pystencils/kerncraft_coupling/generate_benchmark.py

-import os
-import subprocess
-
-from jinja2 import Template
-
-from pystencils.astnodes import PragmaBlock
-from pystencils.backends.cbackend import generate_c, get_headers
-from pystencils.cpu.cpujit import get_compiler_config, run_compile_step
-from pystencils.data_types import get_base_type
-from pystencils.include import get_pystencils_include_path
-from pystencils.sympyextensions import prod
-
-benchmark_template = Template("""
-#include "kerncraft.h"
-#include <stdlib.h>
-#include <stdint.h>
-#include <stdbool.h>
-#include <math.h>
-#include <stdio.h>
-
-{{ includes }}
-
-{%- if likwid %}
-#include <likwid.h>
-{%- endif %}
-
-#define RESTRICT __restrict__
-#define FUNC_PREFIX
-void dummy(void *);
-void timing(double* wcTime, double* cpuTime);
-extern int var_false;
-
-
-{{kernel_code}}
-
-
-int main(int argc, char **argv)
-{
-  {%- if likwid %}
-  likwid_markerInit();
-  {%- endif %}
-
-  {%- for field_name, dataType, size in fields %}
-
-  // Initialization {{field_name}}
-  double * {{field_name}} = (double *) aligned_malloc(sizeof({{dataType}}) * {{size}}, 64);
-  for (unsigned long long i = 0; i < {{size}}; ++i)
-    {{field_name}}[i] = 0.23;
-
-  if(var_false)
-    dummy({{field_name}});
-
-  {%- endfor %}
-
-
-
-  {%- for constantName, dataType in constants %}
-
-  // Constant {{constantName}}
-  {{dataType}} {{constantName}};
-  {{constantName}} = 0.23;
-  if(var_false)
-      dummy(& {{constantName}});
-
-  {%- endfor %}
-
-  {%- if likwid and openmp %}
-  #pragma omp parallel
-  {
-  likwid_markerRegisterRegion("loop");
-  #pragma omp barrier
-  {%- elif likwid %}
-  likwid_markerRegisterRegion("loop");
-  {%- endif %}
-
-  for(int warmup = 1; warmup >= 0; --warmup) {
-    int repeat = 2;
-    if(warmup == 0) {
-      repeat = atoi(argv[1]);
-      {%- if likwid %}
-      likwid_markerStartRegion("loop");
-      {%- endif %}
-    }
-    
-    {%- if timing %}
-    double wcStartTime, cpuStartTime, wcEndTime, cpuEndTime;
-    timing(&wcStartTime, &cpuStartTime);
-    {%- endif %}
-    
-    for (; repeat > 0; --repeat)
-    {
-      {{kernelName}}({{call_argument_list}});
-
-      // Dummy calls
-      {%- for field_name, dataType, size in fields %}
-      if(var_false) dummy((void*){{field_name}});
-      {%- endfor %}
-      {%- for constantName, dataType in constants %}
-      if(var_false) dummy((void*)&{{constantName}});
-      {%- endfor %}
-    }
-    {%- if timing %}
-    timing(&wcEndTime, &cpuEndTime);
-    if( warmup == 0)
-        printf("%e\\n", (wcEndTime - wcStartTime) / atoi(argv[1]) );
-    {%- endif %}
-
-  }
-
-  {%- if likwid %}
-  likwid_markerStopRegion("loop");
-  {%- if openmp %}
-  }
-  {%- endif %}
-  {%- endif %}
-
-  {%- if likwid %}
-  likwid_markerClose();
-  {%- endif %}
-}
-""")
-
-
-def generate_benchmark(ast, likwid=False, openmp=False, timing=False):
-    """Return C code of a benchmark program for the given kernel.
-
-    Args:
-        ast: the pystencils AST object as returned by create_kernel
-        likwid: if True likwid markers are added to the code
-        openmp: relevant only if likwid=True, to generated correct likwid initialization code
-        timing: add timing output to the code, prints time per iteration to stdout
-
-    Returns:
-        C code as string
-    """
-    accessed_fields = {f.name: f for f in ast.fields_accessed}
-    constants = []
-    fields = []
-    call_parameters = []
-    for p in ast.get_parameters():
-        if not p.is_field_parameter:
-            constants.append((p.symbol.name, str(p.symbol.dtype)))
-            call_parameters.append(p.symbol.name)
-        else:
-            assert p.is_field_pointer, "Benchmark implemented only for kernels with fixed loop size"
-            field = accessed_fields[p.field_name]
-            dtype = str(get_base_type(p.symbol.dtype))
-            fields.append((p.field_name, dtype, prod(field.shape)))
-            call_parameters.append(p.field_name)
-
-    header_list = get_headers(ast)
-    includes = "\n".join(["#include %s" % (include_file,) for include_file in header_list])
-
-    # Strip "#pragma omp parallel" from within kernel, because main function takes care of that
-    # when likwid and openmp are enabled
-    if likwid and openmp:
-        if len(ast.body.args) > 0 and isinstance(ast.body.args[0], PragmaBlock):
-            ast.body.args[0].pragma_line = ''
-
-    args = {
-        'likwid': likwid,
-        'openmp': openmp,
-        'kernel_code': generate_c(ast, dialect='c'),
-        'kernelName': ast.function_name,
-        'fields': fields,
-        'constants': constants,
-        'call_argument_list': ",".join(call_parameters),
-        'includes': includes,
-        'timing': timing,
-    }
-    return benchmark_template.render(**args)
-
-
-def run_c_benchmark(ast, inner_iterations, outer_iterations=3):
-    """Runs the given kernel with outer loop in C
-
-    Args:
-        ast:
-        inner_iterations: timings are recorded around this many iterations
-        outer_iterations: number of timings recorded
-
-    Returns:
-        list of times per iterations for each outer iteration
-    """
-    import kerncraft
-
-    benchmark_code = generate_benchmark(ast, timing=True)
-    with open('bench.c', 'w') as f:
-        f.write(benchmark_code)
-
-    kerncraft_path = os.path.dirname(kerncraft.__file__)
-
-    extra_flags = ['-I' + get_pystencils_include_path(),
-                   '-I' + os.path.join(kerncraft_path, 'headers')]
-
-    compiler_config = get_compiler_config()
-    compile_cmd = [compiler_config['command']] + compiler_config['flags'].split()
-    compile_cmd += [*extra_flags,
-                    os.path.join(kerncraft_path, 'headers', 'timing.c'),
-                    os.path.join(kerncraft_path, 'headers', 'dummy.c'),
-                    'bench.c',
-                    '-o', 'bench',
-                    ]
-    run_compile_step(compile_cmd)
-
-    results = []
-    for _ in range(outer_iterations):
-        benchmark_time = float(subprocess.check_output(['./bench', str(inner_iterations)]))
-        results.append(benchmark_time)
-    return results

pystencils/llvm/__init__.py

-from .kernelcreation import create_kernel
-from .llvmjit import make_python_function
-
-__all__ = ['create_kernel', 'make_python_function']

pystencils/llvm/control_flow.py

-import llvmlite.ir as ir
-
-
-class Loop(object):
-    def __init__(self, builder, start_val, stop_val, step_val=1, loop_name='loop', phi_name="_phi"):
-        self.builder = builder
-        self.start_val = start_val
-        self.stop_val = stop_val
-        self.step_val = step_val
-        self.loop_name = loop_name
-        self.phi_name = phi_name
-
-    def __enter__(self):
-        self.loop_end, self.after, phi = self._for_loop(self.start_val, self.stop_val, self.step_val, self.loop_name,
-                                                        self.phi_name)
-        return phi
-
-    def _for_loop(self, start_val, stop_val, step_val, loop_name, phi_name):
-        # TODO size of int??? unisgned???
-        integer = ir.IntType(64)
-
-        # Loop block
-        pre_loop_bb = self.builder.block
-        loop_bb = self.builder.append_basic_block(name='loop_' + loop_name)
-
-        self.builder.branch(loop_bb)
-
-        # Insert an explicit fall through from the current block to loop_bb
-        self.builder.position_at_start(loop_bb)
-
-        # Add phi
-        phi = self.builder.phi(integer, name=phi_name)
-        phi.add_incoming(start_val, pre_loop_bb)
-
-        loop_end_bb = self.builder.append_basic_block(name=loop_name + "_end_bb")
-        self.builder.position_at_start(loop_end_bb)
-
-        next_var = self.builder.add(phi, step_val, name=loop_name + '_next_it')
-        cond = self.builder.icmp_unsigned('<', next_var, stop_val, name=loop_name + "_cond")
-
-        after_bb = self.builder.append_basic_block(name=loop_name + "_after_bb")
-
-        self.builder.cbranch(cond, loop_bb, after_bb)
-        phi.add_incoming(next_var, loop_end_bb)
-
-        self.builder.position_at_end(loop_bb)
-
-        return loop_end_bb, after_bb, phi
-
-    def __exit__(self, exc_type, exc, exc_tb):
-        self.builder.branch(self.loop_end)
-        self.builder.position_at_end(self.after)

pystencils/llvm/kernelcreation.py

-from pystencils.llvm.llvmjit import make_python_function
-from pystencils.transformations import insert_casts
-
-
-def create_kernel(assignments, function_name="kernel", type_info=None, split_groups=(),
-                  iteration_slice=None, ghost_layers=None):
-    """
-    Creates an abstract syntax tree for a kernel function, by taking a list of update rules.
-
-    Loops are created according to the field accesses in the equations.
-
-    Args:
-    assignments: list of sympy equations, containing accesses to :class:`pystencils.field.Field`.
-                 Defining the update rules of the kernel
-    function_name: name of the generated function - only important if generated code is written out
-                   type_info: a map from symbol name to a C type specifier. If not specified all symbols are assumed to
-                   be of type 'double' except symbols which occur on the left hand side of equations where the
-                   right hand side is a sympy Boolean which are assumed to be 'bool' .
-    split_groups: Specification on how to split up inner loop into multiple loops. For details see
-                  transformation :func:`pystencils.transformation.split_inner_loop`
-    iteration_slice: if not None, iteration is done only over this slice of the field
-    ghost_layers: a sequence of pairs for each coordinate with lower and upper nr of ghost layers
-                 if None, the number of ghost layers is determined automatically and assumed to be equal for a
-                 all dimensions
-
-    :return: :class:`pystencils.ast.KernelFunction` node
-    """
-    from pystencils.cpu import create_kernel
-    code = create_kernel(assignments, function_name, type_info, split_groups, iteration_slice, ghost_layers)
-    code.body = insert_casts(code.body)
-    code._compile_function = make_python_function
-    code._backend = 'llvm'
-    return code

pystencils/sympy_gmpy_bug_workaround.py

-# Disable gmpy backend until this bug is resolved if joblib serialize
-# See https://github.com/sympy/sympy/pull/13530
-import os
-import warnings
-
-os.environ['MPMATH_NOGMPY'] = '1'
-try:
-    import mpmath.libmp
-    # In case the user has imported sympy first, then pystencils
-    if mpmath.libmp.BACKEND == 'gmpy':
-        warnings.warn("You are using the gmpy backend. You might encounter an error 'argument is not an mpz sympy'. "
-                      "This is due to a known bug in sympy/gmpy library. "
-                      "To prevent this, import pystencils first then sympy or set the environment variable "
-                      "MPMATH_NOGMPY=1")
-except ImportError:
-    pass
-
-__all__ = []

pystencils_tests/benchmark/generate.py

-import numpy as np
-import sympy as sp
-
-from pystencils import Assignment, Field, create_kernel
-
-
-def meassure():
-    size = [30, 50, 3]
-    arr = np.zeros(size)
-    a = Field.create_from_numpy_array('a', arr, index_dimensions=1)
-    b = Field.create_from_numpy_array('b', arr, index_dimensions=1)
-    s = sp.Symbol("s")
-    rhs = a[0, -1](0) + a[0, 1] + a[-1, 0] + a[1, 0]
-    updateRule = Assignment(b[0, 0], s * rhs)
-    print(updateRule)
-
-    ast = create_kernel([updateRule])
-
-    # benchmark = generate_benchmark(ast)
-    # main = benchmark[0]
-    # kernel = benchmark[1]
-    # with open('src/main.cpp', 'w') as file:
-    #     file.write(main)
-    # with open('src/kernel.cpp', 'w') as file:
-    #     file.write(kernel)
-
-    func = ast.compile({'omega': 2/3})
-
-    from pystencils.kerncraft_coupling.generate_benchmark import generate_benchmark
-    from pystencils.kerncraft_coupling import BenchmarkAnalysis
-    from pystencils.kerncraft_coupling.kerncraft_interface import PyStencilsKerncraftKernel, KerncraftParameters
-    from kerncraft.machinemodel import MachineModel
-    from kerncraft.models import ECMData
-
-
-    machineFilePath = "../pystencils_tests/kerncraft_inputs/default_machine_file.yaml"
-    machine = MachineModel(path_to_yaml=machineFilePath)
-
-
-    benchmark = BenchmarkAnalysis(ast, machine)
-    #TODO what do i want to do with benchmark?
-
-    kernel = PyStencilsKerncraftKernel(ast)
-    model = ECMData(kernel, machine, KerncraftParameters())
-    model.analyze()
-    model.report()
-
-
-if __name__ == "__main__":
-    meassure()

pystencils_tests/benchmark/main.c

-#include "iacaMarks.h"
-
-int main(int argc, char * argv[]){
-	int a = 0;
-	for(int i = 0; i < argc+100000; i++){
-		IACA_START
-		a += i;
-	}
-	IACA_END
-	return a;
-}

pystencils_tests/kerncraft_inputs/2d-5pt.c

-double a[30][50][3];
-double b[30][50][3];
-double s;
-
-for(int j=1; j<30-1; ++j)
-    for(int i=1; i<50-1; ++i)
-        b[j][i] = ( a[j][i-1] + a[j][i+1]
-                  + a[j-1][i] + a[j+1][i]) * s;

pystencils_tests/kerncraft_inputs/3d-7pt.c

-double a[M][N][N];
-double b[M][N][N];
-double s;
-
-for(int k=1; k<M-1; ++k)
-    for(int j=1; j<N-1; ++j)
-        for(int i=1; i<N-1; ++i)
-            b[k][j][i] = ( a[k][j][i-1] + a[k][j][i+1]
-                         + a[k][j-1][i] + a[k][j+1][i]
-                         + a[k-1][j][i] + a[k+1][j][i]) * s;