NASP/cuda-samples
85
0
Fork 0
mirror of https://github.com/NVIDIA/cuda-samples.git synced 2026-10-26 14:48:17 +08:00
Code Issues Packages Projects Releases Wiki Activity

Apply consistent code formatting across the repo. Add clang-format and pre-commit hooks.

Browse Source
This commit is contained in:
Rob Armstrong
2025-03-27 10:30:07 -07:00
parent 2cd58fbc9a
commit ceab6e8bcc
782 changed files with 107230 additions and 106548 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
Samples/8_Platform_Specific/Tegra/fluidsGLES/README.md
View File

@@ -34,4 +34,3 @@ Download and install the [CUDA Toolkit 12.5](https://developer.nvidia.com/cuda-d
Make sure the dependencies mentioned in [Dependencies]() section above are installed.
## References (for more details)

27
Samples/8_Platform_Specific/Tegra/fluidsGLES/defines.h
View File

@@ -28,21 +28,20 @@
#ifndef DEFINES_H
#define DEFINES_H
#define DIM 512 // Square size of solver domain
#define DS (DIM * DIM) // Total domain size
#define CPADW (DIM / 2 + 1) // Padded width for real->complex in-place FFT
#define RPADW \
(2 * (DIM / 2 + 1)) // Padded width for real->complex in-place FFT
#define PDS (DIM * CPADW) // Padded total domain size
#define DIM 512 // Square size of solver domain
#define DS (DIM * DIM) // Total domain size
#define CPADW (DIM / 2 + 1) // Padded width for real->complex in-place FFT
#define RPADW (2 * (DIM / 2 + 1)) // Padded width for real->complex in-place FFT
#define PDS (DIM * CPADW) // Padded total domain size
#define DT 0.09f // Delta T for interative solver
#define VIS 0.0025f // Viscosity constant
#define FORCE (5.8f * DIM) // Force scale factor
#define FR 4 // Force update radius
#define DT 0.09f // Delta T for interative solver
#define VIS 0.0025f // Viscosity constant
#define FORCE (5.8f * DIM) // Force scale factor
#define FR 4 // Force update radius
#define TILEX 64 // Tile width
#define TILEY 64 // Tile height
#define TIDSX 64 // Tids in X
#define TIDSY 4 // Tids in Y
#define TILEX 64 // Tile width
#define TILEY 64 // Tile height
#define TIDSX 64 // Tids in X
#define TIDSY 4 // Tids in Y
#endif

381
Samples/8_Platform_Specific/Tegra/fluidsGLES/fluidsGLES.cpp
View File

@@ -10,22 +10,20 @@
*/
// Includes
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <stdarg.h>
#include <unistd.h>
#include <X11/Xlib.h>
#include <X11/Xutil.h>
#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
void error_exit(const char* format, ... )
void error_exit(const char *format, ...)
{
va_list args;
va_start( args, format );
vfprintf( stderr, format, args );
va_end( args );
va_start(args, format);
vfprintf(stderr, format, args);
va_end(args);
exit(1);
}
@@ -33,19 +31,19 @@ void error_exit(const char* format, ... )
#include "graphics_interface.h"
// CUDA standard includes
#include <cuda_runtime.h>
#include <cuda_gl_interop.h>
#include <cuda_runtime.h>
// CUDA FFT Libraries
#include <cufft.h>
// CUDA helper functions
#include "helper_functions.h"
#include <rendercheck_gles.h>
#include <helper_cuda.h>
#include <rendercheck_gles.h>
#include "defines.h"
#include "fluidsGLES_kernels.h"
#include "helper_functions.h"
typedef float matrix4[4][4];
typedef float vector3[3];
@@ -65,20 +63,20 @@ void cleanup(void);
void reshape(int x, int y);
// CUFFT plan handle
cufftHandle planr2c;
cufftHandle planc2r;
cufftHandle planr2c;
cufftHandle planc2r;
static cData *vxfield = NULL;
static cData *vyfield = NULL;
cData *hvfield = NULL;
cData *dvfield = NULL;
cData *hvfield = NULL;
cData *dvfield = NULL;
static int wWidth = MAX(512, DIM);
static int wHeight = MAX(512, DIM);
static int clicked = 0;
static int fpsCount = 0;
static int fpsLimit = 1;
StopWatchInterface *timer = NULL;
static int clicked = 0;
static int fpsCount = 0;
static int fpsLimit = 1;
StopWatchInterface *timer = NULL;
int gui_mode; // For X window
// Rotate & translate variable temp., will remove and use shaders.
@@ -87,19 +85,19 @@ float translate_z = -3.0;
// Particle data
GLuint vbo = 0,vao = 0; // OpenGLES vertex buffer object
GLuint m_texture = 0;
GLuint vbo = 0, vao = 0; // OpenGLES vertex buffer object
GLuint m_texture = 0;
struct cudaGraphicsResource *cuda_vbo_resource; // handles OpenGLES-CUDA exchange
static cData *particles = NULL; // particle positions in host memory
static int lastx = 0, lasty = 0;
static cData *particles = NULL; // particle positions in host memory
static int lastx = 0, lasty = 0;
// Texture pitch
size_t tPitch = 0; // Now this is compatible with gcc in 64-bit
char *ref_file = NULL;
bool g_bQAAddTestForce = true;
int g_iFrameToCompare = 100;
int g_TotalErrors = 0;
char *ref_file = NULL;
bool g_bQAAddTestForce = true;
int g_iFrameToCompare = 100;
int g_TotalErrors = 0;
bool g_bExitESC = false;
@@ -107,7 +105,7 @@ const unsigned int window_width = 512;
const unsigned int window_height = 512;
// CheckFBO/BackBuffer class objects
CheckRender *g_CheckRender = NULL;
CheckRender *g_CheckRender = NULL;
void autoTest(char **);
void displayFrame();
@@ -132,8 +130,8 @@ GLuint mesh_shader = 0;
void mat_identity(matrix4 m)
{
m[0][1] = m[0][2] = m[0][3] = m[1][0] = m[1][2] = m[1][3] = m[2][0] =
m[2][1] = m[2][3] = m[3][0] = m[3][1] = m[3][2] = 0.0f;
m[0][1] = m[0][2] = m[0][3] = m[1][0] = m[1][2] = m[1][3] = m[2][0] = m[2][1] = m[2][3] = m[3][0] = m[3][1] =
m[3][2] = 0.0f;
m[0][0] = m[1][1] = m[2][2] = m[3][3] = 1.0f;
}
@@ -141,18 +139,14 @@ void mat_identity(matrix4 m)
void mat_multiply(matrix4 m0, matrix4 m1)
{
float m[4];
for(int r = 0; r < 4; r++)
{
for (int r = 0; r < 4; r++) {
m[0] = m[1] = m[2] = m[3] = 0.0f;
for(int c = 0; c < 4; c++)
{
for(int i = 0; i < 4; i++)
{
for (int c = 0; c < 4; c++) {
for (int i = 0; i < 4; i++) {
m[c] += m0[i][r] * m1[c][i];
}
}
for(int c = 0; c < 4; c++)
{
for (int c = 0; c < 4; c++) {
m0[c][r] = m[c];
}
}
@@ -163,52 +157,52 @@ void mat4f_Ortho(float left, float right, float bottom, float top, float near, f
float r_l = right - left;
float t_b = top - bottom;
float f_n = far - near;
float tx = - (right + left) / (right - left);
float ty = - (top + bottom) / (top - bottom);
float tz = - (far + near) / (far - near);
float tx = -(right + left) / (right - left);
float ty = -(top + bottom) / (top - bottom);
float tz = -(far + near) / (far - near);
matrix4 m2;
m2[0][0] = 2.0f/ r_l;
m2[0][0] = 2.0f / r_l;
m2[0][1] = 0.0f;
m2[0][2] = 0.0f;
m2[0][3] = 0.0f;
m2[1][0] = 0.0f;
m2[1][0] = 0.0f;
m2[1][1] = 2.0f / t_b;
m2[1][2] = 0.0f;
m2[1][3] = 0.0f;
m2[2][0] = 0.0f;
m2[2][1] = 0.0f;
m2[2][1] = 0.0f;
m2[2][2] = -2.0f / f_n;
m2[2][3] = 0.0f;
m2[3][0] = tx;
m2[3][1] = ty;
m2[3][1] = ty;
m2[3][2] = tz;
m2[3][3] = 1.0f;
mat_multiply(m, m2);
mat_multiply(m, m2);
}
void readAndCompileShaderFromGLSLFile(GLuint new_shaderprogram, const char *filename, GLenum shaderType)
{
FILE *file = fopen(filename,"rb"); // open shader text file
if (!file)
FILE *file = fopen(filename, "rb"); // open shader text file
if (!file)
error_exit("Filename %s does not exist\n", filename);
/* get the size of the file and read it */
fseek(file,0,SEEK_END);
fseek(file, 0, SEEK_END);
GLint size = ftell(file);
char *data = (char*)malloc(sizeof(char)*(size + 1));
memset(data, 0, sizeof(char)*(size + 1));
fseek(file,0,SEEK_SET);
size_t res = fread(data,1,size,file);
char *data = (char *)malloc(sizeof(char) * (size + 1));
memset(data, 0, sizeof(char) * (size + 1));
fseek(file, 0, SEEK_SET);
size_t res = fread(data, 1, size, file);
fclose(file);
GLuint shader = glCreateShader(shaderType);
glShaderSource(shader, 1, (const GLchar**)&data, &size);
glShaderSource(shader, 1, (const GLchar **)&data, &size);
glCompileShader(shader);
GET_GLERROR(0);
@@ -216,19 +210,18 @@ void readAndCompileShaderFromGLSLFile(GLuint new_shaderprogram, const char *file
glGetShaderiv(shader, GL_COMPILE_STATUS, &compile_success);
GET_GLERROR(0);
if (compile_success == GL_FALSE)
{
if (compile_success == GL_FALSE) {
printf("Compilation of %s failed!\n Reason:\n", filename);
GLint maxLength = 0;
glGetShaderiv(shader, GL_INFO_LOG_LENGTH, &maxLength);
char errorLog[maxLength];
glGetShaderInfoLog(shader, maxLength, &maxLength, &errorLog[0]);
printf("%s", errorLog);
glDeleteShader(shader);
glDeleteShader(shader);
exit(1);
}
@@ -259,8 +252,7 @@ GLuint ShaderCreate(const char *vshader_filename, const char *fshader_filename)
GLint link_success;
glGetProgramiv(new_shaderprogram, GL_LINK_STATUS, &link_success);
if (link_success == GL_FALSE)
{
if (link_success == GL_FALSE) {
printf("Linking of %s with %s failed!\n Reason:\n", vshader_filename, fshader_filename);
GLint maxLength = 0;
@@ -282,17 +274,16 @@ void motion(int x, int y)
// Convert motion coordinates to domain
float fx = (lastx / (float)wWidth);
float fy = (lasty / (float)wHeight);
int nx = (int)(fx * DIM);
int ny = (int)(fy * DIM);
int nx = (int)(fx * DIM);
int ny = (int)(fy * DIM);
if (clicked && nx < DIM-FR && nx > FR-1 && ny < DIM-FR && ny > FR-1)
{
if (clicked && nx < DIM - FR && nx > FR - 1 && ny < DIM - FR && ny > FR - 1) {
int ddx = x - lastx;
int ddy = y - lasty;
fx = ddx / (float)wWidth;
fy = ddy / (float)wHeight;
int spy = ny-FR;
int spx = nx-FR;
fx = ddx / (float)wWidth;
fy = ddy / (float)wHeight;
int spy = ny - FR;
int spx = nx - FR;
addForces(dvfield, DIM, DIM, spx, spy, FORCE * DT * fx, FORCE * DT * fy, FR);
lastx = x;
lasty = y;
@@ -302,12 +293,12 @@ void motion(int x, int y)
//===========================================================================
// InitGraphicsState() - initialize OpenGLES
//===========================================================================
static void InitGraphicsState(int argc, char** argv)
static void InitGraphicsState(int argc, char **argv)
{
char *GL_version = (char *)glGetString(GL_VERSION);
char *GL_vendor = (char *)glGetString(GL_VENDOR);
char *GL_renderer = (char *)glGetString(GL_RENDERER);
printf("Version: %s\n", GL_version);
printf("Vendor: %s\n", GL_vendor);
printf("Renderer: %s\n", GL_renderer);
@@ -316,26 +307,24 @@ static void InitGraphicsState(int argc, char** argv)
GLint bsize;
// initialize buffer object
glGenBuffers(1, &vbo);
glGenBuffers(1, &vbo);
glBindBuffer(GL_ARRAY_BUFFER, vbo);
glBufferData(GL_ARRAY_BUFFER, sizeof(cData) * DS, particles, GL_DYNAMIC_DRAW);
glGetBufferParameteriv(GL_ARRAY_BUFFER, GL_BUFFER_SIZE, &bsize);
if (bsize != (sizeof(cData) * DS))
{
if (bsize != (sizeof(cData) * DS)) {
printf("Failed to initialize GL extensions.\n");
exit(EXIT_FAILURE);
}
checkCudaErrors(cudaGraphicsGLRegisterBuffer(&cuda_vbo_resource, vbo, cudaGraphicsMapFlagsNone));
// GLSL stuff
char *vertex_shader_path = sdkFindFilePath("mesh.vert.glsl", argv[0]);
char *vertex_shader_path = sdkFindFilePath("mesh.vert.glsl", argv[0]);
char *fragment_shader_path = sdkFindFilePath("mesh.frag.glsl", argv[0]);
if (vertex_shader_path == NULL || fragment_shader_path == NULL)
{
if (vertex_shader_path == NULL || fragment_shader_path == NULL) {
printf("Error finding shader file\n");
exit(EXIT_FAILURE);
}
@@ -345,19 +334,18 @@ static void InitGraphicsState(int argc, char** argv)
free(vertex_shader_path);
free(fragment_shader_path);
glUseProgram(mesh_shader);
}
void displayFrame(void)
{
if (!ref_file)
{
if (!ref_file) {
sdkStartTimer(&timer);
simulateFluids();
}
GLint view_arr[4];
GLint view_arr[4];
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
@@ -366,20 +354,20 @@ void displayFrame(void)
glDepthMask(GL_FALSE);
glUseProgram(mesh_shader);
// Set modelview and projection matricies
GLint h_ModelViewMatrix = glGetUniformLocation(mesh_shader, "modelview");
GLint h_ProjectionMatrix = glGetUniformLocation(mesh_shader, "projection");
GLint h_ModelViewMatrix = glGetUniformLocation(mesh_shader, "modelview");
GLint h_ProjectionMatrix = glGetUniformLocation(mesh_shader, "projection");
matrix4 modelview;
matrix4 projection;
mat_identity(modelview);
mat_identity(projection);
// (float left, float right, float bottom, float top, float near, float far, matrix4 m)
// (float left, float right, float bottom, float top, float near, float far, matrix4 m)
mat4f_Ortho(0.0, 1.0, 1.0, 0.0, 0.0, 1.0, projection);
glUniformMatrix4fv(h_ModelViewMatrix, 1, GL_FALSE, (GLfloat*)modelview);
glUniformMatrix4fv(h_ProjectionMatrix, 1, GL_FALSE, (GLfloat*)projection);
glUniformMatrix4fv(h_ModelViewMatrix, 1, GL_FALSE, (GLfloat *)modelview);
glUniformMatrix4fv(h_ProjectionMatrix, 1, GL_FALSE, (GLfloat *)projection);
// Set position coords
GLint h_position = glGetAttribLocation(mesh_shader, "a_position");
@@ -388,7 +376,7 @@ void displayFrame(void)
glBindBuffer(GL_ARRAY_BUFFER, vbo);
glDrawArrays(GL_POINTS, 0, DS*sizeof(cData));
glDrawArrays(GL_POINTS, 0, DS * sizeof(cData));
glDisableVertexAttribArray(h_position);
glDisable(GL_DEPTH_TEST);
@@ -396,8 +384,7 @@ void displayFrame(void)
glDisable(GL_BLEND);
glDepthMask(GL_TRUE);
if (ref_file)
{
if (ref_file) {
return;
}
@@ -408,9 +395,8 @@ void displayFrame(void)
fpsCount++;
if (fpsCount == fpsLimit)
{
char fps[256];
if (fpsCount == fpsLimit) {
char fps[256];
float ifps = 1.f / (sdkGetAverageTimerValue(&timer) / 1000.f);
sprintf(fps, "Cuda/GL Stable Fluids (%d x %d): %3.1f fps", DIM, DIM, ifps);
graphics_set_windowtitle(fps);
@@ -423,7 +409,7 @@ void displayFrame(void)
void autoTest(char **argv)
{
CFrameBufferObject *fbo = new CFrameBufferObject(wWidth, wHeight, 4, false, GL_TEXTURE_2D);
g_CheckRender = new CheckFBO(wWidth, wHeight, 4, fbo);
g_CheckRender = new CheckFBO(wWidth, wHeight, 4, fbo);
g_CheckRender->setPixelFormat(GL_RGBA);
g_CheckRender->setExecPath(argv[0]);
@@ -431,26 +417,24 @@ void autoTest(char **argv)
fbo->bindRenderPath();
for (int count=0; count<g_iFrameToCompare; count++)
{
for (int count = 0; count < g_iFrameToCompare; count++) {
simulateFluids();
// add in a little force so the automated testing is interesing.
if (ref_file)
{
int x = wWidth/(count+1);
int y = wHeight/(count+1);
if (ref_file) {
int x = wWidth / (count + 1);
int y = wHeight / (count + 1);
float fx = (x / (float)wWidth);
float fy = (y / (float)wHeight);
int nx = (int)(fx * DIM);
int ny = (int)(fy * DIM);
int nx = (int)(fx * DIM);
int ny = (int)(fy * DIM);
int ddx = 35;
int ddy = 35;
fx = ddx / (float)wWidth;
fy = ddy / (float)wHeight;
int spy = ny-FR;
int spx = nx-FR;
fx = ddx / (float)wWidth;
fy = ddy / (float)wHeight;
int spy = ny - FR;
int spx = nx - FR;
addForces(dvfield, DIM, DIM, spx, spy, FORCE * DT * fx, FORCE * DT * fy, FR);
@@ -469,8 +453,7 @@ void autoTest(char **argv)
g_CheckRender->readback(wWidth, wHeight);
g_CheckRender->savePPM("fluidsGLES.ppm", true, NULL);
if (!g_CheckRender->PPMvsPPM("fluidsGLES.ppm", ref_file, MAX_EPSILON_ERROR, 0.25f))
{
if (!g_CheckRender->PPMvsPPM("fluidsGLES.ppm", ref_file, MAX_EPSILON_ERROR, 0.25f)) {
g_TotalErrors++;
}
}
@@ -482,18 +465,16 @@ bool runFluidsSimulation(int argc, char **argv, char *ref_file)
sdkCreateTimer(&timer);
if (ref_file != NULL)
{
// command line mode only - auto test
graphics_setup_window(0,0, wWidth, wHeight, sSDKname);
if (ref_file != NULL) {
// command line mode only - auto test
graphics_setup_window(0, 0, wWidth, wHeight, sSDKname);
InitGraphicsState(argc, argv); // set up GLES stuff
autoTest(argv);
cleanup();
}
else
{
else {
// create X11 window and set up associated OpenGL ES context
graphics_setup_window(0,0, wWidth, wHeight, sSDKname);
graphics_setup_window(0, 0, wWidth, wHeight, sSDKname);
InitGraphicsState(argc, argv); // set up GLES stuff
@@ -501,60 +482,50 @@ bool runFluidsSimulation(int argc, char **argv, char *ref_file)
graphics_swap_buffers();
XEvent event;
KeySym key;
char text[255];
char text[255];
while (1)
{
while (XPending(display) > 0)
{
while (1) {
while (XPending(display) > 0) {
XNextEvent(display, &event);
if (event.type==Expose && event.xexpose.count==0)
{
if (event.type == Expose && event.xexpose.count == 0) {
printf("Redraw requested!\n");
}
if (event.type==KeyPress && XLookupString(&event.xkey,text,255,&key,0)==1)
{
if (text[0] == 27 || text[0] == 'q' || text[0] == 'Q')
{
keyboard(text[0], 0, 0, argc, argv);
return true;
}
if (text[0] == 114)
{
keyboard(text[0], 0, 0, argc, argv);
}
printf("You pressed the %c key!\n",text[0]);
}
if (event.type==ButtonPress)
{
lastx = event.xbutton.x;
lasty = event.xbutton.y;
if (event.type == KeyPress && XLookupString(&event.xkey, text, 255, &key, 0) == 1) {
if (text[0] == 27 || text[0] == 'q' || text[0] == 'Q') {
keyboard(text[0], 0, 0, argc, argv);
return true;
}
if (text[0] == 114) {
keyboard(text[0], 0, 0, argc, argv);
}
printf("You pressed the %c key!\n", text[0]);
}
if (event.type == ButtonPress) {
lastx = event.xbutton.x;
lasty = event.xbutton.y;
clicked = !clicked;
}
if (event.type==ButtonRelease)
{
lastx = event.xbutton.x;
lasty = event.xbutton.y;
if (event.type == ButtonRelease) {
lastx = event.xbutton.x;
lasty = event.xbutton.y;
clicked = !clicked;
}
if (event.type == MotionNotify)
{
motion(event.xmotion.x, event.xmotion.y);
if (event.type == MotionNotify) {
motion(event.xmotion.x, event.xmotion.y);
}
else
{
else {
XFlush(display);
}
}
displayFrame();
usleep(1000); // need not take full CPU and GPU
usleep(1000); // need not take full CPU and GPU
}
}
@@ -567,70 +538,65 @@ bool runFluidsSimulation(int argc, char **argv, char *ref_file)
float myrand(void)
{
static int seed = 72191;
char sq[22];
char sq[22];
if (ref_file)
{
if (ref_file) {
seed *= seed;
sprintf(sq, "%010d", seed);
// pull the middle 5 digits out of sq
sq[8] = 0;
seed = atoi(&sq[3]);
seed = atoi(&sq[3]);
return seed/99999.f;
return seed / 99999.f;
}
else
{
return rand()/(float)RAND_MAX;
else {
return rand() / (float)RAND_MAX;
}
}
void initParticles(cData *p, int dx, int dy)
{
int i, j;
for (i = 0; i < dy; i++)
{
for (j = 0; j < dx; j++)
{
p[i*dx+j].x = (j+0.5f+(myrand() - 0.5f))/dx;
p[i*dx+j].y = (i+0.5f+(myrand() - 0.5f))/dy;
for (i = 0; i < dy; i++) {
for (j = 0; j < dx; j++) {
p[i * dx + j].x = (j + 0.5f + (myrand() - 0.5f)) / dx;
p[i * dx + j].y = (i + 0.5f + (myrand() - 0.5f)) / dy;
}
}
}
void keyboard(unsigned char key, int x, int y, int argc, char **argv)
{
switch (key)
{
case 'q':
case 'Q':
case 27:
g_bExitESC = true;
cleanup();
graphics_close_window(); // close window and destroy OpenGL ES context
return;
break;
case 'r':
printf("\nResetting\n");
memset(hvfield, 0, sizeof(cData) * DS);
cudaMemcpy(dvfield, hvfield, sizeof(cData) * DS, cudaMemcpyHostToDevice);
switch (key) {
case 'q':
case 'Q':
case 27:
g_bExitESC = true;
cleanup();
graphics_close_window(); // close window and destroy OpenGL ES context
return;
break;
case 'r':
printf("\nResetting\n");
memset(hvfield, 0, sizeof(cData) * DS);
cudaMemcpy(dvfield, hvfield, sizeof(cData) * DS, cudaMemcpyHostToDevice);
initParticles(particles, DIM, DIM);
initParticles(particles, DIM, DIM);
checkCudaErrors(cudaGraphicsUnregisterResource(cuda_vbo_resource));
checkCudaErrors(cudaGraphicsUnregisterResource(cuda_vbo_resource));
getLastCudaError("cudaGraphicsUnregisterBuffer failed");
getLastCudaError("cudaGraphicsUnregisterBuffer failed");
glBindBuffer(GL_ARRAY_BUFFER, 0);
glDeleteBuffers(1, &vbo);
InitGraphicsState(argc, argv); // set up GLES stuff
graphics_swap_buffers();
glBindBuffer(GL_ARRAY_BUFFER, 0);
glDeleteBuffers(1, &vbo);
InitGraphicsState(argc, argv); // set up GLES stuff
graphics_swap_buffers();
getLastCudaError("cudaGraphicsGLRegisterBuffer failed");
break;
getLastCudaError("cudaGraphicsGLRegisterBuffer failed");
break;
default:
break;
default:
break;
}
}
@@ -657,18 +623,19 @@ void cleanup(void)
int main(int argc, char **argv)
{
int devID;
int devID;
cudaDeviceProp deviceProps;
#if defined(__linux__)
setenv ("DISPLAY", ":0", 0);
setenv("DISPLAY", ":0", 0);
#endif
printf("%s Starting...\n\n", sSDKname);
printf("NOTE: The CUDA Samples are not meant for performance measurements. Results may vary when GPU Boost is enabled.\n\n");
printf("NOTE: The CUDA Samples are not meant for performance measurements. Results may vary when GPU Boost is "
"enabled.\n\n");
#if defined (__aarch64__) || defined(__arm__)
#if defined(__aarch64__) || defined(__arm__)
// find iGPU on the system which is compute capable which will perform GLES-CUDA interop
devID = findIntegratedGPU();
#else
@@ -678,12 +645,10 @@ int main(int argc, char **argv)
// get number of SMs on this GPU
checkCudaErrors(cudaGetDeviceProperties(&deviceProps, devID));
printf("CUDA device [%s] has %d Multi-Processors\n",
deviceProps.name, deviceProps.multiProcessorCount);
printf("CUDA device [%s] has %d Multi-Processors\n", deviceProps.name, deviceProps.multiProcessorCount);
// automated build testing harness
if (checkCmdLineFlag(argc, (const char **)argv, "file"))
{
if (checkCmdLineFlag(argc, (const char **)argv, "file")) {
getCmdLineArgumentString(argc, (const char **)argv, "file", &ref_file);
}
@@ -694,7 +659,7 @@ int main(int argc, char **argv)
memset(hvfield, 0, sizeof(cData) * DS);
// Allocate and initialize device data
checkCudaErrors(cudaMallocPitch((void **)&dvfield, &tPitch, sizeof(cData)*DIM, DIM));
checkCudaErrors(cudaMallocPitch((void **)&dvfield, &tPitch, sizeof(cData) * DIM, DIM));
checkCudaErrors(cudaMemcpy(dvfield, hvfield, sizeof(cData) * DS, cudaMemcpyHostToDevice));
// Temporary complex velocity field data
@@ -715,16 +680,14 @@ int main(int argc, char **argv)
runFluidsSimulation(argc, argv, ref_file);
if (ref_file)
{
if (ref_file) {
printf("[fluidsGLES] - Test Results: %d Failures\n", g_TotalErrors);
exit(g_TotalErrors == 0 ? EXIT_SUCCESS : EXIT_FAILURE);
}
sdkDeleteTimer(&timer);
if (!ref_file)
{
if (!ref_file) {
exit(EXIT_SUCCESS);
}

430
Samples/8_Platform_Specific/Tegra/fluidsGLES/fluidsGLES_kernels.cu
View File

@@ -25,13 +25,12 @@
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <cuda_runtime.h>
#include <cufft.h> // CUDA FFT Libraries
#include <helper_cuda.h> // Helper functions for CUDA Error handling
#include <stdio.h>
#include <stdlib.h>
#include <cuda_runtime.h>
#include <cufft.h> // CUDA FFT Libraries
#include <helper_cuda.h> // Helper functions for CUDA Error handling
// OpenGL Graphics includes
#include <GLES3/gl31.h>
@@ -40,51 +39,52 @@
// Texture object for reading velocity field
cudaTextureObject_t texObj;
static cudaArray *array = NULL;
static cudaArray *array = NULL;
// Particle data
extern GLuint vbo; // OpenGL vertex buffer object
extern struct cudaGraphicsResource
*cuda_vbo_resource; // handles OpenGL-CUDA exchange
extern GLuint vbo; // OpenGL vertex buffer object
extern struct cudaGraphicsResource *cuda_vbo_resource; // handles OpenGL-CUDA exchange
// Texture pitch
extern size_t tPitch;
extern size_t tPitch;
extern cufftHandle planr2c;
extern cufftHandle planc2r;
cData *vxfield = NULL;
cData *vyfield = NULL;
cData *vxfield = NULL;
cData *vyfield = NULL;
void setupTexture(int x, int y) {
cudaChannelFormatDesc desc = cudaCreateChannelDesc<float2>();
void setupTexture(int x, int y)
{
cudaChannelFormatDesc desc = cudaCreateChannelDesc<float2>();
cudaMallocArray(&array, &desc, y, x);
getLastCudaError("cudaMalloc failed");
cudaMallocArray(&array, &desc, y, x);
getLastCudaError("cudaMalloc failed");
cudaResourceDesc texRes;
memset(&texRes, 0, sizeof(cudaResourceDesc));
cudaResourceDesc texRes;
memset(&texRes, 0, sizeof(cudaResourceDesc));
texRes.resType = cudaResourceTypeArray;
texRes.res.array.array = array;
texRes.resType = cudaResourceTypeArray;
texRes.res.array.array = array;
cudaTextureDesc texDescr;
memset(&texDescr, 0, sizeof(cudaTextureDesc));
cudaTextureDesc texDescr;
memset(&texDescr, 0, sizeof(cudaTextureDesc));
texDescr.normalizedCoords = false;
texDescr.filterMode = cudaFilterModeLinear;
texDescr.addressMode[0] = cudaAddressModeWrap;
texDescr.readMode = cudaReadModeElementType;
texDescr.normalizedCoords = false;
texDescr.filterMode = cudaFilterModeLinear;
texDescr.addressMode[0] = cudaAddressModeWrap;
texDescr.readMode = cudaReadModeElementType;
checkCudaErrors(cudaCreateTextureObject(&texObj, &texRes, &texDescr, NULL));
checkCudaErrors(cudaCreateTextureObject(&texObj, &texRes, &texDescr, NULL));
}
void updateTexture(cData *data, size_t wib, size_t h, size_t pitch) {
checkCudaErrors(cudaMemcpy2DToArray(array, 0, 0, data, pitch, wib, h,
cudaMemcpyDeviceToDevice));
void updateTexture(cData *data, size_t wib, size_t h, size_t pitch)
{
checkCudaErrors(cudaMemcpy2DToArray(array, 0, 0, data, pitch, wib, h, cudaMemcpyDeviceToDevice));
}
void deleteTexture(void) {
checkCudaErrors(cudaDestroyTextureObject(texObj));
checkCudaErrors(cudaFreeArray(array));
void deleteTexture(void)
{
checkCudaErrors(cudaDestroyTextureObject(texObj));
checkCudaErrors(cudaFreeArray(array));
}
// Note that these kernels are designed to work with arbitrary
@@ -98,54 +98,61 @@ void deleteTexture(void) {
// This method adds constant force vectors to the velocity field
// stored in 'v' according to v(x,t+1) = v(x,t) + dt * f.
__global__ void addForces_k(cData *v, int dx, int dy, int spx, int spy,
float fx, float fy, int r, size_t pitch) {
int tx = threadIdx.x;
int ty = threadIdx.y;
cData *fj = (cData *)((char *)v + (ty + spy) * pitch) + tx + spx;
__global__ void addForces_k(cData *v, int dx, int dy, int spx, int spy, float fx, float fy, int r, size_t pitch)
{
int tx = threadIdx.x;
int ty = threadIdx.y;
cData *fj = (cData *)((char *)v + (ty + spy) * pitch) + tx + spx;
cData vterm = *fj;
tx -= r;
ty -= r;
float s = 1.f / (1.f + tx * tx * tx * tx + ty * ty * ty * ty);
vterm.x += s * fx;
vterm.y += s * fy;
*fj = vterm;
cData vterm = *fj;
tx -= r;
ty -= r;
float s = 1.f / (1.f + tx * tx * tx * tx + ty * ty * ty * ty);
vterm.x += s * fx;
vterm.y += s * fy;
*fj = vterm;
}
// This method performs the velocity advection step, where we
// trace velocity vectors back in time to update each grid cell.
// That is, v(x,t+1) = v(p(x,-dt),t). Here we perform bilinear
// interpolation in the velocity space.
__global__ void advectVelocity_k(cData *v, float *vx, float *vy, int dx,
int pdx, int dy, float dt, int lb,
cudaTextureObject_t texObject) {
int gtidx = blockIdx.x * blockDim.x + threadIdx.x;
int gtidy = blockIdx.y * (lb * blockDim.y) + threadIdx.y * lb;
int p;
__global__ void advectVelocity_k(cData *v,
float *vx,
float *vy,
int dx,
int pdx,
int dy,
float dt,
int lb,
cudaTextureObject_t texObject)
{
int gtidx = blockIdx.x * blockDim.x + threadIdx.x;
int gtidy = blockIdx.y * (lb * blockDim.y) + threadIdx.y * lb;
int p;
cData vterm, ploc;
float vxterm, vyterm;
cData vterm, ploc;
float vxterm, vyterm;
// gtidx is the domain location in x for this thread
if (gtidx < dx) {
for (p = 0; p < lb; p++) {
// fi is the domain location in y for this thread
int fi = gtidy + p;
// gtidx is the domain location in x for this thread
if (gtidx < dx) {
for (p = 0; p < lb; p++) {
// fi is the domain location in y for this thread
int fi = gtidy + p;
if (fi < dy) {
int fj = fi * pdx + gtidx;
vterm = tex2D<cData>(texObject, (float)gtidx, (float)fi);
ploc.x = (gtidx + 0.5f) - (dt * vterm.x * dx);
ploc.y = (fi + 0.5f) - (dt * vterm.y * dy);
vterm = tex2D<cData>(texObject, ploc.x, ploc.y);
vxterm = vterm.x;
vyterm = vterm.y;
vx[fj] = vxterm;
vy[fj] = vyterm;
}
if (fi < dy) {
int fj = fi * pdx + gtidx;
vterm = tex2D<cData>(texObject, (float)gtidx, (float)fi);
ploc.x = (gtidx + 0.5f) - (dt * vterm.x * dx);
ploc.y = (fi + 0.5f) - (dt * vterm.y * dy);
vterm = tex2D<cData>(texObject, ploc.x, ploc.y);
vxterm = vterm.x;
vyterm = vterm.y;
vx[fj] = vxterm;
vy[fj] = vyterm;
}
}
}
}
}
// This method performs velocity diffusion and forces mass conservation
@@ -156,201 +163,194 @@ __global__ void advectVelocity_k(cData *v, float *vx, float *vy, int dx,
// and k is the wavenumber. The projection step forces the Fourier
// velocity vectors to be orthogonal to the vectors for each
// wavenumber: v(k,t) = v(k,t) - ((k dot v(k,t) * k) / k^2.
__global__ void diffuseProject_k(cData *vx, cData *vy, int dx, int dy, float dt,
float visc, int lb) {
int gtidx = blockIdx.x * blockDim.x + threadIdx.x;
int gtidy = blockIdx.y * (lb * blockDim.y) + threadIdx.y * lb;
int p;
__global__ void diffuseProject_k(cData *vx, cData *vy, int dx, int dy, float dt, float visc, int lb)
{
int gtidx = blockIdx.x * blockDim.x + threadIdx.x;
int gtidy = blockIdx.y * (lb * blockDim.y) + threadIdx.y * lb;
int p;
cData xterm, yterm;
cData xterm, yterm;
// gtidx is the domain location in x for this thread
if (gtidx < dx) {
for (p = 0; p < lb; p++) {
// fi is the domain location in y for this thread
int fi = gtidy + p;
// gtidx is the domain location in x for this thread
if (gtidx < dx) {
for (p = 0; p < lb; p++) {
// fi is the domain location in y for this thread
int fi = gtidy + p;
if (fi < dy) {
int fj = fi * dx + gtidx;
xterm = vx[fj];
yterm = vy[fj];
if (fi < dy) {
int fj = fi * dx + gtidx;
xterm = vx[fj];
yterm = vy[fj];
// Compute the index of the wavenumber based on the
// data order produced by a standard NN FFT.
int iix = gtidx;
int iiy = (fi > dy / 2) ? (fi - (dy)) : fi;
// Compute the index of the wavenumber based on the
// data order produced by a standard NN FFT.
int iix = gtidx;
int iiy = (fi > dy / 2) ? (fi - (dy)) : fi;
// Velocity diffusion
float kk = (float)(iix * iix + iiy * iiy); // k^2
float diff = 1.f / (1.f + visc * dt * kk);
xterm.x *= diff;
xterm.y *= diff;
yterm.x *= diff;
yterm.y *= diff;
// Velocity diffusion
float kk = (float)(iix * iix + iiy * iiy); // k^2
float diff = 1.f / (1.f + visc * dt * kk);
xterm.x *= diff;
xterm.y *= diff;
yterm.x *= diff;
yterm.y *= diff;
// Velocity projection
if (kk > 0.f) {
float rkk = 1.f / kk;
// Real portion of velocity projection
float rkp = (iix * xterm.x + iiy * yterm.x);
// Imaginary portion of velocity projection
float ikp = (iix * xterm.y + iiy * yterm.y);
xterm.x -= rkk * rkp * iix;
xterm.y -= rkk * ikp * iix;
yterm.x -= rkk * rkp * iiy;
yterm.y -= rkk * ikp * iiy;
// Velocity projection
if (kk > 0.f) {
float rkk = 1.f / kk;
// Real portion of velocity projection
float rkp = (iix * xterm.x + iiy * yterm.x);
// Imaginary portion of velocity projection
float ikp = (iix * xterm.y + iiy * yterm.y);
xterm.x -= rkk * rkp * iix;
xterm.y -= rkk * ikp * iix;
yterm.x -= rkk * rkp * iiy;
yterm.y -= rkk * ikp * iiy;
}
vx[fj] = xterm;
vy[fj] = yterm;
}
}
vx[fj] = xterm;
vy[fj] = yterm;
}
}
}
}
// This method updates the velocity field 'v' using the two complex
// arrays from the previous step: 'vx' and 'vy'. Here we scale the
// real components by 1/(dx*dy) to account for an unnormalized FFT.
__global__ void updateVelocity_k(cData *v, float *vx, float *vy, int dx,
int pdx, int dy, int lb, size_t pitch) {
int gtidx = blockIdx.x * blockDim.x + threadIdx.x;
int gtidy = blockIdx.y * (lb * blockDim.y) + threadIdx.y * lb;
int p;
__global__ void updateVelocity_k(cData *v, float *vx, float *vy, int dx, int pdx, int dy, int lb, size_t pitch)
{
int gtidx = blockIdx.x * blockDim.x + threadIdx.x;
int gtidy = blockIdx.y * (lb * blockDim.y) + threadIdx.y * lb;
int p;
float vxterm, vyterm;
cData nvterm;
float vxterm, vyterm;
cData nvterm;
// gtidx is the domain location in x for this thread
if (gtidx < dx) {
for (p = 0; p < lb; p++) {
// fi is the domain location in y for this thread
int fi = gtidy + p;
// gtidx is the domain location in x for this thread
if (gtidx < dx) {
for (p = 0; p < lb; p++) {
// fi is the domain location in y for this thread
int fi = gtidy + p;
if (fi < dy) {
int fjr = fi * pdx + gtidx;
vxterm = vx[fjr];
vyterm = vy[fjr];
if (fi < dy) {
int fjr = fi * pdx + gtidx;
vxterm = vx[fjr];
vyterm = vy[fjr];
// Normalize the result of the inverse FFT
float scale = 1.f / (dx * dy);
nvterm.x = vxterm * scale;
nvterm.y = vyterm * scale;
// Normalize the result of the inverse FFT
float scale = 1.f / (dx * dy);
nvterm.x = vxterm * scale;
nvterm.y = vyterm * scale;
cData *fj = (cData *)((char *)v + fi * pitch) + gtidx;
*fj = nvterm;
}
} // If this thread is inside the domain in Y
} // If this thread is inside the domain in X
cData *fj = (cData *)((char *)v + fi * pitch) + gtidx;
*fj = nvterm;
}
} // If this thread is inside the domain in Y
} // If this thread is inside the domain in X
}
// This method updates the particles by moving particle positions
// according to the velocity field and time step. That is, for each
// particle: p(t+1) = p(t) + dt * v(p(t)).
__global__ void advectParticles_k(cData *part, cData *v, int dx, int dy,
float dt, int lb, size_t pitch) {
int gtidx = blockIdx.x * blockDim.x + threadIdx.x;
int gtidy = blockIdx.y * (lb * blockDim.y) + threadIdx.y * lb;
int p;
__global__ void advectParticles_k(cData *part, cData *v, int dx, int dy, float dt, int lb, size_t pitch)
{
int gtidx = blockIdx.x * blockDim.x + threadIdx.x;
int gtidy = blockIdx.y * (lb * blockDim.y) + threadIdx.y * lb;
int p;
// gtidx is the domain location in x for this thread
cData pterm, vterm;
// gtidx is the domain location in x for this thread
cData pterm, vterm;
if (gtidx < dx) {
for (p = 0; p < lb; p++) {
// fi is the domain location in y for this thread
int fi = gtidy + p;
if (gtidx < dx) {
for (p = 0; p < lb; p++) {
// fi is the domain location in y for this thread
int fi = gtidy + p;
if (fi < dy) {
int fj = fi * dx + gtidx;
pterm = part[fj];
if (fi < dy) {
int fj = fi * dx + gtidx;
pterm = part[fj];
int xvi = ((int)(pterm.x * dx));
int yvi = ((int)(pterm.y * dy));
vterm = *((cData *)((char *)v + yvi * pitch) + xvi);
int xvi = ((int)(pterm.x * dx));
int yvi = ((int)(pterm.y * dy));
vterm = *((cData *)((char *)v + yvi * pitch) + xvi);
pterm.x += dt * vterm.x;
pterm.x = pterm.x - (int)pterm.x;
pterm.x += 1.f;
pterm.x = pterm.x - (int)pterm.x;
pterm.y += dt * vterm.y;
pterm.y = pterm.y - (int)pterm.y;
pterm.y += 1.f;
pterm.y = pterm.y - (int)pterm.y;
pterm.x += dt * vterm.x;
pterm.x = pterm.x - (int)pterm.x;
pterm.x += 1.f;
pterm.x = pterm.x - (int)pterm.x;
pterm.y += dt * vterm.y;
pterm.y = pterm.y - (int)pterm.y;
pterm.y += 1.f;
pterm.y = pterm.y - (int)pterm.y;
part[fj] = pterm;
}
} // If this thread is inside the domain in Y
} // If this thread is inside the domain in X
part[fj] = pterm;
}
} // If this thread is inside the domain in Y
} // If this thread is inside the domain in X
}
// These are the external function calls necessary for launching fluid simuation
extern "C" void addForces(cData *v, int dx, int dy, int spx, int spy, float fx,
float fy, int r) {
dim3 tids(2 * r + 1, 2 * r + 1);
extern "C" void addForces(cData *v, int dx, int dy, int spx, int spy, float fx, float fy, int r)
{
dim3 tids(2 * r + 1, 2 * r + 1);
addForces_k<<<1, tids>>>(v, dx, dy, spx, spy, fx, fy, r, tPitch);
getLastCudaError("addForces_k failed.");
addForces_k<<<1, tids>>>(v, dx, dy, spx, spy, fx, fy, r, tPitch);
getLastCudaError("addForces_k failed.");
}
extern "C" void advectVelocity(cData *v, float *vx, float *vy, int dx, int pdx,
int dy, float dt) {
dim3 grid((dx / TILEX) + (!(dx % TILEX) ? 0 : 1),
(dy / TILEY) + (!(dy % TILEY) ? 0 : 1));
extern "C" void advectVelocity(cData *v, float *vx, float *vy, int dx, int pdx, int dy, float dt)
{
dim3 grid((dx / TILEX) + (!(dx % TILEX) ? 0 : 1), (dy / TILEY) + (!(dy % TILEY) ? 0 : 1));
dim3 tids(TIDSX, TIDSY);
dim3 tids(TIDSX, TIDSY);
updateTexture(v, DIM * sizeof(cData), DIM, tPitch);
advectVelocity_k<<<grid, tids>>>(v, vx, vy, dx, pdx, dy, dt, TILEY / TIDSY,
texObj);
getLastCudaError("advectVelocity_k failed.");
updateTexture(v, DIM * sizeof(cData), DIM, tPitch);
advectVelocity_k<<<grid, tids>>>(v, vx, vy, dx, pdx, dy, dt, TILEY / TIDSY, texObj);
getLastCudaError("advectVelocity_k failed.");
}
extern "C" void diffuseProject(cData *vx, cData *vy, int dx, int dy, float dt,
float visc) {
// Forward FFT
checkCudaErrors(cufftExecR2C(planr2c, (cufftReal *)vx, (cufftComplex *)vx));
checkCudaErrors(cufftExecR2C(planr2c, (cufftReal *)vy, (cufftComplex *)vy));
extern "C" void diffuseProject(cData *vx, cData *vy, int dx, int dy, float dt, float visc)
{
// Forward FFT
checkCudaErrors(cufftExecR2C(planr2c, (cufftReal *)vx, (cufftComplex *)vx));
checkCudaErrors(cufftExecR2C(planr2c, (cufftReal *)vy, (cufftComplex *)vy));
uint3 grid = make_uint3((dx / TILEX) + (!(dx % TILEX) ? 0 : 1),
(dy / TILEY) + (!(dy % TILEY) ? 0 : 1), 1);
uint3 tids = make_uint3(TIDSX, TIDSY, 1);
uint3 grid = make_uint3((dx / TILEX) + (!(dx % TILEX) ? 0 : 1), (dy / TILEY) + (!(dy % TILEY) ? 0 : 1), 1);
uint3 tids = make_uint3(TIDSX, TIDSY, 1);
diffuseProject_k<<<grid, tids>>>(vx, vy, dx, dy, dt, visc, TILEY / TIDSY);
getLastCudaError("diffuseProject_k failed.");
diffuseProject_k<<<grid, tids>>>(vx, vy, dx, dy, dt, visc, TILEY / TIDSY);
getLastCudaError("diffuseProject_k failed.");
// Inverse FFT
checkCudaErrors(cufftExecC2R(planc2r, (cufftComplex *)vx, (cufftReal *)vx));
checkCudaErrors(cufftExecC2R(planc2r, (cufftComplex *)vy, (cufftReal *)vy));
// Inverse FFT
checkCudaErrors(cufftExecC2R(planc2r, (cufftComplex *)vx, (cufftReal *)vx));
checkCudaErrors(cufftExecC2R(planc2r, (cufftComplex *)vy, (cufftReal *)vy));
}
extern "C" void updateVelocity(cData *v, float *vx, float *vy, int dx, int pdx,
int dy) {
dim3 grid((dx / TILEX) + (!(dx % TILEX) ? 0 : 1),
(dy / TILEY) + (!(dy % TILEY) ? 0 : 1));
dim3 tids(TIDSX, TIDSY);
extern "C" void updateVelocity(cData *v, float *vx, float *vy, int dx, int pdx, int dy)
{
dim3 grid((dx / TILEX) + (!(dx % TILEX) ? 0 : 1), (dy / TILEY) + (!(dy % TILEY) ? 0 : 1));
dim3 tids(TIDSX, TIDSY);
updateVelocity_k<<<grid, tids>>>(v, vx, vy, dx, pdx, dy, TILEY / TIDSY,
tPitch);
getLastCudaError("updateVelocity_k failed.");
updateVelocity_k<<<grid, tids>>>(v, vx, vy, dx, pdx, dy, TILEY / TIDSY, tPitch);
getLastCudaError("updateVelocity_k failed.");
}
extern "C" void advectParticles(GLuint vbo, cData *v, int dx, int dy,
float dt) {
dim3 grid((dx / TILEX) + (!(dx % TILEX) ? 0 : 1),
(dy / TILEY) + (!(dy % TILEY) ? 0 : 1));
dim3 tids(TIDSX, TIDSY);
extern "C" void advectParticles(GLuint vbo, cData *v, int dx, int dy, float dt)
{
dim3 grid((dx / TILEX) + (!(dx % TILEX) ? 0 : 1), (dy / TILEY) + (!(dy % TILEY) ? 0 : 1));
dim3 tids(TIDSX, TIDSY);
cData *p;
checkCudaErrors(cudaGraphicsMapResources(1, &cuda_vbo_resource, 0));
getLastCudaError("cudaGraphicsMapResources failed");
cData *p;
checkCudaErrors(cudaGraphicsMapResources(1, &cuda_vbo_resource, 0));
getLastCudaError("cudaGraphicsMapResources failed");
size_t num_bytes;
checkCudaErrors(cudaGraphicsResourceGetMappedPointer((void **)&p, &num_bytes,
cuda_vbo_resource));
getLastCudaError("cudaGraphicsResourceGetMappedPointer failed");
size_t num_bytes;
checkCudaErrors(cudaGraphicsResourceGetMappedPointer((void **)&p, &num_bytes, cuda_vbo_resource));
getLastCudaError("cudaGraphicsResourceGetMappedPointer failed");
advectParticles_k<<<grid, tids>>>(p, v, dx, dy, dt, TILEY / TIDSY, tPitch);
getLastCudaError("advectParticles_k failed.");
advectParticles_k<<<grid, tids>>>(p, v, dx, dy, dt, TILEY / TIDSY, tPitch);
getLastCudaError("advectParticles_k failed.");
checkCudaErrors(cudaGraphicsUnmapResources(1, &cuda_vbo_resource, 0));
checkCudaErrors(cudaGraphicsUnmapResources(1, &cuda_vbo_resource, 0));
}

17
Samples/8_Platform_Specific/Tegra/fluidsGLES/fluidsGLES_kernels.cuh
View File

@@ -39,16 +39,14 @@ void deleteTexture(void);
// This method adds constant force vectors to the velocity field
// stored in 'v' according to v(x,t+1) = v(x,t) + dt * f.
__global__ void addForces_k(cData *v, int dx, int dy, int spx, int spy,
float fx, float fy, int r, size_t pitch);
__global__ void addForces_k(cData *v, int dx, int dy, int spx, int spy, float fx, float fy, int r, size_t pitch);
// This method performs the velocity advection step, where we
// trace velocity vectors back in time to update each grid cell.
// That is, v(x,t+1) = v(p(x,-dt),t). Here we perform bilinear
// interpolation in the velocity space.
__global__ void advectVelocity_k(cData *v, float *vx, float *vy, int dx,
int pdx, int dy, float dt, int lb,
cudaTextureObject_t tex);
__global__ void
advectVelocity_k(cData *v, float *vx, float *vy, int dx, int pdx, int dy, float dt, int lb, cudaTextureObject_t tex);
// This method performs velocity diffusion and forces mass conservation
// in the frequency domain. The inputs 'vx' and 'vy' are complex-valued
@@ -58,19 +56,16 @@ __global__ void advectVelocity_k(cData *v, float *vx, float *vy, int dx,
// and k is the wavenumber. The projection step forces the Fourier
// velocity vectors to be orthogonal to the wave wave vectors for each
// wavenumber: v(k,t) = v(k,t) - ((k dot v(k,t) * k) / k^2.
__global__ void diffuseProject_k(cData *vx, cData *vy, int dx, int dy, float dt,
float visc, int lb);
__global__ void diffuseProject_k(cData *vx, cData *vy, int dx, int dy, float dt, float visc, int lb);
// This method updates the velocity field 'v' using the two complex
// arrays from the previous step: 'vx' and 'vy'. Here we scale the
// real components by 1/(dx*dy) to account for an unnormalized FFT.
__global__ void updateVelocity_k(cData *v, float *vx, float *vy, int dx,
int pdx, int dy, int lb, size_t pitch);
__global__ void updateVelocity_k(cData *v, float *vx, float *vy, int dx, int pdx, int dy, int lb, size_t pitch);
// This method updates the particles by moving particle positions
// according to the velocity field and time step. That is, for each
// particle: p(t+1) = p(t) + dt * v(p(t)).
__global__ void advectParticles_k(cData *part, cData *v, int dx, int dy,
float dt, int lb, size_t pitch);
__global__ void advectParticles_k(cData *part, cData *v, int dx, int dy, float dt, int lb, size_t pitch);
#endif

19
Samples/8_Platform_Specific/Tegra/fluidsGLES/fluidsGLES_kernels.h
View File

@@ -20,16 +20,14 @@ void deleteTexture(void);
// This method adds constant force vectors to the velocity field
// stored in 'v' according to v(x,t+1) = v(x,t) + dt * f.
__global__ void
addForces_k(cData *v, int dx, int dy, int spx, int spy, float fx, float fy, int r, size_t pitch);
__global__ void addForces_k(cData *v, int dx, int dy, int spx, int spy, float fx, float fy, int r, size_t pitch);
// This method performs the velocity advection step, where we
// trace velocity vectors back in time to update each grid cell.
// That is, v(x,t+1) = v(p(x,-dt),t). Here we perform bilinear
// interpolation in the velocity space.
__global__ void
advectVelocity_k(cData *v, float *vx, float *vy,
int dx, int pdx, int dy, float dt, int lb, cudaTextureObject_t tex);
advectVelocity_k(cData *v, float *vx, float *vy, int dx, int pdx, int dy, float dt, int lb, cudaTextureObject_t tex);
// This method performs velocity diffusion and forces mass conservation
// in the frequency domain. The inputs 'vx' and 'vy' are complex-valued
@@ -39,23 +37,16 @@ advectVelocity_k(cData *v, float *vx, float *vy,
// and k is the wavenumber. The projection step forces the Fourier
// velocity vectors to be orthogonal to the wave wave vectors for each
// wavenumber: v(k,t) = v(k,t) - ((k dot v(k,t) * k) / k^2.
__global__ void
diffuseProject_k(cData *vx, cData *vy, int dx, int dy, float dt,
float visc, int lb);
__global__ void diffuseProject_k(cData *vx, cData *vy, int dx, int dy, float dt, float visc, int lb);
// This method updates the velocity field 'v' using the two complex
// arrays from the previous step: 'vx' and 'vy'. Here we scale the
// real components by 1/(dx*dy) to account for an unnormalized FFT.
__global__ void
updateVelocity_k(cData *v, float *vx, float *vy,
int dx, int pdx, int dy, int lb, size_t pitch);
__global__ void updateVelocity_k(cData *v, float *vx, float *vy, int dx, int pdx, int dy, int lb, size_t pitch);
// This method updates the particles by moving particle positions
// according to the velocity field and time step. That is, for each
// particle: p(t+1) = p(t) + dt * v(p(t)).
__global__ void
advectParticles_k(cData *part, cData *v, int dx, int dy,
float dt, int lb, size_t pitch);
__global__ void advectParticles_k(cData *part, cData *v, int dx, int dy, float dt, int lb, size_t pitch);
#endif

252
Samples/8_Platform_Specific/Tegra/fluidsGLES/graphics_interface.h
View File

@@ -31,165 +31,183 @@
#include <stdlib.h>
Display *display;
int screen;
Window win = 0;
int screen;
Window win = 0;
#include <GLES3/gl31.h>
#include <EGL/egl.h>
#include <EGL/eglext.h>
#include <GLES3/gl31.h>
#define GET_GLERROR(ret) \
\
{ \
GLenum err = glGetError(); \
if (err != GL_NO_ERROR) { \
fprintf(stderr, "[%s line %d] OpenGL Error: 0x%x\n", __FILE__, __LINE__, \
err); \
fflush(stderr); \
\
switch (err) { \
case GL_INVALID_ENUM: \
printf("GL_INVALID_ENUM\n"); \
break; \
case GL_INVALID_VALUE: \
printf("GL_INVALID_VALUE\n"); \
break; \
case GL_INVALID_OPERATION: \
printf("GL_INVALID_OPERATION\n"); \
break; \
case GL_OUT_OF_MEMORY: \
printf("GL_OUT_OF_MEMORY\n"); \
break; \
case GL_INVALID_FRAMEBUFFER_OPERATION: \
printf("GL_INVALID_FRAMEBUFFER_OPERATION\n"); \
break; \
default: \
printf("UKNOWN OPENGL ERROR CODE 0x%x\n", err); \
}; \
} \
\
}
#define GET_GLERROR(ret) \
\
{ \
GLenum err = glGetError(); \
if (err != GL_NO_ERROR) { \
fprintf(stderr, "[%s line %d] OpenGL Error: 0x%x\n", __FILE__, __LINE__, err); \
fflush(stderr); \
\
switch (err) { \
case GL_INVALID_ENUM: \
printf("GL_INVALID_ENUM\n"); \
break; \
case GL_INVALID_VALUE: \
printf("GL_INVALID_VALUE\n"); \
break; \
case GL_INVALID_OPERATION: \
printf("GL_INVALID_OPERATION\n"); \
break; \
case GL_OUT_OF_MEMORY: \
printf("GL_OUT_OF_MEMORY\n"); \
break; \
case GL_INVALID_FRAMEBUFFER_OPERATION: \
printf("GL_INVALID_FRAMEBUFFER_OPERATION\n"); \
break; \
default: \
printf("UKNOWN OPENGL ERROR CODE 0x%x\n", err); \
}; \
} \
}
EGLDisplay eglDisplay = EGL_NO_DISPLAY;
EGLSurface eglSurface = EGL_NO_SURFACE;
EGLContext eglContext = EGL_NO_CONTEXT;
int graphics_setup_window(int xpos, int ypos, int width, int height,
const char *windowname) {
// OpenGL ES 3.1
EGLint configAttrs[] = {
EGL_RED_SIZE, 1, EGL_GREEN_SIZE, 1, EGL_BLUE_SIZE, 1, EGL_DEPTH_SIZE, 16,
EGL_SAMPLE_BUFFERS, 0, EGL_SAMPLES, 0,
// EGL_CONFORMANT, EGL_OPENGL_BIT,
EGL_RENDERABLE_TYPE, EGL_OPENGL_ES2_BIT, // 3_BIT_KHR,
EGL_NONE};
EGLint contextAttrs[] = {EGL_CONTEXT_CLIENT_VERSION, 3, EGL_NONE};
int graphics_setup_window(int xpos, int ypos, int width, int height, const char *windowname)
{
// OpenGL ES 3.1
EGLint configAttrs[] = {EGL_RED_SIZE,
1,
EGL_GREEN_SIZE,
1,
EGL_BLUE_SIZE,
1,
EGL_DEPTH_SIZE,
16,
EGL_SAMPLE_BUFFERS,
0,
EGL_SAMPLES,
0,
// EGL_CONFORMANT, EGL_OPENGL_BIT,
EGL_RENDERABLE_TYPE,
EGL_OPENGL_ES2_BIT, // 3_BIT_KHR,
EGL_NONE};
EGLint contextAttrs[] = {EGL_CONTEXT_CLIENT_VERSION, 3, EGL_NONE};
EGLConfig *configList = NULL;
EGLint configCount;
EGLConfig *configList = NULL;
EGLint configCount;
display = XOpenDisplay(NULL);
if (!display) error_exit("Error opening X display.\n");
display = XOpenDisplay(NULL);
if (!display)
error_exit("Error opening X display.\n");
screen = DefaultScreen(display);
screen = DefaultScreen(display);
eglDisplay = eglGetDisplay(0);
eglDisplay = eglGetDisplay(0);
if (eglDisplay == EGL_NO_DISPLAY)
error_exit("EGL failed to obtain display\n");
if (eglDisplay == EGL_NO_DISPLAY)
error_exit("EGL failed to obtain display\n");
if (!eglInitialize(eglDisplay, 0, 0))
error_exit("EGL failed to initialize\n");
if (!eglInitialize(eglDisplay, 0, 0))
error_exit("EGL failed to initialize\n");
if (!eglChooseConfig(eglDisplay, configAttrs, NULL, 0, &configCount) ||
!configCount)
error_exit("EGL failed to return any matching configurations\n");
if (!eglChooseConfig(eglDisplay, configAttrs, NULL, 0, &configCount) || !configCount)
error_exit("EGL failed to return any matching configurations\n");
configList = (EGLConfig *)malloc(configCount * sizeof(EGLConfig));
configList = (EGLConfig *)malloc(configCount * sizeof(EGLConfig));
if (!eglChooseConfig(eglDisplay, configAttrs, configList, configCount,
&configCount) ||
!configCount)
error_exit("EGL failed to populate configuration list\n");
if (!eglChooseConfig(eglDisplay, configAttrs, configList, configCount, &configCount) || !configCount)
error_exit("EGL failed to populate configuration list\n");
Window xRootWindow = DefaultRootWindow(display);
XSetWindowAttributes xCreateWindowAttributes;
xCreateWindowAttributes.event_mask = ExposureMask;
win = XCreateWindow(display, xRootWindow, 0, 0, width, height, 0,
CopyFromParent, InputOutput, CopyFromParent, CWEventMask,
&xCreateWindowAttributes);
XMapWindow(display, win);
Atom netWmStateAtom = XInternAtom(display, "_NET_WM_STATE", false);
XEvent xEvent;
memset(&xEvent, 0, sizeof(xEvent));
xEvent.type = ClientMessage;
xEvent.xclient.window = win;
xEvent.xclient.message_type = netWmStateAtom;
xEvent.xclient.format = 32;
xEvent.xclient.data.l[0] = 1;
xEvent.xclient.data.l[1] = false;
XSendEvent(display, xRootWindow, false, SubstructureNotifyMask, &xEvent);
Window xRootWindow = DefaultRootWindow(display);
XSetWindowAttributes xCreateWindowAttributes;
xCreateWindowAttributes.event_mask = ExposureMask;
win = XCreateWindow(display,
xRootWindow,
0,
0,
width,
height,
0,
CopyFromParent,
InputOutput,
CopyFromParent,
CWEventMask,
&xCreateWindowAttributes);
XMapWindow(display, win);
Atom netWmStateAtom = XInternAtom(display, "_NET_WM_STATE", false);
XEvent xEvent;
memset(&xEvent, 0, sizeof(xEvent));
xEvent.type = ClientMessage;
xEvent.xclient.window = win;
xEvent.xclient.message_type = netWmStateAtom;
xEvent.xclient.format = 32;
xEvent.xclient.data.l[0] = 1;
xEvent.xclient.data.l[1] = false;
XSendEvent(display, xRootWindow, false, SubstructureNotifyMask, &xEvent);
XStoreName(display, win, windowname);
XStoreName(display, win, windowname);
XSelectInput(display, win, ExposureMask | KeyPressMask | ButtonPressMask |
ButtonReleaseMask | KeyReleaseMask |
VisibilityChangeMask | PointerMotionMask);
XSelectInput(display,
win,
ExposureMask | KeyPressMask | ButtonPressMask | ButtonReleaseMask | KeyReleaseMask
| VisibilityChangeMask | PointerMotionMask);
EGLint windowAttrs[] = {EGL_NONE};
EGLint windowAttrs[] = {EGL_NONE};
eglSurface = eglCreateWindowSurface(eglDisplay, configList[0],
(EGLNativeWindowType)win, windowAttrs);
eglSurface = eglCreateWindowSurface(eglDisplay, configList[0], (EGLNativeWindowType)win, windowAttrs);
if (!eglSurface) error_exit("EGL couldn't create window\n");
if (!eglSurface)
error_exit("EGL couldn't create window\n");
eglBindAPI(EGL_OPENGL_ES_API);
eglBindAPI(EGL_OPENGL_ES_API);
eglContext = eglCreateContext(eglDisplay, configList[0], NULL, contextAttrs);
if (!eglContext) error_exit("EGL couldn't create context\n");
eglContext = eglCreateContext(eglDisplay, configList[0], NULL, contextAttrs);
if (!eglContext)
error_exit("EGL couldn't create context\n");
if (!eglMakeCurrent(eglDisplay, eglSurface, eglSurface, eglContext))
error_exit("EGL couldn't make context/surface current\n");
if (!eglMakeCurrent(eglDisplay, eglSurface, eglSurface, eglContext))
error_exit("EGL couldn't make context/surface current\n");
EGLint Context_RendererType;
eglQueryContext(eglDisplay, eglContext, EGL_CONTEXT_CLIENT_TYPE,
&Context_RendererType);
EGLint Context_RendererType;
eglQueryContext(eglDisplay, eglContext, EGL_CONTEXT_CLIENT_TYPE, &Context_RendererType);
switch (Context_RendererType) {
switch (Context_RendererType) {
case EGL_OPENGL_API:
printf("Using OpenGL API is not supported\n");
exit(EXIT_FAILURE);
break;
printf("Using OpenGL API is not supported\n");
exit(EXIT_FAILURE);
break;
case EGL_OPENGL_ES_API:
printf("Using OpenGL ES API");
break;
printf("Using OpenGL ES API");
break;
case EGL_OPENVG_API:
error_exit("Context Query Returned OpenVG. This is Unsupported\n");
error_exit("Context Query Returned OpenVG. This is Unsupported\n");
default:
error_exit("Unknown Context Type. %04X\n", Context_RendererType);
}
error_exit("Unknown Context Type. %04X\n", Context_RendererType);
}
return 1;
return 1;
}
void graphics_set_windowtitle(const char *windowname) {
XStoreName(display, win, windowname);
}
void graphics_set_windowtitle(const char *windowname) { XStoreName(display, win, windowname); }
void graphics_swap_buffers() { eglSwapBuffers(eglDisplay, eglSurface); }
void graphics_close_window() {
if (eglDisplay != EGL_NO_DISPLAY) {
eglMakeCurrent(eglDisplay, EGL_NO_SURFACE, EGL_NO_SURFACE, EGL_NO_CONTEXT);
void graphics_close_window()
{
if (eglDisplay != EGL_NO_DISPLAY) {
eglMakeCurrent(eglDisplay, EGL_NO_SURFACE, EGL_NO_SURFACE, EGL_NO_CONTEXT);
if (eglContext != EGL_NO_CONTEXT) eglDestroyContext(eglDisplay, eglContext);
if (eglContext != EGL_NO_CONTEXT)
eglDestroyContext(eglDisplay, eglContext);
if (eglSurface != EGL_NO_SURFACE) eglDestroySurface(eglDisplay, eglSurface);
if (eglSurface != EGL_NO_SURFACE)
eglDestroySurface(eglDisplay, eglSurface);
eglTerminate(eglDisplay);
}
eglTerminate(eglDisplay);
}
if (win) XDestroyWindow(display, win);
if (win)
XDestroyWindow(display, win);
XCloseDisplay(display);
XCloseDisplay(display);
}