[x264-devel] [PATCH 3/4] x264_intra_sad_x3_8x8c_neon

George Stephanos gaf.stephanos at gmail.com
Thu Feb 2 13:56:43 CET 2012


---
 common/arm/pixel-a.S |   78 ++++++++++++++++++++++++++++++++++++++++++++++++++
 common/arm/pixel.h   |    1 +
 common/pixel.c       |    1 +
 3 files changed, 80 insertions(+), 0 deletions(-)

diff --git a/common/arm/pixel-a.S b/common/arm/pixel-a.S
index 995049b..db32671 100644
--- a/common/arm/pixel-a.S
+++ b/common/arm/pixel-a.S
@@ -1336,3 +1336,81 @@ function x264_intra_sad_x3_8x8_neon
 
     bx          lr
 .endfunc
+
+function x264_intra_sad_x3_8x8c_neon
+    push        {r4, lr}
+    vmov.i8     q8, #0
+    vmov.i8     q11, #0
+    add         r2, #8
+    sub         lr, r1, #FDEC_STRIDE
+    mov         r3, #FENC_STRIDE
+    vld1.8      {d0}, [lr]
+    mov         r4, #FDEC_STRIDE
+    sub         lr, r1, #1
+
+.irp Y,24,25,26,27,28,29,30,31
+    vld1.8      {d\Y}, [r0], r3
+    vld1.8      {d18[]}, [lr], r4
+.if \Y == 24
+    vabdl.u8    q2, d0, d\Y
+    vabdl.u8    q10, d18, d\Y
+.else
+    vabal.u8    q2, d0, d\Y
+    vabal.u8    q10, d18, d\Y
+.endif
+.if \Y < 28
+    vaddw.u8    q8, d18
+.else
+    vaddw.u8    q11, d18
+.endif
+.endr
+    
+    vmovl.u8    q3, d0
+    vmov.i8     d17, #0
+    vadd.u16    d20, d21
+    vadd.u16    d4, d5
+    vpadd.u16   d20, d17
+    vpadd.u16   d4, d17
+    vpadd.u16   d20, d17
+    vpadd.u16   d4, d17
+    vpadd.u16   d6, d17
+    vpadd.u16   d7, d17
+    vpadd.u16   d6, d17
+    vpadd.u16   d7, d17
+
+    vst1.32     {d4[0]}, [r2,:32]
+    sub         r2, #4
+    vst1.32     {d20[0]}, [r2,:32]
+
+    vadd.u16    d3, d7, d22
+    vadd.u16    d0, d6, d16
+    vrshr.u16   d1, d7, #2
+    vrshr.u16   d3, #3
+    vrshr.u16   d0, #3
+    vrshr.u16   d2, d22, #2
+    vdup.8      d1, d1[0]
+    vdup.8      d3, d3[0]
+    vdup.8      d0, d0[0]
+    vdup.8      d2, d2[0]
+
+    vext.8      d0, d0, d1, #4
+    vext.8      d1, d2, d3, #4
+
+    vabdl.u8    q11, d0, d24
+    vabdl.u8    q10, d0, d25
+    vabal.u8    q11, d0, d26
+    vabal.u8    q10, d0, d27
+    vabal.u8    q11, d1, d28
+    vabal.u8    q10, d1, d29
+    vabal.u8    q11, d1, d30
+    vabal.u8    q10, d1, d31
+
+    sub         r2, #4
+    vadd.u16    q11, q10
+    vadd.u16    d22, d23
+    vpadd.u16   d22, d17
+    vpadd.u16   d22, d17
+    vst1.32     {d22[0]}, [r2,:32]
+
+    pop        {r4, pc}
+.endfunc
diff --git a/common/arm/pixel.h b/common/arm/pixel.h
index 07a72c2..506cf59 100644
--- a/common/arm/pixel.h
+++ b/common/arm/pixel.h
@@ -71,4 +71,5 @@ float x264_pixel_ssim_end4_neon( int sum0[5][4], int sum1[5][4], int width );
 
 void x264_intra_sad_x3_4x4_armv6( uint8_t *, uint8_t *, int * );
 void x264_intra_sad_x3_8x8_neon( uint8_t *, uint8_t *, int * );
+void x264_intra_sad_x3_8x8c_neon( uint8_t *, uint8_t *, int * );
 #endif
diff --git a/common/pixel.c b/common/pixel.c
index af7006f..f6d6a04 100644
--- a/common/pixel.c
+++ b/common/pixel.c
@@ -1213,6 +1213,7 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
     if( cpu&X264_CPU_NEON )
     {
         pixf->intra_sad_x3_8x8  = x264_intra_sad_x3_8x8_neon;
+        pixf->intra_sad_x3_8x8c   = x264_intra_sad_x3_8x8c_neon;
         INIT5( sad, _neon );
         INIT5( sad_aligned, _neon );
         INIT7( sad_x3, _neon );
-- 
1.7.4.1



More information about the x264-devel mailing list