ARM Skia NEON patches - 29 - Xfermode: SkFourByteInterp

src/opts/SkColor_opts_neon.h

 #ifndef SkColor_opts_neon_DEFINED
 #define SkColor_opts_neon_DEFINED
 
 #include "SkTypes.h"
 #include "SkColorPriv.h"
 
 #include <arm_neon.h>
 
 #define NEON_A (SK_A32_SHIFT / 8)
 #define NEON_R (SK_R32_SHIFT / 8)
@@ skipping 64 matching lines @@
     src_wide = vreinterpretq_s16_u16(vmovl_u8(src));
     dst_wide = vreinterpretq_s16_u16(vmovl_u8(dst));
 
     src_wide = (src_wide - dst_wide) * vreinterpretq_s16_u16(scale);
 
     dst_wide += vshrq_n_s16(src_wide, 5);
 
     return vmovn_u16(vreinterpretq_u16_s16(dst_wide));
 }
 
+static inline SkPMColor SkFourByteInterp256_neon(SkPMColor src, SkPMColor dst,
+                                                 unsigned scale) {

[reed1, 2013/12/02 16:25:42]

Can we add SkASSERT(scale <= 256); at the beginning, just to help document the
subtle difference between this guy and the next.

Perhaps rename scale to srcWeight or srcScale, for clarity?

[kevin.petit.not.used.account, 2013/12/02 16:59:45]

Done.

+    int16x8_t vscale = vdupq_n_s16(scale);
+    int16x8_t vsrc_wide, vdst_wide, vdiff;
+    uint8x8_t res;
+
+    vsrc_wide = vreinterpretq_s16_u16(vmovl_u8(vreinterpret_u8_u32(vdup_n_u32(src))));
+    vdst_wide = vreinterpretq_s16_u16(vmovl_u8(vreinterpret_u8_u32(vdup_n_u32(dst))));
+
+    vdiff = vsrc_wide - vdst_wide;
+    vdiff *= vscale;
+
+    vdiff = vshrq_n_s16(vdiff, 8);
+
+    vdst_wide += vdiff;
+
+    res = vmovn_u16(vreinterpretq_u16_s16(vdst_wide));
+
+    return vget_lane_u32(vreinterpret_u32_u8(res), 0);
+}
+
+static inline SkPMColor SkFourByteInterp_neon(SkPMColor src, SkPMColor dst,
+                                              U8CPU srcWeight) {

[reed1, 2013/12/02 16:25:42]

SkASSERT(srcWeight <= 255);

[kevin.petit.not.used.account, 2013/12/02 16:59:45]

Done.

+    unsigned scale = SkAlpha255To256(srcWeight);
+    return SkFourByteInterp256_neon(src, dst, scale);
+}
+
 #endif /* #ifndef SkColor_opts_neon_DEFINED */