vec_kern_unr_pref.h

Go to the documentation of this file.

#ifndef TBCI_VEC_KERN_UNR_PREF_H
#define TBCI_VEC_KERN_UNR_PREF_H  
#include "tbci/basics.h"

/* "unroll_prefetch_def.h" is #include'd from basics.h */

NAMESPACE_TBCI

#define ADD3(r,v1,v2,f1,f2)     r = v1 + v2
VKERN_TEMPL_3V(do_vec_vec_add, ADD3);

#define SUB3(r,v1,v2,f1,f2)     r = v1 - v2
VKERN_TEMPL_3V(do_vec_vec_sub, SUB3);

#define MUL3(r,v1,v2,f1,f2)     r = v1 * v2
VKERN_TEMPL_3V(do_vec_vec_mul, MUL3);

#define CMUL3(r,v1,v2,f1,f2)    r = CPLX__ conj(v1) * v2
VKERN_TEMPL_3V(do_vec_vec_cmul, CMUL3);

#define DIV3(r,v1,v2,f1,f2)     r = v1 / v2
VKERN_TEMPL_3V(do_vec_vec_div, DIV3);

#define CDIV3(r,v1,v2,f1,f2)    r = CPLX__ conj(v1) / v2
VKERN_TEMPL_3V(do_vec_vec_cdiv, CDIV3);


#define ADD2(r,v1,f1,f2)        r += v1
VKERN_TEMPL_2V(do_vec_add_vec, ADD2);

#define SUB2(r,v1,f1,f2)        r -= v1
VKERN_TEMPL_2V(do_vec_sub_vec, SUB2);

#define SUB2I(r,v1,f1,f2)       r  = v1 - r
VKERN_TEMPL_2V(do_vec_sub_vec_inv, SUB2I);

#define MUL2(r,v1,f1,f2)        r *= v1
VKERN_TEMPL_2V(do_vec_mul_vec, MUL2);

#define CMUL2(r,v1,f1,f2)       r  = CPLX__ conj(r) * v1
VKERN_TEMPL_2V(do_vec_cmul_vec, CMUL2);

#define CMUL2I(r,v1,f1,f2)      r *= CPLX__ conj(v1)
VKERN_TEMPL_2V(do_vec_cmul_vec_inv, CMUL2I);

#define DIV2(r,v1,f1,f2)        r /= v1
VKERN_TEMPL_2V(do_vec_div_vec, DIV2);

#define DIV2I(r,v1,f1,f2)       r = v1 / r
VKERN_TEMPL_2V(do_vec_div_vec_inv, DIV2I);

#define CDIV2(r,v1,f1,f2)       r  = CPLX__ conj(r) / v1
VKERN_TEMPL_2V(do_vec_cdiv_vec, CDIV2);

#define CDIV2I(r,v1,f1,f2)      r  = CPLX__ conj(v1) / r
VKERN_TEMPL_2V(do_vec_cdiv_vec_inv, CDIV2I);



#define ADD2NV(r,v1,f1,f2)      r = v1 + f2
VKERN_TEMPL_2V_C(do_vec_val_add, ADD2NV);

#define SUB2NV(r,v1,f1,f2)      r = v1 - f2
VKERN_TEMPL_2V_C(do_vec_val_sub, SUB2NV);

#define MUL2NV(r,v1,f1,f2)      r = v1 * f2
VKERN_TEMPL_2V_C(do_vec_val_mul, MUL2NV);


#define ADD2RV(r,v1,f1,f2)      r = f2 + v1
VKERN_TEMPL_2V_C(do_val_vec_add, ADD2RV);

#define SUB2RV(r,v1,f1,f2)      r = f2 - v1
VKERN_TEMPL_2V_C(do_val_vec_sub, SUB2RV);

#define MUL2RV(r,v1,f1,f2)      r = f2 * v1
VKERN_TEMPL_2V_C(do_val_vec_mul, MUL2RV);

#define DIV2RV(r,v1,f1,f2)      r = f2 / v1
VKERN_TEMPL_2V_C(do_val_vec_div, DIV2RV);


#define ADD1NV(r,f1,f2)         r += f2
VKERN_TEMPL_1V_C(do_vec_add_val, ADD1NV);

#define SUB1NV(r,f1,f2)         r -= f2
VKERN_TEMPL_1V_C(do_vec_sub_val, SUB1NV);

#define SUB1RV(r,f1,f2)         r  = f2 - r
VKERN_TEMPL_1V_C(do_val_sub_vec, SUB1RV);

#define MUL1NV(r,f1,f2)         r *= f2
VKERN_TEMPL_1V_C(do_vec_mul_val, MUL1NV);

#define DIV1NV(r,f1,f2)         r /= f2
VKERN_TEMPL_1V_C(do_vec_div_val, DIV1NV);

#define DIV1RV(r,f1,f2)         r = f2 / r
VKERN_TEMPL_1V_C(do_val_div_vec, DIV1RV);

#define ADD1RV(r,f1,f2)         r  = f2 + r
VKERN_TEMPL_1V_C(do_val_add_vec, ADD1RV);

// unused ...
#define MUL1RV(r,f1,f2)         r  = f2 * r;


/* TSVector stuff */

#define ADD2NS(r,v1,f1,f2)      r += f2*v1
VKERN_TEMPL_2V_C(do_vec_add_svc, ADD2NS);

#define SUB2NS(r,v1,f1,f2)      r -= f2*v1
VKERN_TEMPL_2V_C(do_vec_sub_svc, SUB2NS);

#define SUB2RS(r,v1,f1,f2)      r  = f2*v1 - r
VKERN_TEMPL_2V_C(do_vec_sub_svc_inv, SUB2RS);


#define ADD3NS(r,v1,v2,f1,f2)   r = v1 + f2*v2
VKERN_TEMPL_3V_C(do_vec_svc_add, ADD3NS);

#define SUB3NS(r,v1,v2,f1,f2)   r = v1 - f2*v2
VKERN_TEMPL_3V_C(do_vec_svc_sub, SUB3NS);


#define ADD3SN(r,v1,v2,f1,f2)   r = f2*v1 + v2
VKERN_TEMPL_3V_C(do_svc_vec_add, ADD3SN);

#define SUB3SN(r,v1,v2,f1,f2)   r = f2*v1 - v2
VKERN_TEMPL_3V_C(do_svc_vec_sub, SUB3SN);


#define ADD3SS(r,v1,v2,f1,f2)   r = f1*v1 + f2*v2
VKERN_TEMPL_3V_CC(do_svc_svc_add, ADD3SS);

#define SUB3SS(r,v1,v2,f1,f2)   r = f1*v1 - f2*v2
VKERN_TEMPL_3V_CC(do_svc_svc_sub, SUB3SS);


#define ADD2SN(r,v1,f1,f2)      r  = f2*r + v1
VKERN_TEMPL_2V_C(do_svc_add_vec, ADD2SN);

#define SUB2SN(r,v1,f1,f2)      r  = f2*r - v1
VKERN_TEMPL_2V_C(do_svc_sub_vec, SUB2SN);


#define ADD2SS(r,v1,f1,f2)      r  = f1*r + f2*v1
VKERN_TEMPL_2V_CC(do_svc_add_svc, ADD2SS);

#define SUB2SS(r,v1,f1,f2)      r  = f1*r - f2*v1
VKERN_TEMPL_2V_CC(do_svc_sub_svc, SUB2SS);


#define ADD2SV(r,v1,f1,f2)      r = f1*v1 + f2
VKERN_TEMPL_2V_CC(do_svc_val_add, ADD2SV);

#define SUB2SV(r,v1,f1,f2)      r = f1*v1 - f2
VKERN_TEMPL_2V_CC(do_svc_val_sub, SUB2SV);


#define ADD1SV(r,f1,f2)         r = f1*r + f2
VKERN_TEMPL_1V_CC(do_svc_add_val, ADD1SV);

#define SUB1SV(r,f1,f2)         r = f1*r - f2
VKERN_TEMPL_1V_CC(do_svc_sub_val, SUB1SV);


#define ADD2VS(r,v1,f1,f2)      r = f1 + f2*v1
VKERN_TEMPL_2V_CC(do_val_svc_add, ADD2VS);

#define SUB2VS(r,v1,f1,f2)      r = f1 - f2*v1
VKERN_TEMPL_2V_CC(do_val_svc_sub, SUB2VS);

#define DIV2VS(r,v1,f1,f2)      r = f1 / (f2*v1)
VKERN_TEMPL_2V_CC(do_val_svc_div, DIV2VS);


#define NEG2(r,v1,f1,f2)        r = -v1
VKERN_TEMPL_2V(do_vec_neg_vec, NEG2);

#define NEG1(r,f1,f2)           r = -r
VKERN_TEMPL_1V(do_vec_neg, NEG1);

#define DOT2(r,v1,f1,f2)        f2 += CPLX__ conj(r) * v1
VKERN_TEMPL_2V_T(do_vec_dot_quick, DOT2, T);

#define XDOT2(r,v1,f1,f2)       { T y = CPLX__ conj(r) * v1; T t = f2+y; f1 += (t-f2)-y; f2 = t; }
VKERN_TEMPL_2V_T(do_vec_dot_exact, XDOT2, T);

#define MULT2(r,v1,f1,f2)       f2 += r * v1
VKERN_TEMPL_2V_T(do_vec_mult_quick, MULT2, T);
VKERN_TEMPL_2V_T_STRIDE(do_vec_mult_stride_quick, MULT2, T);

#define XMULT2(r,v1,f1,f2)      { T y = r * v1; T t = f2+y; f1 += (t-f2)-y; f2 = t; }
VKERN_TEMPL_2V_T(do_vec_mult_exact, XMULT2, T);
VKERN_TEMPL_2V_T_STRIDE(do_vec_mult_stride_exact, XMULT2, T);

#define FABS1(r,f1,f2)          f2 += fabssqr(r)
VKERN_TEMPL_1V_T_LD(do_vec_fabssqr_quick, FABS1, double);

#define XFABS1(r,f1,f2)         { double y = fabssqr(r); double t = f2+y; f1 += (t-f2)-y; f2 = t; }
VKERN_TEMPL_1V_T(do_vec_fabssqr_exact, XFABS1, double);

#define SQR1(r,f1,f2)           f2 += r*r
VKERN_TEMPL_1V_T(do_vec_sumsqr_quick, SQR1, T);

#define XSQR1(r,f1,f2)          { T y = r*r; T t = f2+y; f1 += (t-f2)-y; f2 = t; }
VKERN_TEMPL_1V_T(do_vec_sumsqr_exact, XSQR1, T);

#define SUM1(r,f1,f2)           f2 += r
VKERN_TEMPL_1V_T(do_vec_sum_quick, SUM1, T);

#define XSUM1(r,f1,f2)          { T t = f2+r; f1 += (t-f2)-r; f2 = t; }
VKERN_TEMPL_1V_T(do_vec_sum_exact, XSUM1, T);


// Used in do_bdmat_vec_mult
#define SUMMULT3(r,v1,v2,f1,f2) r += v1*v2
VKERN_TEMPL_3V(do_add_vec_vec_mul, SUMMULT3);
#define SUMCMULT3(r,v1,v2,f1,f2)        r += CPLX__ conj(v1)*v2
VKERN_TEMPL_3V(do_add_vec_vec_cmul, SUMCMULT3);


NAMESPACE_END

#endif  /* TBCI_VEC_KERN_UNR_PREF_H */