class Simd4Float
{
- public:
- Simd4Float() {}
+public:
+ Simd4Float() {}
- // gcc-4.9 does not recognize that we use the parameter
- Simd4Float(float gmx_unused f) : simdInternal_(vec_splats(f)) {}
+ // gcc-4.9 does not recognize that we use the parameter
+ Simd4Float(float gmx_unused f) : simdInternal_(vec_splats(f)) {}
- // Internal utility constructor to simplify return statements
- Simd4Float(__vector float simd) : simdInternal_(simd) {}
+ // Internal utility constructor to simplify return statements
+ Simd4Float(__vector float simd) : simdInternal_(simd) {}
- __vector float simdInternal_;
+ __vector float simdInternal_;
};
class Simd4FBool
{
- public:
- Simd4FBool() {}
+public:
+ Simd4FBool() {}
- //! \brief Construct from scalar bool
- Simd4FBool(bool b) : simdInternal_(reinterpret_cast<__vector vsxBool int>(vec_splats( b ? 0xFFFFFFFF : 0))) {}
+ //! \brief Construct from scalar bool
+ Simd4FBool(bool b) :
+ simdInternal_(reinterpret_cast<__vector vsxBool int>(vec_splats(b ? 0xFFFFFFFF : 0)))
+ {
+ }
- // Internal utility constructor to simplify return statements
- Simd4FBool(__vector vsxBool int simd) : simdInternal_(simd) {}
+ // Internal utility constructor to simplify return statements
+ Simd4FBool(__vector vsxBool int simd) : simdInternal_(simd) {}
- __vector vsxBool int simdInternal_;
+ __vector vsxBool int simdInternal_;
};
// The VSX load & store operations are a bit of a mess. The interface is different
// revisit things once xlc is a bit more stable - for now you probably want
// to stick to gcc...
-static inline Simd4Float gmx_simdcall
-load4(const float *m)
+static inline Simd4Float gmx_simdcall load4(const float* m)
{
- return {
- *reinterpret_cast<const __vector float *>(m)
- };
+ return { *reinterpret_cast<const __vector float*>(m) };
}
-static inline void gmx_simdcall
-store4(float *m, Simd4Float a)
+static inline void gmx_simdcall store4(float* m, Simd4Float a)
{
- *reinterpret_cast<__vector float *>(m) = a.simdInternal_;
+ *reinterpret_cast<__vector float*>(m) = a.simdInternal_;
}
-static inline Simd4Float gmx_simdcall
-load4U(const float *m)
+static inline Simd4Float gmx_simdcall load4U(const float* m)
{
- return {
+ return
+ {
#if __GNUC__ < 7
- *reinterpret_cast<const __vector float *>(m)
+ *reinterpret_cast<const __vector float*>(m)
#else
- vec_xl(0, m)
+ vec_xl(0, m)
#endif
};
}
-static inline void gmx_simdcall
-store4U(float *m, Simd4Float a)
+static inline void gmx_simdcall store4U(float* m, Simd4Float a)
{
#if __GNUC__ < 7
- *reinterpret_cast<__vector float *>(m) = a.simdInternal_;
+ *reinterpret_cast<__vector float*>(m) = a.simdInternal_;
#else
vec_xst(a.simdInternal_, 0, m);
#endif
}
-static inline Simd4Float gmx_simdcall
-simd4SetZeroF()
+static inline Simd4Float gmx_simdcall simd4SetZeroF()
{
- return {
- vec_splats(0.0F)
- };
+ return { vec_splats(0.0F) };
}
-static inline Simd4Float gmx_simdcall
-operator&(Simd4Float a, Simd4Float b)
+static inline Simd4Float gmx_simdcall operator&(Simd4Float a, Simd4Float b)
{
- return {
- vec_and(a.simdInternal_, b.simdInternal_)
- };
+ return { vec_and(a.simdInternal_, b.simdInternal_) };
}
-static inline Simd4Float gmx_simdcall
-andNot(Simd4Float a, Simd4Float b)
+static inline Simd4Float gmx_simdcall andNot(Simd4Float a, Simd4Float b)
{
- return {
- vec_andc(b.simdInternal_, a.simdInternal_)
- };
+ return { vec_andc(b.simdInternal_, a.simdInternal_) };
}
-static inline Simd4Float gmx_simdcall
-operator|(Simd4Float a, Simd4Float b)
+static inline Simd4Float gmx_simdcall operator|(Simd4Float a, Simd4Float b)
{
- return {
- vec_or(a.simdInternal_, b.simdInternal_)
- };
+ return { vec_or(a.simdInternal_, b.simdInternal_) };
}
-static inline Simd4Float gmx_simdcall
-operator^(Simd4Float a, Simd4Float b)
+static inline Simd4Float gmx_simdcall operator^(Simd4Float a, Simd4Float b)
{
- return {
- vec_xor(a.simdInternal_, b.simdInternal_)
- };
+ return { vec_xor(a.simdInternal_, b.simdInternal_) };
}
-static inline Simd4Float gmx_simdcall
-operator+(Simd4Float a, Simd4Float b)
+static inline Simd4Float gmx_simdcall operator+(Simd4Float a, Simd4Float b)
{
- return {
- vec_add(a.simdInternal_, b.simdInternal_)
- };
+ return { vec_add(a.simdInternal_, b.simdInternal_) };
}
-static inline Simd4Float gmx_simdcall
-operator-(Simd4Float a, Simd4Float b)
+static inline Simd4Float gmx_simdcall operator-(Simd4Float a, Simd4Float b)
{
- return {
- vec_sub(a.simdInternal_, b.simdInternal_)
- };
+ return { vec_sub(a.simdInternal_, b.simdInternal_) };
}
-static inline Simd4Float gmx_simdcall
-operator-(Simd4Float x)
+static inline Simd4Float gmx_simdcall operator-(Simd4Float x)
{
- return {
- -x.simdInternal_
- };
+ return { -x.simdInternal_ };
}
-static inline Simd4Float gmx_simdcall
-operator*(Simd4Float a, Simd4Float b)
+static inline Simd4Float gmx_simdcall operator*(Simd4Float a, Simd4Float b)
{
- return {
- vec_mul(a.simdInternal_, b.simdInternal_)
- };
+ return { vec_mul(a.simdInternal_, b.simdInternal_) };
}
-static inline Simd4Float gmx_simdcall
-fma(Simd4Float a, Simd4Float b, Simd4Float c)
+static inline Simd4Float gmx_simdcall fma(Simd4Float a, Simd4Float b, Simd4Float c)
{
- return {
- vec_madd(a.simdInternal_, b.simdInternal_, c.simdInternal_)
- };
+ return { vec_madd(a.simdInternal_, b.simdInternal_, c.simdInternal_) };
}
-static inline Simd4Float gmx_simdcall
-fms(Simd4Float a, Simd4Float b, Simd4Float c)
+static inline Simd4Float gmx_simdcall fms(Simd4Float a, Simd4Float b, Simd4Float c)
{
- return {
- vec_msub(a.simdInternal_, b.simdInternal_, c.simdInternal_)
- };
+ return { vec_msub(a.simdInternal_, b.simdInternal_, c.simdInternal_) };
}
-static inline Simd4Float gmx_simdcall
-fnma(Simd4Float a, Simd4Float b, Simd4Float c)
+static inline Simd4Float gmx_simdcall fnma(Simd4Float a, Simd4Float b, Simd4Float c)
{
- return {
- vec_nmsub(a.simdInternal_, b.simdInternal_, c.simdInternal_)
- };
+ return { vec_nmsub(a.simdInternal_, b.simdInternal_, c.simdInternal_) };
}
-static inline Simd4Float gmx_simdcall
-fnms(Simd4Float a, Simd4Float b, Simd4Float c)
+static inline Simd4Float gmx_simdcall fnms(Simd4Float a, Simd4Float b, Simd4Float c)
{
- return {
- vec_nmadd(a.simdInternal_, b.simdInternal_, c.simdInternal_)
- };
+ return { vec_nmadd(a.simdInternal_, b.simdInternal_, c.simdInternal_) };
}
-static inline Simd4Float gmx_simdcall
-rsqrt(Simd4Float x)
+static inline Simd4Float gmx_simdcall rsqrt(Simd4Float x)
{
- return {
- vec_rsqrte(x.simdInternal_)
- };
+ return { vec_rsqrte(x.simdInternal_) };
}
-static inline Simd4Float gmx_simdcall
-abs(Simd4Float x)
+static inline Simd4Float gmx_simdcall abs(Simd4Float x)
{
- return {
- vec_abs( x.simdInternal_ )
- };
+ return { vec_abs(x.simdInternal_) };
}
-static inline Simd4Float gmx_simdcall
-max(Simd4Float a, Simd4Float b)
+static inline Simd4Float gmx_simdcall max(Simd4Float a, Simd4Float b)
{
- return {
- vec_max(a.simdInternal_, b.simdInternal_)
- };
+ return { vec_max(a.simdInternal_, b.simdInternal_) };
}
-static inline Simd4Float gmx_simdcall
-min(Simd4Float a, Simd4Float b)
+static inline Simd4Float gmx_simdcall min(Simd4Float a, Simd4Float b)
{
- return {
- vec_min(a.simdInternal_, b.simdInternal_)
- };
+ return { vec_min(a.simdInternal_, b.simdInternal_) };
}
-static inline Simd4Float gmx_simdcall
-round(Simd4Float x)
+static inline Simd4Float gmx_simdcall round(Simd4Float x)
{
- return {
- vec_round( x.simdInternal_ )
- };
+ return { vec_round(x.simdInternal_) };
}
-static inline Simd4Float gmx_simdcall
-trunc(Simd4Float x)
+static inline Simd4Float gmx_simdcall trunc(Simd4Float x)
{
- return {
- vec_trunc( x.simdInternal_ )
- };
+ return { vec_trunc(x.simdInternal_) };
}
-static inline float gmx_simdcall
-dotProduct(Simd4Float a, Simd4Float b)
+static inline float gmx_simdcall dotProduct(Simd4Float a, Simd4Float b)
{
const __vector unsigned char perm1 = { 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7 };
const __vector unsigned char perm2 = { 4, 5, 6, 7, 0, 1, 2, 3, 4, 5, 6, 7, 0, 1, 2, 3 };
return vec_extract(sum, 0);
}
-static inline void gmx_simdcall
-transpose(Simd4Float * v0, Simd4Float * v1,
- Simd4Float * v2, Simd4Float * v3)
+static inline void gmx_simdcall transpose(Simd4Float* v0, Simd4Float* v1, Simd4Float* v2, Simd4Float* v3)
{
__vector float t0 = vec_mergeh(v0->simdInternal_, v2->simdInternal_);
__vector float t1 = vec_mergel(v0->simdInternal_, v2->simdInternal_);
v3->simdInternal_ = vec_mergel(t1, t3);
}
-static inline Simd4FBool gmx_simdcall
-operator==(Simd4Float a, Simd4Float b)
+static inline Simd4FBool gmx_simdcall operator==(Simd4Float a, Simd4Float b)
{
- return {
- vec_cmpeq(a.simdInternal_, b.simdInternal_)
- };
+ return { vec_cmpeq(a.simdInternal_, b.simdInternal_) };
}
-static inline Simd4FBool gmx_simdcall
-operator!=(Simd4Float a, Simd4Float b)
+static inline Simd4FBool gmx_simdcall operator!=(Simd4Float a, Simd4Float b)
{
- return {
- vec_or(vec_cmpgt(a.simdInternal_, b.simdInternal_),
- vec_cmplt(a.simdInternal_, b.simdInternal_))
- };
+ return { vec_or(vec_cmpgt(a.simdInternal_, b.simdInternal_),
+ vec_cmplt(a.simdInternal_, b.simdInternal_)) };
}
-static inline Simd4FBool gmx_simdcall
-operator<(Simd4Float a, Simd4Float b)
+static inline Simd4FBool gmx_simdcall operator<(Simd4Float a, Simd4Float b)
{
- return {
- vec_cmplt(a.simdInternal_, b.simdInternal_)
- };
+ return { vec_cmplt(a.simdInternal_, b.simdInternal_) };
}
-static inline Simd4FBool gmx_simdcall
-operator<=(Simd4Float a, Simd4Float b)
+static inline Simd4FBool gmx_simdcall operator<=(Simd4Float a, Simd4Float b)
{
- return {
- vec_cmple(a.simdInternal_, b.simdInternal_)
- };
+ return { vec_cmple(a.simdInternal_, b.simdInternal_) };
}
-static inline Simd4FBool gmx_simdcall
-operator&&(Simd4FBool a, Simd4FBool b)
+static inline Simd4FBool gmx_simdcall operator&&(Simd4FBool a, Simd4FBool b)
{
- return {
- vec_and(a.simdInternal_, b.simdInternal_)
- };
+ return { vec_and(a.simdInternal_, b.simdInternal_) };
}
-static inline Simd4FBool gmx_simdcall
-operator||(Simd4FBool a, Simd4FBool b)
+static inline Simd4FBool gmx_simdcall operator||(Simd4FBool a, Simd4FBool b)
{
- return {
- vec_or(a.simdInternal_, b.simdInternal_)
- };
+ return { vec_or(a.simdInternal_, b.simdInternal_) };
}
-static inline bool gmx_simdcall
-anyTrue(Simd4FBool a)
+static inline bool gmx_simdcall anyTrue(Simd4FBool a)
{
return vec_any_ne(a.simdInternal_, reinterpret_cast<__vector vsxBool int>(vec_splats(0)));
}
-static inline Simd4Float gmx_simdcall
-selectByMask(Simd4Float a, Simd4FBool m)
+static inline Simd4Float gmx_simdcall selectByMask(Simd4Float a, Simd4FBool m)
{
- return {
- vec_and(a.simdInternal_, reinterpret_cast<__vector float>(m.simdInternal_))
- };
+ return { vec_and(a.simdInternal_, reinterpret_cast<__vector float>(m.simdInternal_)) };
}
-static inline Simd4Float gmx_simdcall
-selectByNotMask(Simd4Float a, Simd4FBool m)
+static inline Simd4Float gmx_simdcall selectByNotMask(Simd4Float a, Simd4FBool m)
{
- return {
- vec_andc(a.simdInternal_, reinterpret_cast<__vector float>(m.simdInternal_))
- };
+ return { vec_andc(a.simdInternal_, reinterpret_cast<__vector float>(m.simdInternal_)) };
}
-static inline Simd4Float gmx_simdcall
-blend(Simd4Float a, Simd4Float b, Simd4FBool sel)
+static inline Simd4Float gmx_simdcall blend(Simd4Float a, Simd4Float b, Simd4FBool sel)
{
- return {
- vec_sel(a.simdInternal_, b.simdInternal_, sel.simdInternal_)
- };
+ return { vec_sel(a.simdInternal_, b.simdInternal_, sel.simdInternal_) };
}
-static inline float gmx_simdcall
-reduce(Simd4Float x)
+static inline float gmx_simdcall reduce(Simd4Float x)
{
const __vector unsigned char perm1 = { 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7 };
const __vector unsigned char perm2 = { 4, 5, 6, 7, 0, 1, 2, 3, 4, 5, 6, 7, 0, 1, 2, 3 };
return vec_extract(x.simdInternal_, 0);
}
-} // namespace gmx
+} // namespace gmx
#endif // GMX_SIMD_IMPLEMENTATION_IBM_VSX_SIMD4_FLOAT_H
biod.pnpi.spb.ru / alexxy / gromacs.git / blobdiff