Revision: 53356
http://brlcad.svn.sourceforge.net/brlcad/?rev=53356&view=rev
Author: n_reed
Date: 2012-10-25 16:55:33 +0000 (Thu, 25 Oct 2012)
Log Message:
-----------
be smarter at picking number of sample points
Modified Paths:
--------------
brlcad/trunk/src/librt/primitives/ell/ell.c
Modified: brlcad/trunk/src/librt/primitives/ell/ell.c
===================================================================
--- brlcad/trunk/src/librt/primitives/ell/ell.c 2012-10-25 15:57:02 UTC (rev
53355)
+++ brlcad/trunk/src/librt/primitives/ell/ell.c 2012-10-25 16:55:33 UTC (rev
53356)
@@ -756,42 +756,62 @@
}
}
+/* choose number of samples to use per ellipse based on view and ell info */
+static int
+ell_ellipse_samples(
+ const struct rt_ell_internal *ell,
+ const struct rt_view_info *info)
+{
+ fastf_t samples_per_mm, mm_per_sample;
+ fastf_t radius, radius_samples;
+
+ samples_per_mm = sqrt(info->view_samples) / info->view_size;
+ mm_per_sample = 1.0 / samples_per_mm;
+
+ radius = fabs((MAGNITUDE(ell->a) + MAGNITUDE(ell->b) + MAGNITUDE(ell->c))
+ / 3.0);
+ radius_samples = radius * samples_per_mm;
+
+ /* (2 * PI * radius_samples) would give us the number of times we expect
+ * our ellipse curves to be sampled by the view. This is sufficient to
+ * produce a very good rasterized image, but it is actually overkill.
+ *
+ * If we reduce the number of lines segments, those segment may
+ * nevertheless pass through almost all of the same pixels in the view,
+ * producing an equivalent rasterized image. Rather than expending the
+ * effort to directly calculate how few segments we can get away with
+ * without reducing the apparant quality of the rasterized curve, we just
+ * use this empirical calculation.
+ */
+ return pow(bn_twopi * radius_samples, .55);
+}
+
int
rt_ell_adaptive_plot(struct rt_db_internal *ip, const struct rt_view_info
*info)
{
struct ell_draw_configuration config;
struct rt_ell_internal *eip;
- fastf_t samples;
BU_CK_LIST_HEAD(info->vhead);
RT_CK_DB_INTERNAL(ip);
eip = (struct rt_ell_internal *)ip->idb_ptr;
RT_ELL_CK_MAGIC(eip);
- samples = primitive_diagonal_samples(ip, info);
-
config.vhead = info->vhead;
VMOVE(config.ell_center, eip->v);
- config.points_per_section = sqrt(samples);
+ config.points_per_section = ell_ellipse_samples(eip, info);
+
if (config.points_per_section < 6) {
- config.points_per_section = 6;
+ return 0;
}
- if (config.points_per_section > 32) {
- config.points_per_section = 32;
- }
-
- config.num_cross_sections = sqrt(samples) / 3.0;
-
- if (config.num_cross_sections < 1) {
+ if (config.points_per_section < 16) {
config.num_cross_sections = 1;
+ } else {
+ config.num_cross_sections = sqrt(config.points_per_section);
}
- if (config.num_cross_sections > 5) {
- config.num_cross_sections = 5;
- }
-
VMOVE(config.ell_travel_vector, eip->a);
VMOVE(config.ell_axis_vector_a, eip->b);
VMOVE(config.ell_axis_vector_b, eip->c);
This was sent by the SourceForge.net collaborative development platform, the
world's largest Open Source development site.
------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_sfd2d_oct
_______________________________________________
BRL-CAD Source Commits mailing list
brlcad-commits@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/brlcad-commits
