diff options
Diffstat (limited to 'Runtime/Math/Simd/sse.h')
| -rw-r--r-- | Runtime/Math/Simd/sse.h | 237 |
1 files changed, 237 insertions, 0 deletions
diff --git a/Runtime/Math/Simd/sse.h b/Runtime/Math/Simd/sse.h new file mode 100644 index 0000000..59796d7 --- /dev/null +++ b/Runtime/Math/Simd/sse.h @@ -0,0 +1,237 @@ +#ifndef SIMD_SSE_H +#define SIMD_SSE_H + +#include <xmmintrin.h> +#include <emmintrin.h> + +typedef __m128 vec4f; // vector 4 packed +typedef __m128 vec4fs; // vector 4 scalar +typedef __m128 vec4b; // vector 4 bool packed +typedef __m128 vec4bs; // vector 4 bool scalar + +#define SWZ_MASK(x, y, z, w) _MM_SHUFFLE(w,z,y,x) + +#define cvec4f(name, x,y,z,w) static const vec4f name = {x,y,z,w} +#define cvec4b(name, x,y,z,w) static const vec4b name = {x,y,z,w} +#define cvec4fs(name, s) static const vec4f name = {s,s,s,s} + +enum simd_mask +{ + kXYZW = SWZ_MASK(0,1,2,3), + kXXXX = SWZ_MASK(0,0,0,0), + kYYYY = SWZ_MASK(1,1,1,1), + kZZZZ = SWZ_MASK(2,2,2,2), + kWWWW = SWZ_MASK(3,3,3,3), + + kXWYZ = SWZ_MASK(0,3,1,2), + kXZWY = SWZ_MASK(0,2,3,1), + + kYZWX = SWZ_MASK(1,2,3,0), + kYXZW = SWZ_MASK(1,0,2,3), + kYWZX = SWZ_MASK(1,3,2,0), + kYZXW = SWZ_MASK(1,2,0,3), + kYXWZ = SWZ_MASK(1,0,3,2), + + kZWXY = SWZ_MASK(2,3,0,1), + kZYXW = SWZ_MASK(2,1,0,3), + kZYWX = SWZ_MASK(2,1,3,0), + kZXYW = SWZ_MASK(2,0,1,3), + + kWYZX = SWZ_MASK(3,1,2,0), + kWXZY = SWZ_MASK(3,0,2,1), + kWYXZ = SWZ_MASK(3,1,0,2), + kWWWZ = SWZ_MASK(3,3,3,2), + kWWZZ = SWZ_MASK(3,3,2,2), + kWZYX = SWZ_MASK(3,2,1,0), +}; + +#define Vzero() _mm_setzero_ps() + +#define Vone() _mm_set1_ps(1.f) + +#define Vpermute(v, mask) _mm_shuffle_ps( (v), (v), (mask) ) + +#define Vmove(l, r) _mm_move_ss( (l), (r) ) + + +template<int SWZ> struct Vswizzle +{ + static MECANIM_FORCE_INLINE vec4f rhs(vec4f r) + { + return Vpermute(r, SWZ); + } + + static MECANIM_FORCE_INLINE vec4f lhs(vec4f l, vec4f r) + { + vec4f m = Vmove(Vswizzle<SWZ>::rhs(l), r); + return Vswizzle<SWZ>::rhs(m); + } +}; + +template<> struct Vswizzle<kXYZW> +{ + static MECANIM_FORCE_INLINE vec4f rhs(vec4f r) + { + return r; + } + static MECANIM_FORCE_INLINE vec4f lhs(vec4f l, vec4f r) + { + return Vmove(l, r); + } +}; + +static MECANIM_FORCE_INLINE float Vstoresf(vec4f r) +{ + float f; _mm_store_ss(&f, r); return f; +} + +#define Vstoresb(r) ( _mm_movemask_ps((r)) & 0x1<<0) != 0 + +// Aligned store +#define Vstorepf(v, base, offset) _mm_store_ps((base)+(offset), (v)) + +static MECANIM_FORCE_INLINE void Vstorepb(vec4f v, bool* r) +{ + r[0] = ( _mm_movemask_ps(v) & 0x1<<0) != 0; + r[1] = ( _mm_movemask_ps(v) & 0x1<<1) != 0; + r[2] = ( _mm_movemask_ps(v) & 0x1<<2) != 0; + r[3] = ( _mm_movemask_ps(v) & 0x1<<3) != 0; +} + +#define Vloadsf(s) _mm_set1_ps(s) + +static MECANIM_FORCE_INLINE vec4f Vloadsb(bool s) +{ + union { + int b[4]; + vec4f v; + } static const false_true[2] = { + {0,0,0,0}, + {~0,~0,~0,~0} + }; + + return false_true[s].v; +} + +#define Vload4sf(x, y, z, w) _mm_set_ps(w, z, y, x) + +static MECANIM_FORCE_INLINE vec4f Vload4sb(bool x, bool y, bool z, bool w) +{ + union { + int b; + float f; + } static const false_true[2] = { + 0,~0 + }; + + return _mm_set_ps(false_true[w].f, false_true[z].f, false_true[y].f, false_true[x].f); +} + +#define Vloadpf(v, offset) _mm_load_ps( (v)+(offset)) + +#define Vadd(l, r) _mm_add_ps((l), (r)) + +#define Vsub( l, r) _mm_sub_ps((l), (r)) + +#define Vmul( l, r) _mm_mul_ps((l), (r)) + +#define Vdiv( l, r) _mm_div_ps((l), (r)) + +#define Vmadd( a, b, c) _mm_add_ps(_mm_mul_ps((a), (b)), (c)) + +#define Vmsub( a, b, c) _mm_sub_ps(_mm_mul_ps((a), (b)), (c)) + + +static MECANIM_FORCE_INLINE vec4f Vneg(vec4f r) +{ + static const vec4f sign_constant = {-0.f,-0.f,-0.f,-0.f}; + return _mm_xor_ps( (r), sign_constant); +} + +// vector sgn: return -1, 1 +static MECANIM_FORCE_INLINE vec4f Vsgn(vec4f r) +{ + static const vec4f sign_constant = {-0.f,-0.f,-0.f,-0.f}; + return _mm_or_ps(Vone(), _mm_and_ps(sign_constant, (r) )); +} + +// vector sgn: return -1, 0, 1 +static MECANIM_FORCE_INLINE vec4f Vsign(vec4f r) +{ + static const vec4f sign_constant = {-0.f,-0.f,-0.f,-0.f}; + return _mm_or_ps( _mm_and_ps( _mm_cmpneq_ps(r, Vzero()), Vone()), _mm_and_ps(sign_constant, r )); +} + +#define Vinc(r) Vadd( (r), Vone()) +#define Vdec(r) Vsub( (r), Vone()) + +static MECANIM_FORCE_INLINE vec4f Vabs(vec4f r) +{ + static const vec4f sign_constant = {-0.f,-0.f,-0.f,-0.f}; + return _mm_andnot_ps(sign_constant, (r)); +} + +#define Vmax( l, r) _mm_max_ps((l), (r)) +#define Vmin( l, r) _mm_min_ps((l), (r)) + +static MECANIM_FORCE_INLINE vec4fs Vlargest(vec4f r) +{ + r = Vmax(r, Vswizzle<kYZWX>::rhs(r)); + r = Vmax(r, Vswizzle<kZWXY>::rhs(r)); + return r; +} + +static MECANIM_FORCE_INLINE vec4fs Vsmallest(vec4f r) +{ + r = Vmin(r, Vswizzle<kYZWX>::rhs(r)); + r = Vmin(r, Vswizzle<kZWXY>::rhs(r)); + return r; +} + +static MECANIM_FORCE_INLINE vec4fs Vsum(vec4f r) +{ + r = Vadd(r, Vswizzle<kYZWX>::rhs(r) ); + r = Vadd(r, Vswizzle<kZWXY>::rhs(r) ); + return Vswizzle<kXXXX>::rhs(r); +} + +#define Vdot( l, r) Vsum( Vmul((l), (r)) ) +#define Vsqrt(r) _mm_sqrt_ps((r)) + +static MECANIM_FORCE_INLINE vec4f Vrsqrt(vec4f r) +{ + vec4f const e = _mm_rsqrt_ps(r); + return Vmul(Vmul(e, Vsub(_mm_set1_ps(3.0f), Vmul(Vmul(e,e),r))), _mm_set1_ps(.5f)); +} + +static MECANIM_FORCE_INLINE vec4f Vrcp(vec4f r) +{ + vec4f e = _mm_rcp_ps( r ); + return Vsub( Vadd(e, e), Vmul(r, Vmul(e, e))); +} + +#define Vcombine(x,y,z,w) _mm_movelh_ps(_mm_unpacklo_ps( (x), (y) ), _mm_unpacklo_ps((z), (w))) + +// Vector comparison +#define Vcmpeq( a, b) _mm_cmpeq_ps((a), (b)) +#define Vcmpneq( a, b) _mm_cmpneq_ps((a), (b)) +#define Vcmpgt( a, b) _mm_cmpgt_ps((a), (b)) +#define Vcmpge( a, b) _mm_cmpge_ps((a), (b)) +#define Vcmplt( a, b) _mm_cmplt_ps((a), (b)) +#define Vcmple( a, b) _mm_cmple_ps((a), (b)) + +static MECANIM_FORCE_INLINE vec4f Vsel( vec4f c, vec4f a, vec4f b) +{ + return _mm_xor_ps(b, _mm_and_ps(_mm_xor_ps(a, b), c)); +} + +// vector logics +#define Vnot(r) _mm_cmpeq_ps( (r), Vzero() ) +#define Vxnor( a, b) Vnot(_mm_xor_ps((a), (b))) +#define Vxor( a, b) _mm_xor_ps((a), (b)) +#define Vand( a, b) _mm_and_ps((a), (b)) +#define Vor( a, b) _mm_or_ps((a), (b)) +#define Vall(a) (_mm_movemask_ps((a)) & 0xf) == 0xf +#define Vany( a) _mm_movemask_ps((a)) != 0 + +#endif |