1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
|
from __future__ import print_function
from ctypes import CFUNCTYPE, c_int, POINTER
import sys
try:
from time import perf_counter as time
except ImportError:
from time import time
import numpy as np
try:
import faulthandler; faulthandler.enable()
except ImportError:
pass
import llvmlite
llvmlite.ir_layer_typed_pointers_enabled = False
import llvmlite.ir as ll
import llvmlite.binding as llvm
llvm.initialize_native_target()
llvm.initialize_native_asmprinter()
t1 = time()
# Pointers are opaque, so we should define them as such here.
fnty = ll.FunctionType(ll.IntType(32), [ll.PointerType(), ll.IntType(32)])
module = ll.Module()
func = ll.Function(module, fnty, name="sum")
bb_entry = func.append_basic_block()
bb_loop = func.append_basic_block()
bb_exit = func.append_basic_block()
builder = ll.IRBuilder()
builder.position_at_end(bb_entry)
builder.branch(bb_loop)
builder.position_at_end(bb_loop)
index = builder.phi(ll.IntType(32))
index.add_incoming(ll.Constant(index.type, 0), bb_entry)
accum = builder.phi(ll.IntType(32))
accum.add_incoming(ll.Constant(accum.type, 0), bb_entry)
# These GEP and load need an excplicit type.
ptr = builder.gep(func.args[0], [index], source_etype=ll.IntType(32))
value = builder.load(ptr, typ=ll.IntType(32))
added = builder.add(accum, value)
accum.add_incoming(added, bb_loop)
indexp1 = builder.add(index, ll.Constant(index.type, 1))
index.add_incoming(indexp1, bb_loop)
cond = builder.icmp_unsigned('<', indexp1, func.args[1])
builder.cbranch(cond, bb_loop, bb_exit)
builder.position_at_end(bb_exit)
builder.ret(added)
strmod = str(module)
t2 = time()
print("-- generate IR:", t2-t1)
t3 = time()
llmod = llvm.parse_assembly(strmod)
t4 = time()
print("-- parse assembly:", t4-t3)
print(llmod)
pmb = llvm.create_pass_manager_builder()
pmb.opt_level = 2
pm = llvm.create_module_pass_manager()
pmb.populate(pm)
t5 = time()
pm.run(llmod)
t6 = time()
print("-- optimize:", t6-t5)
t7 = time()
target_machine = llvm.Target.from_default_triple().create_target_machine()
with llvm.create_mcjit_compiler(llmod, target_machine) as ee:
ee.finalize_object()
cfptr = ee.get_function_address("sum")
t8 = time()
print("-- JIT compile:", t8 - t7)
print(target_machine.emit_assembly(llmod))
cfunc = CFUNCTYPE(c_int, POINTER(c_int), c_int)(cfptr)
A = np.arange(10, dtype=np.int32)
res = cfunc(A.ctypes.data_as(POINTER(c_int)), A.size)
print(res, A.sum())
|
