def main(device, loops=1):
gpuGFLOPS = device.functions["gpuGFLOPS"]
cuFuncSetBlockShape(gpuGFLOPS, BLOCK_SIZE_G, 1, 1)
t0 = time()
for i in range(loops):
cuCtxSynchronize()
cuLaunchGrid(gpuGFLOPS, GRID_SIZE_G, 1)
cuCtxSynchronize()
t0 = time() - t0
flopsc = 4096. * ITERATIONS_C * BLOCK_SIZE_C
flopsg = 4096. * ITERATIONS_G * BLOCK_SIZE_G * GRID_SIZE_G
flopsc *= 1.e-9 * float(loops)
flopsg *= 1.e-9 * float(loops)
t1 = time()
for i in range(loops):
cpuGFLOPS()
t1 = time() - t1
# peakg = 4.*8.*2.*1.458 # 4MP*8SP/MP*2flops/SP/clock*clock[GHz] (8600GTS)
peakg = 14. * 8. * 2. * 1.512 # 14MP*8SP/MP*2flops/SP/clock*clock[GHz] (9800GT)
print "%8.3f%8.2f%8.3f%8.2f [%.2f]" % (t1, flopsc / t1, t0, flopsg / t0,
peakg)
def main(loops = 1):
blockDim = dim3(BLOCK_SIZE,1,1)
gridDim = dim3(GRID_SIZE,1,1)
t0 = time()
cudaThreadSynchronize()
for i in range(loops):
cudaConfigureCall(gridDim,blockDim,0,0)
gpuGFLOPS()
cudaThreadSynchronize()
t0 = time()-t0
cudaThreadExit()
flopsc = 4096.*ITERATIONS*BLOCK_SIZE
flopsg = flopsc*GRID_SIZE
flopsc *= 1.e-9*float(loops)
flopsg *= 1.e-9*float(loops)
t1 = time()
for i in range(loops):
cpuGFLOPS()
t1 = time()-t1
peakg = 4.*8.*2.*1.458 # 2MP*8SP/MP*2flops/SP/clock*clock[GHz]
print "%8.3f%8.2f%8.3f%8.2f [%.2f]" % (
t1,flopsc/t1,t0,flopsg/t0,peakg)
def main(device,loops = 1):
gpuGFLOPS = device.functions["gpuGFLOPS"]
cuFuncSetBlockShape(gpuGFLOPS,BLOCK_SIZE_G,1,1)
t0 = time()
for i in range(loops):
cuCtxSynchronize()
cuLaunchGrid(gpuGFLOPS,GRID_SIZE_G,1)
cuCtxSynchronize()
t0 = time()-t0
flopsc = 4096.*ITERATIONS_C*BLOCK_SIZE_C
flopsg = 4096.*ITERATIONS_G*BLOCK_SIZE_G*GRID_SIZE_G
flopsc *= 1.e-9*float(loops)
flopsg *= 1.e-9*float(loops)
t1 = time()
for i in range(loops):
cpuGFLOPS()
t1 = time()-t1
# peakg = 4.*8.*2.*1.458 # 4MP*8SP/MP*2flops/SP/clock*clock[GHz] (8600GTS)
peakg = 14.*8.*2.*1.512 # 14MP*8SP/MP*2flops/SP/clock*clock[GHz] (9800GT)
print "%8.3f%8.2f%8.3f%8.2f [%.2f]" % (
t1,flopsc/t1,t0,flopsg/t0,peakg)
def main(loops = 1):
blockDim = dim3(BLOCK_SIZE_G,1,1)
gridDim = dim3(GRID_SIZE_G,1,1)
t0 = time()
cudaThreadSynchronize()
for i in range(loops):
cudaConfigureCall(gridDim,blockDim,0,0)
gpuGFLOPS()
cudaThreadSynchronize()
t0 = time()-t0
cudaThreadExit()
flopsc = 4096.*ITERATIONS_C*BLOCK_SIZE_C
flopsg = 4096.*ITERATIONS_G*BLOCK_SIZE_G*GRID_SIZE_G
flopsc *= 1.e-9*float(loops)
flopsg *= 1.e-9*float(loops)
t1 = time()
for i in range(loops):
cpuGFLOPS()
t1 = time()-t1
# peakg = 4.*8.*2.*1.458 # 4MP*8SP/MP*2flops/SP/clock*clock[GHz] (8600GTS)
peakg = 14.*8.*2.*1.512 # 14MP*8SP/MP*2flops/SP/clock*clock[GHz] (9800GT)
print "%8.3f%8.2f%8.3f%8.2f [%.2f]" % (t1,flopsc/t1,t0,flopsg/t0,peakg)
print "%8.3f%8.2f" % (flopsc/t1*2.8,flopsg/t0*1.512/112)
def main(loops=1):
blockDim = dim3(BLOCK_SIZE_G, 1, 1)
gridDim = dim3(GRID_SIZE_G, 1, 1)
t0 = time()
cudaThreadSynchronize()
for i in range(loops):
cudaConfigureCall(gridDim, blockDim, 0, 0)
gpuGFLOPS()
cudaThreadSynchronize()
t0 = time() - t0
cudaThreadExit()
flopsc = 4096. * ITERATIONS_C * BLOCK_SIZE_C
flopsg = 4096. * ITERATIONS_G * BLOCK_SIZE_G * GRID_SIZE_G
flopsc *= 1.e-9 * float(loops)
flopsg *= 1.e-9 * float(loops)
t1 = time()
for i in range(loops):
cpuGFLOPS()
t1 = time() - t1
# peakg = 4.*8.*2.*1.458 # 4MP*8SP/MP*2flops/SP/clock*clock[GHz] (8600GTS)
peakg = 14. * 8. * 2. * 1.512 # 14MP*8SP/MP*2flops/SP/clock*clock[GHz] (9800GT)
print "%8.3f%8.2f%8.3f%8.2f [%.2f]" % (t1, flopsc / t1, t0, flopsg / t0,
peakg)
print "%8.3f%8.2f" % (flopsc / t1 * 2.8, flopsg / t0 * 1.512 / 112)
