always_inline VecFloat ldexp_float(VecFloat const & x, typename VecFloat::int_vec const & n)
{
typedef typename VecFloat::int_vec int_vec;
const VecFloat exponent_mask = VecFloat::gen_exp_mask();
const VecFloat exponent = exponent_mask & x;
const VecFloat x_wo_x = andnot(exponent_mask, x); // clear exponent
int_vec new_exp = slli(n, 16+7) + int_vec(exponent); // new exponent
VecFloat new_exp_float(new_exp);
VecFloat ret = x_wo_x | new_exp_float;
return ret;
}
always_inline VecType vec_sin_float(VecType const & arg)
{
typedef typename VecType::int_vec int_vec;
const typename VecType::float_type four_over_pi = 1.27323954473516268615107010698011489627567716592367;
VecType sign = arg & VecType::gen_sign_mask();
VecType abs_arg = arg & VecType::gen_abs_mask();
VecType y = abs_arg * four_over_pi;
int_vec j = y.truncate_to_int();
/* cephes: j=(j+1) & (~1) */
j = (j + int_vec(1)) & int_vec(~1);
y = j.convert_to_float();
/* sign based on quadrant */
VecType swap_sign_bit = slli(j & int_vec(4), 29);
sign = sign ^ swap_sign_bit;
/* polynomial mask */
VecType poly_mask = VecType (mask_eq(j & int_vec(2), int_vec(0)));
/* black magic */
static float DP1 = 0.78515625;
static float DP2 = 2.4187564849853515625e-4;
static float DP3 = 3.77489497744594108e-8;
VecType base = ((abs_arg - y * DP1) - y * DP2) - y * DP3;
/* [0..pi/4] */
VecType z = base * base;
VecType p1 = (( 2.443315711809948E-005 * z
- 1.388731625493765E-003) * z
+ 4.166664568298827E-002) * z * z
-0.5f * z + 1.0
;
/* [pi/4..pi/2] */
VecType p2 = ((-1.9515295891E-4 * z
+ 8.3321608736E-3) * z
- 1.6666654611E-1) * z * base + base;
VecType approximation = select(p1, p2, poly_mask);
return approximation ^ sign;
}
int encode_op(char *opcode, char *op_data)
{
int rd,rs,rt,imm,funct,shaft,target;
char tmp[256];
const char *fi = "%s %d";
const char *fg = "%s %%g%d";
const char *ff = "%s %%f%d";
const char *fl = "%s %s";
const char *fgi = "%s %%g%d, %d";
const char *fgl = "%s %%g%d, %s";
const char *fgg = "%s %%g%d, %%g%d";
const char *fggl = "%s %%g%d, %%g%d, %s";
const char *fggi = "%s %%g%d, %%g%d, %d";
const char *fggg = "%s %%g%d, %%g%d, %%g%d";
const char *fff = "%s %%f%d, %%f%d";
const char *fgf = "%s %%g%d, %%f%d";
const char *ffg = "%s %%f%d, %%g%d";
const char *fffl = "%s %%f%d, %%f%d, %s";
const char *ffff = "%s %%f%d, %%f%d, %%f%d";
const char *ffgi = "%s %%f%d, %%g%d, %d";
const char *ffgg = "%s %%f%d, %%g%d, %%g%d";
char lname[256];
shaft = funct = target = 0;
if(strcmp(opcode, "mvhi") == 0){
if(sscanf(op_data, fgi, tmp, &rs, &imm) == 3)
return mvhi(rs,0,imm);
}
if(strcmp(opcode, "mvlo") == 0){
if(sscanf(op_data, fgi, tmp, &rs, &imm) == 3)
return mvlo(rs,0,imm);
}
if(strcmp(opcode, "add") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return add(rs,rt,rd,0);
}
if(strcmp(opcode, "nor") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return nor(rs,rt,rd,0);
}
if(strcmp(opcode, "sub") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return sub(rs,rt,rd,0);
}
if(strcmp(opcode, "mul") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return mul(rs,rt,rd,0);
}
if(strcmp(opcode, "addi") == 0){
if(sscanf(op_data, fggi, tmp, &rt, &rs, &imm) == 4)
return addi(rs,rt,imm);
}
if(strcmp(opcode, "subi") == 0){
if(sscanf(op_data, fggi, tmp, &rt, &rs, &imm) == 4)
return subi(rs,rt,imm);
}
if(strcmp(opcode, "muli") == 0){
if(sscanf(op_data, fggi, tmp, &rt, &rs, &imm) == 4)
return muli(rs,rt,imm);
}
if(strcmp(opcode, "input") == 0){
if(sscanf(op_data, fg, tmp, &rd) == 2)
return input(0,0,rd,0);
}
if(strcmp(opcode, "inputw") == 0){
if(sscanf(op_data, fg, tmp, &rd) == 2)
return inputw(0,0,rd,0);
}
if(strcmp(opcode, "inputf") == 0){
if(sscanf(op_data, ff, tmp, &rd) == 2)
return inputf(0,0,rd,0);
}
if(strcmp(opcode, "output") == 0){
if(sscanf(op_data, fg, tmp, &rs) == 2)
return output(rs,0,0,0);
}
if(strcmp(opcode, "outputw") == 0){
if(sscanf(op_data, fg, tmp, &rs) == 2)
return outputw(rs,0,0,0);
}
if(strcmp(opcode, "outputf") == 0){
if(sscanf(op_data, ff, tmp, &rs) == 2)
return outputf(rs,0,0,0);
}
if(strcmp(opcode, "and") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return _and(rs,rt,rd,0);
}
if(strcmp(opcode, "or") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return _or(rs,rt,rd,0);
}
if(strcmp(opcode, "sll") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return sll(rs,rt,rd,0);
}
if(strcmp(opcode, "srl") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return srl(rs,rt,rd,0);
}
if(strcmp(opcode, "slli") == 0){
if(sscanf(op_data, fggi, tmp, &rt, &rs, &imm) == 4)
return slli(rs,rt,imm);
}
if(strcmp(opcode, "srli") == 0){
if(sscanf(op_data, fggi, tmp, &rt, &rs, &imm) == 4)
return srli(rs,rt,imm);
}
if(strcmp(opcode, "b") == 0){
if(sscanf(op_data, fg, tmp, &rs) == 2)
return b(rs,0,0,0);
}
if(strcmp(opcode, "jmp") == 0){
if(sscanf(op_data, fl, tmp, lname) == 2) {
strcpy(label_name[label_cnt],lname);
return jmp(label_cnt++);
}
}
if(strcmp(opcode, "jeq") == 0){
if(sscanf(op_data, fggl, tmp, &rs, &rt, lname) == 4) {
strcpy(label_name[label_cnt],lname);
return jeq(rs,rt,label_cnt++);
}
}
if(strcmp(opcode, "jne") == 0){
if(sscanf(op_data, fggl, tmp, &rs, &rt, lname) == 4) {
strcpy(label_name[label_cnt],lname);
return jne(rs,rt,label_cnt++);
}
}
if(strcmp(opcode, "jlt") == 0){
if(sscanf(op_data, fggl, tmp, &rs, &rt, lname) == 4) {
strcpy(label_name[label_cnt],lname);
return jlt(rs,rt,label_cnt++);
}
}
if(strcmp(opcode, "jle") == 0){
if(sscanf(op_data, fggl, tmp, &rs, &rt, lname) == 4) {
strcpy(label_name[label_cnt],lname);
return jle(rs,rt,label_cnt++);
}
}
if(strcmp(opcode, "call") == 0){
if(sscanf(op_data, fl, tmp, lname) == 2) {
strcpy(label_name[label_cnt],lname);
return call(label_cnt++);
}
}
if(strcmp(opcode, "callR") == 0){
if(sscanf(op_data, fg, tmp, &rs) == 2)
return callr(rs,0,0,0);
}
if(strcmp(opcode, "return") == 0){
return _return(0);
}
if(strcmp(opcode, "ld") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return ld(rs,rt,rd,0);
}
if(strcmp(opcode, "ldi") == 0){
if(sscanf(op_data, fggi, tmp, &rt, &rs, &imm) == 4)
return ldi(rs,rt,imm);
}
if(strcmp(opcode, "ldlr") == 0){
if(sscanf(op_data, fgi, tmp, &rs, &imm) == 3)
return ldlr(rs,0,imm);
}
if(strcmp(opcode, "fld") == 0){
if(sscanf(op_data, ffgg, tmp, &rd, &rs,&rt) == 4)
return fld(rs,rt,rd,0);
}
if(strcmp(opcode, "st") == 0){
if(sscanf(op_data, fggg, tmp, &rd, &rs,&rt) == 4)
return st(rs,rt,rd,0);
}
if(strcmp(opcode, "sti") == 0){
if(sscanf(op_data, fggi, tmp, &rt, &rs, &imm) == 4)
return sti(rs,rt,imm);
}
if(strcmp(opcode, "stlr") == 0){
if(sscanf(op_data, fgi, tmp, &rs, &imm) == 3)
return stlr(rs,0,imm);
}
if(strcmp(opcode, "fst") == 0){
if(sscanf(op_data, ffgg, tmp, &rd, &rs,&rt) == 4)
return fst(rs,rt,rd,0);
}
if(strcmp(opcode, "fadd") == 0){
if(sscanf(op_data, ffff, tmp, &rd, &rs, &rt) == 4)
return fadd(rs,rt,rd,0);
}
if(strcmp(opcode, "fsub") == 0){
if(sscanf(op_data, ffff, tmp, &rd, &rs, &rt) == 4)
return fsub(rs,rt,rd,0);
}
if(strcmp(opcode, "fmul") == 0){
if(sscanf(op_data, ffff, tmp, &rd, &rs, &rt) == 4)
return fmul(rs,rt,rd,0);
}
if(strcmp(opcode, "fdiv") == 0){
if(sscanf(op_data, ffff, tmp, &rd, &rs, &rt) == 4)
return fdiv(rs,rt,rd,0);
}
if(strcmp(opcode, "fsqrt") == 0){
if(sscanf(op_data, fff, tmp, &rd, &rs) == 3)
return fsqrt(rs,0,rd,0);
}
if(strcmp(opcode, "fabs") == 0){
if(sscanf(op_data, fff, tmp, &rd, &rs) == 3)
return _fabs(rs,0,rd,0);
}
if(strcmp(opcode, "fmov") == 0){
if(sscanf(op_data, fff, tmp, &rd, &rs) == 3)
return fmov(rs,0,rd,0);
}
if(strcmp(opcode, "fneg") == 0){
if(sscanf(op_data, fff, tmp, &rd, &rs) == 3)
return fneg(rs,0,rd,0);
}
if(strcmp(opcode, "fldi") == 0){
if(sscanf(op_data, ffgi, tmp, &rt, &rs, &imm) == 4)
return fldi(rs,rt,imm);
}
if(strcmp(opcode, "fsti") == 0){
if(sscanf(op_data, ffgi, tmp, &rt, &rs, &imm) == 4)
return fsti(rs,rt,imm);
}
if(strcmp(opcode, "fjeq") == 0){
if(sscanf(op_data, fffl, tmp, &rs, &rt, lname) == 4) {
strcpy(label_name[label_cnt],lname);
return fjeq(rs,rt,label_cnt++);
}
}
if(strcmp(opcode, "fjlt") == 0){
if(sscanf(op_data, fffl, tmp, &rs, &rt, lname) == 4) {
strcpy(label_name[label_cnt],lname);
return fjlt(rs,rt,label_cnt++);
}
}
if(strcmp(opcode, "halt") == 0){
return halt(0,0,0,0);
}
if(strcmp(opcode, "setL") == 0){
if(sscanf(op_data, fgl, tmp, &rd, lname) == 3) {
strcpy(label_name[label_cnt],lname);
return setl(0,rd,label_cnt++);
}
}
if(strcmp(opcode, "padd") == 0){
if(sscanf(op_data, fgi, tmp, &rt, &imm) == 3) {
return padd(0,rt,imm);
}
}
if(strcmp(opcode, "link") == 0){
if(sscanf(op_data, fi, tmp, &imm) == 2) {
return link(0,0,imm);
}
}
if(strcmp(opcode, "movlr") == 0){
return movlr(0,0,0,0);
}
if(strcmp(opcode, "btmplr") == 0){
return btmplr(0,0,0,0);
}
/*
if(strcmp(opcode, "padd") == 0){
if(sscanf(op_data, fgg, tmp, &rd, &rt) == 3) {
return padd(0,rt,d,0);
}
}
*/
return -1;
}
