portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-09-20 09:15:38 +01:00
Code Releases Activity

NAMESPACE and formatting

Browse Source
This commit is contained in:
paboyle 2018-01-12 18:11:04 +00:00
parent bbb657da5c
commit 6ab744c720
1 changed files with 43 additions and 42 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

85
lib/simd/Grid_generic_types.h
View File

@ -1,4 +1,4 @@
/*************************************************************************************
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
@ -24,8 +24,8 @@ Author: Antonin Portelli <antonin.portelli@me.com>
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
See the full license in the file "LICENSE" in the top level distribution directory
*************************************************************************************/
/* END LEGAL */
*************************************************************************************/
/* END LEGAL */
static_assert(GEN_SIMD_WIDTH % 16u == 0, "SIMD vector size is not an integer multiple of 16 bytes");
@ -34,52 +34,53 @@ static_assert(GEN_SIMD_WIDTH % 16u == 0, "SIMD vector size is not an integer mul
// playing with compiler pragmas
#ifdef VECTOR_LOOPS
#ifdef __clang__
#define VECTOR_FOR(i, w, inc)\
_Pragma("clang loop unroll(full) vectorize(enable) interleave(enable) vectorize_width(w)")\
for (unsigned int i = 0; i < w; i += inc)
#define VECTOR_FOR(i, w, inc) \
_Pragma("clang loop unroll(full) vectorize(enable) interleave(enable) vectorize_width(w)") \
for (unsigned int i = 0; i < w; i += inc)
#elif defined __INTEL_COMPILER
#define VECTOR_FOR(i, w, inc)\
_Pragma("simd vectorlength(w*8)")\
for (unsigned int i = 0; i < w; i += inc)
#define VECTOR_FOR(i, w, inc) \
_Pragma("simd vectorlength(w*8)") \
for (unsigned int i = 0; i < w; i += inc)
#else
#define VECTOR_FOR(i, w, inc)\
for (unsigned int i = 0; i < w; i += inc)
#define VECTOR_FOR(i, w, inc) \
for (unsigned int i = 0; i < w; i += inc)
#endif
#else
#define VECTOR_FOR(i, w, inc)\
for (unsigned int i = 0; i < w; i += inc)
#define VECTOR_FOR(i, w, inc) \
for (unsigned int i = 0; i < w; i += inc)
#endif
namespace Grid {
namespace Optimization {
NAMESPACE_BEGIN(Grid);
NAMESPACE_BEGIN(Optimization);
// type traits giving the number of elements for each vector type
template <typename T> struct W;
template <> struct W<double> {
constexpr static unsigned int c = GEN_SIMD_WIDTH/16u;
constexpr static unsigned int r = GEN_SIMD_WIDTH/8u;
};
template <> struct W<float> {
constexpr static unsigned int c = GEN_SIMD_WIDTH/8u;
constexpr static unsigned int r = GEN_SIMD_WIDTH/4u;
};
template <> struct W<Integer> {
constexpr static unsigned int r = GEN_SIMD_WIDTH/4u;
};
template <> struct W<uint16_t> {
constexpr static unsigned int c = GEN_SIMD_WIDTH/4u;
constexpr static unsigned int r = GEN_SIMD_WIDTH/2u;
};
// type traits giving the number of elements for each vector type
template <typename T> struct W;
template <> struct W<double> {
constexpr static unsigned int c = GEN_SIMD_WIDTH/16u;
constexpr static unsigned int r = GEN_SIMD_WIDTH/8u;
};
template <> struct W<float> {
constexpr static unsigned int c = GEN_SIMD_WIDTH/8u;
constexpr static unsigned int r = GEN_SIMD_WIDTH/4u;
};
template <> struct W<Integer> {
constexpr static unsigned int r = GEN_SIMD_WIDTH/4u;
};
template <> struct W<uint16_t> {
constexpr static unsigned int c = GEN_SIMD_WIDTH/4u;
constexpr static unsigned int r = GEN_SIMD_WIDTH/2u;
};
// SIMD vector types
template <typename T>
struct vec {
alignas(GEN_SIMD_WIDTH) T v[W<T>::r];
};
// SIMD vector types
template <typename T>
struct vec {
alignas(GEN_SIMD_WIDTH) T v[W<T>::r];
};
typedef vec<float> vecf;
typedef vec<double> vecd;
typedef vec<uint16_t> vech; // half precision comms
typedef vec<Integer> veci;
typedef vec<float> vecf;
typedef vec<double> vecd;
typedef vec<uint16_t> vech; // half precision comms
typedef vec<Integer> veci;
}}
NAMESPACE_END(Optimization);
NAMESPACE_END(Grid);