def TestSynIterInc():
SIZE = 64
# build and run the kernel
prgm = env.Program()
code = prgm.get_stream()
code.add(ptx.dcl_output(reg.o0, USAGE=ptx.usage.pos))
ones = prgm.acquire_register((1, 1, 1, 1))
counter = prgm.acquire_register()
code.add(ptx.mov(counter, ones))
for i in syn_iter(code, 4, step=1, mode=INC):
code.add(ptx.iadd(counter, counter, ones))
code.add(ptx.mov(reg.o0, counter.x))
domain = (0, 0, SIZE, SIZE)
proc = env.Processor(0)
ext_output=proc.alloc_remote('i', 1, SIZE)
prgm.set_binding(reg.o0, ext_output)
prgm.add(code)
proc.execute(prgm, domain)
passed = True
for i in xrange(0, SIZE):
if ext_output[i] != 5:
passed = False
print "Passed == ", passed
proc.free(ext_output)
return
def TestSynIterInc():
SIZE = 64
# build and run the kernel
prgm = env.Program()
code = prgm.get_stream()
code.add(ptx.dcl_output(reg.o0, USAGE=ptx.usage.pos))
ones = prgm.acquire_register((1, 1, 1, 1))
counter = prgm.acquire_register()
code.add(ptx.mov(counter, ones))
for i in syn_iter(code, 4, step=1, mode=INC):
code.add(ptx.iadd(counter, counter, ones))
code.add(ptx.mov(reg.o0, counter.x))
domain = (0, 0, SIZE, SIZE)
proc = env.Processor(0)
ext_output = proc.alloc_remote('i', 1, SIZE)
prgm.set_binding(reg.o0, ext_output)
prgm.add(code)
proc.execute(prgm, domain)
passed = True
for i in xrange(0, SIZE):
if ext_output[i] != 5:
passed = False
print "Passed == ", passed
proc.free(ext_output)
return
def TestSimpleKernelNPy():
import corepy.arch.cal.isa as isa
SIZE = 128
proc = Processor(0)
arr_input = proc.alloc_remote_npy('f', 4, SIZE, SIZE)
arr_output = proc.alloc_remote_npy('f', 4, SIZE, SIZE)
#for i in xrange(0, SIZE * SIZE * 4):
# arr_input[i] = float(i + 1)
# arr_output[i] = 0.0
#print arr_input.shape
#print arr_output.shape
#print type(arr_input.data)
val = 0.0
for i in xrange(0, SIZE):
for j in xrange(0, SIZE):
for k in xrange(0, 4):
arr_input[i][j][k] = val
arr_output[i][j][k] = 0.0
val += 1.0
# build and run the kernel
prgm = Program()
code = prgm.get_stream()
#code.add(isa.dcl_input('v0', USAGE=isa.usage.pos, INTERP='linear_noperspective'))
code.add("dcl_input_position_interp(constant) v0.xy__")
code.add(isa.dcl_output('o0', USAGE=isa.usage.generic))
code.add(isa.dcl_resource(0, '2d', isa.fmt.float, UNNORM=True))
code.add(isa.sample(0, 0, 'o0', 'v0.xy'))
#code.add(isa.load(0, 'o0', 'v0.g'))
domain = (0, 0, SIZE, SIZE)
prgm.set_binding("o0", arr_output)
prgm.set_binding("i0", arr_input)
prgm.add(code)
prgm.cache_code()
prgm.print_code()
proc.execute(prgm, domain)
# Check the output
val = 0.0
for i in xrange(0, SIZE):
for j in xrange(0, SIZE):
for k in xrange(0, 4):
if arr_output[i][j][k] != val:
print "ERROR index %d is %f, should be %f" % (i, arr_output[i], val)
val += 1.0
return
def TestSimpleKernelNPy():
import corepy.arch.cal.isa as isa
SIZE = 128
proc = Processor(0)
arr_input = proc.alloc_remote_npy("f", 4, SIZE, SIZE)
arr_output = proc.alloc_remote_npy("f", 4, SIZE, SIZE)
# for i in xrange(0, SIZE * SIZE * 4):
# arr_input[i] = float(i + 1)
# arr_output[i] = 0.0
# print arr_input.shape
# print arr_output.shape
# print type(arr_input.data)
val = 0.0
for i in xrange(0, SIZE):
for j in xrange(0, SIZE):
for k in xrange(0, 4):
arr_input[i][j][k] = val
arr_output[i][j][k] = 0.0
val += 1.0
# build and run the kernel
prgm = Program()
code = prgm.get_stream()
# code.add(isa.dcl_input('v0', USAGE=isa.usage.pos, INTERP='linear_noperspective'))
code.add("dcl_input_position_interp(constant) v0.xy__")
code.add(isa.dcl_output("o0", USAGE=isa.usage.generic))
code.add(isa.dcl_resource(0, "2d", isa.fmt.float, UNNORM=True))
code.add(isa.sample(0, 0, "o0", "v0.xy"))
# code.add(isa.load(0, 'o0', 'v0.g'))
domain = (0, 0, SIZE, SIZE)
prgm.set_binding("o0", arr_output)
prgm.set_binding("i0", arr_input)
prgm.add(code)
prgm.cache_code()
prgm.print_code()
proc.execute(prgm, domain)
# Check the output
val = 0.0
for i in xrange(0, SIZE):
for j in xrange(0, SIZE):
for k in xrange(0, 4):
if arr_output[i][j][k] != val:
print "ERROR index %d is %f, should be %f" % (i, arr_output[i], val)
val += 1.0
return
def TestSynIterIncFloatExtStopExtStart():
SIZE = 64
# build and run the kernel
prgm = env.Program()
code = prgm.get_stream()
code.add(ptx.dcl_output(reg.o0, USAGE=ptx.usage.pos))
ones = prgm.acquire_register((1, 1, 1, 1))
counter = prgm.acquire_register()
code.add(ptx.mov(counter, ones))
stop = prgm.acquire_register((4.0, 4.0, 4.0, 4.0))
start = prgm.acquire_register((2.0, 2.0, 2.0, 2.0))
step = prgm.acquire_register((1.0, 1.0, 1.0, 1.0))
fiter = syn_iter_float(code, stop, step=step, mode=INC)
fiter.set_start_reg(start)
for i in fiter:
code.add(ptx.iadd(counter, counter, ones))
code.add(ptx.mov(reg.o0, counter.x))
domain = (0, 0, SIZE, SIZE)
proc = env.Processor(0)
ext_output=proc.alloc_remote('i', 1, SIZE, 1)
prgm.set_binding(reg.o0, ext_output)
prgm.add(code)
proc.execute(prgm, domain)
passed = True
for i in xrange(0, SIZE):
if ext_output[i] != 3:
passed = False
print "Passed == ", passed
proc.free(ext_output)
return
def TestSynIterIncFloatExtStopExtStart():
SIZE = 64
# build and run the kernel
prgm = env.Program()
code = prgm.get_stream()
code.add(ptx.dcl_output(reg.o0, USAGE=ptx.usage.pos))
ones = prgm.acquire_register((1, 1, 1, 1))
counter = prgm.acquire_register()
code.add(ptx.mov(counter, ones))
stop = prgm.acquire_register((4.0, 4.0, 4.0, 4.0))
start = prgm.acquire_register((2.0, 2.0, 2.0, 2.0))
step = prgm.acquire_register((1.0, 1.0, 1.0, 1.0))
fiter = syn_iter_float(code, stop, step=step, mode=INC)
fiter.set_start_reg(start)
for i in fiter:
code.add(ptx.iadd(counter, counter, ones))
code.add(ptx.mov(reg.o0, counter.x))
domain = (0, 0, SIZE, SIZE)
proc = env.Processor(0)
ext_output = proc.alloc_remote('i', 1, SIZE, 1)
prgm.set_binding(reg.o0, ext_output)
prgm.add(code)
proc.execute(prgm, domain)
passed = True
for i in xrange(0, SIZE):
if ext_output[i] != 3:
passed = False
print "Passed == ", passed
proc.free(ext_output)
return