inoryy

Fixing MKL on AMD Zen CPU

As per discussion on Reddit, it seems a workaround for the Intel MKL's notorious SIMD throttling of AMD Zen CPUs is as simple a setting MKL_DEBUG_CPU_TYPE=5 environment variable.

Benchmarks

All three scripts are executed in the same Python 3.7 environment on a first-gen AMD Zen CPU (1950x).
The difference will be even bigger on newer models as first-gen Zen resolves 256-bit AVX2 in two 128-bit instructions.

inoryy

Node* mcts(Node* root) {
    save();
    while (true) {
        if (TIME >= TIME_LIMIT) break;

        // select
        Node* node = root;
        while (!isTerminal() && node->isExpanded()) {
            node = node->uctChild();
            apply(node->action);

inoryy

import tensorflow as tf


class TfCategorical:
    def __init__(self, logits):
        self.logits = logits
        self.probs = tf.nn.softmax(logits)

    def sample(self):
        u = tf.random_uniform(tf.shape(self.logits))

inoryy

#!/usr/bin/env python

import sys, re, os

DEFAULT_MAIN = 'main.cpp'
DEFAULT_OUT  = 'arena.cpp'
BUNDLE_SYM   = '// ###'

def bundle(flname, main = False):
    base = os.path.dirname(flname)

inoryy

from pydoc import locate
import tensorflow as tf
import numpy as np
from seq2seq import tasks, models
from seq2seq.training import utils as training_utils
from seq2seq.tasks.inference_task import InferenceTask, unbatch_dict


class DecodeOnce(InferenceTask):
  '''

inoryy

from heapq import heappop, heappush

def dijkstra(g, s, f):
    h = [(0,s)]
    cost, prev = {}, {}
    cost[s], prev[s] = 0, None

    while len(h) > 0:
        _,v = heappop(h)
        if v == f:

inoryy

#!/home/inoryy/anaconda3/bin/python

import sys, time
from math import sqrt, acos, pi, cos, sin
from decimal import Decimal, ROUND_HALF_UP

class Point(object):
    __slots__ = 'x', 'y'

    def __init__(self, x, y):

inoryy

from random import random
from math import log, cos, sin, sqrt, pi

# Implementation of Box-Muller transform to
# generate two random numbers from N(mu, sigma) distribution
def pnorm(mu = 0, sigma = 1):
    r = sqrt(-2*log(random()))
    a = 2*pi*random()
    return sigma*r*cos(a) + mu, sigma*r*sin(a) + mu

inoryy

import pandas
from sklearn.linear_model import LogisticRegression
from sklearn import cross_validation

def sanitize(dataset):
    dataset["Age"] = dataset["Age"].fillna(dataset["Age"].median())
    dataset["Fare"] = dataset["Fare"].fillna(dataset["Fare"].median())

    dataset.loc[dataset["Sex"] == "male", "Sex"] = 0 
    dataset.loc[dataset["Sex"] == "female", "Sex"] = 1

inoryy

<?php

final class FormTypes {
  const HIDDEN = 'hidden';
  const TEXT   = 'text';
  // ...
  
  private __constructor() {}
}