Databases
StackOverflow2013
Postsdbo
POWhy would introducing useless MOV instructions speed up a tight loop in x86_64 assembly?
 

Note that there are some explanatory texts on larger screens.

plurals
  1. POWhy would introducing useless MOV instructions speed up a tight loop in x86_64 assembly?
    primarykey
    data
    text
    <p><strong>Background:</strong></p> <p>While optimizing some <a href="http://en.wikipedia.org/wiki/Pascal_%28programming_language%29">Pascal</a> code with embedded assembly language, I noticed an unnecessary <code>MOV</code> instruction, and removed it.</p> <p>To my surprise, removing the un-necessary instruction caused my program to <em>slow down</em>.</p> <p>I found that <strong>adding arbitrary, useless <code>MOV</code> instructions increased performance</strong> even further.</p> <p>The effect is erratic, and changes based on execution order: <strong>the same junk instructions transposed</strong> up or down by a single line <strong>produce a slowdown</strong>.</p> <p>I understand that the CPU does all kinds of optimizations and streamlining, but, this seems more like black magic.</p> <p><strong>The data:</strong></p> <p>A version of my code conditionally compiles <strong>three junk operations</strong> in the middle of a loop that runs <code>2**20==1048576</code> times. (The surrounding program just calculates <a href="http://en.wikipedia.org/wiki/SHA-2">SHA-256</a> hashes).</p> <p>The results on my rather old machine (Intel(R) Core(TM)2 CPU 6400 @ 2.13&nbsp;GHz):</p> <pre><code>avg time (ms) with -dJUNKOPS: 1822.84 ms avg time (ms) without: 1836.44 ms </code></pre> <p>The programs were run 25 times in a loop, with the run order changing randomly each time.</p> <p><strong>Excerpt:</strong></p> <pre><code>{$asmmode intel} procedure example_junkop_in_sha256; var s1, t2 : uint32; begin // Here are parts of the SHA-256 algorithm, in Pascal: // s0 {r10d} := ror(a, 2) xor ror(a, 13) xor ror(a, 22) // s1 {r11d} := ror(e, 6) xor ror(e, 11) xor ror(e, 25) // Here is how I translated them (side by side to show symmetry): asm MOV r8d, a ; MOV r9d, e ROR r8d, 2 ; ROR r9d, 6 MOV r10d, r8d ; MOV r11d, r9d ROR r8d, 11 {13 total} ; ROR r9d, 5 {11 total} XOR r10d, r8d ; XOR r11d, r9d ROR r8d, 9 {22 total} ; ROR r9d, 14 {25 total} XOR r10d, r8d ; XOR r11d, r9d // Here is the extraneous operation that I removed, causing a speedup // s1 is the uint32 variable declared at the start of the Pascal code. // // I had cleaned up the code, so I no longer needed this variable, and // could just leave the value sitting in the r11d register until I needed // it again later. // // Since copying to RAM seemed like a waste, I removed the instruction, // only to discover that the code ran slower without it. {$IFDEF JUNKOPS} MOV s1, r11d {$ENDIF} // The next part of the code just moves on to another part of SHA-256, // maj { r12d } := (a and b) xor (a and c) xor (b and c) mov r8d, a mov r9d, b mov r13d, r9d // Set aside a copy of b and r9d, r8d mov r12d, c and r8d, r12d { a and c } xor r9d, r8d and r12d, r13d { c and b } xor r12d, r9d // Copying the calculated value to the same s1 variable is another speedup. // As far as I can tell, it doesn't actually matter what register is copied, // but moving this line up or down makes a huge difference. {$IFDEF JUNKOPS} MOV s1, r9d // after mov r12d, c {$ENDIF} // And here is where the two calculated values above are actually used: // T2 {r12d} := S0 {r10d} + Maj {r12d}; ADD r12d, r10d MOV T2, r12d end end; </code></pre> <p><strong>Try it yourself:</strong></p> <p>The code is online <a href="https://github.com/tangentstorm/coinops/tree/junkops">at GitHub</a> if you want to try it out yourself.</p> <p><strong>My questions:</strong></p> <ul> <li>Why would uselessly copying a register's contents to <a href="http://en.wikipedia.org/wiki/Random-access_memory">RAM</a> ever increase performance?</li> <li>Why would the same useless instruction provide a speedup on some lines, and a slowdown on others?</li> <li>Is this behavior something that could be exploited predictably by a compiler?</li> </ul>
    singulars
    1. PO
      singulars
      1. PTAnswer
    1. This table or related slice is empty.
    1. PTQuestion
    1. USRaymond Hettinger
    1. UStangentstorm
    plurals
    1. PL
      singulars
      1. LTLinked
    2. PL
      singulars
      1. LTLinked
    3. PL
      singulars
      1. LTLinked
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. PO
      singulars
      1. PTAnswer
    2. PO
      singulars
      1. PTAnswer
    3. PO
      singulars
      1. PTAnswer
    1. VO
      singulars
      1. POWhy would introducing useless MOV instructions speed up a tight loop in x86_64 assembly?
      1. This table or related slice is empty.
      1. VTUpMod
    2. VO
      singulars
      1. POWhy would introducing useless MOV instructions speed up a tight loop in x86_64 assembly?
      1. This table or related slice is empty.
      1. VTUpMod
    3. VO
      singulars
      1. POWhy would introducing useless MOV instructions speed up a tight loop in x86_64 assembly?
      1. This table or related slice is empty.
      1. VTUpMod
    1. COThere are all sorts of 'useless' instructions that can actually serve to break dependency chains, mark physical registers as retired, etc. Exploiting these operations requires some knowledge of the [microarchitecture](http://www.agner.org/optimize/microarchitecture.pdf). Your question should provide a short sequence of instructions as a minimal example, rather than directing people to github.
      singulars
      1. POWhy would introducing useless MOV instructions speed up a tight loop in x86_64 assembly?
      1. USBrett Hale
    2. CO@BrettHale good point, thanks. I added a code excerpt with some commentary. Would copying a register's value to ram mark the register as retired, even if the value in it is used later?
      singulars
      1. POWhy would introducing useless MOV instructions speed up a tight loop in x86_64 assembly?
      1. UStangentstorm
    3. COtry to exchange the order of "mov r8d, a" and "mov r9d, b" without "mov s1, r11d". r8d is used in the instruction "xor r10d, r8d" and "mov r8d, something" can not be executed before the end of the "xor".
      singulars
      1. POWhy would introducing useless MOV instructions speed up a tight loop in x86_64 assembly?
      1. USjohnfound
 

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