Fixed WAF

This commit is contained in:
André R. Brodtkorb 2018-07-25 16:39:50 +02:00
parent d94daeae7e
commit a0f429148c
2 changed files with 74 additions and 66 deletions

View File

@ -22,7 +22,11 @@ along with this program. If not, see <http://www.gnu.org/licenses/>.
#Import packages we need #Import packages we need
import numpy as np import numpy as np
import pyopencl as cl #OpenCL in Python
import pycuda.compiler as cuda_compiler
import pycuda.gpuarray
import pycuda.driver as cuda
from SWESimulators import Common from SWESimulators import Common
@ -47,24 +51,24 @@ class WAF:
g: Gravitational accelleration (9.81 m/s^2) g: Gravitational accelleration (9.81 m/s^2)
""" """
def __init__(self, \ def __init__(self, \
cl_ctx, \ context, \
h0, hu0, hv0, \ h0, hu0, hv0, \
nx, ny, \ nx, ny, \
dx, dy, dt, \ dx, dy, dt, \
g, \ g, \
block_width=16, block_height=16): block_width=16, block_height=16):
self.cl_ctx = cl_ctx #Create a CUDA stream
self.stream = cuda.Stream()
#Create an OpenCL command queue
self.cl_queue = cl.CommandQueue(self.cl_ctx)
#Get kernels #Get kernels
self.kernel = Common.get_kernel(self.cl_ctx, "WAF_kernel.opencl", block_width, block_height) self.waf_module = context.get_kernel("WAF_kernel.cu", block_width, block_height)
self.waf_kernel = self.waf_module.get_function("WAFKernel")
self.waf_kernel.prepare("iiffffiPiPiPiPiPiPi")
#Create data by uploading to device #Create data by uploading to device
ghost_cells_x = 2 ghost_cells_x = 2
ghost_cells_y = 2 ghost_cells_y = 2
self.cl_data = Common.SWEDataArkawaA(self.cl_ctx, nx, ny, ghost_cells_x, ghost_cells_y, h0, hu0, hv0) self.data = Common.SWEDataArakawaA(self.stream, nx, ny, ghost_cells_x, ghost_cells_y, h0, hu0, hv0)
#Save input parameters #Save input parameters
#Notice that we need to specify them in the correct dataformat for the #Notice that we need to specify them in the correct dataformat for the
@ -80,14 +84,16 @@ class WAF:
self.t = np.float32(0.0) self.t = np.float32(0.0)
#Compute kernel launch parameters #Compute kernel launch parameters
self.local_size = (block_width, block_height) self.local_size = (block_width, block_height, 1)
self.global_size = ( \ self.global_size = ( \
int(np.ceil(self.nx / float(self.local_size[0])) * self.local_size[0]), \ int(np.ceil(self.nx / float(self.local_size[0]))), \
int(np.ceil(self.ny / float(self.local_size[1])) * self.local_size[1]) \ int(np.ceil(self.ny / float(self.local_size[1]))) \
) )
def __str__(self):
return "Weighted average flux"
""" """
Function which steps n timesteps Function which steps n timesteps
@ -104,32 +110,30 @@ class WAF:
break break
#Along X, then Y #Along X, then Y
self.kernel.swe_2D(self.cl_queue, self.global_size, self.local_size, \ self.waf_kernel.prepared_async_call(self.global_size, self.local_size, self.stream, \
self.nx, self.ny, \ self.nx, self.ny, \
self.dx, self.dy, local_dt, \ self.dx, self.dy, local_dt, \
self.g, \ self.g, \
np.int32(0), \ np.int32(0), \
self.cl_data.h0.data, self.cl_data.h0.pitch, \ self.data.h0.data.gpudata, self.data.h0.pitch, \
self.cl_data.hu0.data, self.cl_data.hu0.pitch, \ self.data.hu0.data.gpudata, self.data.hu0.pitch, \
self.cl_data.hv0.data, self.cl_data.hv0.pitch, \ self.data.hv0.data.gpudata, self.data.hv0.pitch, \
self.cl_data.h1.data, self.cl_data.h1.pitch, \ self.data.h1.data.gpudata, self.data.h1.pitch, \
self.cl_data.hu1.data, self.cl_data.hu1.pitch, \ self.data.hu1.data.gpudata, self.data.hu1.pitch, \
self.cl_data.hv1.data, self.cl_data.hv1.pitch) self.data.hv1.data.gpudata, self.data.hv1.pitch)
self.cl_data.swap()
#Along Y, then X #Along Y, then X
self.kernel.swe_2D(self.cl_queue, self.global_size, self.local_size, \ self.waf_kernel.prepared_async_call(self.global_size, self.local_size, self.stream, \
self.nx, self.ny, \ self.nx, self.ny, \
self.dx, self.dy, local_dt, \ self.dx, self.dy, local_dt, \
self.g, \ self.g, \
np.int32(1), \ np.int32(1), \
self.cl_data.h0.data, self.cl_data.h0.pitch, \ self.data.h1.data.gpudata, self.data.h1.pitch, \
self.cl_data.hu0.data, self.cl_data.hu0.pitch, \ self.data.hu1.data.gpudata, self.data.hu1.pitch, \
self.cl_data.hv0.data, self.cl_data.hv0.pitch, \ self.data.hv1.data.gpudata, self.data.hv1.pitch, \
self.cl_data.h1.data, self.cl_data.h1.pitch, \ self.data.h0.data.gpudata, self.data.h0.pitch, \
self.cl_data.hu1.data, self.cl_data.hu1.pitch, \ self.data.hu0.data.gpudata, self.data.hu0.pitch, \
self.cl_data.hv1.data, self.cl_data.hv1.pitch) self.data.hv0.data.gpudata, self.data.hv0.pitch)
self.cl_data.swap()
self.t += local_dt self.t += local_dt
@ -140,5 +144,5 @@ class WAF:
def download(self): def download(self):
return self.cl_data.download(self.cl_queue) return self.data.download(self.stream)

View File

@ -24,30 +24,32 @@ along with this program. If not, see <http://www.gnu.org/licenses/>.
#include "common.opencl" #include "common.cu"
/** /**
* Computes the flux along the x axis for all faces * Computes the flux along the x axis for all faces
*/ */
void computeFluxF(__local float Q[3][block_height+4][block_width+4], __device__
__local float F[3][block_height+1][block_width+1], void computeFluxF(float Q[3][block_height+4][block_width+4],
float F[3][block_height+1][block_width+1],
const float g_, const float dx_, const float dt_) { const float g_, const float dx_, const float dt_) {
//Index of thread within block //Index of thread within block
const int tx = get_local_id(0); const int tx = get_local_id(0);
const int ty = get_local_id(1); const int ty = get_local_id(1);
for (int j=ty; j<block_height; j+=get_local_size(1)) { {
int j=ty;
const int l = j + 2; //Skip ghost cells const int l = j + 2; //Skip ghost cells
for (int i=tx; i<block_width+1; i+=get_local_size(0)) { for (int i=tx; i<block_width+1; i+=block_width) {
const int k = i + 1; const int k = i + 1;
// Q at interface from the right and left // Q at interface from the right and left
const float3 Ql2 = (float3)(Q[0][l][k-1], Q[1][l][k-1], Q[2][l][k-1]); const float3 Ql2 = make_float3(Q[0][l][k-1], Q[1][l][k-1], Q[2][l][k-1]);
const float3 Ql1 = (float3)(Q[0][l][k ], Q[1][l][k ], Q[2][l][k ]); const float3 Ql1 = make_float3(Q[0][l][k ], Q[1][l][k ], Q[2][l][k ]);
const float3 Qr1 = (float3)(Q[0][l][k+1], Q[1][l][k+1], Q[2][l][k+1]); const float3 Qr1 = make_float3(Q[0][l][k+1], Q[1][l][k+1], Q[2][l][k+1]);
const float3 Qr2 = (float3)(Q[0][l][k+2], Q[1][l][k+2], Q[2][l][k+2]); const float3 Qr2 = make_float3(Q[0][l][k+2], Q[1][l][k+2], Q[2][l][k+2]);
// Computed flux // Computed flux
const float3 flux = WAF_1D_flux(Ql2, Ql1, Qr1, Qr2, g_, dx_, dt_); const float3 flux = WAF_1D_flux(Ql2, Ql1, Qr1, Qr2, g_, dx_, dt_);
@ -68,24 +70,26 @@ void computeFluxF(__local float Q[3][block_height+4][block_width+4],
/** /**
* Computes the flux along the y axis for all faces * Computes the flux along the y axis for all faces
*/ */
void computeFluxG(__local float Q[3][block_height+4][block_width+4], __device__
__local float G[3][block_height+1][block_width+1], void computeFluxG(float Q[3][block_height+4][block_width+4],
float G[3][block_height+1][block_width+1],
const float g_, const float dy_, const float dt_) { const float g_, const float dy_, const float dt_) {
//Index of thread within block //Index of thread within block
const int tx = get_local_id(0); const int tx = get_local_id(0);
const int ty = get_local_id(1); const int ty = get_local_id(1);
//Compute fluxes along the y axis //Compute fluxes along the y axis
for (int j=ty; j<block_height+1; j+=get_local_size(1)) { for (int j=ty; j<block_height+1; j+=block_height) {
const int l = j + 1; const int l = j + 1;
for (int i=tx; i<block_width; i+=get_local_size(0)) { {
int i=tx;
const int k = i + 2; //Skip ghost cells const int k = i + 2; //Skip ghost cells
// Q at interface from the right and left // Q at interface from the right and left
// Note that we swap hu and hv // Note that we swap hu and hv
const float3 Ql2 = (float3)(Q[0][l-1][k], Q[2][l-1][k], Q[1][l-1][k]); const float3 Ql2 = make_float3(Q[0][l-1][k], Q[2][l-1][k], Q[1][l-1][k]);
const float3 Ql1 = (float3)(Q[0][l ][k], Q[2][l ][k], Q[1][l ][k]); const float3 Ql1 = make_float3(Q[0][l ][k], Q[2][l ][k], Q[1][l ][k]);
const float3 Qr1 = (float3)(Q[0][l+1][k], Q[2][l+1][k], Q[1][l+1][k]); const float3 Qr1 = make_float3(Q[0][l+1][k], Q[2][l+1][k], Q[1][l+1][k]);
const float3 Qr2 = (float3)(Q[0][l+2][k], Q[2][l+2][k], Q[1][l+2][k]); const float3 Qr2 = make_float3(Q[0][l+2][k], Q[2][l+2][k], Q[1][l+2][k]);
// Computed flux // Computed flux
// Note that we swap back // Note that we swap back
@ -110,23 +114,23 @@ void computeFluxG(__local float Q[3][block_height+4][block_width+4],
__kernel void swe_2D( __global__ void WAFKernel(
int nx_, int ny_, int nx_, int ny_,
float dx_, float dy_, float dt_, float dx_, float dy_, float dt_,
float g_, int step_, float g_, int step_,
//Input h^n //Input h^n
__global float* h0_ptr_, int h0_pitch_, float* h0_ptr_, int h0_pitch_,
__global float* hu0_ptr_, int hu0_pitch_, float* hu0_ptr_, int hu0_pitch_,
__global float* hv0_ptr_, int hv0_pitch_, float* hv0_ptr_, int hv0_pitch_,
//Output h^{n+1} //Output h^{n+1}
__global float* h1_ptr_, int h1_pitch_, float* h1_ptr_, int h1_pitch_,
__global float* hu1_ptr_, int hu1_pitch_, float* hu1_ptr_, int hu1_pitch_,
__global float* hv1_ptr_, int hv1_pitch_) { float* hv1_ptr_, int hv1_pitch_) {
//Shared memory variables //Shared memory variables
__local float Q[3][block_height+4][block_width+4]; __shared__ float Q[3][block_height+4][block_width+4];
__local float F[3][block_height+1][block_width+1]; __shared__ float F[3][block_height+1][block_width+1];
@ -135,12 +139,12 @@ __kernel void swe_2D(
hu0_ptr_, hu0_pitch_, hu0_ptr_, hu0_pitch_,
hv0_ptr_, hv0_pitch_, hv0_ptr_, hv0_pitch_,
Q, nx_, ny_); Q, nx_, ny_);
barrier(CLK_LOCAL_MEM_FENCE); __syncthreads();
//Set boundary conditions //Set boundary conditions
noFlowBoundary2(Q, nx_, ny_); noFlowBoundary2(Q, nx_, ny_);
barrier(CLK_LOCAL_MEM_FENCE); __syncthreads();
@ -148,37 +152,37 @@ __kernel void swe_2D(
if (step_ == 0) { if (step_ == 0) {
//Compute fluxes along the x axis and evolve //Compute fluxes along the x axis and evolve
computeFluxF(Q, F, g_, dx_, dt_); computeFluxF(Q, F, g_, dx_, dt_);
barrier(CLK_LOCAL_MEM_FENCE); __syncthreads();
evolveF2(Q, F, nx_, ny_, dx_, dt_); evolveF2(Q, F, nx_, ny_, dx_, dt_);
barrier(CLK_LOCAL_MEM_FENCE); __syncthreads();
//Fix boundary conditions //Fix boundary conditions
noFlowBoundary2(Q, nx_, ny_); noFlowBoundary2(Q, nx_, ny_);
barrier(CLK_LOCAL_MEM_FENCE); __syncthreads();
//Compute fluxes along the y axis and evolve //Compute fluxes along the y axis and evolve
computeFluxG(Q, F, g_, dy_, dt_); computeFluxG(Q, F, g_, dy_, dt_);
barrier(CLK_LOCAL_MEM_FENCE); __syncthreads();
evolveG2(Q, F, nx_, ny_, dy_, dt_); evolveG2(Q, F, nx_, ny_, dy_, dt_);
barrier(CLK_LOCAL_MEM_FENCE); __syncthreads();
} }
//Step 1 => evolve y first, then x //Step 1 => evolve y first, then x
else { else {
//Compute fluxes along the y axis and evolve //Compute fluxes along the y axis and evolve
computeFluxG(Q, F, g_, dy_, dt_); computeFluxG(Q, F, g_, dy_, dt_);
barrier(CLK_LOCAL_MEM_FENCE); __syncthreads();
evolveG2(Q, F, nx_, ny_, dy_, dt_); evolveG2(Q, F, nx_, ny_, dy_, dt_);
barrier(CLK_LOCAL_MEM_FENCE); __syncthreads();
//Fix boundary conditions //Fix boundary conditions
noFlowBoundary2(Q, nx_, ny_); noFlowBoundary2(Q, nx_, ny_);
barrier(CLK_LOCAL_MEM_FENCE); __syncthreads();
//Compute fluxes along the x axis and evolve //Compute fluxes along the x axis and evolve
computeFluxF(Q, F, g_, dx_, dt_); computeFluxF(Q, F, g_, dx_, dt_);
barrier(CLK_LOCAL_MEM_FENCE); __syncthreads();
evolveF2(Q, F, nx_, ny_, dx_, dt_); evolveF2(Q, F, nx_, ny_, dx_, dt_);
barrier(CLK_LOCAL_MEM_FENCE); __syncthreads();
} }