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD32 src;
- CARD16 *dstLine, *dst;
- CARD16 w;
- FbStride dstStride;
- __m64 vsrc, vsrca;
-
- CHECKPOINT();
-
- fbComposeGetSolid(pSrc, src, pDst->format);
-
- if (src >> 24 == 0)
- return;
-
- fbComposeGetStart (pDst, xDst, yDst, CARD16, dstStride, dstLine, 1);
-
- vsrc = load8888 (src);
- vsrca = expand_alpha (vsrc);
-
- while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- w = width;
-
- CHECKPOINT();
-
- while (w && (unsigned long)dst & 7)
- {
- ullong d = *dst;
- __m64 vdest = expand565 ((__m64)d, 0);
- vdest = pack565(over(vsrc, vsrca, vdest), vdest, 0);
- *dst = (ullong)vdest;
-
- w--;
- dst++;
- }
-
- while (w >= 4)
- {
- __m64 vdest;
-
- vdest = *(__m64 *)dst;
-
- vdest = pack565 (over(vsrc, vsrca, expand565(vdest, 0)), vdest, 0);
- vdest = pack565 (over(vsrc, vsrca, expand565(vdest, 1)), vdest, 1);
- vdest = pack565 (over(vsrc, vsrca, expand565(vdest, 2)), vdest, 2);
- vdest = pack565 (over(vsrc, vsrca, expand565(vdest, 3)), vdest, 3);
-
- *(__m64 *)dst = vdest;
-
- dst += 4;
- w -= 4;
- }
-
- CHECKPOINT();
-
- while (w)
- {
- ullong d = *dst;
- __m64 vdest = expand565 ((__m64)d, 0);
- vdest = pack565(over(vsrc, vsrca, vdest), vdest, 0);
- *dst = (ullong)vdest;
-
- w--;
- dst++;
- }
- }
-
- _mm_empty();
-}
-
-void
-fbCompositeSolidMask_nx8888x8888Cmmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD32 src, srca;
- CARD32 *dstLine;
- CARD32 *maskLine;
- FbStride dstStride, maskStride;
- __m64 vsrc, vsrca;
-
- CHECKPOINT();
-
- fbComposeGetSolid(pSrc, src, pDst->format);
-
- srca = src >> 24;
- if (srca == 0)
- return;
-
- fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1);
- fbComposeGetStart (pMask, xMask, yMask, CARD32, maskStride, maskLine, 1);
-
- vsrc = load8888(src);
- vsrca = expand_alpha(vsrc);
-
- while (height--)
- {
- int twidth = width;
- CARD32 *p = (CARD32 *)maskLine;
- CARD32 *q = (CARD32 *)dstLine;
-
- while (twidth && (unsigned long)q & 7)
- {
- CARD32 m = *(CARD32 *)p;
-
- if (m)
- {
- __m64 vdest = load8888(*q);
- vdest = in_over(vsrc, vsrca, load8888(m), vdest);
- *q = store8888(vdest);
- }
-
- twidth--;
- p++;
- q++;
- }
-
- while (twidth >= 2)
- {
- CARD32 m0, m1;
- m0 = *p;
- m1 = *(p + 1);
-
- if (m0 | m1)
- {
- __m64 dest0, dest1;
- __m64 vdest = *(__m64 *)q;
-
- dest0 = in_over(vsrc, vsrca, load8888(m0),
- expand8888 (vdest, 0));
- dest1 = in_over(vsrc, vsrca, load8888(m1),
- expand8888 (vdest, 1));
-
- *(__m64 *)q = pack8888(dest0, dest1);
- }
-
- p += 2;
- q += 2;
- twidth -= 2;
- }
-
- while (twidth)
- {
- CARD32 m = *(CARD32 *)p;
-
- if (m)
- {
- __m64 vdest = load8888(*q);
- vdest = in_over(vsrc, vsrca, load8888(m), vdest);
- *q = store8888(vdest);
- }
-
- twidth--;
- p++;
- q++;
- }
-
- dstLine += dstStride;
- maskLine += maskStride;
- }
-
- _mm_empty();
-}
-
-void
-fbCompositeSrc_8888x8x8888mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD32 *dstLine, *dst;
- CARD32 *srcLine, *src;
- CARD32 mask;
- __m64 vmask;
- FbStride dstStride, srcStride;
- CARD16 w;
- __m64 srca;
-
- CHECKPOINT();
-
- fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1);
- fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, srcStride, srcLine, 1);
-
- fbComposeGetSolid (pMask, mask, pDst->format);
- mask = mask | mask >> 8 | mask >> 16 | mask >> 24;
- vmask = load8888 (mask);
- srca = MC(4x00ff);
-
- while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- src = srcLine;
- srcLine += srcStride;
- w = width;
-
- while (w && (unsigned long)dst & 7)
- {
- __m64 s = load8888 (*src);
- __m64 d = load8888 (*dst);
-
- *dst = store8888 (in_over (s, expand_alpha (s), vmask, d));
-
- w--;
- dst++;
- src++;
- }
-
- while (w >= 2)
- {
- __m64 vs = *(__m64 *)src;
- __m64 vd = *(__m64 *)dst;
- __m64 vsrc0 = expand8888 (vs, 0);
- __m64 vsrc1 = expand8888 (vs, 1);
-
- *(__m64 *)dst = pack8888 (
- in_over (vsrc0, expand_alpha (vsrc0), vmask, expand8888 (vd, 0)),
- in_over (vsrc1, expand_alpha (vsrc1), vmask, expand8888 (vd, 1)));
-
- w -= 2;
- dst += 2;
- src += 2;
- }
-
- while (w)
- {
- __m64 s = load8888 (*src);
- __m64 d = load8888 (*dst);
-
- *dst = store8888 (in_over (s, expand_alpha (s), vmask, d));
-
- w--;
- dst++;
- src++;
- }
- }
-
- _mm_empty();
-}
-
-void
-fbCompositeSrc_x888x8x8888mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD32 *dstLine, *dst;
- CARD32 *srcLine, *src;
- CARD32 mask;
- __m64 vmask;
- FbStride dstStride, srcStride;
- CARD16 w;
- __m64 srca;
-
- CHECKPOINT();
-
- fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1);
- fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, srcStride, srcLine, 1);
- fbComposeGetSolid (pMask, mask, pDst->format);
-
- mask = mask | mask >> 8 | mask >> 16 | mask >> 24;
- vmask = load8888 (mask);
- srca = MC(4x00ff);
-
- while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- src = srcLine;
- srcLine += srcStride;
- w = width;
-
- while (w && (unsigned long)dst & 7)
- {
- __m64 s = load8888 (*src | 0xff000000);
- __m64 d = load8888 (*dst);
-
- *dst = store8888 (in_over (s, srca, vmask, d));
-
- w--;
- dst++;
- src++;
- }
-
- while (w >= 2)
- {
-
- __m64 vd0 = *(__m64 *)(dst);
- __m64 vs0 = *(__m64 *)(src);
-
- vd0 = pack8888 (
- in_over_full_src_alpha (expand8888 (vs0, 0), vmask, expand8888 (vd0, 0)),
- in_over_full_src_alpha (expand8888 (vs0, 1), vmask, expand8888 (vd0, 1)));
-
- *(__m64 *)(dst) = vd0;
-
- w -= 2;
- dst += 2;
- src += 2;
- }
-
- while (w)
- {
- __m64 s = load8888 (*src | 0xff000000);
- __m64 d = load8888 (*dst);
-
- *dst = store8888 (in_over (s, srca, vmask, d));
-
- w--;
- dst++;
- src++;
- }
- }
-
- _mm_empty();
-}
-
-void
-fbCompositeSrc_8888x8888mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD32 *dstLine, *dst;
- CARD32 *srcLine, *src;
- FbStride dstStride, srcStride;
- CARD16 w;
- __m64 srca;
-
- CHECKPOINT();
-
- fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1);
- fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, srcStride, srcLine, 1);
-
- srca = MC (4x00ff);
-
- while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- src = srcLine;
- srcLine += srcStride;
- w = width;
-
- while (w && (unsigned long)dst & 7)
- {
- __m64 s = load8888 (*src);
- __m64 d = load8888 (*dst);
-
- *dst = store8888 (over (s, expand_alpha (s), d));
-
- w--;
- dst++;
- src++;
- }
-
- while (w >= 2)
- {
- __m64 vd = *(__m64 *)(dst + 0);
- __m64 vs = *(__m64 *)(src + 0);
- __m64 vs0 = expand8888 (vs, 0);
- __m64 vs1 = expand8888 (vs, 1);
-
- *(__m64 *)dst = (__m64)pack8888 (
- over (vs0, expand_alpha (vs0), expand8888 (vd, 0)),
- over (vs1, expand_alpha (vs1), expand8888 (vd, 1)));
-
- w -= 2;
- dst += 2;
- src += 2;
- }
-
- while (w)
- {
- __m64 s = load8888 (*src);
- __m64 d = load8888 (*dst);
-
- *dst = store8888 (over (s, expand_alpha (s), d));
-
- w--;
- dst++;
- src++;
- }
- }
-
- _mm_empty();
-}
-
-void
-fbCompositeSrc_8888x0565mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD16 *dstLine, *dst;
- CARD32 *srcLine, *src;
- FbStride dstStride, srcStride;
- CARD16 w;
-
- CHECKPOINT();
-
- fbComposeGetStart (pDst, xDst, yDst, CARD16, dstStride, dstLine, 1);
- fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, srcStride, srcLine, 1);
-
- assert (pSrc->pDrawable == pMask->pDrawable);
-
- while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- src = srcLine;
- srcLine += srcStride;
- w = width;
-
- CHECKPOINT();
-
- while (w && (unsigned long)dst & 7)
- {
- __m64 vsrc = load8888 (*src);
- ullong d = *dst;
- __m64 vdest = expand565 ((__m64)d, 0);
-
- vdest = pack565(over(vsrc, expand_alpha(vsrc), vdest), vdest, 0);
-
- *dst = (ullong)vdest;
-
- w--;
- dst++;
- src++;
- }
-
- CHECKPOINT();
-
- while (w >= 4)
- {
- __m64 vsrc0, vsrc1, vsrc2, vsrc3;
- __m64 vdest;
-
- vsrc0 = load8888(*(src + 0));
- vsrc1 = load8888(*(src + 1));
- vsrc2 = load8888(*(src + 2));
- vsrc3 = load8888(*(src + 3));
-
- vdest = *(__m64 *)dst;
-
- vdest = pack565(over(vsrc0, expand_alpha(vsrc0), expand565(vdest, 0)), vdest, 0);
- vdest = pack565(over(vsrc1, expand_alpha(vsrc1), expand565(vdest, 1)), vdest, 1);
- vdest = pack565(over(vsrc2, expand_alpha(vsrc2), expand565(vdest, 2)), vdest, 2);
- vdest = pack565(over(vsrc3, expand_alpha(vsrc3), expand565(vdest, 3)), vdest, 3);
-
- *(__m64 *)dst = vdest;
-
- w -= 4;
- dst += 4;
- src += 4;
- }
-
- CHECKPOINT();
-
- while (w)
- {
- __m64 vsrc = load8888 (*src);
- ullong d = *dst;
- __m64 vdest = expand565 ((__m64)d, 0);
-
- vdest = pack565(over(vsrc, expand_alpha(vsrc), vdest), vdest, 0);
-
- *dst = (ullong)vdest;
-
- w--;
- dst++;
- src++;
- }
- }
-
- _mm_empty();
-}
-
-void
-fbCompositeSolidMask_nx8x8888mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD32 src, srca;
- CARD32 *dstLine, *dst;
- CARD8 *maskLine, *mask;
- FbStride dstStride, maskStride;
- CARD16 w;
- __m64 vsrc, vsrca;
- ullong srcsrc;
-
- CHECKPOINT();
-
- fbComposeGetSolid(pSrc, src, pDst->format);
-
- srca = src >> 24;
- if (srca == 0)
- return;
-
- srcsrc = (unsigned long long)src << 32 | src;
-
- fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1);
- fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1);
-
- vsrc = load8888 (src);
- vsrca = expand_alpha (vsrc);
-
- while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- mask = maskLine;
- maskLine += maskStride;
- w = width;
-
- CHECKPOINT();
-
- while (w && (unsigned long)dst & 7)
- {
- ullong m = *mask;
-
- if (m)
- {
- __m64 vdest = in_over(vsrc, vsrca, expand_alpha_rev ((__m64)m), load8888(*dst));
- *dst = store8888(vdest);
- }
-
- w--;
- mask++;
- dst++;
- }
-
- CHECKPOINT();
-
- while (w >= 2)
- {
- ullong m0, m1;
- m0 = *mask;
- m1 = *(mask + 1);
-
- if (srca == 0xff && (m0 & m1) == 0xff)
- {
- *(unsigned long long *)dst = srcsrc;
- }
- else if (m0 | m1)
- {
- __m64 vdest;
- __m64 dest0, dest1;
-
- vdest = *(__m64 *)dst;
-
- dest0 = in_over(vsrc, vsrca, expand_alpha_rev ((__m64)m0), expand8888(vdest, 0));
- dest1 = in_over(vsrc, vsrca, expand_alpha_rev ((__m64)m1), expand8888(vdest, 1));
-
- *(__m64 *)dst = pack8888(dest0, dest1);
- }
-
- mask += 2;
- dst += 2;
- w -= 2;
- }
-
- CHECKPOINT();
-
- while (w)
- {
- ullong m = *mask;
-
- if (m)
- {
- __m64 vdest = load8888(*dst);
- vdest = in_over(vsrc, vsrca, expand_alpha_rev ((__m64)m), vdest);
- *dst = store8888(vdest);
- }
-
- w--;
- mask++;
- dst++;
- }
- }
-
- _mm_empty();
-}
-
-
-Bool
-fbSolidFillmmx (DrawablePtr pDraw,
- int x,
- int y,
- int width,
- int height,
- FbBits xor)
-{
- FbStride stride;
- int bpp;
- ullong fill;
- __m64 vfill;
- CARD32 byte_width;
- CARD8 *byte_line;
- FbBits *bits;
- int xoff, yoff;
-#ifdef __GNUC__
- __m64 v1, v2, v3, v4, v5, v6, v7;
-#endif
-
- CHECKPOINT();
-
- fbGetDrawable(pDraw, bits, stride, bpp, xoff, yoff);
-
- if (bpp == 16 && (xor >> 16 != (xor & 0xffff)))
- return FALSE;
-
- if (bpp != 16 && bpp != 32)
- return FALSE;
-
- if (bpp == 16)
- {
- stride = stride * sizeof (FbBits) / 2;
- byte_line = (CARD8 *)(((CARD16 *)bits) + stride * (y + yoff) + (x + xoff));
- byte_width = 2 * width;
- stride *= 2;
- }
- else
- {
- stride = stride * sizeof (FbBits) / 4;
- byte_line = (CARD8 *)(((CARD32 *)bits) + stride * (y + yoff) + (x + xoff));
- byte_width = 4 * width;
- stride *= 4;
- }
-
- fill = ((ullong)xor << 32) | xor;
- vfill = (__m64)fill;
-
-#ifdef __GNUC__
- __asm__ (
- "movq %7, %0\n"
- "movq %7, %1\n"
- "movq %7, %2\n"
- "movq %7, %3\n"
- "movq %7, %4\n"
- "movq %7, %5\n"
- "movq %7, %6\n"
- : "=y" (v1), "=y" (v2), "=y" (v3),
- "=y" (v4), "=y" (v5), "=y" (v6), "=y" (v7)
- : "y" (vfill));
-#endif
-
- while (height--)
- {
- int w;
- CARD8 *d = byte_line;
- byte_line += stride;
- w = byte_width;
-
- while (w >= 2 && ((unsigned long)d & 3))
- {
- *(CARD16 *)d = xor;
- w -= 2;
- d += 2;
- }
-
- while (w >= 4 && ((unsigned long)d & 7))
- {
- *(CARD32 *)d = xor;
-
- w -= 4;
- d += 4;
- }
-
- while (w >= 64)
- {
-#ifdef __GNUC__
- __asm__ (
- "movq %1, (%0)\n"
- "movq %2, 8(%0)\n"
- "movq %3, 16(%0)\n"
- "movq %4, 24(%0)\n"
- "movq %5, 32(%0)\n"
- "movq %6, 40(%0)\n"
- "movq %7, 48(%0)\n"
- "movq %8, 56(%0)\n"
- :
- : "r" (d),
- "y" (vfill), "y" (v1), "y" (v2), "y" (v3),
- "y" (v4), "y" (v5), "y" (v6), "y" (v7)
- : "memory");
-#else
- *(__m64*) (d + 0) = vfill;
- *(__m64*) (d + 8) = vfill;
- *(__m64*) (d + 16) = vfill;
- *(__m64*) (d + 24) = vfill;
- *(__m64*) (d + 32) = vfill;
- *(__m64*) (d + 40) = vfill;
- *(__m64*) (d + 48) = vfill;
- *(__m64*) (d + 56) = vfill;
-#endif
- w -= 64;
- d += 64;
- }
-
- while (w >= 4)
- {
- *(CARD32 *)d = xor;
-
- w -= 4;
- d += 4;
- }
- if (w >= 2)
- {
- *(CARD16 *)d = xor;
- w -= 2;
- d += 2;
- }
- }
-
- _mm_empty();
- return TRUE;
-}
-
-void
-fbCompositeSolidMaskSrc_nx8x8888mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD32 src, srca;
- CARD32 *dstLine, *dst;
- CARD8 *maskLine, *mask;
- FbStride dstStride, maskStride;
- CARD16 w;
- __m64 vsrc, vsrca;
- ullong srcsrc;
-
- CHECKPOINT();
-
- fbComposeGetSolid(pSrc, src, pDst->format);
-
- srca = src >> 24;
- if (srca == 0)
- {
- fbSolidFillmmx (pDst->pDrawable, xDst, yDst, width, height, 0);
- return;
- }
-
- srcsrc = (ullong)src << 32 | src;
-
- fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1);
- fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1);
-
- vsrc = load8888 (src);
- vsrca = expand_alpha (vsrc);
-
- while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- mask = maskLine;
- maskLine += maskStride;
- w = width;
-
- CHECKPOINT();
-
- while (w && (unsigned long)dst & 7)
- {
- ullong m = *mask;
-
- if (m)
- {
- __m64 vdest = in(vsrc, expand_alpha_rev ((__m64)m));
- *dst = store8888(vdest);
- }
- else
- {
- *dst = 0;
- }
-
- w--;
- mask++;
- dst++;
- }
-
- CHECKPOINT();
-
- while (w >= 2)
- {
- ullong m0, m1;
- m0 = *mask;
- m1 = *(mask + 1);
-
- if (srca == 0xff && (m0 & m1) == 0xff)
- {
- *(ullong *)dst = srcsrc;
- }
- else if (m0 | m1)
- {
- __m64 vdest;
- __m64 dest0, dest1;
-
- vdest = *(__m64 *)dst;
-
- dest0 = in(vsrc, expand_alpha_rev ((__m64)m0));
- dest1 = in(vsrc, expand_alpha_rev ((__m64)m1));
-
- *(__m64 *)dst = pack8888(dest0, dest1);
- }
- else
- {
- *(ullong *)dst = 0;
- }
-
- mask += 2;
- dst += 2;
- w -= 2;
- }
-
- CHECKPOINT();
-
- while (w)
- {
- ullong m = *mask;
-
- if (m)
- {
- __m64 vdest = load8888(*dst);
- vdest = in(vsrc, expand_alpha_rev ((__m64)m));
- *dst = store8888(vdest);
- }
- else
- {
- *dst = 0;
- }
-
- w--;
- mask++;
- dst++;
- }
- }
-
- _mm_empty();
-}
-
-void
-fbCompositeSolidMask_nx8x0565mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD32 src, srca;
- CARD16 *dstLine, *dst;
- CARD8 *maskLine, *mask;
- FbStride dstStride, maskStride;
- CARD16 w;
- __m64 vsrc, vsrca;
- unsigned long long srcsrcsrcsrc, src16;
-
- CHECKPOINT();
-
- fbComposeGetSolid(pSrc, src, pDst->format);
-
- srca = src >> 24;
- if (srca == 0)
- return;
-
- fbComposeGetStart (pDst, xDst, yDst, CARD16, dstStride, dstLine, 1);
- fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1);
-
- vsrc = load8888 (src);
- vsrca = expand_alpha (vsrc);
-
- src16 = (ullong)pack565(vsrc, _mm_setzero_si64(), 0);
-
- srcsrcsrcsrc = (ullong)src16 << 48 | (ullong)src16 << 32 |
- (ullong)src16 << 16 | (ullong)src16;
-
- while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- mask = maskLine;
- maskLine += maskStride;
- w = width;
-
- CHECKPOINT();
-
- while (w && (unsigned long)dst & 7)
- {
- ullong m = *mask;
-
- if (m)
- {
- ullong d = *dst;
- __m64 vd = (__m64)d;
- __m64 vdest = in_over(vsrc, vsrca, expand_alpha_rev ((__m64)m), expand565(vd, 0));
- *dst = (ullong)pack565(vdest, _mm_setzero_si64(), 0);
- }
-
- w--;
- mask++;
- dst++;
- }
-
- CHECKPOINT();
-
- while (w >= 4)
- {
- ullong m0, m1, m2, m3;
- m0 = *mask;
- m1 = *(mask + 1);
- m2 = *(mask + 2);
- m3 = *(mask + 3);
-
- if (srca == 0xff && (m0 & m1 & m2 & m3) == 0xff)
- {
- *(unsigned long long *)dst = srcsrcsrcsrc;
- }
- else if (m0 | m1 | m2 | m3)
- {
- __m64 vdest;
- __m64 vm0, vm1, vm2, vm3;
-
- vdest = *(__m64 *)dst;
-
- vm0 = (__m64)m0;
- vdest = pack565(in_over(vsrc, vsrca, expand_alpha_rev(vm0), expand565(vdest, 0)), vdest, 0);
- vm1 = (__m64)m1;
- vdest = pack565(in_over(vsrc, vsrca, expand_alpha_rev(vm1), expand565(vdest, 1)), vdest, 1);
- vm2 = (__m64)m2;
- vdest = pack565(in_over(vsrc, vsrca, expand_alpha_rev(vm2), expand565(vdest, 2)), vdest, 2);
- vm3 = (__m64)m3;
- vdest = pack565(in_over(vsrc, vsrca, expand_alpha_rev(vm3), expand565(vdest, 3)), vdest, 3);
-
- *(__m64 *)dst = vdest;
- }
-
- w -= 4;
- mask += 4;
- dst += 4;
- }
-
- CHECKPOINT();
-
- while (w)
- {
- ullong m = *mask;
-
- if (m)
- {
- ullong d = *dst;
- __m64 vd = (__m64)d;
- __m64 vdest = in_over(vsrc, vsrca, expand_alpha_rev ((__m64)m), expand565(vd, 0));
- *dst = (ullong)pack565(vdest, _mm_setzero_si64(), 0);
- }
-
- w--;
- mask++;
- dst++;
- }
- }
-
- _mm_empty();
-}
-
-void
-fbCompositeSrc_8888RevNPx0565mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD16 *dstLine, *dst;
- CARD32 *srcLine, *src;
- FbStride dstStride, srcStride;
- CARD16 w;
-
- CHECKPOINT();
-
- fbComposeGetStart (pDst, xDst, yDst, CARD16, dstStride, dstLine, 1);
- fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, srcStride, srcLine, 1);
-
- assert (pSrc->pDrawable == pMask->pDrawable);
-
- while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- src = srcLine;
- srcLine += srcStride;
- w = width;
-
- CHECKPOINT();
-
- while (w && (unsigned long)dst & 7)
- {
- __m64 vsrc = load8888 (*src);
- ullong d = *dst;
- __m64 vdest = expand565 ((__m64)d, 0);
-
- vdest = pack565(over_rev_non_pre(vsrc, vdest), vdest, 0);
-
- *dst = (ullong)vdest;
-
- w--;
- dst++;
- src++;
- }
-
- CHECKPOINT();
-
- while (w >= 4)
- {
- CARD32 s0, s1, s2, s3;
- unsigned char a0, a1, a2, a3;
-
- s0 = *src;
- s1 = *(src + 1);
- s2 = *(src + 2);
- s3 = *(src + 3);
-
- a0 = (s0 >> 24);
- a1 = (s1 >> 24);
- a2 = (s2 >> 24);
- a3 = (s3 >> 24);
-
- if ((a0 & a1 & a2 & a3) == 0xFF)
- {
- __m64 vdest;
- vdest = pack565(invert_colors(load8888(s0)), _mm_setzero_si64(), 0);
- vdest = pack565(invert_colors(load8888(s1)), vdest, 1);
- vdest = pack565(invert_colors(load8888(s2)), vdest, 2);
- vdest = pack565(invert_colors(load8888(s3)), vdest, 3);
-
- *(__m64 *)dst = vdest;
- }
- else if (a0 | a1 | a2 | a3)
- {
- __m64 vdest = *(__m64 *)dst;
-
- vdest = pack565(over_rev_non_pre(load8888(s0), expand565(vdest, 0)), vdest, 0);
- vdest = pack565(over_rev_non_pre(load8888(s1), expand565(vdest, 1)), vdest, 1);
- vdest = pack565(over_rev_non_pre(load8888(s2), expand565(vdest, 2)), vdest, 2);
- vdest = pack565(over_rev_non_pre(load8888(s3), expand565(vdest, 3)), vdest, 3);
-
- *(__m64 *)dst = vdest;
- }
-
- w -= 4;
- dst += 4;
- src += 4;
- }
-
- CHECKPOINT();
-
- while (w)
- {
- __m64 vsrc = load8888 (*src);
- ullong d = *dst;
- __m64 vdest = expand565 ((__m64)d, 0);
-
- vdest = pack565(over_rev_non_pre(vsrc, vdest), vdest, 0);
-
- *dst = (ullong)vdest;
-
- w--;
- dst++;
- src++;
- }
- }
-
- _mm_empty();
-}
-
-/* "8888RevNP" is GdkPixbuf's format: ABGR, non premultiplied */
+typedef unsigned long long ullong;
-void
-fbCompositeSrc_8888RevNPx8888mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD32 *dstLine, *dst;
- CARD32 *srcLine, *src;
- FbStride dstStride, srcStride;
- CARD16 w;
-
- CHECKPOINT();
-
- fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1);
- fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, srcStride, srcLine, 1);
-
- assert (pSrc->pDrawable == pMask->pDrawable);
-
- while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- src = srcLine;
- srcLine += srcStride;
- w = width;
-
- while (w && (unsigned long)dst & 7)
- {
- __m64 s = load8888 (*src);
- __m64 d = load8888 (*dst);
-
- *dst = store8888 (over_rev_non_pre (s, d));
-
- w--;
- dst++;
- src++;
- }
-
- while (w >= 2)
- {
- ullong s0, s1;
- unsigned char a0, a1;
- __m64 d0, d1;
-
- s0 = *src;
- s1 = *(src + 1);
-
- a0 = (s0 >> 24);
- a1 = (s1 >> 24);
-
- if ((a0 & a1) == 0xFF)
- {
- d0 = invert_colors(load8888(s0));
- d1 = invert_colors(load8888(s1));
-
- *(__m64 *)dst = pack8888 (d0, d1);
- }
- else if (a0 | a1)
- {
- __m64 vdest = *(__m64 *)dst;
-
- d0 = over_rev_non_pre (load8888(s0), expand8888 (vdest, 0));
- d1 = over_rev_non_pre (load8888(s1), expand8888 (vdest, 1));
-
- *(__m64 *)dst = pack8888 (d0, d1);
- }
-
- w -= 2;
- dst += 2;
- src += 2;
- }
-
- while (w)
- {
- __m64 s = load8888 (*src);
- __m64 d = load8888 (*dst);
-
- *dst = store8888 (over_rev_non_pre (s, d));
-
- w--;
- dst++;
- src++;
- }
- }
-
- _mm_empty();
-}
+#ifdef __GNUC__
+typedef ullong mmxdatafield;
+#endif
+#ifdef _MSC_VER
+typedef unsigned __int64 ullong;
+typedef __m64 mmxdatafield;
+#endif
-void
-fbCompositeSolidMask_nx8888x0565Cmmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD32 src, srca;
- CARD16 *dstLine;
- CARD32 *maskLine;
- FbStride dstStride, maskStride;
- __m64 vsrc, vsrca;
+Bool
+fbSolidFillmmx (DrawablePtr pDraw,
+ int x,
+ int y,
+ int width,
+ int height,
+ FbBits xor)
+{
+ FbStride stride;
+ int bpp;
+ ullong fill;
+ __m64 vfill;
+ CARD32 byte_width;
+ CARD8 *byte_line;
+ FbBits *bits;
+ int xoff, yoff;
+#ifdef __GNUC__
+ __m64 v1, v2, v3, v4, v5, v6, v7;
+#endif
CHECKPOINT();
- fbComposeGetSolid(pSrc, src, pDst->format);
-
- srca = src >> 24;
- if (srca == 0)
- return;
-
- fbComposeGetStart (pDst, xDst, yDst, CARD16, dstStride, dstLine, 1);
- fbComposeGetStart (pMask, xMask, yMask, CARD32, maskStride, maskLine, 1);
+ fbGetDrawable(pDraw, bits, stride, bpp, xoff, yoff);
- vsrc = load8888 (src);
- vsrca = expand_alpha (vsrc);
+ if (bpp == 16 && (xor >> 16 != (xor & 0xffff)))
+ return FALSE;
- while (height--)
- {
- int twidth = width;
- CARD32 *p = (CARD32 *)maskLine;
- CARD16 *q = (CARD16 *)dstLine;
-
- while (twidth && ((unsigned long)q & 7))
- {
- CARD32 m = *(CARD32 *)p;
-
- if (m)
- {
- ullong d = *q;
- __m64 vdest = expand565 ((__m64)d, 0);
- vdest = pack565 (in_over (vsrc, vsrca, load8888 (m), vdest), vdest, 0);
- *q = (ullong)vdest;
- }
-
- twidth--;
- p++;
- q++;
- }
-
- while (twidth >= 4)
- {
- CARD32 m0, m1, m2, m3;
-
- m0 = *p;
- m1 = *(p + 1);
- m2 = *(p + 2);
- m3 = *(p + 3);
-
- if ((m0 | m1 | m2 | m3))
- {
- __m64 vdest = *(__m64 *)q;
-
- vdest = pack565(in_over(vsrc, vsrca, load8888(m0), expand565(vdest, 0)), vdest, 0);
- vdest = pack565(in_over(vsrc, vsrca, load8888(m1), expand565(vdest, 1)), vdest, 1);
- vdest = pack565(in_over(vsrc, vsrca, load8888(m2), expand565(vdest, 2)), vdest, 2);
- vdest = pack565(in_over(vsrc, vsrca, load8888(m3), expand565(vdest, 3)), vdest, 3);
-
- *(__m64 *)q = vdest;
- }
- twidth -= 4;
- p += 4;
- q += 4;
- }
-
- while (twidth)
- {
- CARD32 m;
-
- m = *(CARD32 *)p;
- if (m)
- {
- ullong d = *q;
- __m64 vdest = expand565((__m64)d, 0);
- vdest = pack565 (in_over(vsrc, vsrca, load8888(m), vdest), vdest, 0);
- *q = (ullong)vdest;
- }
-
- twidth--;
- p++;
- q++;
- }
-
- maskLine += maskStride;
- dstLine += dstStride;
- }
+ if (bpp != 16 && bpp != 32)
+ return FALSE;
- _mm_empty ();
-}
-
-void
-fbCompositeIn_nx8x8mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD8 *dstLine, *dst;
- CARD8 *maskLine, *mask;
- FbStride dstStride, maskStride;
- CARD16 w;
- CARD32 src;
- CARD8 sa;
- __m64 vsrc, vsrca;
-
- fbComposeGetStart (pDst, xDst, yDst, CARD8, dstStride, dstLine, 1);
- fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1);
-
- fbComposeGetSolid(pSrc, src, pDst->format);
-
- sa = src >> 24;
- if (sa == 0)
- return;
-
- vsrc = load8888(src);
- vsrca = expand_alpha(vsrc);
-
- while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- mask = maskLine;
- maskLine += maskStride;
- w = width;
-
- if ((((unsigned long)pDst & 3) == 0) &&
- (((unsigned long)pSrc & 3) == 0))
- {
- while (w >= 4)
- {
- CARD32 m;
- __m64 vmask;
- __m64 vdest;
-
- m = 0;
-
- vmask = load8888 (*(CARD32 *)mask);
- vdest = load8888 (*(CARD32 *)dst);
-
- *(CARD32 *)dst = store8888 (in (in (vsrca, vmask), vdest));
-
- dst += 4;
- mask += 4;
- w -= 4;
- }
- }
-
- while (w--)
- {
- CARD16 tmp;
- CARD8 a;
- CARD32 m, d;
- CARD32 r;
-
- a = *mask++;
- d = *dst;
-
- m = FbInU (sa, 0, a, tmp);
- r = FbInU (m, 0, d, tmp);
-
- *dst++ = r;
- }
- }
-
- _mm_empty();
-}
-
-void
-fbCompositeIn_8x8mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD8 *dstLine, *dst;
- CARD8 *srcLine, *src;
- FbStride srcStride, dstStride;
- CARD16 w;
-
- fbComposeGetStart (pDst, xDst, yDst, CARD8, dstStride, dstLine, 1);
- fbComposeGetStart (pSrc, xSrc, ySrc, CARD8, srcStride, srcLine, 1);
-
- while (height--)
+ if (bpp == 16)
{
- dst = dstLine;
- dstLine += dstStride;
- src = srcLine;
- srcLine += srcStride;
- w = width;
-
- if ((((unsigned long)pDst & 3) == 0) &&
- (((unsigned long)pSrc & 3) == 0))
- {
- while (w >= 4)
- {
- CARD32 *s = (CARD32 *)src;
- CARD32 *d = (CARD32 *)dst;
-
- *d = store8888 (in (load8888 (*s), load8888 (*d)));
-
- w -= 4;
- dst += 4;
- src += 4;
- }
- }
-
- while (w--)
- {
- CARD8 s, d;
- CARD16 tmp;
-
- s = *src;
- d = *dst;
-
- *dst = FbInU (s, 0, d, tmp);
-
- src++;
- dst++;
- }
+ stride = stride * sizeof (FbBits) / 2;
+ byte_line = (CARD8 *)(((CARD16 *)bits) + stride * (y + yoff) + (x + xoff));
+ byte_width = 2 * width;
+ stride *= 2;
}
-
- _mm_empty ();
-}
-
-void
-fbCompositeSrcAdd_8888x8x8mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD8 *dstLine, *dst;
- CARD8 *maskLine, *mask;
- FbStride dstStride, maskStride;
- CARD16 w;
- CARD32 src;
- CARD8 sa;
- __m64 vsrc, vsrca;
-
- fbComposeGetStart (pDst, xDst, yDst, CARD8, dstStride, dstLine, 1);
- fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1);
-
- fbComposeGetSolid(pSrc, src, pDst->format);
-
- sa = src >> 24;
- if (sa == 0)
- return;
-
- vsrc = load8888(src);
- vsrca = expand_alpha(vsrc);
-
- while (height--)
+ else
{
- dst = dstLine;
- dstLine += dstStride;
- mask = maskLine;
- maskLine += maskStride;
- w = width;
-
- if ((((unsigned long)pMask & 3) == 0) &&
- (((unsigned long)pDst & 3) == 0))
- {
- while (w >= 4)
- {
- __m64 vmask = load8888 (*(CARD32 *)mask);
- __m64 vdest = load8888 (*(CARD32 *)dst);
-
- *(CARD32 *)dst = store8888 (_mm_adds_pu8 (in (vsrca, vmask), vdest));
-
- w -= 4;
- dst += 4;
- mask += 4;
- }
- }
-
- while (w--)
- {
- CARD16 tmp;
- CARD16 a;
- CARD32 m, d;
- CARD32 r;
-
- a = *mask++;
- d = *dst;
-
- m = FbInU (sa, 0, a, tmp);
- r = FbAdd (m, d, 0, tmp);
-
- *dst++ = r;
- }
+ stride = stride * sizeof (FbBits) / 4;
+ byte_line = (CARD8 *)(((CARD32 *)bits) + stride * (y + yoff) + (x + xoff));
+ byte_width = 4 * width;
+ stride *= 4;
}
-
- _mm_empty();
-}
-
-void
-fbCompositeSrcAdd_8000x8000mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD8 *dstLine, *dst;
- CARD8 *srcLine, *src;
- FbStride dstStride, srcStride;
- CARD16 w;
- CARD8 s, d;
- CARD16 t;
- CHECKPOINT();
+ fill = ((ullong)xor << 32) | xor;
+ vfill = (__m64)fill;
- fbComposeGetStart (pSrc, xSrc, ySrc, CARD8, srcStride, srcLine, 1);
- fbComposeGetStart (pDst, xDst, yDst, CARD8, dstStride, dstLine, 1);
+#ifdef __GNUC__
+ __asm__ (
+ "movq %7, %0\n"
+ "movq %7, %1\n"
+ "movq %7, %2\n"
+ "movq %7, %3\n"
+ "movq %7, %4\n"
+ "movq %7, %5\n"
+ "movq %7, %6\n"
+ : "=y" (v1), "=y" (v2), "=y" (v3),
+ "=y" (v4), "=y" (v5), "=y" (v6), "=y" (v7)
+ : "y" (vfill));
+#endif
while (height--)
{
- dst = dstLine;
- dstLine += dstStride;
- src = srcLine;
- srcLine += srcStride;
- w = width;
-
- while (w && (unsigned long)dst & 7)
- {
- s = *src;
- d = *dst;
- t = d + s;
- s = t | (0 - (t >> 8));
- *dst = s;
-
- dst++;
- src++;
- w--;
- }
+ int w;
+ CARD8 *d = byte_line;
+ byte_line += stride;
+ w = byte_width;
- while (w >= 8)
+ while (w >= 2 && ((unsigned long)d & 3))
{
- *(__m64*)dst = _mm_adds_pu8(*(__m64*)src, *(__m64*)dst);
- dst += 8;
- src += 8;
- w -= 8;
+ *(CARD16 *)d = xor;
+ w -= 2;
+ d += 2;
}
- while (w)
+ while (w >= 4 && ((unsigned long)d & 7))
{
- s = *src;
- d = *dst;
- t = d + s;
- s = t | (0 - (t >> 8));
- *dst = s;
+ *(CARD32 *)d = xor;
- dst++;
- src++;
- w--;
+ w -= 4;
+ d += 4;
}
- }
-
- _mm_empty();
-}
-void
-fbCompositeSrcAdd_8888x8888mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD32 *dstLine, *dst;
- CARD32 *srcLine, *src;
- FbStride dstStride, srcStride;
- CARD16 w;
-
- CHECKPOINT();
-
- fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, srcStride, srcLine, 1);
- fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1);
-
- while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- src = srcLine;
- srcLine += srcStride;
- w = width;
-
- while (w && (unsigned long)dst & 7)
+ while (w >= 64)
{
- *dst = _mm_cvtsi64_si32(_mm_adds_pu8(_mm_cvtsi32_si64(*src),
- _mm_cvtsi32_si64(*dst)));
- dst++;
- src++;
- w--;
+#ifdef __GNUC__
+ __asm__ (
+ "movq %1, (%0)\n"
+ "movq %2, 8(%0)\n"
+ "movq %3, 16(%0)\n"
+ "movq %4, 24(%0)\n"
+ "movq %5, 32(%0)\n"
+ "movq %6, 40(%0)\n"
+ "movq %7, 48(%0)\n"
+ "movq %8, 56(%0)\n"
+ :
+ : "r" (d),
+ "y" (vfill), "y" (v1), "y" (v2), "y" (v3),
+ "y" (v4), "y" (v5), "y" (v6), "y" (v7)
+ : "memory");
+#else
+ *(__m64*) (d + 0) = vfill;
+ *(__m64*) (d + 8) = vfill;
+ *(__m64*) (d + 16) = vfill;
+ *(__m64*) (d + 24) = vfill;
+ *(__m64*) (d + 32) = vfill;
+ *(__m64*) (d + 40) = vfill;
+ *(__m64*) (d + 48) = vfill;
+ *(__m64*) (d + 56) = vfill;
+#endif
+ w -= 64;
+ d += 64;
}
- while (w >= 2)
+ while (w >= 4)
{
- *(ullong*)dst = (ullong) _mm_adds_pu8(*(__m64*)src, *(__m64*)dst);
- dst += 2;
- src += 2;
- w -= 2;
+ *(CARD32 *)d = xor;
+
+ w -= 4;
+ d += 4;
}
-
- if (w)
+ if (w >= 2)
{
- *dst = _mm_cvtsi64_si32(_mm_adds_pu8(_mm_cvtsi32_si64(*src),
- _mm_cvtsi32_si64(*dst)));
-
+ *(CARD16 *)d = xor;
+ w -= 2;
+ d += 2;
}
}
_mm_empty();
+ return TRUE;
}
Bool
diff --git a/fb/fbmmx.h b/fb/fbmmx.h
index ca9c7e7..43079e9 100644
--- a/fb/fbmmx.h
+++ b/fb/fbmmx.h
@@ -44,226 +44,6 @@ Bool fbHaveMMX(void);
#ifdef USE_MMX
-void fbComposeSetupMMX(void);
-
-void fbCompositeSolidMask_nx8888x0565Cmmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
-void fbCompositeSrcAdd_8888x8888mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
-void fbCompositeSrc_8888x8888mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
-void
-fbCompositeSolidMaskSrc_nx8x8888mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
-void
-fbCompositeSrc_x888x8x8888mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
-void fbCompositeSolidMask_nx8888x8888Cmmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
-void fbCompositeSolidMask_nx8x8888mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
-void fbCompositeIn_nx8x8mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
-void fbCompositeIn_8x8mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
-void fbCompositeSrcAdd_8888x8x8mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
-void fbCompositeSrcAdd_8000x8000mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
-void fbCompositeSrc_8888RevNPx8888mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
-void fbCompositeSrc_8888x0565mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
-void fbCompositeSrc_8888RevNPx0565mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
-void fbCompositeSolid_nx8888mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
-void fbCompositeSolid_nx0565mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
-void fbCompositeSolidMask_nx8x0565mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
-void fbCompositeSrc_8888x8x8888mmx (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height);
Bool fbCopyAreammx (DrawablePtr pSrc,
DrawablePtr pDst,
int src_x,
diff --git a/render/mipict.c b/render/mipict.c
index 3d6c1ae..dff7f5e 100644
--- a/render/mipict.c
+++ b/render/mipict.c
@@ -1,4 +1,4 @@
- /*
+/*
*
* Copyright © 1999 Keith Packard
*
More information about the xorg-commit
mailing list