MORTAL2000 · January 29, 2020 11:29 · Feb 12, 2016
diff --git a/vfc5.cpp b/vfc5.cpp
@@ -0,0 +1,156 @@
+#include <stdbool.h>
+#include <spu_intrinsics.h>
+
+// shuffle helpers
+#define L0 0x00010203
+#define L1 0x04050607
+#define L2 0x08090a0b
+#define L3 0x0c0d0e0f
+
+#define R0 0x10111213
+#define R1 0x14151617
+#define R2 0x18191a1b
+#define R3 0x1c1d1e1f
+
+#define ZERO 0x80808080
+
+#define SHUFFLE(l, r, x, y, z, w) si_shufb(l, r, ((qword)(vec_uint4){x, y, z, w}))
+
+// splat helper
+#define SPLAT(v, idx) si_shufb(v, v, (qword)(vec_uint4)(L ## idx))
+
+struct matrix_t
+{
+    vec_float4 row0;
+    vec_float4 row1;
+    vec_float4 row2;
+    vec_float4 row3;
+};
+
+struct aabb_t
+{
+    vec_float4 min;
+    vec_float4 max;
+};
+
+static inline void transform_points_4(qword* dest, qword x, qword y, qword z, const struct matrix_t* mat)
+{
+#define COMP(c) \
+    qword res_ ## c = SPLAT((qword)mat->row3, c); \
+    res_ ## c = si_fma(z, SPLAT((qword)mat->row2, c), res_ ## c); \
+    res_ ## c = si_fma(y, SPLAT((qword)mat->row1, c), res_ ## c); \
+    res_ ## c = si_fma(x, SPLAT((qword)mat->row0, c), res_ ## c); \
+    dest[c] = res_ ## c;
+
+    COMP(0);
+    COMP(1);
+    COMP(2);
+    COMP(3);
+
+#undef COMP
+}
+
+static inline void transform_matrix(struct matrix_t* dest, const struct matrix_t* lhs, const struct matrix_t* rhs)
+{
+#define COMP_0(c) \
+    qword res_ ## c = si_fm((qword)lhs->row2, SPLAT((qword)rhs->row ## c, 2)); \
+    res_ ## c = si_fma((qword)lhs->row1, SPLAT((qword)rhs->row ## c, 1), res_ ## c); \
+    res_ ## c = si_fma((qword)lhs->row0, SPLAT((qword)rhs->row ## c, 0), res_ ## c); \
+    dest->row ## c = (vec_float4)res_ ## c;
+
+#define COMP_1(c) \
+    qword res_ ## c = si_fma((qword)lhs->row2, SPLAT((qword)rhs->row ## c, 2), (qword)lhs->row3); \
+    res_ ## c = si_fma((qword)lhs->row1, SPLAT((qword)rhs->row ## c, 1), res_ ## c); \
+    res_ ## c = si_fma((qword)lhs->row0, SPLAT((qword)rhs->row ## c, 0), res_ ## c); \
+    dest->row ## c = (vec_float4)res_ ## c;
+
+    COMP_0(0);
+    COMP_0(1);
+    COMP_0(2);
+    COMP_1(3);
+
+#undef COMP_0
+#undef COMP_1
+}
+
+__attribute__((noinline)) unsigned int is_visible(const struct matrix_t* transform, const struct aabb_t* aabb, const struct matrix_t* frustum)
+{
+    qword min = (qword)aabb->min;
+    qword max = (qword)aabb->max;
+
+    // get aabb points (SoA)
+    qword minmax_x = SHUFFLE(min, max, L0, R0, L0, R0); // x X x X
+    qword minmax_y = SHUFFLE(min, max, L1, L1, R1, R1); // y y Y Y
+    qword minmax_z_0 = SPLAT(min, 2); // z z z z
+    qword minmax_z_1 = SPLAT(max, 2); // Z Z Z Z
+
+    // get clipping matrix
+    struct matrix_t clip;
+
+    transform_matrix(&clip, frustum, transform);
+
+    // transform points to clip space
+    qword points_cs_0[4];
+    qword points_cs_1[4];
+
+    transform_points_4(points_cs_0, minmax_x, minmax_y, minmax_z_0, &clip);
+    transform_points_4(points_cs_1, minmax_x, minmax_y, minmax_z_1, &clip);
+
+    // calculate -w
+    qword points_cs_0_negw = si_xor(points_cs_0[3], (qword)(vec_uint4)(0x80000000));
+    qword points_cs_1_negw = si_xor(points_cs_1[3], (qword)(vec_uint4)(0x80000000));
+
+    // for each plane...
+    #define NOUT(a, b, c, d) si_orx(si_or(si_fcgt(a, b), si_fcgt(c, d)))
+
+    qword nout0 = NOUT(points_cs_0[0], points_cs_0_negw, points_cs_1[0], points_cs_1_negw);
+    qword nout1 = NOUT(points_cs_0[3], points_cs_0[0], points_cs_1[3], points_cs_1[0]);
+    qword nout2 = NOUT(points_cs_0[1], points_cs_0_negw, points_cs_1[1], points_cs_1_negw);
+    qword nout3 = NOUT(points_cs_0[3], points_cs_0[1], points_cs_1[3], points_cs_1[1]);
+    qword nout4 = NOUT(points_cs_0[2], (qword)(0), points_cs_1[2], (qword)(0));
+    qword nout5 = NOUT(points_cs_0[3], points_cs_0[2], points_cs_1[3], points_cs_1[2]);
+
+    #undef NOUT
+
+    // merge "not outside" flags
+    qword nout01 = si_and(nout0, nout1);
+    qword nout012 = si_and(nout01, nout2);
+
+    qword nout34 = si_and(nout3, nout4);
+    qword nout345 = si_and(nout34, nout5);
+
+    qword nout = si_and(nout012, nout345);
+
+    return si_to_uint(nout);
+}
+
+// simple ortho frustum
+struct matrix_t frustum =
+{
+    { 0.1f, 0, 0, 0 },
+    { 0, 0.1f, 0, 0 },
+    { 0, 0, 0.1f, 0 },
+    { 0, 0, 0, 1 }
+};
+
+// small box
+struct aabb_t aabb =
+{
+    { -1, -2, -3 },
+    { 1, 2, 3 }
+};
+
+// and some weird matrix
+struct matrix_t transform =
+{
+    { 0.123f, 0.456f, 0.789f },
+    { 0.456f, 0.123f, 0.789f },
+    { 0.789f, 0.123f, 0.456f },
+    { 1.f, -1.f, 1.f }
+};
+
+int main()
+{
+    is_visible(&transform, &aabb, &frustum);
+    si_stop(0);
+}