[Pixman] [PATCH 1/2] mmx: Use Loongson pinsrh instruction in pack_565

[Matt Turner]

The pinsrh instruction is analogous to MMX EXT's pinsrw, except like
other Loongson vector instructions it cannot access the general purpose
registers. In the cases of other Loongson vector instructions, this is a
headache, but it is actually a good thing here. Since the instruction is
different from MMX, I've named the intrinsic loongson_insert_pi16.

 text	data	bss	dec	 hex filename
25976	1952	  0   27928	6d18 .libs/libpixman_loongson_mmi_la-pixman-mmx.o
25336	1952	  0   27288	6a98 .libs/libpixman_loongson_mmi_la-pixman-mmx.o

-and: 181
+and: 147
-dsll: 143
+dsll: 95
-dsrl: 87
+dsrl: 135
-ldc1: 523
+ldc1: 462
-lw: 767
+lw: 721
+pinsrh: 35
---
 pixman/loongson-mmintrin.h |   11 +++++++++++
 pixman/pixman-mmx.c        |   14 ++++++++++++++
 2 files changed, 25 insertions(+), 0 deletions(-)

diff --git a/pixman/loongson-mmintrin.h b/pixman/loongson-mmintrin.h
index 44d30f5..a67e45a 100644
--- a/pixman/loongson-mmintrin.h
+++ b/pixman/loongson-mmintrin.h
@@ -216,3 +216,14 @@ _mm_xor_si64 (__m64 __m1, __m64 __m2)
 	);
 	return ret;
 }
+
+extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
+loongson_insert_pi16 (__m64 __m1, __m64 __m2, int __pos)
+{
+	__m64 ret;
+	asm("pinsrh_%3 %0, %1, %2\n\t"
+	   : "=f" (ret)
+	   : "f" (__m1), "f" (__m2), "i" (__pos)
+	);
+	return ret;
+}
diff --git a/pixman/pixman-mmx.c b/pixman/pixman-mmx.c
index 08eb25d..a547e23 100644
--- a/pixman/pixman-mmx.c
+++ b/pixman/pixman-mmx.c
@@ -177,10 +177,12 @@ typedef struct
     mmxdatafield mmx_565_r;
     mmxdatafield mmx_565_g;
     mmxdatafield mmx_565_b;
+#ifndef USE_LOONGSON_MMI
     mmxdatafield mmx_mask_0;
     mmxdatafield mmx_mask_1;
     mmxdatafield mmx_mask_2;
     mmxdatafield mmx_mask_3;
+#endif
     mmxdatafield mmx_full_alpha;
     mmxdatafield mmx_4x0101;
 } mmx_data_t;
@@ -202,10 +204,12 @@ static const mmx_data_t c =
     MMXDATA_INIT (.mmx_565_r,                    0x000000f800000000),
     MMXDATA_INIT (.mmx_565_g,                    0x0000000000fc0000),
     MMXDATA_INIT (.mmx_565_b,                    0x00000000000000f8),
+#ifndef USE_LOONGSON_MMI
     MMXDATA_INIT (.mmx_mask_0,                   0xffffffffffff0000),
     MMXDATA_INIT (.mmx_mask_1,                   0xffffffff0000ffff),
     MMXDATA_INIT (.mmx_mask_2,                   0xffff0000ffffffff),
     MMXDATA_INIT (.mmx_mask_3,                   0x0000ffffffffffff),
+#endif
     MMXDATA_INIT (.mmx_full_alpha,               0x00ff000000000000),
     MMXDATA_INIT (.mmx_4x0101,                   0x0101010101010101),
 };
@@ -523,6 +527,15 @@ pack_565 (__m64 pixel, __m64 target, int pos)
     g = _mm_and_si64 (p, MC (565_g));
     b = _mm_and_si64 (p, MC (565_b));
 
+#ifdef USE_LOONGSON_MMI
+    r = shift (r, -(32 - 8));
+    g = shift (g, -(16 - 3));
+    b = shift (b, -(0  + 3));
+
+    p = _mm_or_si64 (r, g);
+    p = _mm_or_si64 (p, b);
+    return loongson_insert_pi16 (t, p, pos);
+#else
     r = shift (r, -(32 - 8) + pos * 16);
     g = shift (g, -(16 - 3) + pos * 16);
     b = shift (b, -(0  + 3) + pos * 16);
@@ -540,6 +553,7 @@ pack_565 (__m64 pixel, __m64 target, int pos)
     p = _mm_or_si64 (g, p);
 
     return _mm_or_si64 (b, p);
+#endif
 }
 
 #ifndef _MSC_VER
-- 
1.7.8.5