From 696eb2d5cdadb31787f19d92ccc5d42d2dbf6e6f Mon Sep 17 00:00:00 2001
From: Martin Bauer <martin.bauer@fau.de>
Date: Fri, 29 Mar 2019 17:51:58 +0100
Subject: [PATCH] Support for conditionals in vectorized loops (by specifying
all/any)
---
pystencils/astnodes.py | 1 -
pystencils/backends/cbackend.py | 22 +++++++++-
pystencils/backends/simd_instruction_sets.py | 4 ++
pystencils/cpu/vectorization.py | 17 +++++++-
pystencils/data_types.py | 5 +++
pystencils/kernelcreation.py | 4 +-
pystencils/transformations.py | 13 +++++-
pystencils_tests/test_conditional_vec.py | 43 ++++++++++++++++++++
8 files changed, 103 insertions(+), 6 deletions(-)
create mode 100644 pystencils_tests/test_conditional_vec.py
diff --git a/pystencils/astnodes.py b/pystencils/astnodes.py
index 992cbf2..31d49c1 100644
--- a/pystencils/astnodes.py
+++ b/pystencils/astnodes.py
@@ -65,7 +65,6 @@ class Conditional(Node):
false_block: Optional['Block'] = None) -> None:
super(Conditional, self).__init__(parent=None)
- assert condition_expr.is_Boolean or condition_expr.is_Relational
self.condition_expr = condition_expr
def handle_child(c):
diff --git a/pystencils/backends/cbackend.py b/pystencils/backends/cbackend.py
index f0787bc..cf96dcd 100644
--- a/pystencils/backends/cbackend.py
+++ b/pystencils/backends/cbackend.py
@@ -4,6 +4,7 @@ from sympy.core import S
from typing import Set
from sympy.printing.ccode import C89CodePrinter
+from pystencils.cpu.vectorization import vec_any, vec_all
from pystencils.fast_approximation import fast_division, fast_sqrt, fast_inv_sqrt
try:
@@ -198,6 +199,9 @@ class CBackend:
return node.get_code(self._dialect, self._vector_instruction_set)
def _print_Conditional(self, node):
+ cond_type = get_type_of_expression(node.condition_expr)
+ if isinstance(cond_type, VectorType):
+ raise ValueError("Problem with Conditional inside vectorized loop - use vec_any or vec_all")
condition_expr = self.sympy_printer.doprint(node.condition_expr)
true_block = self._print_Block(node.true_block)
result = "if (%s)\n%s " % (condition_expr, true_block)
@@ -274,6 +278,8 @@ class CustomSympyPrinter(CCodePrinter):
return "__fsqrt_rn(%s)" % tuple(self._print(a) for a in expr.args)
else:
return "({})".format(self._print(sp.sqrt(expr.args[0])))
+ elif isinstance(expr, vec_any) or isinstance(expr, vec_all):
+ return self._print(expr.args[0])
elif isinstance(expr, fast_inv_sqrt):
if self._dialect == "cuda":
return "__frsqrt_rn(%s)" % tuple(self._print(a) for a in expr.args)
@@ -328,7 +334,8 @@ class VectorizedCustomSympyPrinter(CustomSympyPrinter):
elif expr.func == fast_division:
result = self._scalarFallback('_print_Function', expr)
if not result:
- return self.instruction_set['/'].format(self._print(expr.args[0]), self._print(expr.args[1]))
+ result = self.instruction_set['/'].format(self._print(expr.args[0]), self._print(expr.args[1]))
+ return result
elif expr.func == fast_sqrt:
return "({})".format(self._print(sp.sqrt(expr.args[0])))
elif expr.func == fast_inv_sqrt:
@@ -338,6 +345,19 @@ class VectorizedCustomSympyPrinter(CustomSympyPrinter):
return self.instruction_set['rsqrt'].format(self._print(expr.args[0]))
else:
return "({})".format(self._print(1 / sp.sqrt(expr.args[0])))
+ elif isinstance(expr, vec_any):
+ expr_type = get_type_of_expression(expr.args[0])
+ if type(expr_type) is not VectorType:
+ return self._print(expr.args[0])
+ else:
+ return self.instruction_set['any'].format(self._print(expr.args[0]))
+ elif isinstance(expr, vec_all):
+ expr_type = get_type_of_expression(expr.args[0])
+ if type(expr_type) is not VectorType:
+ return self._print(expr.args[0])
+ else:
+ return self.instruction_set['all'].format(self._print(expr.args[0]))
+
return super(VectorizedCustomSympyPrinter, self)._print_Function(expr)
def _print_And(self, expr):
diff --git a/pystencils/backends/simd_instruction_sets.py b/pystencils/backends/simd_instruction_sets.py
index 0220750..f72b0fe 100644
--- a/pystencils/backends/simd_instruction_sets.py
+++ b/pystencils/backends/simd_instruction_sets.py
@@ -98,11 +98,15 @@ def get_vector_instruction_set(data_type='double', instruction_set='avx'):
result['bool'] = "__m%dd" % (bit_width,)
result['headers'] = headers[instruction_set]
+ result['any'] = "%s_movemask_%s({0}) > 0" % (pre, suf)
+ result['all'] = "%s_movemask_%s({0}) == 0xF" % (pre, suf)
if instruction_set == 'avx512':
size = 8 if data_type == 'double' else 16
result['&'] = '_kand_mask%d({0}, {1})' % (size,)
result['|'] = '_kor_mask%d({0}, {1})' % (size,)
+ result['any'] = '!_ktestz_mask%d_u8({0}, {0})' % (size, )
+ result['all'] = '_kortestc_mask%d_u8({0}, {0})' % (size, )
result['blendv'] = '%s_mask_blend_%s({2}, {0}, {1})' % (pre, suf)
result['rsqrt'] = "_mm512_rsqrt14_%s({0})" % (suf,)
result['bool'] = "__mmask%d" % (size,)
diff --git a/pystencils/cpu/vectorization.py b/pystencils/cpu/vectorization.py
index 6a55b69..2790fe3 100644
--- a/pystencils/cpu/vectorization.py
+++ b/pystencils/cpu/vectorization.py
@@ -12,6 +12,16 @@ from pystencils.transformations import cut_loop, filtered_tree_iteration, replac
from pystencils.field import Field
+# noinspection PyPep8Naming
+class vec_any(sp.Function):
+ nargs = (1, )
+
+
+# noinspection PyPep8Naming
+class vec_all(sp.Function):
+ nargs = (1, )
+
+
def vectorize(kernel_ast: ast.KernelFunction, instruction_set: str = 'avx',
assume_aligned: bool = False, nontemporal: Union[bool, Container[Union[str, Field]]] = False,
assume_inner_stride_one: bool = False, assume_sufficient_line_padding: bool = True):
@@ -119,7 +129,7 @@ def vectorize_inner_loops_and_adapt_load_stores(ast_node, vector_width, assume_a
def insert_vector_casts(ast_node):
"""Inserts necessary casts from scalar values to vector values."""
- handled_functions = (sp.Add, sp.Mul, fast_division, fast_sqrt, fast_inv_sqrt)
+ handled_functions = (sp.Add, sp.Mul, fast_division, fast_sqrt, fast_inv_sqrt, vec_any, vec_all)
def visit_expr(expr):
@@ -182,6 +192,11 @@ def insert_vector_casts(ast_node):
lhs_type = assignment.lhs.args[1]
if type(lhs_type) is VectorType and type(rhs_type) is not VectorType:
assignment.rhs = cast_func(assignment.rhs, lhs_type)
+ elif isinstance(arg, ast.Conditional):
+ arg.condition_expr = fast_subs(arg.condition_expr, substitution_dict,
+ skip=lambda e: isinstance(e, ast.ResolvedFieldAccess))
+ arg.condition_expr = visit_expr(arg.condition_expr)
+ visit_node(arg, substitution_dict)
else:
visit_node(arg, substitution_dict)
diff --git a/pystencils/data_types.py b/pystencils/data_types.py
index 56d19d4..7bdc9d3 100644
--- a/pystencils/data_types.py
+++ b/pystencils/data_types.py
@@ -1,6 +1,7 @@
import ctypes
import sympy as sp
import numpy as np
+
try:
import llvmlite.ir as ir
except ImportError as e:
@@ -311,6 +312,8 @@ def collate_types(types):
@memorycache(maxsize=2048)
def get_type_of_expression(expr):
from pystencils.astnodes import ResolvedFieldAccess
+ from pystencils.cpu.vectorization import vec_all, vec_any
+
expr = sp.sympify(expr)
if isinstance(expr, sp.Integer):
return create_type("int")
@@ -324,6 +327,8 @@ def get_type_of_expression(expr):
raise ValueError("All symbols inside this expression have to be typed! ", str(expr))
elif isinstance(expr, cast_func):
return expr.args[1]
+ elif isinstance(expr, vec_any) or isinstance(expr, vec_all):
+ return create_type("bool")
elif hasattr(expr, 'func') and expr.func == sp.Piecewise:
collated_result_type = collate_types(tuple(get_type_of_expression(a[0]) for a in expr.args))
collated_condition_type = collate_types(tuple(get_type_of_expression(a[1]) for a in expr.args))
diff --git a/pystencils/kernelcreation.py b/pystencils/kernelcreation.py
index 5044049..024ee4a 100644
--- a/pystencils/kernelcreation.py
+++ b/pystencils/kernelcreation.py
@@ -6,7 +6,8 @@ from pystencils.astnodes import LoopOverCoordinate, Conditional, Block, SympyAss
from pystencils.cpu.vectorization import vectorize
from pystencils.simp.assignment_collection import AssignmentCollection
from pystencils.gpucuda.indexing import indexing_creator_from_params
-from pystencils.transformations import remove_conditionals_in_staggered_kernel, loop_blocking
+from pystencils.transformations import remove_conditionals_in_staggered_kernel, loop_blocking, \
+ move_constants_before_loop
def create_kernel(assignments, target='cpu', data_type="double", iteration_slice=None, ghost_layers=None,
@@ -248,6 +249,7 @@ def create_staggered_kernel(staggered_field, expressions, subexpressions=(), tar
if target == 'cpu':
remove_conditionals_in_staggered_kernel(ast)
+ move_constants_before_loop(ast)
if blocking:
loop_blocking(ast, blocking)
if cpu_vectorize_info is True:
diff --git a/pystencils/transformations.py b/pystencils/transformations.py
index 6fc9f83..6411d6a 100644
--- a/pystencils/transformations.py
+++ b/pystencils/transformations.py
@@ -7,7 +7,6 @@ import hashlib
import sympy as sp
from sympy.logic.boolalg import Boolean
from sympy.tensor import IndexedBase
-
from pystencils.simp.assignment_collection import AssignmentCollection
from pystencils.assignment import Assignment
from pystencils.field import Field, FieldType
@@ -514,6 +513,13 @@ def resolve_field_accesses(ast_node, read_only_field_names=set(),
assert type(enclosing_block) is ast.Block
sub_ast.lhs = visit_sympy_expr(sub_ast.lhs, enclosing_block, sub_ast)
sub_ast.rhs = visit_sympy_expr(sub_ast.rhs, enclosing_block, sub_ast)
+ elif isinstance(sub_ast, ast.Conditional):
+ enclosing_block = sub_ast.parent
+ assert type(enclosing_block) is ast.Block
+ sub_ast.condition_expr = visit_sympy_expr(sub_ast.condition_expr, enclosing_block, sub_ast)
+ visit_node(sub_ast.true_block)
+ if sub_ast.false_block:
+ visit_node(sub_ast.false_block)
else:
for i, a in enumerate(sub_ast.args):
visit_node(a)
@@ -545,7 +551,7 @@ def move_constants_before_loop(ast_node):
last_block_child = prev_element
if isinstance(element, ast.Conditional):
- critical_symbols = element.condition_expr.atoms(sp.Symbol)
+ break
else:
critical_symbols = element.symbols_defined
if node.undefined_symbols.intersection(critical_symbols):
@@ -1000,6 +1006,9 @@ def typing_from_sympy_inspection(eqs, default_type="double"):
elif isinstance(eq, ast.Node) and not isinstance(eq, ast.SympyAssignment):
continue
else:
+ from pystencils.cpu.vectorization import vec_all, vec_any
+ if isinstance(eq.rhs, vec_all) or isinstance(eq.rhs, vec_any):
+ result[eq.lhs.name] = "bool"
# problematic case here is when rhs is a symbol: then it is impossible to decide here without
# further information what type the left hand side is - default fallback is the dict value then
if isinstance(eq.rhs, Boolean) and not isinstance(eq.rhs, sp.Symbol):
diff --git a/pystencils_tests/test_conditional_vec.py b/pystencils_tests/test_conditional_vec.py
new file mode 100644
index 0000000..c1d11f1
--- /dev/null
+++ b/pystencils_tests/test_conditional_vec.py
@@ -0,0 +1,43 @@
+import pystencils as ps
+import sympy as sp
+import numpy as np
+from pystencils.astnodes import Conditional, Block
+from pystencils.cpu.vectorization import vec_all, vec_any
+
+
+def test_vec_any():
+ data_arr = np.zeros((15, 15))
+
+ data_arr[3:9, 2:7] = 1.0
+ data = ps.fields("data: double[2D]", data=data_arr)
+
+ c = [
+ ps.Assignment(sp.Symbol("t1"), vec_any(data.center() > 0.0)),
+ Conditional(vec_any(data.center() > 0.0), Block([
+ ps.Assignment(data.center(), 2.0)
+ ]))
+ ]
+ ast = ps.create_kernel(c, target='cpu',
+ cpu_vectorize_info={'instruction_set': 'avx'})
+ kernel = ast.compile()
+ kernel(data=data_arr)
+ np.testing.assert_equal(data_arr[3:9, 0:8], 2.0)
+
+
+def test_vec_all():
+ data_arr = np.zeros((15, 15))
+
+ data_arr[3:9, 2:7] = 1.0
+ data = ps.fields("data: double[2D]", data=data_arr)
+
+ c = [
+ Conditional(vec_all(data.center() > 0.0), Block([
+ ps.Assignment(data.center(), 2.0)
+ ]))
+ ]
+ ast = ps.create_kernel(c, target='cpu',
+ cpu_vectorize_info={'instruction_set': 'avx'})
+ kernel = ast.compile()
+ before = data_arr.copy()
+ kernel(data=data_arr)
+ np.testing.assert_equal(data_arr, before)
--
GitLab