ComplexPair BandStopTransform::transform (complex_t c)
{
if (c == infinity())
c = -1;
else
c = (1. + c) / (1. - c); // bilinear
complex_t u (0);
u = addmul (u, 4 * (b2 + a2 - 1), c);
u += 8 * (b2 - a2 + 1);
u *= c;
u += 4 * (a2 + b2 - 1);
u = std::sqrt (u);
complex_t v = u * -.5;
v += a;
v = addmul (v, -a, c);
u *= .5;
u += a;
u = addmul (u, -a, c);
complex_t d (b + 1);
d = addmul (d, b-1, c);
return ComplexPair (u/d, v/d);
}
complex_t BiquadBase::response (double normalizedFrequency) const
{
const double a0 = getA0 ();
const double a1 = getA1 ();
const double a2 = getA2 ();
const double b0 = getB0 ();
const double b1 = getB1 ();
const double b2 = getB2 ();
const double w = 2 * doublePi * normalizedFrequency;
const complex_t czn1 = std::polar (1., -w);
const complex_t czn2 = std::polar (1., -2 * w);
complex_t ch (1);
complex_t cbot (1);
complex_t ct (b0/a0);
complex_t cb (1);
ct = addmul (ct, b1/a0, czn1);
ct = addmul (ct, b2/a0, czn2);
cb = addmul (cb, a1/a0, czn1);
cb = addmul (cb, a2/a0, czn2);
ch *= ct;
cbot *= cb;
return ch / cbot;
}
ComplexPair BandPassTransform::transform (complex_t c)
{
if (c == infinity())
return ComplexPair (-1, 1);
c = (1. + c) / (1. - c); // bilinear
complex_t v = 0;
v = addmul (v, 4 * (b2 * (a2 - 1) + 1), c);
v += 8 * (b2 * (a2 - 1) - 1);
v *= c;
v += 4 * (b2 * (a2 - 1) + 1);
v = std::sqrt (v);
complex_t u = -v;
u = addmul (u, ab_2, c);
u += ab_2;
v = addmul (v, ab_2, c);
v += ab_2;
complex_t d = 0;
d = addmul (d, 2 * (b - 1), c) + 2 * (1 + b);
return ComplexPair (u/d, v/d);
}
void fun(ulong a, ulong b)
{
LOCAL_HIREMAINDER;
LOCAL_OVERFLOW;
addll(a,b);
addllx(a,b);
mulll(a,b);
addmul(a,b);
#if 0
bfffo(a);
#endif
}
complex_t Cascade::response (double normalizedFrequency) const
{
double w = 2 * doublePi * normalizedFrequency;
const complex_t czn1 = std::polar (1., -w);
const complex_t czn2 = std::polar (1., -2 * w);
complex_t ch (1);
complex_t cbot (1);
const Biquad* stage = m_stageArray;
for (int i = m_numStages; --i >=0; ++stage)
{
complex_t cb (1);
complex_t ct (stage->getB0()/stage->getA0());
ct = addmul (ct, stage->getB1()/stage->getA0(), czn1);
ct = addmul (ct, stage->getB2()/stage->getA0(), czn2);
cb = addmul (cb, stage->getA1()/stage->getA0(), czn1);
cb = addmul (cb, stage->getA2()/stage->getA0(), czn2);
ch *= ct;
cbot *= cb;
}
return ch / cbot;
}
/*
Handle additional data, same for en/de crypt
This does not use ChaCha data so need not be rerun
when you have done auth & want to reset ChaCha
for the decryption
*/
static void do_additional( const byte *ad, u64 adlen )
{
u64 n ;
assert( ad != NULL ) ;
memset( accum, 0, AES_BYTES ) ;
while (adlen > 0) {
n = (u64) ((adlen >= AES_BYTES)? AES_BYTES : adlen) ;
addmul((byte *) accum,ad,n,(byte *) H);
ad += n ;
adlen -= n;
}
#ifdef TESTING
printf("accum after ad: %08x %08x %08x %08x\n",
accum[0], accum[1], accum[2], accum[3] ) ;
#endif
}
/*
simple interface to auth.c
call addmul() with full block of data
operate on global accum[]
*/
static void mix_in( u32 *data )
{
addmul( (byte *) accum, (byte *) data,
AES_BYTES, (byte *) H ) ;
}
int crypto_aead_decrypt(
unsigned char *m,unsigned long long *outputmlen,
unsigned char *nsec,
const unsigned char *c,unsigned long long clen,
const unsigned char *ad,unsigned long long adlen,
const unsigned char *npub,
const unsigned char *k
)
{
u64 mlen, n, origmlen, start ;
const byte *ptr, *origc;
byte temp[AES_BYTES], *p ;
int i ;
assert( m != NULL ) ;
assert( outputmlen != NULL ) ;
assert( c != NULL ) ;
assert( ad != NULL ) ;
assert( npub != NULL ) ;
assert( k != NULL ) ;
if( clen < AUTH_BYTES) return -1;
/* where auth data should be */
mlen = clen - AUTH_BYTES ;
ptr = c + mlen ;
*outputmlen = mlen;
setup(mlen, adlen, npub, k) ;
do_additional( ad, adlen ) ;
origc = c;
origmlen = mlen;
#ifdef TESTING
printf("decrypt: total %llu text %llu ad %llu\n", clen, mlen, adlen ) ;
printf("decrypt: ciphertext starts with %08xu\n", ((u32 *) origc)[0] ) ;
printf("decrypt: starting authentication check %08x\n", ((u32 *) ptr)[0] ) ;
#endif
/*
authenticate
*/
start = mlen ;
while( (mlen > 0) && (mlen <= start) ) {
n = (u64) ((mlen >= AES_BYTES) ? AES_BYTES : mlen) ;
addmul( (byte *) accum, c, n, (byte *) H) ;
c += n;
mlen -= n;
}
if( mlen > start ) {
#ifdef TESTING
printf("decrypt: mlen out of bounds\n" ) ;
#endif
return -2 ;
}
#ifdef TESTING
printf("decrypt: auth before finalisation %08x %08x\n", accum[0], accum[1] ) ;
#endif
mix_in(finalblock) ;
xor_128( accum, I ) ;
/* see if we have a match */
if( memcmp((void *) accum, (void *) ptr, AUTH_BYTES) != 0) {
#ifdef TESTING
printf("decrypt: auth failed, result here %08x\n", accum[0] ) ;
#endif
return -3;
}
#ifdef TESTING
printf("decrypt: auth matched, going to decrypt\n" ) ;
printf("decrypt: 1st ciphertext word %08x\n", *((u32 *) c) ) ;
#endif
/*
reset things
start encryption with correct ChaCha state
*/
c = origc;
mlen = origmlen;
setup(mlen, adlen, npub, k) ;
/*
decrypt, simpler than encryption
no authentication to do
*/
start = mlen ;
while( (mlen > 0) && (mlen <= start) ) {
n = (u64) ((mlen >= AES_BYTES)? AES_BYTES : mlen) ;
white_aes( temp ) ;
for( i = 0, p = temp ; i < n ; i++, c++, m++, p++ )
*m = *c ^ *p ;
mlen -= n;
}
if( mlen > start ) {
#ifdef TESTING
printf("decrypt: mlen out of bounds\n" ) ;
#endif
return -2 ;
}
#ifdef TESTING
printf("decrypt: reached end, showing first words\n" ) ;
printf("plaintext %08x ciphertext %08x auth %08x\n",*((u32 *) m), *((u32 *) c), accum[0] ) ;
#endif
memset( temp, 0, AES_BYTES ) ;
cleanup() ;
return 0;
}
/*
declarations for this and decrypt
copied from Bernstein's example AES-GCM code
to be sure they match test framework
*/
int crypto_aead_encrypt(
unsigned char *c,unsigned long long *clen,
const unsigned char *m, unsigned long long mlen,
const unsigned char *ad,unsigned long long adlen,
const unsigned char *nsec,
const unsigned char *npub,
const unsigned char *k
)
{
u64 n, start ;
byte temp[AES_BYTES] ;
byte *p, *q, *ptr ;
int i ;
#ifdef TESTING
byte *origm, *origc ;
assert( c != NULL ) ;
/* are framework declarations compatible with mine? */
assert( sizeof(long long) == sizeof( u64) ) ;
printf("encrypt: plaintext %llu ad %llu\n", mlen, adlen ) ;
origm = m ;
origc = c ;
#endif
/*
Bernstein's example aes-gcm code sets *clen
I copy that, assuming the interface needs it
*/
*clen = mlen+AUTH_BYTES ;
/* address for authentication data */
ptr = c + mlen ;
/* key schedule */
setup( mlen, (u64) adlen, npub, k) ;
/* data authenticated but not encrypted */
do_additional( ad, (u64) adlen ) ;
/*
encryption
mlen is unsigned so always > = 0
use start for an error check
*/
start = mlen ;
while( (mlen != 0) && (mlen <= start) ) {
n = (u64) ((mlen >= AES_BYTES)? AES_BYTES : mlen) ;
white_aes( temp ) ;
for( i = 0, p = temp, q = c ; i < n ; i++, q++, m++, p++ )
*q = *m ^ *p ;
addmul( (byte *) accum, c, n, (byte *) H) ;
mlen -= n;
c += n ;
}
if( mlen > start ) {
#ifdef TESTING
printf("encrypt: mlen out of bounds\n" ) ;
#endif
return -2 ;
}
#ifdef TESTING
printf("encrypt: auth before finalisation %08x %08x\n", accum[0], accum[1] ) ;
#endif
/*
finish authentication
*/
mix_in(finalblock) ;
xor_128( accum, I ) ;
memcpy( ptr, (void *) accum, AES_BYTES ) ;
#ifdef TESTING
printf("encrypt: reached end, showing first words\n" ) ;
printf("plaintext %08x ciphertext %08x auth %08x\n", *((u32 *) origm), *((u32 *) origc), accum[0] ) ;
#endif
memset( temp, 0, AES_BYTES ) ;
cleanup() ;
return 0 ;
}
int NewtonCuda::oneStep()
{
//Timer timer;
std::vector<Vector3f> force = m->getForce();
float E = m->getEnergy();
std::vector<float> val;
//timer.start();
if (m_fixRigidMotion)
{
m->getStiffnessSparse(val, false,true,true);
}
else
{
m->getStiffnessSparse(val, false,true);
}
//timer.end();
//std::cout << "assembly time: " << timer.getSeconds() << "\n";
int ndof = 3 * (int)m->x.size();
std::vector<float> bb(ndof);
for (unsigned int ii = 0; ii<m->x.size(); ii++){
for (int jj = 0; jj<3; jj++){
int row = 3 * ii + jj;
if (m->fixed[ii]){
bb[row] = 0;
}
else{
bb[row] = force[ii][jj];
}
}
}
if (m_fixRigidMotion)
{
for(int ii =0;ii<6;ii++){
bb.push_back(0.f);
}
}
//timer.start();
solver.solve(val, &(bb[0]));
//timer.end();
//std::cout << "lin solve time: " << timer.getSeconds() << "\n";
//timer.start();
double totalMag = 0;
for (int ii = 0; ii<ndof; ii++){
totalMag += std::abs(bb[ii]);
}
totalMag /= ndof;
for (unsigned int ii = 0; ii<m->x.size(); ii++){
for (int jj = 0; jj<3; jj++){
force[ii][jj] = (float)bb[3 * ii + jj];
}
}
//std::cout << "total disp: " << totalMag << " ,h: " << h << "\n";
std::cout<< "E: " << E << "\n";
//line search
std::vector<Vector3f> x0 = m->x;
h = 1.0f;
float E1 = E;
bool valid = true;
while (1){
if (totalMag * h<dx_tol){
return -1;
}
m->x = x0;
addmul(m->x, h, force);
E1 = m->getEnergy();
if (E1>E || fem_error){
fem_error = 0;
h = 0.5f* h;
if (h<1e-12){
valid = false;
break;
}
}
else{
break;
}
}
//timer.end();
//std::cout << "line search time: " << timer.getSeconds() << "\n";
if (!valid){
return 1;
}
return 0;
}
int main() {
cELF *elf_a = new cELF((char *)"testlib_a.elf");
// get text segment
cSection *text = elf_a->getSectionByName((char *)".text");
printf(".text section: fileoffset=%p\n", text->getFileOffset());
unsigned char *textptr = elf_a->image + text->getFileOffset();
// apply relocation records for .text inside our "image" (textptr points into it)
int test_a = 10;
int test_b = 100;
//test_a = test_b = 0;
elf_a->importSymbol((int)text->getFileOffset(), (char *)"_test_a", &test_a);
elf_a->importSymbol((int)text->getFileOffset(), (char *)"_test_b", &test_b);
elf_a->importSymbol((int)text->getFileOffset(), (char *)"_printf", (void *)printf);
elf_a->importSymbol((int)text->getFileOffset(), (char *)"_puts", (void *)puts);
elf_a->importSymbol((int)text->getFileOffset(), (char *)"_vprintf", (void *)vprintf);
cSymbol *d1 = elf_a->getSymbolByName((char *)"_someData1");
cSymbol *d2 = elf_a->getSymbolByName((char *)"_someData2");
cSymbol *d3 = elf_a->getSymbolByName((char *)"_someData3");
printf("someData1 symbol segment offset: %llp symbol->getFileOffset()=%llp text=%s\n", d1->getSegmentOffset(), d1->getFileOffset(), d1->getAbsoluteFileOffset() );
printf("someData2 symbol segment offset: %llp symbol->getFileOffset()=%llp text=%s\n", d2->getSegmentOffset(), d2->getFileOffset(), d2->getAbsoluteFileOffset() );
printf("someData3 symbol segment offset: %llp symbol->getFileOffset()=%llp text=%s\n", d3->getSegmentOffset(), d3->getFileOffset(), d3->getAbsoluteFileOffset() );
printf("should be .text : file_offset=%p\n", elf_a->getSectionByInternalID(0)->getFileOffset());
printf("should be .data : file_offset=%p\n", elf_a->getSectionByInternalID(1)->getFileOffset());
printf("should be .bss : file_offset=%p\n", elf_a->getSectionByInternalID(2)->getFileOffset());
printf("should be .rdata: file_offset=%p\n", elf_a->getSectionByInternalID(3)->getFileOffset());
// get the function pointers
int (*add)(int a, int b);
int (*mul)(int a, int b);
int (*printsomething)(char *msg, ...);
void (*printglobals)();
*(int *)&add = elf_a->getProcAddress((char *)"_add");
*(int *)&mul = elf_a->getProcAddress((char *)"_mul");
*(int *)&printsomething = elf_a->getProcAddress((char *)"_printsomething");
*(int *)&printglobals = elf_a->getProcAddress((char *)"_printglobals");
// call it
printf("add=%p mul=%p\n", add, mul);
printf("lib.c add(2,2) = %d\n", add(2,2));
printf("lib.c mul(3,3) = %d\n", mul(3,3));
printsomething((char *)"haiiiii! 123=%d\n", 123);
elf_a->relocateTextForRdata();
elf_a->relocateTextForData();
printglobals();
printglobals();
printglobals();
printf("End!\n");
#if 0
cELF *elf_b = new cELF((char *)"testlib_b.elf");
int (*addmul)(int a, int b, int factor);
elf_b->importSymbol((char *)"_add", (void *)add);
elf_b->importSymbol((char *)"_mul", (void *)mul);
*(int *)&addmul = elf_b->getProcAddress((char *)"_addmul");
printf("addmul: %p\n", addmul);
printf("lib.c addmul(3,3, 2) = %d\n", addmul(3, 3, 2));
printf("Address of printf=%p real printf=%p\n", GetProcAddress(NULL, "printf"), printf);
#endif
return 0;
}