Databases
StackOverflow2013
Postsdbo
PO
 

Note that there are some explanatory texts on larger screens.

plurals
  1. PO
    primarykey
    data
    text
    <p>Here are some concrete numbers, on Windows using GCC 4.7.2:</p> <pre><code>#include &lt;stdio.h&gt; #include &lt;stdlib.h&gt; int main() { unsigned int result = 0; for (int n = -500000000; n != 500000000; n++) { int d = -1; for (int i = 0; i != ITERATIONS; i++) d &amp;= rand(); #if CHECK == 0 if (d == 0) result++; #elif CHECK == 1 result += n / d; #elif CHECK == 2 result += n / (d + !d); #elif CHECK == 3 result += d == 0 ? 0 : n / d; #elif CHECK == 4 result += d == 0 ? 1 : n / d; #elif CHECK == 5 if (d != 0) result += n / d; #endif } printf("%u\n", result); } </code></pre> <p>Note that I am intentionally not calling <code>srand()</code>, so that <code>rand()</code> always returns exactly the same results. Note also that <code>-DCHECK=0</code> merely counts the zeroes, so that it is obvious how often appeared.</p> <p>Now, compiling and timing it various ways:</p> <pre><code>$ for it in 0 1 2 3 4 5; do for ch in 0 1 2 3 4 5; do gcc test.cc -o test -O -DITERATIONS=$it -DCHECK=$ch &amp;&amp; { time=`time ./test`; echo "Iterations $it, check $ch: exit status $?, output $time"; }; done; done </code></pre> <p>shows output that can be summarised in a table:</p> <pre><code>Iterations → | 0 | 1 | 2 | 3 | 4 | 5 -------------+------------------------------------------------------------------- Zeroes | 0 | 1 | 133173 | 1593376 | 135245875 | 373728555 Check 1 | 0m0.612s | - | - | - | - | - Check 2 | 0m0.612s | 0m6.527s | 0m9.718s | 0m13.464s | 0m18.422s | 0m22.871s Check 3 | 0m0.616s | 0m5.601s | 0m8.954s | 0m13.211s | 0m19.579s | 0m25.389s Check 4 | 0m0.611s | 0m5.570s | 0m9.030s | 0m13.544s | 0m19.393s | 0m25.081s Check 5 | 0m0.612s | 0m5.627s | 0m9.322s | 0m14.218s | 0m19.576s | 0m25.443s </code></pre> <p>If zeroes are rare, the <code>-DCHECK=2</code> version performs badly. As zeroes start appearing more, the <code>-DCHECK=2</code> case starts performing significantly better. Out of the other options, there really isn't much difference.</p> <p>For <code>-O3</code>, though, it is a different story:</p> <pre><code>Iterations → | 0 | 1 | 2 | 3 | 4 | 5 -------------+------------------------------------------------------------------- Zeroes | 0 | 1 | 133173 | 1593376 | 135245875 | 373728555 Check 1 | 0m0.646s | - | - | - | - | - Check 2 | 0m0.654s | 0m5.670s | 0m9.905s | 0m14.238s | 0m17.520s | 0m22.101s Check 3 | 0m0.647s | 0m5.611s | 0m9.085s | 0m13.626s | 0m18.679s | 0m25.513s Check 4 | 0m0.649s | 0m5.381s | 0m9.117s | 0m13.692s | 0m18.878s | 0m25.354s Check 5 | 0m0.649s | 0m6.178s | 0m9.032s | 0m13.783s | 0m18.593s | 0m25.377s </code></pre> <p>There, check 2 has no drawback compared the other checks, and it does keep the benefits as zeroes become more common.</p> <p>You should really measure to see what happens with your compiler and your representative sample data, though.</p>
    singulars
    1. This table or related slice is empty.
    1. POWhat is the fastest integer division supporting division by zero no matter what the result is?
      singulars
      1. PTQuestion
    1. PTAnswer
    1. UShvd
    1. UShvd
    plurals
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. VO
      singulars
      1. PO
      1. This table or related slice is empty.
      1. VTUpMod
    2. VO
      singulars
      1. PO
      1. This table or related slice is empty.
      1. VTUpMod
    3. VO
      singulars
      1. PO
      1. This table or related slice is empty.
      1. VTUpMod
    1. COMake 50% of the entries be `d=0` randomly, instead of making it almost always `d!=0`, and you'll see more branch prediction failures. Branch prediction is great if one branch is almost always followed, or if the following of one branch or the other is really clumpy...
      singulars
      1. PO
      1. USYakk - Adam Nevraumont
    2. CO@Yakk The `d` iteration is the inner loop, so the `d == 0` cases are distributed evenly. And is making 50% of the cases `d == 0` realistic?
      singulars
      1. PO
      1. UShvd
    3. COis making `0.002%` of the cases `d==0` realistic? They are distributed throughout, every 65000 iterations you hit your `d==0` case. While `50%` might not happen often, `10%` or `1%` could easily happen, or even `90%` or `99%`. The test as displayed only really tests "if you basically never, ever go down a branch, does branch prediction make removing the branch pointless?", to which the answer is "yes, but that isn't interesting".
      singulars
      1. PO
      1. USYakk - Adam Nevraumont
 

Querying!

 
Guidance
A row detail

Detail views are divided into sections. All the information in the data section comes from columns in the selected row. The other sections display data from other, related rows.

Related data can be related in a to-one or a to-many fashion. Captions of data related in a to-many fashion link to a list view showing a filtered view of the table.

Try moving around until you find a non-empty to-many entry and click on the label to get to one. You can move back to the root by clicking on the database name in the header.

SQuiL has stopped working due to an internal error.

If you are curious you may find further information in the browser console, which is accessible through the devtools (F12).

Reload