From dffde065b1b19f65e396c5008f8ff1495db68429 Mon Sep 17 00:00:00 2001 From: Yuxuan Shui Date: Sun, 9 Dec 2018 02:04:00 +0000 Subject: [PATCH] Make the shadow actually gaussian The old shadow kernel formula is not "wrong" per se, as the only judge is if it looks good. However, the formula used is not really gaussian. Also, the kernel size calculation doesn't really make sense to me. This commit change the kernel formula to actually use a gaussian distribution. As far as I can tell, there is no perceivable difference. Except now the effective shadow radius is smaller for the same config. Also, made that part of code a bit more mordern. Signed-off-by: Yuxuan Shui --- src/common.h | 2 +- src/compton.c | 38 ++++++++++++++++++++++---------------- 2 files changed, 23 insertions(+), 17 deletions(-) diff --git a/src/common.h b/src/common.h index 9da7fde235..403e2972be 100644 --- a/src/common.h +++ b/src/common.h @@ -426,7 +426,7 @@ typedef uint32_t glx_prog_main_t; typedef struct conv { int size; - double *data; + double data[]; } conv; /// Linked list type of atoms. diff --git a/src/compton.c b/src/compton.c index f1dd0160ab..f77d17e712 100644 --- a/src/compton.c +++ b/src/compton.c @@ -569,36 +569,42 @@ check_fade_fin(session_t *ps, win **_w) { static double __attribute__((const)) gaussian(double r, double x, double y) { - return ((1 / (sqrt(2 * M_PI * r))) * - exp((- (x * x + y * y)) / (2 * r * r))); + // Formula can be found here: + // https://en.wikipedia.org/wiki/Gaussian_blur#Mathematics + // Except a special case for r == 0 to produce sharp shadows + if (r == 0) + return 1; + return exp(-0.5*(x*x+y*y)/(r*r))/(2*M_PI*r*r); } static conv * make_gaussian_map(double r) { conv *c; - int size = ((int) ceil((r * 3)) + 1) & ~1; - int center = size / 2; - int x, y; + int size = r*2+1; + int center = size/2; double t; - double g; - c = malloc(sizeof(conv) + size * size * sizeof(double)); + c = malloc(sizeof(conv)+size*size*sizeof(double)); c->size = size; - c->data = (double *) (c + 1); t = 0.0; - for (y = 0; y < size; y++) { - for (x = 0; x < size; x++) { - g = gaussian(r, x - center, y - center); + /*printf_errf("(): %f", r);*/ + for (int y = 0; y < size; y++) { + for (int x = 0; x < size; x++) { + double g = gaussian(r, x-center, y-center); t += g; - c->data[y * size + x] = g; + c->data[y*size+x] = g; + /*printf("%f ", c->data[y*size+x]);*/ } + /*printf("\n");*/ } - for (y = 0; y < size; y++) { - for (x = 0; x < size; x++) { - c->data[y * size + x] /= t; + for (int y = 0; y < size; y++) { + for (int x = 0; x < size; x++) { + c->data[y*size+x] /= t; + /*printf("%f ", c->data[y*size+x]);*/ } + /*printf("\n");*/ } return c; @@ -4227,7 +4233,7 @@ get_cfg(session_t *ps, int argc, char *const *argv, bool first_pass) { // Range checking and option assignments ps->o.fade_delta = max_i(ps->o.fade_delta, 1); - ps->o.shadow_radius = max_i(ps->o.shadow_radius, 1); + ps->o.shadow_radius = max_i(ps->o.shadow_radius, 0); ps->o.shadow_red = normalize_d(ps->o.shadow_red); ps->o.shadow_green = normalize_d(ps->o.shadow_green); ps->o.shadow_blue = normalize_d(ps->o.shadow_blue);