static void sampling_apply(const linop_data_t* _data, complex float* dst, const complex float* src)
{
const auto data = CAST_DOWN(sampling_data_s, _data);
#ifdef USE_CUDA
const complex float* pattern = get_pat(data, cuda_ondevice(src));
#else
const complex float* pattern = data->pattern;
#endif
md_zmul2(DIMS, data->dims, data->strs, dst, data->strs, src, data->pat_strs, pattern);
}
int arop(int count, int op) {
float32 f1, f2, f3, f4;
int i;
int fp;
char *mode;
int add=0;
int sub=0;
int mul=0;
int div=0;
int oper;
int err;
int err_count=0;
if(!quiet) printf("\nGenerating %0d Arithmetic test vectors ...\n",count);
if(append) mode = "a";
else mode = "w";
if(ofile==0) ofile = "ar.hex";
fp = fopen(ofile,mode);
if(fp == 0) {
printf("ERROR: Could not create file '%s'.\n",ofile);
return(-1);
}
if(!quiet) {
if(op & 0x1) printf("Add OP\n");
if(op & 0x2) printf("Sub OP\n");
if(op & 0x4) printf("Mul OP\n");
if(op & 0x8) printf("Div OP\n");
}
if(op & 0x1) add=1;
if(op & 0x2) sub=1;
if(op & 0x4) mul=1;
if(op & 0x8) div=1;
f1 = get_pat(0); // Initialize pattern generator ...
for(i=0;i<count;i++) {
err = 0;
if(pat>0) {
f1 = get_pat(1);
f2 = get_pat(2);
} else {
f1 = get_fp();
f2 = get_fp();
}
if(rall) float_rounding_mode = (rand() % 4);
oper = -1;
while(oper == -1) {
float_exception_flags = 0; // Reset Exceptions
if( (rand() % 4)==3 & div) {
oper = 8;
f3 = float32_div( f1, f2);
float_exception_flags = 0; // Reset Exceptions
f3 = float32_div( f1, f2);
//*( (float *) &f4 ) = *( (float *) &f1 ) / *( (float *) &f2 );
//if( f4 != f3) {
// err = 1;
// printf("FP Div Error: %x - %x: System: %x Lib: %x\n",f1, f2, f4, f3);
// }
}
if( (rand() % 4)==2 & mul) {
oper = 4;
f3 = float32_mul( f1, f2);
float_exception_flags = 0; // Reset Exceptions
f3 = float32_mul( f1, f2);
//*( (float *) &f4 ) = *( (float *) &f1 ) * *( (float *) &f2 );
//if( f4 != f3) {
// err = 1;
// printf("FP Mul Error: %x - %x: System: %x Lib: %x\n",f1, f2, f4, f3);
// }
}
if( (rand() % 4)==1 & sub) {
oper = 2;
f3 = float32_sub( f1, f2);
float_exception_flags = 0; // Reset Exceptions
f3 = float32_sub( f1, f2);
//*( (float *) &f4 ) = *( (float *) &f1 ) - *( (float *) &f2 );
//if( f4 != f3) {
// err = 1;
// printf("FP Sub Error: %x - %x: System: %x Lib: %x\n",f1, f2, f4, f3);
// }
}
if( (rand() % 4)==0 & add) {
oper = 1;
f3 = float32_add( f1, f2);
float_exception_flags = 0; // Reset Exceptions
f3 = float32_add( f1, f2);
//*( (float *) &f4 ) = *( (float *) &f1 ) + *( (float *) &f2 );
//if( f4 != f3) {
// err = 1;
// printf("FP Add Error: %x - %x: System: %x Lib: %x\n",f1, f2, f4, f3);
// }
}
}
if(err) err_count++;
if(!err) {
//if(float_exception_flags != 0)
// printf("Exceptions: %x\n",float_exception_flags);
if(verb) printf("rmode: %01x, except: %02x, oper: %01x opa: %08x, opb: %08x res: %08x\n", float_rounding_mode, float_exception_flags, oper, f1, f2, f3);
fprintf(fp,"%01x%02x%01x%08x%08x%08x\n", float_rounding_mode, float_exception_flags, oper, f1, f2, f3);
}
else {
printf("\t Vecor mismatch between library and system calculations. This Vector\n");
printf("\t will not be placed in to vector file ...\n");
}
}
close(fp);
if(!quiet) {
printf("Found %d errors\n",err_count);
printf("Wrote %d vectors from total %d specified.\n", (count-err_count), count);
printf("\n ... f2i done.\n");
}
return(0);
}
void get_4x4(int x, int y) {
get_pat(0+(x*4),0+(y*4));
get_pat(0+(x*4),1+(y*4));
get_pat(0+(x*4),2+(y*4));
get_pat(0+(x*4),3+(y*4));
get_pat(1+(x*4),0+(y*4));
get_pat(1+(x*4),1+(y*4));
get_pat(1+(x*4),2+(y*4));
get_pat(1+(x*4),3+(y*4));
get_pat(2+(x*4),0+(y*4));
get_pat(2+(x*4),1+(y*4));
get_pat(2+(x*4),2+(y*4));
get_pat(2+(x*4),3+(y*4));
get_pat(3+(x*4),0+(y*4));
get_pat(3+(x*4),1+(y*4));
get_pat(3+(x*4),2+(y*4));
get_pat(3+(x*4),3+(y*4));
}
void get_2x2(int x) {
get_pat(0+(x*2),0);
get_pat(0+(x*2),1);
get_pat(1+(x*2),0);
get_pat(1+(x*2),1);
}
