void fwt1(unsigned int N, unsigned int d, const long dims[N], const long ostr[N], complex float* low, complex float* hgh, const long istr[N], const complex float* in, const long flen, const float filter[2][2][flen])
{
debug_printf(DP_DEBUG4, "fwt1: %d/%d\n", d, N);
debug_print_dims(DP_DEBUG4, N, dims);
assert(dims[d] >= 2);
long odims[N];
md_copy_dims(N, odims, dims);
odims[d] = bandsize(dims[d], flen);
debug_print_dims(DP_DEBUG4, N, odims);
long o = d + 1;
long u = N - o;
// 0 1 2 3 4 5 6|7
// --d-- * --u--|N
// ---o---
assert(d == md_calc_blockdim(d, dims + 0, istr + 0, CFL_SIZE));
assert(u == md_calc_blockdim(u, dims + o, istr + o, CFL_SIZE * md_calc_size(o, dims)));
assert(d == md_calc_blockdim(d, odims + 0, ostr + 0, CFL_SIZE));
assert(u == md_calc_blockdim(u, odims + o, ostr + o, CFL_SIZE * md_calc_size(o, odims)));
// merge dims
long wdims[3] = { md_calc_size(d, dims), dims[d], md_calc_size(u, dims + o) };
long wistr[3] = { CFL_SIZE, istr[d], CFL_SIZE * md_calc_size(o, dims) };
long wostr[3] = { CFL_SIZE, ostr[d], CFL_SIZE * md_calc_size(o, odims) };
#ifdef USE_CUDA
if (cuda_ondevice(in)) {
assert(cuda_ondevice(low));
assert(cuda_ondevice(hgh));
float* flow = md_gpu_move(1, MD_DIMS(flen), filter[0][0], FL_SIZE);
float* fhgh = md_gpu_move(1, MD_DIMS(flen), filter[0][1], FL_SIZE);
wl3_cuda_down3(wdims, wostr, low, wistr, in, flen, flow);
wl3_cuda_down3(wdims, wostr, hgh, wistr, in, flen, fhgh);
md_free(flow);
md_free(fhgh);
return;
}
#endif
// no clear needed
wavelet_down3(wdims, wostr, low, wistr, in, flen, filter[0][0]);
wavelet_down3(wdims, wostr, hgh, wistr, in, flen, filter[0][1]);
}
void iwt1(unsigned int N, unsigned int d, const long dims[N], const long ostr[N], complex float* out, const long istr[N], const complex float* low, const complex float* hgh, const long flen, const float filter[2][2][flen])
{
debug_printf(DP_DEBUG4, "ifwt1: %d/%d\n", d, N);
debug_print_dims(DP_DEBUG4, N, dims);
assert(dims[d] >= 2);
long idims[N];
md_copy_dims(N, idims, dims);
idims[d] = bandsize(dims[d], flen);
debug_print_dims(DP_DEBUG4, N, idims);
long o = d + 1;
long u = N - o;
// 0 1 2 3 4 5 6|7
// --d-- * --u--|N
// ---o---
assert(d == md_calc_blockdim(d, dims + 0, ostr + 0, CFL_SIZE));
assert(u == md_calc_blockdim(u, dims + o, ostr + o, CFL_SIZE * md_calc_size(o, dims)));
assert(d == md_calc_blockdim(d, idims + 0, istr + 0, CFL_SIZE));
assert(u == md_calc_blockdim(u, idims + o, istr + o, CFL_SIZE * md_calc_size(o, idims)));
long wdims[3] = { md_calc_size(d, dims), dims[d], md_calc_size(u, dims + o) };
long wistr[3] = { CFL_SIZE, istr[d], CFL_SIZE * md_calc_size(o, idims) };
long wostr[3] = { CFL_SIZE, ostr[d], CFL_SIZE * md_calc_size(o, dims) };
md_clear(3, wdims, out, CFL_SIZE); // we cannot clear because we merge outputs
#ifdef USE_CUDA
if (cuda_ondevice(out)) {
assert(cuda_ondevice(low));
assert(cuda_ondevice(hgh));
float* flow = md_gpu_move(1, MD_DIMS(flen), filter[1][0], FL_SIZE);
float* fhgh = md_gpu_move(1, MD_DIMS(flen), filter[1][1], FL_SIZE);
wl3_cuda_up3(wdims, wostr, out, wistr, low, flen, flow);
wl3_cuda_up3(wdims, wostr, out, wistr, hgh, flen, fhgh);
md_free(flow);
md_free(fhgh);
return;
}
#endif
wavelet_up3(wdims, wostr, out, wistr, low, flen, filter[1][0]);
wavelet_up3(wdims, wostr, out, wistr, hgh, flen, filter[1][1]);
}
static int siemens_adc_read(bool vd, int fd, bool linectr, bool partctr, const long dims[DIMS], long pos[DIMS], complex float* buf)
{
char scan_hdr[vd ? 192 : 0];
xread(fd, scan_hdr, sizeof(scan_hdr));
for (pos[COIL_DIM] = 0; pos[COIL_DIM] < dims[COIL_DIM]; pos[COIL_DIM]++) {
char chan_hdr[vd ? 32 : 128];
xread(fd, chan_hdr, sizeof(chan_hdr));
struct mdh2 mdh;
memcpy(&mdh, vd ? (scan_hdr + 40) : (chan_hdr + 20), sizeof(mdh));
if (0 == pos[COIL_DIM]) {
// TODO: rethink this
pos[PHS1_DIM] = mdh.sLC[0] + (linectr ? mdh.linectr : 0);
pos[AVG_DIM] = mdh.sLC[1];
pos[SLICE_DIM] = mdh.sLC[2];
pos[PHS2_DIM] = mdh.sLC[3] + (partctr ? mdh.partctr : 0);
pos[TE_DIM] = mdh.sLC[4];
pos[TIME_DIM] = mdh.sLC[6];
pos[TIME2_DIM] = mdh.sLC[7];
}
debug_print_dims(DP_DEBUG1, DIMS, pos);
if (dims[READ_DIM] != mdh.samples) {
debug_printf(DP_WARN, "Wrong number of samples: %d != %d.\n", dims[READ_DIM], mdh.samples);
return -1;
}
if ((0 != mdh.channels) && (dims[COIL_DIM] != mdh.channels)) {
debug_printf(DP_WARN, "Wrong number of channels: %d != %d.\n", dims[COIL_DIM], mdh.channels);
return -1;
}
xread(fd, buf + pos[COIL_DIM] * dims[READ_DIM], dims[READ_DIM] * CFL_SIZE);
}
pos[COIL_DIM] = 0;
return 0;
}
/**
* Efficiently chain two matrix linops by multiplying the actual matrices together.
* Stores a copy of the new matrix.
* Returns: C = B A
*
* @param a first matrix (applied to input)
* @param b second matrix (applied to output of first matrix)
*/
struct linop_s* linop_matrix_chain(const struct linop_s* a, const struct linop_s* b)
{
const struct operator_matrix_s* a_data = CAST_DOWN(operator_matrix_s, linop_get_data(a));
const struct operator_matrix_s* b_data = CAST_DOWN(operator_matrix_s, linop_get_data(b));
// check compatibility
assert(linop_codomain(a)->N == linop_domain(b)->N);
assert(md_check_compat(linop_codomain(a)->N, 0u, linop_codomain(a)->dims, linop_domain(b)->dims));
unsigned int D = linop_domain(a)->N;
unsigned long outB_flags = md_nontriv_dims(D, linop_codomain(b)->dims);
unsigned long inB_flags = md_nontriv_dims(D, linop_domain(b)->dims);
unsigned long delB_flags = inB_flags & ~outB_flags;
unsigned int N = a_data->N;
assert(N == 2 * D);
long in_dims[N];
md_copy_dims(N, in_dims, a_data->in_dims);
long matA_dims[N];
md_copy_dims(N, matA_dims, a_data->mat_dims);
long matB_dims[N];
md_copy_dims(N, matB_dims, b_data->mat_dims);
long out_dims[N];
md_copy_dims(N, out_dims, b_data->out_dims);
for (unsigned int i = 0; i < D; i++) {
if (MD_IS_SET(delB_flags, i)) {
matA_dims[2 * i + 0] = a_data->mat_dims[2 * i + 1];
matA_dims[2 * i + 1] = a_data->mat_dims[2 * i + 0];
in_dims[2 * i + 0] = a_data->in_dims[2 * i + 1];
in_dims[2 * i + 1] = a_data->in_dims[2 * i + 0];
}
}
long matrix_dims[N];
md_singleton_dims(N, matrix_dims);
unsigned long iflags = md_nontriv_dims(N, in_dims);
unsigned long oflags = md_nontriv_dims(N, out_dims);
unsigned long flags = iflags | oflags;
// we combine a and b and sum over dims not in input or output
md_max_dims(N, flags, matrix_dims, matA_dims, matB_dims);
debug_printf(DP_DEBUG1, "tensor chain: %ld x %ld -> %ld\n",
md_calc_size(N, matA_dims), md_calc_size(N, matB_dims), md_calc_size(N, matrix_dims));
complex float* matrix = md_alloc(N, matrix_dims, CFL_SIZE);
debug_print_dims(DP_DEBUG2, N, matrix_dims);
debug_print_dims(DP_DEBUG2, N, in_dims);
debug_print_dims(DP_DEBUG2, N, matA_dims);
debug_print_dims(DP_DEBUG2, N, matB_dims);
debug_print_dims(DP_DEBUG2, N, out_dims);
md_ztenmul(N, matrix_dims, matrix, matA_dims, a_data->mat, matB_dims, b_data->mat);
// priv2 takes our doubled dimensions
struct operator_matrix_s* data = linop_matrix_priv2(N, out_dims, in_dims, matrix_dims, matrix);
/* although we internally use different dimensions we define the
* correct interface
*/
struct linop_s* c = linop_create(linop_codomain(b)->N, linop_codomain(b)->dims,
linop_domain(a)->N, linop_domain(a)->dims, CAST_UP(data),
linop_matrix_apply, linop_matrix_apply_adjoint,
linop_matrix_apply_normal, NULL, linop_matrix_del);
md_free(matrix);
return c;
}
int main_twixread(int argc, char* argv[argc])
{
long adcs = 0;
bool autoc = false;
bool linectr = false;
bool partctr = false;
long dims[DIMS];
md_singleton_dims(DIMS, dims);
struct opt_s opts[] = {
OPT_LONG('x', &(dims[READ_DIM]), "X", "number of samples (read-out)"),
OPT_LONG('y', &(dims[PHS1_DIM]), "Y", "phase encoding steps"),
OPT_LONG('z', &(dims[PHS2_DIM]), "Z", "partition encoding steps"),
OPT_LONG('s', &(dims[SLICE_DIM]), "S", "number of slices"),
OPT_LONG('v', &(dims[AVG_DIM]), "V", "number of averages"),
OPT_LONG('c', &(dims[COIL_DIM]), "C", "number of channels"),
OPT_LONG('n', &(dims[TIME_DIM]), "N", "number of repetitions"),
OPT_LONG('a', &adcs, "A", "total number of ADCs"),
OPT_SET('A', &autoc, "automatic [guess dimensions]"),
OPT_SET('L', &linectr, "use linectr offset"),
OPT_SET('P', &partctr, "use partctr offset"),
};
cmdline(&argc, argv, 2, 2, usage_str, help_str, ARRAY_SIZE(opts), opts);
if (0 == adcs)
adcs = dims[PHS1_DIM] * dims[PHS2_DIM] * dims[SLICE_DIM] * dims[TIME_DIM];
debug_print_dims(DP_DEBUG1, DIMS, dims);
int ifd;
if (-1 == (ifd = open(argv[1], O_RDONLY)))
error("error opening file.");
struct hdr_s hdr;
bool vd = siemens_meas_setup(ifd, &hdr);
long off[DIMS] = { 0 };
if (autoc) {
long max[DIMS] = { [COIL_DIM] = 1000 };
long min[DIMS] = { 0 }; // min is always 0
adcs = 0;
while (true) {
if (-1 == siemens_bounds(vd, ifd, min, max))
break;
debug_print_dims(DP_DEBUG3, DIMS, max);
adcs++;
}
for (unsigned int i = 0; i < DIMS; i++) {
off[i] = -min[i];
dims[i] = max[i] + off[i];
}
debug_printf(DP_DEBUG2, "Dimensions: ");
debug_print_dims(DP_DEBUG2, DIMS, dims);
debug_printf(DP_DEBUG2, "Offset: ");
debug_print_dims(DP_DEBUG2, DIMS, off);
siemens_meas_setup(ifd, &hdr); // reset
}
complex float* out = create_cfl(argv[2], DIMS, dims);
md_clear(DIMS, dims, out, CFL_SIZE);
long adc_dims[DIMS];
md_select_dims(DIMS, READ_FLAG|COIL_FLAG, adc_dims, dims);
void* buf = md_alloc(DIMS, adc_dims, CFL_SIZE);
while (adcs--) {
long pos[DIMS] = { [0 ... DIMS - 1] = 0 };
if (-1 == siemens_adc_read(vd, ifd, linectr, partctr, dims, pos, buf)) {
debug_printf(DP_WARN, "Stopping.\n");
break;
}
for (unsigned int i = 0; i < DIMS; i++)
pos[i] += off[i];
debug_print_dims(DP_DEBUG1, DIMS, pos);
if (!md_is_index(DIMS, pos, dims)) {
debug_printf(DP_WARN, "Index out of bounds.\n");
continue;
}
md_copy_block(DIMS, pos, dims, out, adc_dims, buf, CFL_SIZE);
}
md_free(buf);
unmap_cfl(DIMS, dims, out);
exit(0);
}
unsigned int optimize_dims(unsigned int D, unsigned int N, long dims[N], long (*strs[D])[N])
{
merge_dims(D, N, dims, strs);
unsigned int ND = remove_empty_dims(D, N, dims, strs);
if (0 == ND) { // atleast return a single dimension
dims[0] = 1;
for (unsigned int j = 0; j < D; j++)
(*strs[j])[0] = 0;
ND = 1;
}
debug_print_dims(DP_DEBUG4, ND, dims);
float blocking[N];
#ifdef BERKELEY_SVN
// actually those are not the blocking factors
// as used below but relative to fast memory
//demmel_factors(D, ND, blocking, strs);
UNUSED(demmel_factors);
#endif
#if 0
debug_printf(DP_DEBUG4, "DB: ");
for (unsigned int i = 0; i < ND; i++)
debug_printf(DP_DEBUG4, "%f\t", blocking[i]);
debug_printf(DP_DEBUG4, "\n");
#endif
#if 1
for (unsigned int i = 0; i < ND; i++)
blocking[i] = 0.5;
// blocking[i] = 1.;
#endif
// try to split dimensions according to blocking factors
// use space up to N
bool split = false;
do {
if (N == ND)
break;
split = split_dims(D, ND, dims, strs, blocking);
if (split)
ND++;
} while(split);
// printf("Split %c :", split ? 'y' : 'n');
// print_dims(ND, dims);
long max_strides[ND];
for (unsigned int i = 0; i < ND; i++) {
max_strides[i] = 0;
for (unsigned int j = 0; j < D; j++)
max_strides[i] = MAX(max_strides[i], (*strs[j])[i]);
}
unsigned int ord[ND];
compute_permutation(ND, ord, max_strides);
// for (unsigned int i = 0; i < ND; i++)
// printf("%d: %ld %d\n", i, max_strides[i], ord[i]);
#if 1
for (unsigned int j = 0; j < D; j++)
reorder_long(ND, ord, *strs[j]);
reorder_long(ND, ord, dims);
#endif
#if 0
printf("opt dims\n");
print_dims(ND, dims);
if (D > 0)
print_dims(ND, *strs[0]);
if (D > 1)
print_dims(ND, *strs[1]);
if (D > 2)
print_dims(ND, *strs[2]);
#endif
return ND;
}
int main_twixread(int argc, char* argv[argc])
{
int c;
long adcs = 0;
bool autoc = false;
bool linectr = false;
bool partctr = false;
long dims[DIMS];
md_singleton_dims(DIMS, dims);
while (-1 != (c = getopt(argc, argv, "x:y:z:s:c:a:n:PLAh"))) {
switch (c) {
case 'x':
dims[READ_DIM] = atoi(optarg);
break;
case 'y':
dims[PHS1_DIM] = atoi(optarg);
break;
case 'z':
dims[PHS2_DIM] = atoi(optarg);
break;
case 's':
dims[SLICE_DIM] = atoi(optarg);
break;
case 'v':
dims[AVG_DIM] = atoi(optarg);
break;
case 'n':
dims[TIME_DIM] = atoi(optarg);
break;
case 'a':
adcs = atoi(optarg);
break;
case 'A':
autoc = true;
break;
case 'c':
dims[COIL_DIM] = atoi(optarg);
break;
case 'P':
partctr = true;
break;
case 'L':
linectr = true;
break;
case 'h':
usage(argv[0], stdout);
help();
exit(0);
default:
usage(argv[0], stderr);
exit(1);
}
}
if (argc - optind != 2) {
usage(argv[0], stderr);
exit(1);
}
if (0 == adcs)
adcs = dims[PHS1_DIM] * dims[PHS2_DIM] * dims[SLICE_DIM] * dims[TIME_DIM];
debug_print_dims(DP_DEBUG1, DIMS, dims);
int ifd;
if (-1 == (ifd = open(argv[optind + 0], O_RDONLY)))
error("error opening file.");
struct hdr_s hdr;
bool vd = siemens_meas_setup(ifd, &hdr);
long off[DIMS] = { 0 };
if (autoc) {
long max[DIMS] = { [COIL_DIM] = 1000 };
long min[DIMS] = { 0 }; // min is always 0
adcs = 0;
while (true) {
if (-1 == siemens_bounds(vd, ifd, min, max))
break;
debug_print_dims(DP_DEBUG3, DIMS, max);
adcs++;
}
for (unsigned int i = 0; i < DIMS; i++) {
off[i] = -min[i];
dims[i] = max[i] + off[i];
}
debug_printf(DP_INFO, "Dimensions: ");
debug_print_dims(DP_INFO, DIMS, dims);
debug_printf(DP_INFO, "Offset: ");
debug_print_dims(DP_INFO, DIMS, off);
siemens_meas_setup(ifd, &hdr); // reset
}
complex float* out = create_cfl(argv[optind + 1], DIMS, dims);
md_clear(DIMS, dims, out, CFL_SIZE);
long adc_dims[DIMS];
md_select_dims(DIMS, READ_FLAG|COIL_FLAG, adc_dims, dims);
void* buf = md_alloc(DIMS, adc_dims, CFL_SIZE);
while (adcs--) {
long pos[DIMS] = { [0 ... DIMS - 1] = 0 };
if (-1 == siemens_adc_read(vd, ifd, linectr, partctr, dims, pos, buf)) {
debug_printf(DP_WARN, "Stopping.\n");
break;
}
for (unsigned int i = 0; i < DIMS; i++)
pos[i] += off[i];
debug_print_dims(DP_DEBUG1, DIMS, pos);
if (!md_is_index(DIMS, pos, dims)) {
debug_printf(DP_WARN, "Index out of bounds.\n");
continue;
}
md_copy_block(DIMS, pos, dims, out, adc_dims, buf, CFL_SIZE);
}
md_free(buf);
unmap_cfl(DIMS, dims, out);
exit(0);
}