smyalygames/FiniteVolumeGPU
1
0
Fork 0
mirror of https://github.com/smyalygames/FiniteVolumeGPU.git synced 2026-01-14 15:48:43 +01:00
Code Issues Actions Packages Projects Releases Wiki Activity

Refactoring

Browse Source
This commit is contained in:
André R. Brodtkorb
2018-11-01 21:34:53 +01:00
parent d9eb72d78c
commit 0671bd747a
6 changed files with 178 additions and 212 deletions
Download Patch File Download Diff File Expand all files Collapse all files

55
GPUSimulators/cuda/SWE2D_FORCE.cu
View File

@@ -28,27 +28,18 @@ along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
__device__
void computeFluxF(float Q[3][BLOCK_HEIGHT+2][BLOCK_WIDTH+2],
float F[3][BLOCK_HEIGHT+1][BLOCK_WIDTH+1],
float F[3][BLOCK_HEIGHT+2][BLOCK_WIDTH+2],
const float g_, const float dx_, const float dt_) {
//Index of thread within block
const int tx = threadIdx.x;
const int ty = threadIdx.y;
//Compute fluxes along the x axis
{
int j=ty;
const int l = j + 1; //Skip ghost cells
for (int i=tx; i<BLOCK_WIDTH+1; i+=BLOCK_WIDTH) {
const int k = i;
for (int j=threadIdx.y; j<BLOCK_HEIGHT+2; j+=BLOCK_HEIGHT) {
for (int i=threadIdx.x; i<BLOCK_WIDTH+1; i+=BLOCK_WIDTH) {
// Q at interface from the right and left
const float3 Qp = make_float3(Q[0][l][k+1],
Q[1][l][k+1],
Q[2][l][k+1]);
const float3 Qm = make_float3(Q[0][l][k],
Q[1][l][k],
Q[2][l][k]);
const float3 Qp = make_float3(Q[0][j][i+1],
Q[1][j][i+1],
Q[2][j][i+1]);
const float3 Qm = make_float3(Q[0][j][i],
Q[1][j][i],
Q[2][j][i]);
// Computed flux
const float3 flux = FORCE_1D_flux(Qm, Qp, g_, dx_, dt_);
@@ -65,27 +56,19 @@ void computeFluxF(float Q[3][BLOCK_HEIGHT+2][BLOCK_WIDTH+2],
*/
__device__
void computeFluxG(float Q[3][BLOCK_HEIGHT+2][BLOCK_WIDTH+2],
float G[3][BLOCK_HEIGHT+1][BLOCK_WIDTH+1],
float G[3][BLOCK_HEIGHT+2][BLOCK_WIDTH+2],
const float g_, const float dy_, const float dt_) {
//Index of thread within block
const int tx = threadIdx.x;
const int ty = threadIdx.y;
//Compute fluxes along the y axis
for (int j=ty; j<BLOCK_HEIGHT+1; j+=BLOCK_HEIGHT) {
const int l = j;
{
int i=tx;
const int k = i + 1; //Skip ghost cells
for (int j=threadIdx.y; j<BLOCK_HEIGHT+1; j+=BLOCK_HEIGHT) {
for (int i=threadIdx.x; i<BLOCK_WIDTH+2; i+=BLOCK_WIDTH) {
// Q at interface from the right and left
// Note that we swap hu and hv
const float3 Qp = make_float3(Q[0][l+1][k],
Q[2][l+1][k],
Q[1][l+1][k]);
const float3 Qm = make_float3(Q[0][l][k],
Q[2][l][k],
Q[1][l][k]);
const float3 Qp = make_float3(Q[0][j+1][i],
Q[2][j+1][i],
Q[1][j+1][i]);
const float3 Qm = make_float3(Q[0][j][i],
Q[2][j][i],
Q[1][j][i]);
// Computed flux
// Note that we swap back
@@ -120,7 +103,7 @@ __global__ void FORCEKernel(
const unsigned int vars = 3;
__shared__ float Q[3][h+2][w+2];
__shared__ float F[3][h+1][w+1];
__shared__ float F[3][h+2][w+2];
//Read into shared memory
readBlock<w, h, gc>( h0_ptr_, h0_pitch_, Q[0], nx_+2, ny_+2);