shithub: libvpx

Info  •  Files  •  Log  •  Branches

Make 4-tap interp filter coefficients even numbers

This CL modified 4-tap interp filter coefficients to be even numbers,
which would help in writing 4-tap filter SIMD optimizations. The coding
performance change was negligible. Speed 1 borg test showed:
        avg_psnr:  ovr_psnr:    ssim:
lowres:  -0.003    -0.012      -0.017
midres:  0.029     0.018        0.043
hdres:   0.024     0.044        0.033

Change-Id: Id7c54bb9a9c1aee19c41bc6f1dc3b9682d158bba

--- a/vp9/common/vp9_filter.c

+++ b/vp9/common/vp9_filter.c

@@ -66,14 +66,14 @@

 // 4-tap filter

 DECLARE_ALIGNED(256, static const InterpKernel,

                 sub_pel_filters_4[SUBPEL_SHIFTS]) = {

-  { 0, 0, 0, 128, 0, 0, 0, 0 },     { 0, 0, -3, 125, 8, -2, 0, 0 },

-  { 0, 0, -6, 120, 18, -4, 0, 0 },  { 0, 0, -8, 115, 27, -6, 0, 0 },

-  { 0, 0, -10, 108, 37, -7, 0, 0 }, { 0, 0, -11, 101, 47, -9, 0, 0 },

-  { 0, 0, -11, 93, 56, -10, 0, 0 }, { 0, 0, -12, 85, 66, -11, 0, 0 },

-  { 0, 0, -11, 75, 75, -11, 0, 0 }, { 0, 0, -11, 66, 85, -12, 0, 0 },

-  { 0, 0, -10, 56, 93, -11, 0, 0 }, { 0, 0, -9, 47, 101, -11, 0, 0 },

-  { 0, 0, -7, 37, 108, -10, 0, 0 }, { 0, 0, -6, 27, 115, -8, 0, 0 },

-  { 0, 0, -4, 18, 120, -6, 0, 0 },  { 0, 0, -2, 8, 125, -3, 0, 0 }

+  { 0, 0, 0, 128, 0, 0, 0, 0 },     { 0, 0, -4, 126, 8, -2, 0, 0 },

+  { 0, 0, -6, 120, 18, -4, 0, 0 },  { 0, 0, -8, 114, 28, -6, 0, 0 },

+  { 0, 0, -10, 108, 36, -6, 0, 0 }, { 0, 0, -12, 102, 46, -8, 0, 0 },

+  { 0, 0, -12, 94, 56, -10, 0, 0 }, { 0, 0, -12, 84, 66, -10, 0, 0 },

+  { 0, 0, -12, 76, 76, -12, 0, 0 }, { 0, 0, -10, 66, 84, -12, 0, 0 },

+  { 0, 0, -10, 56, 94, -12, 0, 0 }, { 0, 0, -8, 46, 102, -12, 0, 0 },

+  { 0, 0, -6, 36, 108, -10, 0, 0 }, { 0, 0, -6, 28, 114, -8, 0, 0 },

+  { 0, 0, -4, 18, 120, -6, 0, 0 },  { 0, 0, -2, 8, 126, -4, 0, 0 }

 };

 const InterpKernel *vp9_filter_kernels[5] = {