tinygrad/test/test_uops_stats.py

225 lines
8.1 KiB
Python

import unittest
from tinygrad import Tensor
from tinygrad.helpers import getenv, GlobalCounters
from tinygrad.engine.schedule import create_schedule
from tinygrad.engine.realize import lower_schedule_item
from tinygrad.codegen.linearize import linearize_uop
from tinygrad.ops import BinaryOps, TernaryOps, flops_mem, UOps, UOp
from tinygrad.dtype import dtypes
from tinygrad.codegen.kernel import Kernel, Opt, OptOps, KernelOptError
# **************** new FlopCounter ****************
def get_stats(x:Tensor):
si = create_schedule([x.lazydata])[-1]
ei = lower_schedule_item(si)
return ei.prg.op_estimate, ei.prg.mem_estimate
class TestMemoryCount(unittest.TestCase):
def test_add(self):
a = Tensor.empty(1024, 1024, dtype=dtypes.uint8)
b = Tensor.empty(1024, 1024, dtype=dtypes.uint8)
_, mem = get_stats(a+b)
self.assertEqual(mem, 1024*1024*3) # 2 reads + 1 write
def test_add_const(self):
a = Tensor.empty(1024, 1024, dtype=dtypes.uint8)
_, mem = get_stats(a+3)
self.assertEqual(mem, 1024*1024*2) # 1 read + 1 write
def test_add_slice(self):
a = Tensor.empty(1024, 1024, dtype=dtypes.uint8)[:512]
_, mem = get_stats(a+3)
self.assertEqual(mem, 512*1024*2) # 1 read + 1 write
def test_expanded(self):
a = Tensor.empty(1024, 1, dtype=dtypes.uint8).expand(1024, 1024)
b = Tensor.empty(1024, 1024, dtype=dtypes.uint8)
_, mem = get_stats(a+b)
self.assertEqual(mem, 1024*1024*2 + 1024) # 1 full read + 1 lil read + 1 write
def test_both_expanded(self):
# TODO: this probably should be a full write
a = Tensor.empty(1024, 1, dtype=dtypes.uint8).expand(1024, 1024)
b = Tensor.empty(1024, 1, dtype=dtypes.uint8).expand(1024, 1024)
_, mem = get_stats(a+b)
self.assertEqual(mem, 1024*1024 + 2*1024) # 2 lil reads + 1 write
def test_self_add(self):
a = Tensor.empty(1024, 1024, dtype=dtypes.uint8)
_, mem = get_stats(a+a)
self.assertEqual(mem, 1024*1024*2) # 1 read + 1 write
def test_self_add_transposed(self):
a = Tensor.empty(1024, 1024, dtype=dtypes.uint8)
_, mem = get_stats(a+a.T)
self.assertEqual(mem, 1024*1024*2) # 1 read + 1 write
def test_self_add_assign(self):
a = Tensor.empty(1024, 1024, dtype=dtypes.uint8).realize()
_, mem = get_stats(a.assign(a+a))
self.assertEqual(mem, 1024*1024*2) # 1 read + 1 write
# NOTE: this still isn't testing unroll using the acc
@unittest.skipUnless(getenv("PYTHON"), "only run test on emulated tensor cores")
class TestUOpsStatsMatmulHalf(unittest.TestCase):
def test_simple_matmul_half(self, N=16):
GlobalCounters.reset()
a, b = Tensor.empty(N, N, dtype=dtypes.half), Tensor.empty(N, N, dtype=dtypes.half)
c = a.matmul(b)
c.realize()
expected_ops = N ** 3 * 2
self.assertEqual(expected_ops, GlobalCounters.global_ops)
def test_bigger_matmul_half(self): self.test_simple_matmul_half(64)
def test_batched_matmul_half(self, N=16):
GlobalCounters.reset()
a, b = Tensor.empty(4, N, N, dtype=dtypes.half), Tensor.empty(1, N, N, dtype=dtypes.half)
c = a.matmul(b)
c.realize()
expected_ops = 4 * N ** 3 * 2
self.assertEqual(expected_ops, GlobalCounters.global_ops)
class TestUOpsStats(unittest.TestCase):
@unittest.skipIf(getenv("PTX"), "wrong in PTX")
def test_simple_add(self):
a = Tensor.empty(100,100)
b = Tensor.empty(100,100)
c = a+b
ops, mem = get_stats(c)
expected_ops = c.numel()
expected_mem = a.nbytes() + b.nbytes() + c.nbytes()
self.assertEqual(mem, expected_mem)
# NOTE; ops also include indexing ops
assert expected_ops <= ops and ops <= expected_ops * 2
@unittest.skipIf(getenv("PTX"), "wrong in PTX")
def test_simple_add_sq(self):
a = Tensor.empty(100,100)
b = Tensor.empty(100,100)
c = (a+b)*(a+b)
ops, mem = get_stats(c)
expected_ops = c.numel()*2
expected_mem = a.nbytes() + b.nbytes() + c.nbytes()
self.assertEqual(mem, expected_mem)
# NOTE; ops also include indexing ops
assert expected_ops <= ops and ops <= expected_ops * 2
def test_simple_matmul(self):
a = Tensor.empty(1024,1024)
b = Tensor.empty(1024,1024)
c = a@b
ops, mem = get_stats(c)
expected_ops = c.numel() * 1024 * 2
required_mem = a.nbytes() + b.nbytes() + c.nbytes()
assert expected_ops <= ops and ops <= expected_ops * 1.2
# NOTE: it's hard to assert on the memory here, all depends on caching
assert required_mem <= mem
#MULACC should have the same stats as MUL + ADD
def test_mulacc(self):
globl = UOp(UOps.DEFINE_GLOBAL, dtypes.int.ptr(), tuple())
o1 = UOp(UOps.CONST, dtypes.int, tuple(), 1)
o2 = UOp(UOps.CONST, dtypes.int, tuple(), 2)
u1 = UOp(UOps.LOAD, dtypes.int, (globl, o1))
u2 = UOp(UOps.LOAD, dtypes.int, (globl, o2))
u3 = UOp(UOps.CONST, dtypes.int, tuple(), 3)
u4 = UOp(UOps.ALU, dtypes.int, (u1,u2), BinaryOps.MUL)
u5 = UOp(UOps.ALU, dtypes.int, (u4,u3), BinaryOps.ADD)
uops = linearize_uop(u5.sink())
globl = UOp(UOps.DEFINE_GLOBAL, dtypes.int.ptr(), tuple())
o1 = UOp(UOps.CONST, dtypes.int, tuple(), 1)
o2 = UOp(UOps.CONST, dtypes.int, tuple(), 2)
u1 = UOp(UOps.LOAD, dtypes.int, (globl, o1))
u2 = UOp(UOps.LOAD, dtypes.int, (globl, o2))
u3 = UOp(UOps.CONST, dtypes.int, tuple(), 3)
u4 = UOp(UOps.ALU, dtypes.int, (u1,u2,u3), TernaryOps.MULACC)
uops_fma = linearize_uop(u4.sink())
self.assertEqual(flops_mem(uops), flops_mem(uops_fma))
N = 100
@unittest.skipIf(getenv("PTX"), "wrong in PTX") # maybe?
class TestStatsOptimized(unittest.TestCase):
@classmethod
def setUpClass(cls):
cls.ast_gemm = (Tensor.empty(N, N) @ Tensor.empty(N, N)).schedule()[-1].ast
cls.ast_reduce = (Tensor.empty(N*N).sum()).schedule()[-1].ast
def check_gemm(self, p, extra_flops=0):
#p.uops.print()
#print(p.src)
print(p.name, p.op_estimate, p.mem_estimate, p.lds_estimate)
self.assertEqual(p.op_estimate, 2*N*N*N + extra_flops) # N**3 mulaccs
self.assertEqual(p.mem_estimate, 3*N*N*4) # 3 NxN mats with floats
def test_gemm(self):
p = Kernel(self.ast_gemm).to_program()
self.check_gemm(p)
self.assertEqual(p.lds_estimate, 2*N*N*N*4 + 4*N*N)
# this is a good lesson about why UPCASTing is a good idea
def test_gemm_one_upcasted(self):
k = Kernel(self.ast_gemm)
k.apply_opt(Opt(OptOps.UPCAST, 0, 4))
p = k.to_program()
self.check_gemm(p)
self.assertEqual(p.lds_estimate, N*N*N*4 + N*N*N*4//4 + 4*N*N)
def test_gemm_upcasted(self):
k = Kernel(self.ast_gemm)
k.apply_opt(Opt(OptOps.UPCAST, 0, 4))
k.apply_opt(Opt(OptOps.UPCAST, 1, 4))
k.apply_opt(Opt(OptOps.UNROLL, 0, 4))
p = k.to_program()
self.check_gemm(p)
self.assertEqual(p.lds_estimate, 2*N*N*N*4//4 + 4*N*N)
def test_gemm_upcasted_locals(self):
k = Kernel(self.ast_gemm)
k.apply_opt(Opt(OptOps.UPCAST, 0, 4))
k.apply_opt(Opt(OptOps.UPCAST, 1, 4))
try:
k.apply_opt(Opt(OptOps.LOCAL, 0, 5))
k.apply_opt(Opt(OptOps.LOCAL, 1, 5))
except KernelOptError:
raise unittest.SkipTest("no locals")
p = k.to_program()
self.check_gemm(p)
self.assertEqual(p.lds_estimate, 2*N*N*N*4//4 + 4*N*N)
def test_gemm_group(self):
k = Kernel(self.ast_gemm)
try:
k.apply_opt(Opt(OptOps.GROUP, 0, 4))
except KernelOptError:
raise unittest.SkipTest("no locals")
SZ = N*N*4
p = k.to_program()
# NOTE: these are sort of wrong. they aren't honoring the IF statement
self.check_gemm(p, extra_flops=SZ*4)
self.assertEqual(p.lds_estimate, 2*N*N*N*4 + SZ*4 + (SZ*4 + 4*N*N)*4)
def test_reduce(self):
k = Kernel(self.ast_reduce)
p = k.to_program()
print(p.name, p.op_estimate, p.mem_estimate, p.lds_estimate)
self.assertEqual(p.op_estimate, N*N)
self.assertEqual(p.mem_estimate, N*N*4 + 4)
def test_reduce_group(self):
k = Kernel(self.ast_reduce)
try:
k.apply_opt(Opt(OptOps.GROUP, 0, 50))
except KernelOptError:
raise unittest.SkipTest("no locals")
p = k.to_program()
# NOTE: these are wrong, they don't respect the if statement
print(p.name, p.op_estimate, p.mem_estimate, p.lds_estimate)
if __name__ == '__main__':
unittest.main(verbosity=2)