The `DefaultMixer` for mixing generic data types has some issues: 1. The full buffer is always zeroed, even if only some is used. 2. Finalizing also works on all values, even if only some are used. 3. "mixing" doesn't allow setting the first value, requiring that everything is cleared beforehand. This commit adds the following functionality: 1. Constructor with the specified `IndexMask` for preliminary zeroing. 2. `set` method to overwrite the value. 3. `finalize` with the specified mask to process a subset of values. This is useful in situations where you want to use the DefaultMixer without having to overwrite all the values many times. A performance improvement was observed for NURBS curve evaluation and attribute interpolation from the point to curve domain of about 15% and 35% respectively (100,000 curves). Differential Revision: https://developer.blender.org/D15434
258 lines
8.2 KiB
C++
258 lines
8.2 KiB
C++
/* SPDX-License-Identifier: GPL-2.0-or-later */
|
|
|
|
/** \file
|
|
* \ingroup bke
|
|
*/
|
|
|
|
#include "BLI_task.hh"
|
|
|
|
#include "BKE_attribute_math.hh"
|
|
#include "BKE_curves.hh"
|
|
|
|
namespace blender::bke::curves::nurbs {
|
|
|
|
bool check_valid_num_and_order(const int points_num,
|
|
const int8_t order,
|
|
const bool cyclic,
|
|
const KnotsMode knots_mode)
|
|
{
|
|
if (points_num < order) {
|
|
return false;
|
|
}
|
|
|
|
if (ELEM(knots_mode, NURBS_KNOT_MODE_BEZIER, NURBS_KNOT_MODE_ENDPOINT_BEZIER)) {
|
|
if (knots_mode == NURBS_KNOT_MODE_BEZIER && points_num <= order) {
|
|
return false;
|
|
}
|
|
return (!cyclic || points_num % (order - 1) == 0);
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
int calculate_evaluated_num(const int points_num,
|
|
const int8_t order,
|
|
const bool cyclic,
|
|
const int resolution,
|
|
const KnotsMode knots_mode)
|
|
{
|
|
if (!check_valid_num_and_order(points_num, order, cyclic, knots_mode)) {
|
|
return points_num;
|
|
}
|
|
return resolution * segments_num(points_num, cyclic);
|
|
}
|
|
|
|
int knots_num(const int points_num, const int8_t order, const bool cyclic)
|
|
{
|
|
if (cyclic) {
|
|
return points_num + order * 2 - 1;
|
|
}
|
|
return points_num + order;
|
|
}
|
|
|
|
void calculate_knots(const int points_num,
|
|
const KnotsMode mode,
|
|
const int8_t order,
|
|
const bool cyclic,
|
|
MutableSpan<float> knots)
|
|
{
|
|
BLI_assert(knots.size() == knots_num(points_num, order, cyclic));
|
|
UNUSED_VARS_NDEBUG(points_num);
|
|
|
|
const bool is_bezier = ELEM(mode, NURBS_KNOT_MODE_BEZIER, NURBS_KNOT_MODE_ENDPOINT_BEZIER);
|
|
const bool is_end_point = ELEM(mode, NURBS_KNOT_MODE_ENDPOINT, NURBS_KNOT_MODE_ENDPOINT_BEZIER);
|
|
/* Inner knots are always repeated once except on Bezier case. */
|
|
const int repeat_inner = is_bezier ? order - 1 : 1;
|
|
/* How many times to repeat 0.0 at the beginning of knot. */
|
|
const int head = is_end_point ? (order - (cyclic ? 1 : 0)) :
|
|
(is_bezier ? min_ii(2, repeat_inner) : 1);
|
|
/* Number of knots replicating widths of the starting knots.
|
|
* Covers both Cyclic and EndPoint cases. */
|
|
const int tail = cyclic ? 2 * order - 1 : (is_end_point ? order : 0);
|
|
|
|
int r = head;
|
|
float current = 0.0f;
|
|
|
|
const int offset = is_end_point && cyclic ? 1 : 0;
|
|
if (offset) {
|
|
knots[0] = current;
|
|
current += 1.0f;
|
|
}
|
|
|
|
for (const int i : IndexRange(offset, knots.size() - offset - tail)) {
|
|
knots[i] = current;
|
|
r--;
|
|
if (r == 0) {
|
|
current += 1.0;
|
|
r = repeat_inner;
|
|
}
|
|
}
|
|
|
|
const int tail_index = knots.size() - tail;
|
|
for (const int i : IndexRange(tail)) {
|
|
knots[tail_index + i] = current + (knots[i] - knots[0]);
|
|
}
|
|
}
|
|
|
|
static void calculate_basis_for_point(const float parameter,
|
|
const int points_num,
|
|
const int degree,
|
|
const Span<float> knots,
|
|
MutableSpan<float> r_weights,
|
|
int &r_start_index)
|
|
{
|
|
const int order = degree + 1;
|
|
|
|
int start = 0;
|
|
int end = 0;
|
|
for (const int i : IndexRange(points_num + degree)) {
|
|
const bool knots_equal = knots[i] == knots[i + 1];
|
|
if (knots_equal || parameter < knots[i] || parameter > knots[i + 1]) {
|
|
continue;
|
|
}
|
|
|
|
start = std::max(i - degree, 0);
|
|
end = i;
|
|
break;
|
|
}
|
|
|
|
Array<float, 12> buffer(order * 2, 0.0f);
|
|
|
|
buffer[end - start] = 1.0f;
|
|
|
|
for (const int i_order : IndexRange(2, degree)) {
|
|
if (end + i_order >= knots.size()) {
|
|
end = points_num + degree - i_order;
|
|
}
|
|
for (const int i : IndexRange(end - start + 1)) {
|
|
const int knot_index = start + i;
|
|
|
|
float new_basis = 0.0f;
|
|
if (buffer[i] != 0.0f) {
|
|
new_basis += ((parameter - knots[knot_index]) * buffer[i]) /
|
|
(knots[knot_index + i_order - 1] - knots[knot_index]);
|
|
}
|
|
|
|
if (buffer[i + 1] != 0.0f) {
|
|
new_basis += ((knots[knot_index + i_order] - parameter) * buffer[i + 1]) /
|
|
(knots[knot_index + i_order] - knots[knot_index + 1]);
|
|
}
|
|
|
|
buffer[i] = new_basis;
|
|
}
|
|
}
|
|
|
|
buffer.as_mutable_span().drop_front(end - start + 1).fill(0.0f);
|
|
r_weights.copy_from(buffer.as_span().take_front(order));
|
|
r_start_index = start;
|
|
}
|
|
|
|
void calculate_basis_cache(const int points_num,
|
|
const int evaluated_num,
|
|
const int8_t order,
|
|
const bool cyclic,
|
|
const Span<float> knots,
|
|
BasisCache &basis_cache)
|
|
{
|
|
BLI_assert(points_num > 0);
|
|
|
|
const int8_t degree = order - 1;
|
|
|
|
basis_cache.weights.resize(evaluated_num * order);
|
|
basis_cache.start_indices.resize(evaluated_num);
|
|
|
|
if (evaluated_num == 0) {
|
|
return;
|
|
}
|
|
|
|
MutableSpan<float> basis_weights(basis_cache.weights);
|
|
MutableSpan<int> basis_start_indices(basis_cache.start_indices);
|
|
|
|
const int last_control_point_index = cyclic ? points_num + degree : points_num;
|
|
const int evaluated_segment_num = segments_num(evaluated_num, cyclic);
|
|
|
|
const float start = knots[degree];
|
|
const float end = knots[last_control_point_index];
|
|
const float step = (end - start) / evaluated_segment_num;
|
|
for (const int i : IndexRange(evaluated_num)) {
|
|
/* Clamp parameter due to floating point inaccuracy. */
|
|
const float parameter = std::clamp(start + step * i, knots[0], knots[points_num + degree]);
|
|
|
|
MutableSpan<float> point_weights = basis_weights.slice(i * order, order);
|
|
|
|
calculate_basis_for_point(
|
|
parameter, last_control_point_index, degree, knots, point_weights, basis_start_indices[i]);
|
|
}
|
|
}
|
|
|
|
template<typename T>
|
|
static void interpolate_to_evaluated(const BasisCache &basis_cache,
|
|
const int8_t order,
|
|
const Span<T> src,
|
|
MutableSpan<T> dst)
|
|
{
|
|
attribute_math::DefaultMixer<T> mixer{dst};
|
|
|
|
threading::parallel_for(dst.index_range(), 128, [&](const IndexRange range) {
|
|
for (const int i : range) {
|
|
Span<float> point_weights = basis_cache.weights.as_span().slice(i * order, order);
|
|
for (const int j : point_weights.index_range()) {
|
|
const int point_index = (basis_cache.start_indices[i] + j) % src.size();
|
|
mixer.mix_in(i, src[point_index], point_weights[j]);
|
|
}
|
|
}
|
|
mixer.finalize(range);
|
|
});
|
|
}
|
|
|
|
template<typename T>
|
|
static void interpolate_to_evaluated_rational(const BasisCache &basis_cache,
|
|
const int8_t order,
|
|
const Span<float> control_weights,
|
|
const Span<T> src,
|
|
MutableSpan<T> dst)
|
|
{
|
|
attribute_math::DefaultMixer<T> mixer{dst};
|
|
|
|
threading::parallel_for(dst.index_range(), 128, [&](const IndexRange range) {
|
|
for (const int i : range) {
|
|
Span<float> point_weights = basis_cache.weights.as_span().slice(i * order, order);
|
|
|
|
for (const int j : point_weights.index_range()) {
|
|
const int point_index = (basis_cache.start_indices[i] + j) % src.size();
|
|
const float weight = point_weights[j] * control_weights[point_index];
|
|
mixer.mix_in(i, src[point_index], weight);
|
|
}
|
|
}
|
|
mixer.finalize(range);
|
|
});
|
|
}
|
|
|
|
void interpolate_to_evaluated(const BasisCache &basis_cache,
|
|
const int8_t order,
|
|
const Span<float> control_weights,
|
|
const GSpan src,
|
|
GMutableSpan dst)
|
|
{
|
|
if (basis_cache.invalid) {
|
|
dst.copy_from(src);
|
|
return;
|
|
}
|
|
|
|
BLI_assert(dst.size() == basis_cache.start_indices.size());
|
|
attribute_math::convert_to_static_type(src.type(), [&](auto dummy) {
|
|
using T = decltype(dummy);
|
|
if constexpr (!std::is_void_v<attribute_math::DefaultMixer<T>>) {
|
|
if (control_weights.is_empty()) {
|
|
interpolate_to_evaluated(basis_cache, order, src.typed<T>(), dst.typed<T>());
|
|
}
|
|
else {
|
|
interpolate_to_evaluated_rational(
|
|
basis_cache, order, control_weights, src.typed<T>(), dst.typed<T>());
|
|
}
|
|
}
|
|
});
|
|
}
|
|
|
|
} // namespace blender::bke::curves::nurbs
|