StaticModelLoader::StaticModelLoader(char* path, char *mpath)
{
m_infile.open(mpath);
if (!m_infile) {
cout << "Cannot open material file!\n";
throw 1;
}
infile.open(path); //"res/car.obj"
if (!infile) {
cout << "Cannot open file!\n";
throw 1;
}
mload();
vload();
infile.clear();
infile.seekg(0, ios::beg);
fload();
infile.close();
m_infile.close();
//debug models
std::cout << " verticies " << v_num
<< " textures " << vt_num
<< " normals " << vn_num
<< " polygones " << f_num
<< std::endl;
}
void repl_driver(void)
{
int k;
k = setjmp(errjmp);
if (k == 2) return;
signal(SIGFPE,handle_sigfpe);
signal(SIGINT,handle_sigint);
close_open_files();
errjmp_ok = 1;
nointerrupt = 0;
if (init_file && (k == 0)) vload(init_file);
repl();
}
std::string generate_impl(std::string const & kernel_prefix, statements_container const & statements, std::vector<mapping_type> const & mappings, unsigned int simd_width) const
{
std::string process_str;
utils::kernel_generation_stream stream;
std::string init0, upper_bound0, inc0, init1, upper_bound1, inc1;
stream << " __attribute__((reqd_work_group_size(" << p_.local_size_0 << "," << p_.local_size_1 << ",1)))" << std::endl;
generate_prototype(stream, kernel_prefix, "unsigned int M, unsigned int N,", mappings, statements);
stream << "{" << std::endl;
stream.inc_tab();
tree_parsing::process(stream, PARENT_NODE_TYPE, "scalar", "#scalartype #namereg = *#pointer;", statements, mappings);
tree_parsing::process(stream, PARENT_NODE_TYPE, "matrix", "#pointer += $OFFSET{#start1, #start2};", statements, mappings);
tree_parsing::process(stream, PARENT_NODE_TYPE, "vector", "#pointer += #start;", statements, mappings);
fetching_loop_info(p_.fetching_policy, "M", stream, init0, upper_bound0, inc0, "get_global_id(0)", "get_global_size(0)");
stream << "for(unsigned int i = " << init0 << "; i < " << upper_bound0 << "; i += " << inc0 << ")" << std::endl;
stream << "{" << std::endl;
stream.inc_tab();
fetching_loop_info(p_.fetching_policy, "N", stream, init1, upper_bound1, inc1, "get_global_id(1)", "get_global_size(1)");
stream << "for(unsigned int j = " << init1 << "; j < " << upper_bound1 << "; j += " << inc1 << ")" << std::endl;
stream << "{" << std::endl;
stream.inc_tab();
process_str = utils::append_width("#scalartype",simd_width) + " #namereg = " + vload(simd_width, "$OFFSET{i*#stride1,j*#stride2}", "#pointer")+ ";";
tree_parsing::process(stream, PARENT_NODE_TYPE, "matrix", process_str, statements, mappings);
tree_parsing::process(stream, PARENT_NODE_TYPE, "vector_diag", "#scalartype #namereg = ((i + ((#diag_offset<0)?#diag_offset:0))!=(j-((#diag_offset>0)?#diag_offset:0)))?0:#pointer[min(i*#stride, j*#stride)];", statements, mappings);
std::map<std::string, std::string> accessors;
accessors["matrix"] = "#namereg";
accessors["vector_diag"] = "#namereg";
accessors["scalar"] = "#namereg";
tree_parsing::evaluate(stream, PARENT_NODE_TYPE, accessors, statements, mappings);
process_str = vstore(simd_width, "#namereg", "$OFFSET{i*#stride1,j*#stride2}", "#pointer")+";";
tree_parsing::process(stream, LHS_NODE_TYPE, "matrix", process_str, statements, mappings);
stream.dec_tab();
stream << "}" << std::endl;
stream.dec_tab();
stream << "}" << std::endl;
stream.dec_tab();
stream << "}" << std::endl;
return stream.str();
}
void getcol(int c, double* x, double** bb, int nth)
{
int ii = c/BL;
int r = c%BL;
int id = ii%nth;
#if TILE_SCHEME == 0
int mynbl = (ii-id+nth-1)/nth;
size_t t = mynbl*((2*id+2+nth*(mynbl-1))/2)*BL*BL;
#else
size_t t = 0;
for(int I=id, i=id*BL; I<ii; I+=nth, i+=nth*BL)
t += (i+BL+RJ-1)/RJ*RJ*BL;
#endif
double* b = bb[id] + t + r*RJ;
for(int j=0; j<=c; j+=RJ){
vstore(x, vload(b));
x += RJ;
b += RJ*BL;
}
}
*
* Copyright (C) 2012-2017 Max-Planck-Society
* \author Martin Reinecke
*/
NOINLINE static void Z(alm2map_kernel) (const Tb cth, Y(Tbri) * restrict p1,
Y(Tbri) * restrict p2, Tb lam_1, Tb lam_2,
const sharp_ylmgen_dbl2 * restrict rf, const dcmplx * restrict alm,
int l, int lmax NJ1)
{
if (njobs>1)
{
while (l<lmax-2)
{
Tb lam_3, lam_4;
Tv r0=vload(rf[l].f[0]),r1=vload(rf[l].f[1]);
for (int i=0; i<nvec; ++i)
lam_3.v[i] = vsub(vmul(vmul(cth.v[i],lam_2.v[i]),r0),vmul(lam_1.v[i],r1));
r0=vload(rf[l+1].f[0]);r1=vload(rf[l+1].f[1]);
for (int i=0; i<nvec; ++i)
lam_4.v[i] = vsub(vmul(vmul(cth.v[i],lam_3.v[i]),r0),vmul(lam_2.v[i],r1));
r0=vload(rf[l+2].f[0]);r1=vload(rf[l+2].f[1]);
for (int i=0; i<nvec; ++i)
lam_1.v[i] = vsub(vmul(vmul(cth.v[i],lam_4.v[i]),r0),vmul(lam_3.v[i],r1));
for (int j=0; j<njobs; ++j)
{
Tv ar2=vload(creal(alm[njobs*l+j])),
ai2=vload(cimag(alm[njobs*l+j])),
ar4=vload(creal(alm[njobs*(l+2)+j])),
ai4=vload(cimag(alm[njobs*(l+2)+j]));
for (int i=0; i<nvec; ++i)
f128tuple histohit(f128tuple xyvec, const colorset pointcolors[4], const i32 th_id){
// don't plot the first 20 iterations
if(threadHits[th_id]++ > 20){
f128 xarr = xyvec.x;
f128 yarr = xyvec.y;
// check to see if any points have escaped to infinity or are NaN
// if they have, reset them and the threadHits counter to 0
if(vvalid(xarr.v) && vvalid(yarr.v)){
// scale the points from fractal-space to histogram-space
f128 scaledX;
scaledX.v = vadd(
vmul(
vadd(xarr.v, xoffsetvec),
hwidShrunkvec),
halfhwidvec);
f128 scaledY;
scaledY.v = vadd(
vmul(
vadd(yarr.v, yoffsetvec),
hheiShrunkvec),
halfhheivec);
// extract each point and plot
for (i32 i = 0; i < FLOATS_PER_VECTOR_REGISTER; i++){
const u32 ix = scaledX.f[i];
const u32 iy = scaledY.f[i];
if(ix < hwid && iy < hhei){
const u64 cell = ix + (iy * hwid);
// lock the row
omp_set_lock(&(locks[iy]));
// load the existing data
// the cache miss here takes up maybe 2/3s of the program's execution time
__m128 histocolor = vload((float *)&(h[cell]));
// add the new color to the old
histocolor = vadd(
histocolor,
pointcolors[i].vec);
// write back
vstore((float *)&(h[cell]), histocolor);
++goodHits;
// unlock the cell
omp_unset_lock(&(locks[iy]));
} else {
++missHits;
}
}
} else {
++badHits;
threadHits[th_id] = 0;
xyvec.x = zerovec;
xyvec.y = zerovec;
}
}
return xyvec;
}
