Couple of refactors and tricks in the kernel to speed things up. ~5FPS average improvement

master
MitchellHansen 8 years ago
parent a40b5545e8
commit ce862feb0b

@ -46,8 +46,7 @@ float4 view_light(float4 in_color, float3 light, float4 light_color, float3 view
float specTmp = max(dot(normalize(convert_float3(mask)), halfwayVector), 0.0f);
in_color += pow(specTmp, 8.0f) * light_color * 0.5f / d;
}
if (in_color.w > 1.0){
if (in_color.w > 1.0f){
in_color.xyz *= in_color.w;
}
@ -96,11 +95,11 @@ bool cast_light_intersection_ray(
float3 delta_t = fabs(1.0f / ray_dir);
// offset is how far we are into a voxel, enables sub voxel movement
float3 offset = ((ray_pos)-floor(ray_pos)) * convert_float3(voxel_step);
// float3 offset = ;
// Intersection T is the collection of the next intersection points
// for all 3 axis XYZ.
float3 intersection_t = delta_t *offset;
float3 intersection_t = delta_t * ((ray_pos)-floor(ray_pos)) * convert_float3(voxel_step);
// for negative values, wrap around the delta_t
intersection_t += delta_t * -convert_float3(isless(intersection_t, 0));
@ -117,23 +116,18 @@ bool cast_light_intersection_ray(
intersection_t += delta_t * fabs(convert_float3(face_mask.xyz));
voxel.xyz += voxel_step.xyz * face_mask.xyz;
// If the ray went out of bounds
int3 overshoot = voxel < *map_dim;
int3 undershoot = voxel >= 0;
if (any(overshoot == (int3)(0, 0, 0)) ||
any(undershoot == (int3)(0, 0, 0))) {
if (any(voxel >= *map_dim) ||
any(voxel < 0)) {
return false;
}
// If we hit a voxel
int index = voxel.x + (*map_dim).x * (voxel.y + (*map_dim).z * (voxel.z));
int voxel_data = map[index];
int voxel_data = map[voxel.x + (*map_dim).x * (voxel.y + (*map_dim).z * (voxel.z))];
if (voxel_data != 0)
return true;
if (length_cutoff > 300)
if (++length_cutoff > 300)
return false;
//} while (any(isless(intersection_t, (float3)(distance_to_light - 1))));
@ -148,6 +142,10 @@ bool cast_light_intersection_ray(
// ====================================== Raycaster entry point =====================================
// ==================================================================================================
constant float4 fog_color = { 0.73f, 0.81f, 0.89f, 0.8f };
constant float4 overshoot_color = { 0.25f, 0.48f, 0.52f, 0.8f };
constant float4 overshoot_color_2 = { 0.25f, 0.1f, 0.52f, 0.8f };
__kernel void raycaster(
global char* map,
global int3* map_dim,
@ -166,19 +164,16 @@ __kernel void raycaster(
int x = get_global_id(0);
int y = get_global_id(1);
int global_id = x * y;
// int global_id = x * y;
// Get and set the random seed from seed memory
int seed = seed_memory[global_id];
int random_number = rand(&seed);
seed_memory[global_id] = seed;
//int seed = seed_memory[global_id];
//int random_number = rand(&seed);
//seed_memory[global_id] = seed;
// Get the pixel on the viewport, and find the view matrix ray that matches it
//int2 pixel = { global_id % (*resolution).x, global_id / (*resolution).x };
int2 pixel = (int2)(x, y);
int2 pixel = (int2)(get_global_id(0), get_global_id(1));
float3 ray_dir = projection_matrix[pixel.x + (*resolution).x * pixel.y];
@ -214,29 +209,18 @@ __kernel void raycaster(
float3 delta_t = fabs(1.0f / ray_dir);
// offset is how far we are into a voxel, enables sub voxel movement
float3 offset = ((*cam_pos) - floor(*cam_pos)) * convert_float3(voxel_step);
// Intersection T is the collection of the next intersection points
// for all 3 axis XYZ.
float3 intersection_t = delta_t * offset;
// delta_t * offset = intersection_t
float3 intersection_t = delta_t * ((*cam_pos) - floor(*cam_pos)) * convert_float3(voxel_step);
// for negative values, wrap around the delta_t
intersection_t += delta_t * -convert_float3(isless(intersection_t, 0));
// Hard cut-off for how far the ray can travel
int max_dist = 800;
int dist = 0;
int3 face_mask = { 0, 0, 0 };
float4 fog_color = { 0.73f, 0.81f, 0.89f, 0.8f };
float4 voxel_color = (float4)(0.0f, 0.0f, 0.0f, 0.001f);
float4 overshoot_color = { 0.25f, 0.48f, 0.52f, 0.8f };
float4 overshoot_color_2 = { 0.25f, 0.1f, 0.52f, 0.8f };
int voxel_data = 0;
// Andrew Woo's raycasting algo
do {
@ -245,33 +229,29 @@ __kernel void raycaster(
intersection_t += delta_t * fabs(convert_float3(face_mask.xyz));
voxel.xyz += voxel_step.xyz * face_mask.xyz;
// If the ray went out of bounds
int3 overshoot = voxel < *map_dim;
int3 undershoot = voxel >= 0;
if (overshoot.x == 0 || overshoot.y == 0 || overshoot.z == 0 || undershoot.x == 0 || undershoot.y == 0){
if (any(voxel >= *map_dim)){
write_imagef(image, pixel, white_light(mix(fog_color, overshoot_color, 1.0 - max(dist / 700.0f, (float)0)), (float3)(lights[7], lights[8], lights[9]), face_mask));
return;
}
if (undershoot.z == 0) {
if (any(voxel < 0)) {
write_imagef(image, pixel, white_light(mix(fog_color, overshoot_color_2, 1.0 - max(dist / 700.0f, (float)0)), (float3)(lights[7], lights[8], lights[9]), face_mask));
return;
}
// If we hit a voxel
int index = voxel.x + (*map_dim).x * (voxel.y + (*map_dim).z * (voxel.z));
int voxel_data = map[index];
voxel_data = map[voxel.x + (*map_dim).x * (voxel.y + (*map_dim).z * (voxel.z))];
// Debug, add the light position
if (all(voxel == convert_int3((float3)(lights[4], lights[5], lights[6]-3))))
voxel_data = 1;
// if (all(voxel == convert_int3((float3)(lights[4], lights[5], lights[6]-3))))
// voxel_data = 1;
if (voxel_data != 0) {
// Determine where on the 2d plane the ray intersected
float4 voxel_color = (float4)(0.0f, 0.0f, 0.0f, 0.001f);
float3 face_position = (float)(0);
float2 tile_face_position = (float)(0);
// Determine where on the 2d plane the ray intersected
float3 face_position = (float3)(0);
float2 tile_face_position = (float2)(0);
float3 sign = (float3)(1.0f, 1.0f, 1.0f);
// First determine the percent of the way the ray is towards the next intersection_t
@ -300,7 +280,7 @@ __kernel void raycaster(
else if (face_mask.z == -1) {
//sign.z *= -1.0;
sign.z *= -1.0;
float x_percent = (intersection_t.x - (intersection_t.z - delta_t.z)) / delta_t.x;
float y_percent = (intersection_t.y - (intersection_t.z - delta_t.z)) / delta_t.y;
@ -318,30 +298,12 @@ __kernel void raycaster(
// in order to cast the intersection ray!!
if (ray_dir.x > 0) {
face_position.x = -face_position.x + 1.0;
//face_position.x = -face_position.x + 1;
//tile_face_position.x = -tile_face_position.x + 1.0;
}
if (ray_dir.x < 0) {
//face_position.x = face_position.x + 0;
// This cures the Z semmetry on the X axis
tile_face_position.x = -tile_face_position.x + 1.0;
}
face_position.x = select((float)(face_position.x), (float)(-face_position.x + 1.0f), (int)(ray_dir.x > 0));
tile_face_position.x = select((float)(tile_face_position.x), (float)(-tile_face_position.x + 1.0f), (int)(ray_dir.x < 0));
if (ray_dir.y > 0){
face_position.y = - face_position.y + 1;
//tile_face_position.y = -tile_face_position.y + 1.0;
}
if (ray_dir.y < 0) {
//face_position.y = face_position.y + 0;
// This cures the Y semmetry on the Z tile faces
} else {
tile_face_position.x = 1.0 - tile_face_position.x;
// We run into the Hairy ball problem, so we need to define
@ -352,59 +314,39 @@ __kernel void raycaster(
}
}
if (ray_dir.z > 0) {
face_position.z = - face_position.z + 1;
//tile_face_position.y = tile_face_position.y + 0.0;
}
face_position.z = select((float)(face_position.z), (float)(-face_position.z + 1.0f), (int)(ray_dir.z > 0));
tile_face_position.y = select((float)(tile_face_position.y), (float)(-tile_face_position.y + 1.0f), (int)(ray_dir.z < 0));
if (ray_dir.z < 0) {
//sign.z *= -1.0;
// face_position.z = - face_position.z + 1;
//face_position.z = face_position.z + 0;
tile_face_position.y = -tile_face_position.y + 1.0;
}
if (voxel_data == 6){
// intersection_t = (1, 1, 1) - intersection_t;
// if (voxel_data == 6){
//
// //float3 ray_pos = (convert_float3(voxel) + face_position);
// //ray_dir *= sign;
// delta_t = fabs(1.0f / ray_dir);
// intersection_t = delta_t * (face_position * convert_float3(voxel_step));
//
// // for negative values, wrap around the delta_t
// intersection_t += delta_t * -convert_float3(isless(intersection_t, 0));
float3 ray_pos = (convert_float3(voxel) + face_position);
//ray_dir *= sign;
delta_t = fabs(1.0f / ray_dir);
float3 offset = ((ray_pos)-floor(ray_pos)) * convert_float3(voxel_step);
intersection_t = delta_t * offset;
// voxel_step = (int3)(1);//convert_int3(sign);
// voxel_step *= (ray_dir > 0) - (ray_dir < 0);
// continue;
// }
// for negative values, wrap around the delta_t
intersection_t += delta_t * -convert_float3(isless(intersection_t, 0));
voxel_step = (1, 1, 1);//convert_int3(sign);
voxel_step *= (ray_dir > 0) - (ray_dir < 0);
continue;
}
// Now either use the face position to retrieve a texture sample, or
// just a plain color for the voxel color
voxel_color = select((float4)voxel_color,
(float4)(0.0f, 0.239f, 0.419f, 0.0f),
(int4)(voxel_data == 6));
if (voxel_data == 6) {
voxel_color = (float4)(0.0f, 0.239f, 0.419f, 0.0f);
}
else if (voxel_data == 5) {
float2 tile_size = convert_float2(*atlas_dim / *tile_dim);
voxel_color = read_imagef(
voxel_color = select((float4)read_imagef(
texture_atlas,
convert_int2(tile_face_position * tile_size) +
convert_int2((float2)(3, 0) * tile_size)
);
convert_int2(tile_face_position * convert_float2(*atlas_dim / *tile_dim)) +
convert_int2((float2)(0, 0) * convert_float2(*atlas_dim / *tile_dim))
),
(float4)(0.0f, 0.239f, 0.419f, 0.0f),
(int4)(voxel_data == 5));
voxel_color.w = 0.0f;
//voxel_color = (float4)(0.25, 0.52, 0.30, 0.1);
}
else if (voxel_data == 1) {
voxel_color = (float4)(0.929f, 0.957f, 0.027f, 0.0f);
}
else {
voxel_color = (float4)(1.0f, 0.0f, 0.0f, 0.0f);
}
if (cast_light_intersection_ray(
map,
@ -416,8 +358,7 @@ __kernel void raycaster(
)) {
// If the light ray intersected an object on the way to the light point
float4 ambient_color = white_light(voxel_color, (float3)(1.0f, 1.0f, 1.0f), face_mask);
write_imagef(image, pixel, ambient_color);
write_imagef(image, pixel, white_light(voxel_color, (float3)(1.0f, 1.0f, 1.0f), face_mask));
return;
}
@ -441,9 +382,7 @@ __kernel void raycaster(
}
dist++;
} while (dist < 700.0f);
} while (++dist < 700.0f);
write_imagef(image, pixel, white_light(mix(fog_color, (float4)(0.40, 0.00, 0.40, 0.2), 1.0 - max(dist / 700.0f, (float)0)), (float3)(lights[7], lights[8], lights[9]), face_mask));

@ -184,15 +184,15 @@ void Old_Map::generate_terrain() {
}
}
for (int x = 100; x < 150; x += 10) {
for (int y = 100; y < 150; y += 10) {
for (int z = 0; z < 10; z += 1) {
voxel_data[x + dimensions.x * (y + dimensions.z * z)] = 6;
//for (int x = 100; x < 150; x += 10) {
// for (int y = 100; y < 150; y += 10) {
// for (int z = 0; z < 10; z += 1) {
//
// voxel_data[x + dimensions.x * (y + dimensions.z * z)] = 6;
}
}
}
// }
// }
//}
@ -225,7 +225,7 @@ void Old_Map::generate_terrain() {
for (int x = 60; x < 65; x++) {
for (int y = 60; y < 65; y++) {
for (int z = 30; z < 35; z++) {
voxel_data[x + dimensions.x * (y + dimensions.z * z)] = 5;
voxel_data[x + dimensions.x * (y + dimensions.z * z)] = 6;
}
}
}
@ -239,13 +239,13 @@ void Old_Map::generate_terrain() {
}
}
for (int x = 30; x < 60; x++) {
//for (int y = 0; y < dimensions.y; y++) {
for (int z = 0; z < 25; z++) {
voxel_data[x + dimensions.x * (50 + dimensions.z * z)] = 6;
}
//for (int x = 30; x < 60; x++) {
// //for (int y = 0; y < dimensions.y; y++) {
// for (int z = 10; z < 25; z++) {
// voxel_data[x + dimensions.x * (50 + dimensions.z * z)] = 6;
// }
// //}
//}
}
// Hand code in some constructions
@ -303,8 +303,8 @@ void Old_Map::generate_terrain() {
// }
//}
set_voxel(sf::Vector3i(45, 70, 5), 1);
set_voxel(sf::Vector3i(47, 70, 5), 1);
set_voxel(sf::Vector3i(45, 70, 6), 6);
set_voxel(sf::Vector3i(47, 70, 6), 6);
set_voxel(sf::Vector3i(100, 100, 50), 1);
}

Loading…
Cancel
Save