Sk4x4f: NEON impl.
Notable tricks:
- v{ld,st}4q_f32 handle transposing loads and stores of floats in one step
- vcvtq_n_{f32_u32,u32_f32} let us do conversion to and from floats without shifts
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1828613002
Review URL: https://codereview.chromium.org/1828613002
diff --git a/src/core/Sk4x4f.h b/src/core/Sk4x4f.h
index d289334..9bd9197 100644
--- a/src/core/Sk4x4f.h
+++ b/src/core/Sk4x4f.h
@@ -17,13 +17,17 @@
static Sk4x4f Transpose(const float[16]);
static Sk4x4f Transpose(const uint8_t[16]);
- void transpose(Sk4f*, Sk4f*, Sk4f*, Sk4f*) const;
+ void transpose(Sk4f* x, Sk4f* y, Sk4f* z, Sk4f* w) const {
+ auto t = Transpose(r,g,b,a);
+ *x = t.r;
+ *y = t.g;
+ *z = t.b;
+ *w = t.a;
+ }
void transpose( float[16]) const;
void transpose(uint8_t[16]) const;
};
-// TODO: NEON
-
#if 1 && !defined(SKNX_NO_SIMD) && SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSE2
inline Sk4x4f Sk4x4f::Transpose(const Sk4f& x, const Sk4f& y, const Sk4f& z, const Sk4f& w) {
@@ -50,18 +54,6 @@
return { r,g,b,a };
}
-inline void Sk4x4f::transpose(Sk4f* x, Sk4f* y, Sk4f* z, Sk4f* w) const {
- auto R = r.fVec,
- G = g.fVec,
- B = b.fVec,
- A = a.fVec;
- _MM_TRANSPOSE4_PS(R,G,B,A);
- *x = R;
- *y = G;
- *z = B;
- *w = A;
-}
-
inline void Sk4x4f::transpose(float fs[16]) const {
Sk4f x,y,z,w;
this->transpose(&x,&y,&z,&w);
@@ -79,6 +71,43 @@
_mm_storeu_si128((__m128i*)bs, _mm_or_si128(A, _mm_or_si128(B, _mm_or_si128(G, R))));
}
+#elif defined(SK_ARM_HAS_NEON)
+
+inline Sk4x4f Sk4x4f::Transpose(const Sk4f& x, const Sk4f& y, const Sk4f& z, const Sk4f& w) {
+ float32x4x2_t xy = vuzpq_f32(x.fVec, y.fVec),
+ zw = vuzpq_f32(z.fVec, w.fVec),
+ rb = vuzpq_f32(xy.val[0], zw.val[0]),
+ ga = vuzpq_f32(xy.val[1], zw.val[1]);
+ return { rb.val[0], ga.val[0], rb.val[1], ga.val[1] };
+}
+
+inline Sk4x4f Sk4x4f::Transpose(const float fs[16]) {
+ float32x4x4_t v = vld4q_f32(fs);
+ return { v.val[0], v.val[1], v.val[2], v.val[3] };
+}
+
+inline Sk4x4f Sk4x4f::Transpose(const uint8_t bs[16]) {
+ auto b16 = vreinterpretq_u32_u8(vld1q_u8(bs));
+ auto r = vcvtq_f32_u32(vandq_u32(vdupq_n_u32(0x000000FF), b16) ),
+ g = vcvtq_n_f32_u32(vandq_u32(vdupq_n_u32(0x0000FF00), b16), 8),
+ b = vcvtq_n_f32_u32(vandq_u32(vdupq_n_u32(0x00FF0000), b16), 16),
+ a = vcvtq_n_f32_u32(vandq_u32(vdupq_n_u32(0xFF000000), b16), 24);
+ return { r,g,b,a };
+}
+
+inline void Sk4x4f::transpose(float fs[16]) const {
+ float32x4x4_t v = {{ r.fVec, g.fVec, b.fVec, a.fVec }};
+ vst4q_f32(fs, v);
+}
+
+inline void Sk4x4f::transpose(uint8_t bs[16]) const {
+ auto R = vandq_u32(vdupq_n_u32(0x000000FF), vcvtq_u32_f32(r.fVec )),
+ G = vandq_u32(vdupq_n_u32(0x0000FF00), vcvtq_n_u32_f32(g.fVec, 8)),
+ B = vandq_u32(vdupq_n_u32(0x00FF0000), vcvtq_n_u32_f32(b.fVec, 16)),
+ A = vandq_u32(vdupq_n_u32(0xFF000000), vcvtq_n_u32_f32(a.fVec, 24));
+ vst1q_u8(bs, vreinterpretq_u8_u32(vorrq_u32(A, vorrq_u32(B, vorrq_u32(G, R)))));
+}
+
#else
inline Sk4x4f Sk4x4f::Transpose(const Sk4f& x, const Sk4f& y, const Sk4f& z, const Sk4f& w) {
@@ -103,13 +132,6 @@
};
}
-inline void Sk4x4f::transpose(Sk4f* x, Sk4f* y, Sk4f* z, Sk4f* w) const {
- *x = { r[0], g[0], b[0], a[0] };
- *y = { r[1], g[1], b[1], a[1] };
- *z = { r[2], g[2], b[2], a[2] };
- *w = { r[3], g[3], b[3], a[3] };
-}
-
inline void Sk4x4f::transpose(float fs[16]) const {
Sk4f x,y,z,w;
this->transpose(&x,&y,&z,&w);