[Pixman] [PATCH 3/3] mmx: Use _mm_shuffle_pi16 when available

[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
---
 pixman/pixman-mmx.c |   25 +++++++++++++++++++++++++
 1 files changed, 25 insertions(+), 0 deletions(-)

diff --git a/pixman/pixman-mmx.c b/pixman/pixman-mmx.c
index 6103474..86bc7eb 100644
--- a/pixman/pixman-mmx.c
+++ b/pixman/pixman-mmx.c
@@ -54,6 +54,10 @@ _mm_empty (void)
 {
 
 }
+
+/* ARM's mmintrin.h also doesn't define the _MM_SHUFFLE macro.  */
+#define _MM_SHUFFLE(fp3,fp2,fp1,fp0) \
+ (((fp3) << 6) | ((fp2) << 4) | ((fp1) << 2) | (fp0))
 #endif
 
 #ifdef USE_X86_MMX_EXT
@@ -65,6 +69,15 @@ _mm_mulhi_pu16 (__m64 __A, __m64 __B)
 {
   return (__m64) __builtin_ia32_pmulhuw ((__v4hi)__A, (__v4hi)__B);
 }
+
+extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_shuffle_pi16 (__m64 __A, int const __N)
+{
+  return (__m64) __builtin_ia32_pshufw ((__v4hi)__A, __N);
+}
+
+#define _MM_SHUFFLE(fp3,fp2,fp1,fp0) \
+ (((fp3) << 6) | ((fp2) << 4) | ((fp1) << 2) | (fp0))
 #endif
 
 /* Notes about writing mmx code
@@ -253,6 +266,9 @@ pix_add (__m64 a, __m64 b)
 static force_inline __m64
 expand_alpha (__m64 pixel)
 {
+#if defined USE_X86_MMX_EXT || defined USE_ARM_IWMMXT
+    return _mm_shuffle_pi16(pixel, _MM_SHUFFLE (3, 3, 3, 3));
+#else
     __m64 t1, t2;
 
     t1 = shift (pixel, -48);
@@ -262,11 +278,15 @@ expand_alpha (__m64 pixel)
     t1 = _mm_or_si64 (t1, t2);
 
     return t1;
+#endif
 }
 
 static force_inline __m64
 expand_alpha_rev (__m64 pixel)
 {
+#if defined USE_X86_MMX_EXT || defined USE_ARM_IWMMXT
+    return _mm_shuffle_pi16(pixel, _MM_SHUFFLE (0, 0, 0, 0));
+#else
     __m64 t1, t2;
 
     /* move alpha to low 16 bits and zero the rest */
@@ -279,11 +299,15 @@ expand_alpha_rev (__m64 pixel)
     t1 = _mm_or_si64 (t1, t2);
 
     return t1;
+#endif
 }
 
 static force_inline __m64
 invert_colors (__m64 pixel)
 {
+#if defined USE_X86_MMX_EXT || defined USE_ARM_IWMMXT
+    return _mm_shuffle_pi16(pixel, _MM_SHUFFLE (3, 0, 1, 2));
+#else
     __m64 x, y, z;
 
     x = y = z = pixel;
@@ -299,6 +323,7 @@ invert_colors (__m64 pixel)
     x = _mm_or_si64 (x, z);
 
     return x;
+#endif
 }
 
 static force_inline __m64
-- 
1.7.3.4