[Pixman] [PATCH v4 2/2] mmx: Use _mm_shuffle_pi16

[Matt Turner]

The pshufw x86 instruction is part of Extended 3DNow! and SSE1. The
equivalent ARM wshufh instruction was available from the first iwMMXt
instrucion set.

This instruction is already used in the SSE2 code.

Reduces code size by ~9%.

amd64
  text    data     bss     dec     hex filename
 29925    2240       0   32165    7da5 .libs/libpixman_mmx_la-pixman-mmx.o
 27237    2240       0   29477    7325 .libs/libpixman_mmx_la-pixman-mmx.o

x86
  text    data     bss     dec     hex filename
 27677    1792       0   29469    731d .libs/libpixman_mmx_la-pixman-mmx.o
 24959    1792       0   26751    687f .libs/libpixman_mmx_la-pixman-mmx.o

arm
  text    data     bss     dec     hex filename
 30176    1792       0   31968    7ce0 .libs/libpixman_iwmmxt_la-pixman-mmx.o
 27384    1792       0   29176    71f8 .libs/libpixman_iwmmxt_la-pixman-mmx.o

v4: use inline assembly since the intrinsic needs SSE or 3DNowA
	- we can't use -msse, since it'll cause more SSE instructions to be
	  generated that we don't want
	- there is no -m3dnowa flag (anymore?)

Signed-off-by: Matt Turner <mattst88 at gmail.com>
---
 pixman/pixman-mmx.c |   55 +++++++++++++++++---------------------------------
 1 files changed, 19 insertions(+), 36 deletions(-)

diff --git a/pixman/pixman-mmx.c b/pixman/pixman-mmx.c
index a31adf8..a74eefd 100644
--- a/pixman/pixman-mmx.c
+++ b/pixman/pixman-mmx.c
@@ -69,7 +69,23 @@ _mm_mulhi_pu16 (__m64 __A, __m64 __B)
     );
     return __A;
 }
+
+extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_shuffle_pi16 (__m64 __A, int8_t const __N)
+{
+    __m64 ret;
+
+    asm("pshufw %2, %1, %0\n\t"
+	: "=y" (ret)
+	: "y" (__A), "K" (__N)
+    );
+
+    return ret;
+}
 #endif
+ 
+#define _MM_SHUFFLE(fp3,fp2,fp1,fp0) \
+ (((fp3) << 6) | ((fp2) << 4) | ((fp1) << 2) | (fp0))
 
 /* Notes about writing mmx code
  *
@@ -249,52 +265,19 @@ pix_add (__m64 a, __m64 b)
 static force_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;
+    return _mm_shuffle_pi16(pixel, _MM_SHUFFLE (3, 3, 3, 3));
 }
 
 static force_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;
+    return _mm_shuffle_pi16(pixel, _MM_SHUFFLE (0, 0, 0, 0));
 }
 
 static force_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;
+    return _mm_shuffle_pi16(pixel, _MM_SHUFFLE (3, 0, 1, 2));
 }
 
 static force_inline __m64
-- 
1.7.3.4