diff options
Diffstat (limited to 'ML_BSSN/src/ML_BSSN_InitGamma.cc')
| -rw-r--r-- | ML_BSSN/src/ML_BSSN_InitGamma.cc | 126 |
1 files changed, 81 insertions, 45 deletions
diff --git a/ML_BSSN/src/ML_BSSN_InitGamma.cc b/ML_BSSN/src/ML_BSSN_InitGamma.cc index 18a9b97..ab44fd1 100644 --- a/ML_BSSN/src/ML_BSSN_InitGamma.cc +++ b/ML_BSSN/src/ML_BSSN_InitGamma.cc @@ -13,13 +13,14 @@ #include "GenericFD.h" #include "Differencing.h" #include "loopcontrol.h" +#include "vectors.h" /* Define macros used in calculations */ #define INITVALUE (42) #define QAD(x) (SQR(SQR(x))) -#define INV(x) ((1.0) / (x)) -#define SQR(x) ((x) * (x)) -#define CUB(x) ((x) * (x) * (x)) +#define INV(x) (kdiv(ToReal(1.0),x)) +#define SQR(x) (kmul(x,x)) +#define CUB(x) (kmul(x,SQR(x))) static void ML_BSSN_InitGamma_Body(cGH const * restrict const cctkGH, int const dir, int const face, CCTK_REAL const normal[3], CCTK_REAL const tangentA[3], CCTK_REAL const tangentB[3], int const min[3], int const max[3], int const n_subblock_gfs, CCTK_REAL * restrict const subblock_gfs[]) { @@ -52,47 +53,48 @@ static void ML_BSSN_InitGamma_Body(cGH const * restrict const cctkGH, int const ptrdiff_t const cdi = sizeof(CCTK_REAL) * di; ptrdiff_t const cdj = sizeof(CCTK_REAL) * dj; ptrdiff_t const cdk = sizeof(CCTK_REAL) * dk; - CCTK_REAL const dx = ToReal(CCTK_DELTA_SPACE(0)); - CCTK_REAL const dy = ToReal(CCTK_DELTA_SPACE(1)); - CCTK_REAL const dz = ToReal(CCTK_DELTA_SPACE(2)); - CCTK_REAL const dt = ToReal(CCTK_DELTA_TIME); - CCTK_REAL const dxi = INV(dx); - CCTK_REAL const dyi = INV(dy); - CCTK_REAL const dzi = INV(dz); - CCTK_REAL const khalf = 0.5; - CCTK_REAL const kthird = 1/3.0; - CCTK_REAL const ktwothird = 2.0/3.0; - CCTK_REAL const kfourthird = 4.0/3.0; - CCTK_REAL const keightthird = 8.0/3.0; - CCTK_REAL const hdxi = 0.5 * dxi; - CCTK_REAL const hdyi = 0.5 * dyi; - CCTK_REAL const hdzi = 0.5 * dzi; + CCTK_REAL_VEC const dx = ToReal(CCTK_DELTA_SPACE(0)); + CCTK_REAL_VEC const dy = ToReal(CCTK_DELTA_SPACE(1)); + CCTK_REAL_VEC const dz = ToReal(CCTK_DELTA_SPACE(2)); + CCTK_REAL_VEC const dt = ToReal(CCTK_DELTA_TIME); + CCTK_REAL_VEC const dxi = INV(dx); + CCTK_REAL_VEC const dyi = INV(dy); + CCTK_REAL_VEC const dzi = INV(dz); + CCTK_REAL_VEC const khalf = ToReal(0.5); + CCTK_REAL_VEC const kthird = ToReal(1.0/3.0); + CCTK_REAL_VEC const ktwothird = ToReal(2.0/3.0); + CCTK_REAL_VEC const kfourthird = ToReal(4.0/3.0); + CCTK_REAL_VEC const keightthird = ToReal(8.0/3.0); + CCTK_REAL_VEC const hdxi = kmul(ToReal(0.5), dxi); + CCTK_REAL_VEC const hdyi = kmul(ToReal(0.5), dyi); + CCTK_REAL_VEC const hdzi = kmul(ToReal(0.5), dzi); /* Initialize predefined quantities */ - CCTK_REAL const p1o12dx = 0.0833333333333333333333333333333*INV(dx); - CCTK_REAL const p1o12dy = 0.0833333333333333333333333333333*INV(dy); - CCTK_REAL const p1o12dz = 0.0833333333333333333333333333333*INV(dz); - CCTK_REAL const p1o144dxdy = 0.00694444444444444444444444444444*INV(dx)*INV(dy); - CCTK_REAL const p1o144dxdz = 0.00694444444444444444444444444444*INV(dx)*INV(dz); - CCTK_REAL const p1o144dydz = 0.00694444444444444444444444444444*INV(dy)*INV(dz); - CCTK_REAL const p1o24dx = 0.0416666666666666666666666666667*INV(dx); - CCTK_REAL const p1o24dy = 0.0416666666666666666666666666667*INV(dy); - CCTK_REAL const p1o24dz = 0.0416666666666666666666666666667*INV(dz); - CCTK_REAL const p1o64dx = 0.015625*INV(dx); - CCTK_REAL const p1o64dy = 0.015625*INV(dy); - CCTK_REAL const p1o64dz = 0.015625*INV(dz); - CCTK_REAL const p1odx = INV(dx); - CCTK_REAL const p1ody = INV(dy); - CCTK_REAL const p1odz = INV(dz); - CCTK_REAL const pm1o12dx2 = -0.0833333333333333333333333333333*INV(SQR(dx)); - CCTK_REAL const pm1o12dy2 = -0.0833333333333333333333333333333*INV(SQR(dy)); - CCTK_REAL const pm1o12dz2 = -0.0833333333333333333333333333333*INV(SQR(dz)); + CCTK_REAL_VEC const p1o12dx = kmul(INV(dx),ToReal(0.0833333333333333333333333333333)); + CCTK_REAL_VEC const p1o12dy = kmul(INV(dy),ToReal(0.0833333333333333333333333333333)); + CCTK_REAL_VEC const p1o12dz = kmul(INV(dz),ToReal(0.0833333333333333333333333333333)); + CCTK_REAL_VEC const p1o144dxdy = kmul(INV(dx),kmul(INV(dy),ToReal(0.00694444444444444444444444444444))); + CCTK_REAL_VEC const p1o144dxdz = kmul(INV(dx),kmul(INV(dz),ToReal(0.00694444444444444444444444444444))); + CCTK_REAL_VEC const p1o144dydz = kmul(INV(dy),kmul(INV(dz),ToReal(0.00694444444444444444444444444444))); + CCTK_REAL_VEC const p1o24dx = kmul(INV(dx),ToReal(0.0416666666666666666666666666667)); + CCTK_REAL_VEC const p1o24dy = kmul(INV(dy),ToReal(0.0416666666666666666666666666667)); + CCTK_REAL_VEC const p1o24dz = kmul(INV(dz),ToReal(0.0416666666666666666666666666667)); + CCTK_REAL_VEC const p1o64dx = kmul(INV(dx),ToReal(0.015625)); + CCTK_REAL_VEC const p1o64dy = kmul(INV(dy),ToReal(0.015625)); + CCTK_REAL_VEC const p1o64dz = kmul(INV(dz),ToReal(0.015625)); + CCTK_REAL_VEC const p1odx = INV(dx); + CCTK_REAL_VEC const p1ody = INV(dy); + CCTK_REAL_VEC const p1odz = INV(dz); + CCTK_REAL_VEC const pm1o12dx2 = kmul(INV(SQR(dx)),ToReal(-0.0833333333333333333333333333333)); + CCTK_REAL_VEC const pm1o12dy2 = kmul(INV(SQR(dy)),ToReal(-0.0833333333333333333333333333333)); + CCTK_REAL_VEC const pm1o12dz2 = kmul(INV(SQR(dz)),ToReal(-0.0833333333333333333333333333333)); /* Loop over the grid points */ #pragma omp parallel - LC_LOOP3 (ML_BSSN_InitGamma, + LC_LOOP3VEC (ML_BSSN_InitGamma, i,j,k, min[0],min[1],min[2], max[0],max[1],max[2], - cctk_lsh[0],cctk_lsh[1],cctk_lsh[2]) + cctk_lsh[0],cctk_lsh[1],cctk_lsh[2], + CCTK_REAL_VEC_SIZE) { ptrdiff_t const index = di*i + dj*j + dk*k; @@ -105,18 +107,52 @@ static void ML_BSSN_InitGamma_Body(cGH const * restrict const cctkGH, int const /* Precompute derivatives */ /* Calculate temporaries and grid functions */ - CCTK_REAL Xt1L = 0; + CCTK_REAL_VEC Xt1L = ToReal(0); - CCTK_REAL Xt2L = 0; + CCTK_REAL_VEC Xt2L = ToReal(0); - CCTK_REAL Xt3L = 0; + CCTK_REAL_VEC Xt3L = ToReal(0); + + /* If necessary, store only partial vectors after the first iteration */ + + if (CCTK_REAL_VEC_SIZE > 2 && CCTK_BUILTIN_EXPECT(i < lc_imin && i+CCTK_REAL_VEC_SIZE > lc_imax, 0)) + { + ptrdiff_t const elt_count_lo = lc_imin-i; + ptrdiff_t const elt_count_hi = lc_imax-i; + vec_store_nta_partial_mid(Xt1[index],Xt1L,elt_count_lo,elt_count_hi); + vec_store_nta_partial_mid(Xt2[index],Xt2L,elt_count_lo,elt_count_hi); + vec_store_nta_partial_mid(Xt3[index],Xt3L,elt_count_lo,elt_count_hi); + break; + } + + /* If necessary, store only partial vectors after the first iteration */ + + if (CCTK_REAL_VEC_SIZE > 1 && CCTK_BUILTIN_EXPECT(i < lc_imin, 0)) + { + ptrdiff_t const elt_count = lc_imin-i; + vec_store_nta_partial_hi(Xt1[index],Xt1L,elt_count); + vec_store_nta_partial_hi(Xt2[index],Xt2L,elt_count); + vec_store_nta_partial_hi(Xt3[index],Xt3L,elt_count); + continue; + } + + /* If necessary, store only partial vectors after the last iteration */ + + if (CCTK_REAL_VEC_SIZE > 1 && CCTK_BUILTIN_EXPECT(i+CCTK_REAL_VEC_SIZE > lc_imax, 0)) + { + ptrdiff_t const elt_count = lc_imax-i; + vec_store_nta_partial_lo(Xt1[index],Xt1L,elt_count); + vec_store_nta_partial_lo(Xt2[index],Xt2L,elt_count); + vec_store_nta_partial_lo(Xt3[index],Xt3L,elt_count); + break; + } /* Copy local copies back to grid functions */ - Xt1[index] = Xt1L; - Xt2[index] = Xt2L; - Xt3[index] = Xt3L; + vec_store_nta(Xt1[index],Xt1L); + vec_store_nta(Xt2[index],Xt2L); + vec_store_nta(Xt3[index],Xt3L); } - LC_ENDLOOP3 (ML_BSSN_InitGamma); + LC_ENDLOOP3VEC (ML_BSSN_InitGamma); } extern "C" void ML_BSSN_InitGamma(CCTK_ARGUMENTS) |