Why is this code 6.5x slower with optimizations enabled?Unit Testing C CodeWith arrays, why is it the case...

I probably found a bug with the sudo apt install function

Why doesn't Newton's third law mean a person bounces back to where they started when they hit the ground?

Compute hash value according to multiplication method

Is it possible to make sharp wind that can cut stuff from afar?

whey we use polarized capacitor?

What are these boxed doors outside store fronts in New York?

What would the Romans have called "sorcery"?

Why is this code 6.5x slower with optimizations enabled?

How is it possible to have an ability score that is less than 3?

What is the command to reset a PC without deleting any files

Draw simple lines in Inkscape

Continuity at a point in terms of closure

GPS Rollover on Android Smartphones

How did the USSR manage to innovate in an environment characterized by government censorship and high bureaucracy?

How to get the available space of $HOME as a variable in shell scripting?

If Manufacturer spice model and Datasheet give different values which should I use?

Shell script can be run only with sh command

Why don't electromagnetic waves interact with each other?

Do airline pilots ever risk not hearing communication directed to them specifically, from traffic controllers?

I’m planning on buying a laser printer but concerned about the life cycle of toner in the machine

Is there really no realistic way for a skeleton monster to move around without magic?

When blogging recipes, how can I support both readers who want the narrative/journey and ones who want the printer-friendly recipe?

DOS, create pipe for stdin/stdout of command.com(or 4dos.com) in C or Batch?

Is there a familial term for apples and pears?



Why is this code 6.5x slower with optimizations enabled?


Unit Testing C CodeWith arrays, why is it the case that a[5] == 5[a]?Why doesn't GCC optimize a*a*a*a*a*a to (a*a*a)*(a*a*a)?Why are elementwise additions much faster in separate loops than in a combined loop?What is “:-!!” in C code?Why is my program slow when looping over exactly 8192 elements?Obfuscated C Code Contest 2006. Please explain sykes2.cWhy does the C preprocessor interpret the word “linux” as the constant “1”?Why does GCC generate 15-20% faster code if I optimize for size instead of speed?How is the linking done for string functions in C?






.everyoneloves__top-leaderboard:empty,.everyoneloves__mid-leaderboard:empty,.everyoneloves__bot-mid-leaderboard:empty{ height:90px;width:728px;box-sizing:border-box;
}







7















I wanted to benchmark glibc's strlen function for some reason and found out it apparently performs much slower with optimizations enabled in GCC and I have no idea why.



Here's my code:



#include <time.h>
#include <string.h>
#include <stdlib.h>
#include <stdio.h>

int main() {
char *s = calloc(1 << 20, 1);
memset(s, 65, 1000000);
clock_t start = clock();
for (int i = 0; i < 128; ++i) {
s[strlen(s)] = 'A';
}
clock_t end = clock();
printf("%lldn", (long long)(end-start));
return 0;
}


On my machine it outputs:



$ gcc test.c && ./a.out
13336
$ gcc -O1 test.c && ./a.out
199004
$ gcc -O2 test.c && ./a.out
83415
$ gcc -O3 test.c && ./a.out
83415


Somehow, enabling optimizations causes it to execute longer.










share|improve this question

























  • With gcc-8.2 debug version takes 51334, release 8246. Release compiler options -O3 -march=native -DNDEBUG

    – Maxim Egorushkin
    2 hours ago













  • Please report it to gcc's bugzilla.

    – Marc Glisse
    1 hour ago











  • Using -fno-builtin makes the problem go away. So presumably the issue is that in this particular instance, GCC's builtin strlen is slower than the library's.

    – David Schwartz
    1 hour ago











  • It is generating repnz scasb for strlen at -O1.

    – Marc Glisse
    1 hour ago













  • @MarcGlisse and for -O2 and -O3, it's loading and comparing the chars as integers. Unfortunately, the naive -O0 uses the library function which uses vector-instructions that beat this optimization easily.

    – EOF
    1 hour ago




















7















I wanted to benchmark glibc's strlen function for some reason and found out it apparently performs much slower with optimizations enabled in GCC and I have no idea why.



Here's my code:



#include <time.h>
#include <string.h>
#include <stdlib.h>
#include <stdio.h>

int main() {
char *s = calloc(1 << 20, 1);
memset(s, 65, 1000000);
clock_t start = clock();
for (int i = 0; i < 128; ++i) {
s[strlen(s)] = 'A';
}
clock_t end = clock();
printf("%lldn", (long long)(end-start));
return 0;
}


On my machine it outputs:



$ gcc test.c && ./a.out
13336
$ gcc -O1 test.c && ./a.out
199004
$ gcc -O2 test.c && ./a.out
83415
$ gcc -O3 test.c && ./a.out
83415


Somehow, enabling optimizations causes it to execute longer.










share|improve this question

























  • With gcc-8.2 debug version takes 51334, release 8246. Release compiler options -O3 -march=native -DNDEBUG

    – Maxim Egorushkin
    2 hours ago













  • Please report it to gcc's bugzilla.

    – Marc Glisse
    1 hour ago











  • Using -fno-builtin makes the problem go away. So presumably the issue is that in this particular instance, GCC's builtin strlen is slower than the library's.

    – David Schwartz
    1 hour ago











  • It is generating repnz scasb for strlen at -O1.

    – Marc Glisse
    1 hour ago













  • @MarcGlisse and for -O2 and -O3, it's loading and comparing the chars as integers. Unfortunately, the naive -O0 uses the library function which uses vector-instructions that beat this optimization easily.

    – EOF
    1 hour ago
















7












7








7








I wanted to benchmark glibc's strlen function for some reason and found out it apparently performs much slower with optimizations enabled in GCC and I have no idea why.



Here's my code:



#include <time.h>
#include <string.h>
#include <stdlib.h>
#include <stdio.h>

int main() {
char *s = calloc(1 << 20, 1);
memset(s, 65, 1000000);
clock_t start = clock();
for (int i = 0; i < 128; ++i) {
s[strlen(s)] = 'A';
}
clock_t end = clock();
printf("%lldn", (long long)(end-start));
return 0;
}


On my machine it outputs:



$ gcc test.c && ./a.out
13336
$ gcc -O1 test.c && ./a.out
199004
$ gcc -O2 test.c && ./a.out
83415
$ gcc -O3 test.c && ./a.out
83415


Somehow, enabling optimizations causes it to execute longer.










share|improve this question
















I wanted to benchmark glibc's strlen function for some reason and found out it apparently performs much slower with optimizations enabled in GCC and I have no idea why.



Here's my code:



#include <time.h>
#include <string.h>
#include <stdlib.h>
#include <stdio.h>

int main() {
char *s = calloc(1 << 20, 1);
memset(s, 65, 1000000);
clock_t start = clock();
for (int i = 0; i < 128; ++i) {
s[strlen(s)] = 'A';
}
clock_t end = clock();
printf("%lldn", (long long)(end-start));
return 0;
}


On my machine it outputs:



$ gcc test.c && ./a.out
13336
$ gcc -O1 test.c && ./a.out
199004
$ gcc -O2 test.c && ./a.out
83415
$ gcc -O3 test.c && ./a.out
83415


Somehow, enabling optimizations causes it to execute longer.







c gcc glibc






share|improve this question















share|improve this question













share|improve this question




share|improve this question








edited 2 hours ago









Fei Xiang

2,1634822




2,1634822










asked 2 hours ago









TsarNTsarN

3814




3814













  • With gcc-8.2 debug version takes 51334, release 8246. Release compiler options -O3 -march=native -DNDEBUG

    – Maxim Egorushkin
    2 hours ago













  • Please report it to gcc's bugzilla.

    – Marc Glisse
    1 hour ago











  • Using -fno-builtin makes the problem go away. So presumably the issue is that in this particular instance, GCC's builtin strlen is slower than the library's.

    – David Schwartz
    1 hour ago











  • It is generating repnz scasb for strlen at -O1.

    – Marc Glisse
    1 hour ago













  • @MarcGlisse and for -O2 and -O3, it's loading and comparing the chars as integers. Unfortunately, the naive -O0 uses the library function which uses vector-instructions that beat this optimization easily.

    – EOF
    1 hour ago





















  • With gcc-8.2 debug version takes 51334, release 8246. Release compiler options -O3 -march=native -DNDEBUG

    – Maxim Egorushkin
    2 hours ago













  • Please report it to gcc's bugzilla.

    – Marc Glisse
    1 hour ago











  • Using -fno-builtin makes the problem go away. So presumably the issue is that in this particular instance, GCC's builtin strlen is slower than the library's.

    – David Schwartz
    1 hour ago











  • It is generating repnz scasb for strlen at -O1.

    – Marc Glisse
    1 hour ago













  • @MarcGlisse and for -O2 and -O3, it's loading and comparing the chars as integers. Unfortunately, the naive -O0 uses the library function which uses vector-instructions that beat this optimization easily.

    – EOF
    1 hour ago



















With gcc-8.2 debug version takes 51334, release 8246. Release compiler options -O3 -march=native -DNDEBUG

– Maxim Egorushkin
2 hours ago







With gcc-8.2 debug version takes 51334, release 8246. Release compiler options -O3 -march=native -DNDEBUG

– Maxim Egorushkin
2 hours ago















Please report it to gcc's bugzilla.

– Marc Glisse
1 hour ago





Please report it to gcc's bugzilla.

– Marc Glisse
1 hour ago













Using -fno-builtin makes the problem go away. So presumably the issue is that in this particular instance, GCC's builtin strlen is slower than the library's.

– David Schwartz
1 hour ago





Using -fno-builtin makes the problem go away. So presumably the issue is that in this particular instance, GCC's builtin strlen is slower than the library's.

– David Schwartz
1 hour ago













It is generating repnz scasb for strlen at -O1.

– Marc Glisse
1 hour ago







It is generating repnz scasb for strlen at -O1.

– Marc Glisse
1 hour ago















@MarcGlisse and for -O2 and -O3, it's loading and comparing the chars as integers. Unfortunately, the naive -O0 uses the library function which uses vector-instructions that beat this optimization easily.

– EOF
1 hour ago







@MarcGlisse and for -O2 and -O3, it's loading and comparing the chars as integers. Unfortunately, the naive -O0 uses the library function which uses vector-instructions that beat this optimization easily.

– EOF
1 hour ago














1 Answer
1






active

oldest

votes


















5














Testing your code on Godbolt's Compiler Explorer provides this explanation:




  • at -O0 or without optimisations, the generated code call the C library function strlen

  • at -O1 the generated code uses a simple inline expansion using a rep scasb instruction.

  • at -O2 and above, the generated code uses a more elaborate inline expansion.


Benchmarking your code repeatedly shows a substantial variation from one run to another, but increasing the number of iterations shows that:




  • the -O1 code is much slower than the C library implementation: 32240 vs 3090

  • the -O2 code is faster than the -O1 but still substantially slower than the C ibrary code: 8570 vs 3090.


This behavior is specific to gcc and the glibc. The same test on OS/X with clang and Apple's Libc does not show a significant difference, which is not a surprise as Godbolt shows that clang generates a call to the C library strlen at all optimisation levels.



This could be considered a bug in gcc/glibc but more extensive benchmarking might show that the overhead of calling strlen has a more important impact than the lack of performance of the inline code for small strings. The strings on which you benchmark are uncommonly large, so focusing the benchmark on ultra-long strings might not give meaningful results.



I updated the benchmark for smaller strings and it shows similar performance for string lengths varying from 0 to 100 at -O0 and -O2 but still a much worse performance at -O1, 3 times slower.



Here is the updated code:



#include <stdlib.h>
#include <string.h>
#include <time.h>

void benchmark(int repeat, int minlen, int maxlen) {
char *s = malloc(maxlen + 1);
memset(s, 'A', minlen);
long long bytes = 0, calls = 0;
clock_t clk = clock();
for (int n = 0; n < repeat; n++) {
for (int i = minlen; i < maxlen; ++i) {
bytes += i + 1;
calls += 1;
s[i] = '';
s[strlen(s)] = 'A';
}
}
clk = clock() - clk;
free(s);
double avglen = (minlen + maxlen - 1) / 2.0;
double ns = (double)clk * 1e9 / CLOCKS_PER_SEC;
printf("average length %7.0f -> avg time: %7.3f ns/byte, %7.3f ns/calln",
avglen, ns / bytes, ns / calls);
}

int main() {
benchmark(10000000, 0, 1);
benchmark(1000000, 0, 10);
benchmark(1000000, 5, 15);
benchmark(100000, 0, 100);
benchmark(100000, 50, 150);
benchmark(10000, 0, 1000);
benchmark(10000, 500, 1500);
benchmark(1000, 0, 10000);
benchmark(1000, 5000, 15000);
benchmark(100, 1000000 - 50, 1000000 + 50);
return 0;
}


Here is the output:




chqrlie> gcc -std=c99 -O0 benchstrlen.c && ./a.out
average length 0 -> avg time: 14.000 ns/byte, 14.000 ns/call
average length 4 -> avg time: 2.364 ns/byte, 13.000 ns/call
average length 10 -> avg time: 1.238 ns/byte, 13.000 ns/call
average length 50 -> avg time: 0.317 ns/byte, 16.000 ns/call
average length 100 -> avg time: 0.169 ns/byte, 17.000 ns/call
average length 500 -> avg time: 0.074 ns/byte, 37.000 ns/call
average length 1000 -> avg time: 0.068 ns/byte, 68.000 ns/call
average length 5000 -> avg time: 0.064 ns/byte, 318.000 ns/call
average length 10000 -> avg time: 0.062 ns/byte, 622.000 ns/call
average length 1000000 -> avg time: 0.062 ns/byte, 62000.000 ns/call
chqrlie> gcc -std=c99 -O1 benchstrlen.c && ./a.out
average length 0 -> avg time: 20.000 ns/byte, 20.000 ns/call
average length 4 -> avg time: 3.818 ns/byte, 21.000 ns/call
average length 10 -> avg time: 2.190 ns/byte, 23.000 ns/call
average length 50 -> avg time: 0.990 ns/byte, 50.000 ns/call
average length 100 -> avg time: 0.816 ns/byte, 82.000 ns/call
average length 500 -> avg time: 0.679 ns/byte, 340.000 ns/call
average length 1000 -> avg time: 0.664 ns/byte, 664.000 ns/call
average length 5000 -> avg time: 0.651 ns/byte, 3254.000 ns/call
average length 10000 -> avg time: 0.649 ns/byte, 6491.000 ns/call
average length 1000000 -> avg time: 0.648 ns/byte, 648000.000 ns/call
chqrlie> gcc -std=c99 -O2 benchstrlen.c && ./a.out
average length 0 -> avg time: 10.000 ns/byte, 10.000 ns/call
average length 4 -> avg time: 2.000 ns/byte, 11.000 ns/call
average length 10 -> avg time: 1.048 ns/byte, 11.000 ns/call
average length 50 -> avg time: 0.337 ns/byte, 17.000 ns/call
average length 100 -> avg time: 0.299 ns/byte, 30.000 ns/call
average length 500 -> avg time: 0.202 ns/byte, 101.000 ns/call
average length 1000 -> avg time: 0.188 ns/byte, 188.000 ns/call
average length 5000 -> avg time: 0.174 ns/byte, 868.000 ns/call
average length 10000 -> avg time: 0.172 ns/byte, 1716.000 ns/call
average length 1000000 -> avg time: 0.172 ns/byte, 172000.000 ns/call





share|improve this answer


























  • Wouldn't it still be better for the inlined version to use the same optimizations as the library strlen, giving the best of both worlds?

    – Daniel H
    1 hour ago






  • 1





    It would, but the hand optimized version in the C library might be larger and more complicated to inline. I have not looked into this recently, but there used to be a mix of complex platform specific macros in <string.h> and hard coded optimisations in the gcc code generator. Definitely still room for improvement on intel targets.

    – chqrlie
    1 hour ago











  • Does it change if you use -march=native -mtune=native?

    – Deduplicator
    10 mins ago











  • Note that the GNU C library function for strlen() is likely optimised for extremely large strings (that no sane programmer will care about) at the expense of performance for small strings (that are extremely common); and the optimisations done by the library version should never be done. The problem here is that the OP's code doesn't keep track of the string's length itself (e.g. with an int len; variable) and should not have used strlen() at all, making the code so bad for performance that "optimised for something no sane programmer would care about" actually helped.

    – Brendan
    7 mins ago














Your Answer






StackExchange.ifUsing("editor", function () {
StackExchange.using("externalEditor", function () {
StackExchange.using("snippets", function () {
StackExchange.snippets.init();
});
});
}, "code-snippets");

StackExchange.ready(function() {
var channelOptions = {
tags: "".split(" "),
id: "1"
};
initTagRenderer("".split(" "), "".split(" "), channelOptions);

StackExchange.using("externalEditor", function() {
// Have to fire editor after snippets, if snippets enabled
if (StackExchange.settings.snippets.snippetsEnabled) {
StackExchange.using("snippets", function() {
createEditor();
});
}
else {
createEditor();
}
});

function createEditor() {
StackExchange.prepareEditor({
heartbeatType: 'answer',
autoActivateHeartbeat: false,
convertImagesToLinks: true,
noModals: true,
showLowRepImageUploadWarning: true,
reputationToPostImages: 10,
bindNavPrevention: true,
postfix: "",
imageUploader: {
brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
allowUrls: true
},
onDemand: true,
discardSelector: ".discard-answer"
,immediatelyShowMarkdownHelp:true
});


}
});














draft saved

draft discarded


















StackExchange.ready(
function () {
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fstackoverflow.com%2fquestions%2f55563598%2fwhy-is-this-code-6-5x-slower-with-optimizations-enabled%23new-answer', 'question_page');
}
);

Post as a guest















Required, but never shown

























1 Answer
1






active

oldest

votes








1 Answer
1






active

oldest

votes









active

oldest

votes






active

oldest

votes









5














Testing your code on Godbolt's Compiler Explorer provides this explanation:




  • at -O0 or without optimisations, the generated code call the C library function strlen

  • at -O1 the generated code uses a simple inline expansion using a rep scasb instruction.

  • at -O2 and above, the generated code uses a more elaborate inline expansion.


Benchmarking your code repeatedly shows a substantial variation from one run to another, but increasing the number of iterations shows that:




  • the -O1 code is much slower than the C library implementation: 32240 vs 3090

  • the -O2 code is faster than the -O1 but still substantially slower than the C ibrary code: 8570 vs 3090.


This behavior is specific to gcc and the glibc. The same test on OS/X with clang and Apple's Libc does not show a significant difference, which is not a surprise as Godbolt shows that clang generates a call to the C library strlen at all optimisation levels.



This could be considered a bug in gcc/glibc but more extensive benchmarking might show that the overhead of calling strlen has a more important impact than the lack of performance of the inline code for small strings. The strings on which you benchmark are uncommonly large, so focusing the benchmark on ultra-long strings might not give meaningful results.



I updated the benchmark for smaller strings and it shows similar performance for string lengths varying from 0 to 100 at -O0 and -O2 but still a much worse performance at -O1, 3 times slower.



Here is the updated code:



#include <stdlib.h>
#include <string.h>
#include <time.h>

void benchmark(int repeat, int minlen, int maxlen) {
char *s = malloc(maxlen + 1);
memset(s, 'A', minlen);
long long bytes = 0, calls = 0;
clock_t clk = clock();
for (int n = 0; n < repeat; n++) {
for (int i = minlen; i < maxlen; ++i) {
bytes += i + 1;
calls += 1;
s[i] = '';
s[strlen(s)] = 'A';
}
}
clk = clock() - clk;
free(s);
double avglen = (minlen + maxlen - 1) / 2.0;
double ns = (double)clk * 1e9 / CLOCKS_PER_SEC;
printf("average length %7.0f -> avg time: %7.3f ns/byte, %7.3f ns/calln",
avglen, ns / bytes, ns / calls);
}

int main() {
benchmark(10000000, 0, 1);
benchmark(1000000, 0, 10);
benchmark(1000000, 5, 15);
benchmark(100000, 0, 100);
benchmark(100000, 50, 150);
benchmark(10000, 0, 1000);
benchmark(10000, 500, 1500);
benchmark(1000, 0, 10000);
benchmark(1000, 5000, 15000);
benchmark(100, 1000000 - 50, 1000000 + 50);
return 0;
}


Here is the output:




chqrlie> gcc -std=c99 -O0 benchstrlen.c && ./a.out
average length 0 -> avg time: 14.000 ns/byte, 14.000 ns/call
average length 4 -> avg time: 2.364 ns/byte, 13.000 ns/call
average length 10 -> avg time: 1.238 ns/byte, 13.000 ns/call
average length 50 -> avg time: 0.317 ns/byte, 16.000 ns/call
average length 100 -> avg time: 0.169 ns/byte, 17.000 ns/call
average length 500 -> avg time: 0.074 ns/byte, 37.000 ns/call
average length 1000 -> avg time: 0.068 ns/byte, 68.000 ns/call
average length 5000 -> avg time: 0.064 ns/byte, 318.000 ns/call
average length 10000 -> avg time: 0.062 ns/byte, 622.000 ns/call
average length 1000000 -> avg time: 0.062 ns/byte, 62000.000 ns/call
chqrlie> gcc -std=c99 -O1 benchstrlen.c && ./a.out
average length 0 -> avg time: 20.000 ns/byte, 20.000 ns/call
average length 4 -> avg time: 3.818 ns/byte, 21.000 ns/call
average length 10 -> avg time: 2.190 ns/byte, 23.000 ns/call
average length 50 -> avg time: 0.990 ns/byte, 50.000 ns/call
average length 100 -> avg time: 0.816 ns/byte, 82.000 ns/call
average length 500 -> avg time: 0.679 ns/byte, 340.000 ns/call
average length 1000 -> avg time: 0.664 ns/byte, 664.000 ns/call
average length 5000 -> avg time: 0.651 ns/byte, 3254.000 ns/call
average length 10000 -> avg time: 0.649 ns/byte, 6491.000 ns/call
average length 1000000 -> avg time: 0.648 ns/byte, 648000.000 ns/call
chqrlie> gcc -std=c99 -O2 benchstrlen.c && ./a.out
average length 0 -> avg time: 10.000 ns/byte, 10.000 ns/call
average length 4 -> avg time: 2.000 ns/byte, 11.000 ns/call
average length 10 -> avg time: 1.048 ns/byte, 11.000 ns/call
average length 50 -> avg time: 0.337 ns/byte, 17.000 ns/call
average length 100 -> avg time: 0.299 ns/byte, 30.000 ns/call
average length 500 -> avg time: 0.202 ns/byte, 101.000 ns/call
average length 1000 -> avg time: 0.188 ns/byte, 188.000 ns/call
average length 5000 -> avg time: 0.174 ns/byte, 868.000 ns/call
average length 10000 -> avg time: 0.172 ns/byte, 1716.000 ns/call
average length 1000000 -> avg time: 0.172 ns/byte, 172000.000 ns/call





share|improve this answer


























  • Wouldn't it still be better for the inlined version to use the same optimizations as the library strlen, giving the best of both worlds?

    – Daniel H
    1 hour ago






  • 1





    It would, but the hand optimized version in the C library might be larger and more complicated to inline. I have not looked into this recently, but there used to be a mix of complex platform specific macros in <string.h> and hard coded optimisations in the gcc code generator. Definitely still room for improvement on intel targets.

    – chqrlie
    1 hour ago











  • Does it change if you use -march=native -mtune=native?

    – Deduplicator
    10 mins ago











  • Note that the GNU C library function for strlen() is likely optimised for extremely large strings (that no sane programmer will care about) at the expense of performance for small strings (that are extremely common); and the optimisations done by the library version should never be done. The problem here is that the OP's code doesn't keep track of the string's length itself (e.g. with an int len; variable) and should not have used strlen() at all, making the code so bad for performance that "optimised for something no sane programmer would care about" actually helped.

    – Brendan
    7 mins ago


















5














Testing your code on Godbolt's Compiler Explorer provides this explanation:




  • at -O0 or without optimisations, the generated code call the C library function strlen

  • at -O1 the generated code uses a simple inline expansion using a rep scasb instruction.

  • at -O2 and above, the generated code uses a more elaborate inline expansion.


Benchmarking your code repeatedly shows a substantial variation from one run to another, but increasing the number of iterations shows that:




  • the -O1 code is much slower than the C library implementation: 32240 vs 3090

  • the -O2 code is faster than the -O1 but still substantially slower than the C ibrary code: 8570 vs 3090.


This behavior is specific to gcc and the glibc. The same test on OS/X with clang and Apple's Libc does not show a significant difference, which is not a surprise as Godbolt shows that clang generates a call to the C library strlen at all optimisation levels.



This could be considered a bug in gcc/glibc but more extensive benchmarking might show that the overhead of calling strlen has a more important impact than the lack of performance of the inline code for small strings. The strings on which you benchmark are uncommonly large, so focusing the benchmark on ultra-long strings might not give meaningful results.



I updated the benchmark for smaller strings and it shows similar performance for string lengths varying from 0 to 100 at -O0 and -O2 but still a much worse performance at -O1, 3 times slower.



Here is the updated code:



#include <stdlib.h>
#include <string.h>
#include <time.h>

void benchmark(int repeat, int minlen, int maxlen) {
char *s = malloc(maxlen + 1);
memset(s, 'A', minlen);
long long bytes = 0, calls = 0;
clock_t clk = clock();
for (int n = 0; n < repeat; n++) {
for (int i = minlen; i < maxlen; ++i) {
bytes += i + 1;
calls += 1;
s[i] = '';
s[strlen(s)] = 'A';
}
}
clk = clock() - clk;
free(s);
double avglen = (minlen + maxlen - 1) / 2.0;
double ns = (double)clk * 1e9 / CLOCKS_PER_SEC;
printf("average length %7.0f -> avg time: %7.3f ns/byte, %7.3f ns/calln",
avglen, ns / bytes, ns / calls);
}

int main() {
benchmark(10000000, 0, 1);
benchmark(1000000, 0, 10);
benchmark(1000000, 5, 15);
benchmark(100000, 0, 100);
benchmark(100000, 50, 150);
benchmark(10000, 0, 1000);
benchmark(10000, 500, 1500);
benchmark(1000, 0, 10000);
benchmark(1000, 5000, 15000);
benchmark(100, 1000000 - 50, 1000000 + 50);
return 0;
}


Here is the output:




chqrlie> gcc -std=c99 -O0 benchstrlen.c && ./a.out
average length 0 -> avg time: 14.000 ns/byte, 14.000 ns/call
average length 4 -> avg time: 2.364 ns/byte, 13.000 ns/call
average length 10 -> avg time: 1.238 ns/byte, 13.000 ns/call
average length 50 -> avg time: 0.317 ns/byte, 16.000 ns/call
average length 100 -> avg time: 0.169 ns/byte, 17.000 ns/call
average length 500 -> avg time: 0.074 ns/byte, 37.000 ns/call
average length 1000 -> avg time: 0.068 ns/byte, 68.000 ns/call
average length 5000 -> avg time: 0.064 ns/byte, 318.000 ns/call
average length 10000 -> avg time: 0.062 ns/byte, 622.000 ns/call
average length 1000000 -> avg time: 0.062 ns/byte, 62000.000 ns/call
chqrlie> gcc -std=c99 -O1 benchstrlen.c && ./a.out
average length 0 -> avg time: 20.000 ns/byte, 20.000 ns/call
average length 4 -> avg time: 3.818 ns/byte, 21.000 ns/call
average length 10 -> avg time: 2.190 ns/byte, 23.000 ns/call
average length 50 -> avg time: 0.990 ns/byte, 50.000 ns/call
average length 100 -> avg time: 0.816 ns/byte, 82.000 ns/call
average length 500 -> avg time: 0.679 ns/byte, 340.000 ns/call
average length 1000 -> avg time: 0.664 ns/byte, 664.000 ns/call
average length 5000 -> avg time: 0.651 ns/byte, 3254.000 ns/call
average length 10000 -> avg time: 0.649 ns/byte, 6491.000 ns/call
average length 1000000 -> avg time: 0.648 ns/byte, 648000.000 ns/call
chqrlie> gcc -std=c99 -O2 benchstrlen.c && ./a.out
average length 0 -> avg time: 10.000 ns/byte, 10.000 ns/call
average length 4 -> avg time: 2.000 ns/byte, 11.000 ns/call
average length 10 -> avg time: 1.048 ns/byte, 11.000 ns/call
average length 50 -> avg time: 0.337 ns/byte, 17.000 ns/call
average length 100 -> avg time: 0.299 ns/byte, 30.000 ns/call
average length 500 -> avg time: 0.202 ns/byte, 101.000 ns/call
average length 1000 -> avg time: 0.188 ns/byte, 188.000 ns/call
average length 5000 -> avg time: 0.174 ns/byte, 868.000 ns/call
average length 10000 -> avg time: 0.172 ns/byte, 1716.000 ns/call
average length 1000000 -> avg time: 0.172 ns/byte, 172000.000 ns/call





share|improve this answer


























  • Wouldn't it still be better for the inlined version to use the same optimizations as the library strlen, giving the best of both worlds?

    – Daniel H
    1 hour ago






  • 1





    It would, but the hand optimized version in the C library might be larger and more complicated to inline. I have not looked into this recently, but there used to be a mix of complex platform specific macros in <string.h> and hard coded optimisations in the gcc code generator. Definitely still room for improvement on intel targets.

    – chqrlie
    1 hour ago











  • Does it change if you use -march=native -mtune=native?

    – Deduplicator
    10 mins ago











  • Note that the GNU C library function for strlen() is likely optimised for extremely large strings (that no sane programmer will care about) at the expense of performance for small strings (that are extremely common); and the optimisations done by the library version should never be done. The problem here is that the OP's code doesn't keep track of the string's length itself (e.g. with an int len; variable) and should not have used strlen() at all, making the code so bad for performance that "optimised for something no sane programmer would care about" actually helped.

    – Brendan
    7 mins ago
















5












5








5







Testing your code on Godbolt's Compiler Explorer provides this explanation:




  • at -O0 or without optimisations, the generated code call the C library function strlen

  • at -O1 the generated code uses a simple inline expansion using a rep scasb instruction.

  • at -O2 and above, the generated code uses a more elaborate inline expansion.


Benchmarking your code repeatedly shows a substantial variation from one run to another, but increasing the number of iterations shows that:




  • the -O1 code is much slower than the C library implementation: 32240 vs 3090

  • the -O2 code is faster than the -O1 but still substantially slower than the C ibrary code: 8570 vs 3090.


This behavior is specific to gcc and the glibc. The same test on OS/X with clang and Apple's Libc does not show a significant difference, which is not a surprise as Godbolt shows that clang generates a call to the C library strlen at all optimisation levels.



This could be considered a bug in gcc/glibc but more extensive benchmarking might show that the overhead of calling strlen has a more important impact than the lack of performance of the inline code for small strings. The strings on which you benchmark are uncommonly large, so focusing the benchmark on ultra-long strings might not give meaningful results.



I updated the benchmark for smaller strings and it shows similar performance for string lengths varying from 0 to 100 at -O0 and -O2 but still a much worse performance at -O1, 3 times slower.



Here is the updated code:



#include <stdlib.h>
#include <string.h>
#include <time.h>

void benchmark(int repeat, int minlen, int maxlen) {
char *s = malloc(maxlen + 1);
memset(s, 'A', minlen);
long long bytes = 0, calls = 0;
clock_t clk = clock();
for (int n = 0; n < repeat; n++) {
for (int i = minlen; i < maxlen; ++i) {
bytes += i + 1;
calls += 1;
s[i] = '';
s[strlen(s)] = 'A';
}
}
clk = clock() - clk;
free(s);
double avglen = (minlen + maxlen - 1) / 2.0;
double ns = (double)clk * 1e9 / CLOCKS_PER_SEC;
printf("average length %7.0f -> avg time: %7.3f ns/byte, %7.3f ns/calln",
avglen, ns / bytes, ns / calls);
}

int main() {
benchmark(10000000, 0, 1);
benchmark(1000000, 0, 10);
benchmark(1000000, 5, 15);
benchmark(100000, 0, 100);
benchmark(100000, 50, 150);
benchmark(10000, 0, 1000);
benchmark(10000, 500, 1500);
benchmark(1000, 0, 10000);
benchmark(1000, 5000, 15000);
benchmark(100, 1000000 - 50, 1000000 + 50);
return 0;
}


Here is the output:




chqrlie> gcc -std=c99 -O0 benchstrlen.c && ./a.out
average length 0 -> avg time: 14.000 ns/byte, 14.000 ns/call
average length 4 -> avg time: 2.364 ns/byte, 13.000 ns/call
average length 10 -> avg time: 1.238 ns/byte, 13.000 ns/call
average length 50 -> avg time: 0.317 ns/byte, 16.000 ns/call
average length 100 -> avg time: 0.169 ns/byte, 17.000 ns/call
average length 500 -> avg time: 0.074 ns/byte, 37.000 ns/call
average length 1000 -> avg time: 0.068 ns/byte, 68.000 ns/call
average length 5000 -> avg time: 0.064 ns/byte, 318.000 ns/call
average length 10000 -> avg time: 0.062 ns/byte, 622.000 ns/call
average length 1000000 -> avg time: 0.062 ns/byte, 62000.000 ns/call
chqrlie> gcc -std=c99 -O1 benchstrlen.c && ./a.out
average length 0 -> avg time: 20.000 ns/byte, 20.000 ns/call
average length 4 -> avg time: 3.818 ns/byte, 21.000 ns/call
average length 10 -> avg time: 2.190 ns/byte, 23.000 ns/call
average length 50 -> avg time: 0.990 ns/byte, 50.000 ns/call
average length 100 -> avg time: 0.816 ns/byte, 82.000 ns/call
average length 500 -> avg time: 0.679 ns/byte, 340.000 ns/call
average length 1000 -> avg time: 0.664 ns/byte, 664.000 ns/call
average length 5000 -> avg time: 0.651 ns/byte, 3254.000 ns/call
average length 10000 -> avg time: 0.649 ns/byte, 6491.000 ns/call
average length 1000000 -> avg time: 0.648 ns/byte, 648000.000 ns/call
chqrlie> gcc -std=c99 -O2 benchstrlen.c && ./a.out
average length 0 -> avg time: 10.000 ns/byte, 10.000 ns/call
average length 4 -> avg time: 2.000 ns/byte, 11.000 ns/call
average length 10 -> avg time: 1.048 ns/byte, 11.000 ns/call
average length 50 -> avg time: 0.337 ns/byte, 17.000 ns/call
average length 100 -> avg time: 0.299 ns/byte, 30.000 ns/call
average length 500 -> avg time: 0.202 ns/byte, 101.000 ns/call
average length 1000 -> avg time: 0.188 ns/byte, 188.000 ns/call
average length 5000 -> avg time: 0.174 ns/byte, 868.000 ns/call
average length 10000 -> avg time: 0.172 ns/byte, 1716.000 ns/call
average length 1000000 -> avg time: 0.172 ns/byte, 172000.000 ns/call





share|improve this answer















Testing your code on Godbolt's Compiler Explorer provides this explanation:




  • at -O0 or without optimisations, the generated code call the C library function strlen

  • at -O1 the generated code uses a simple inline expansion using a rep scasb instruction.

  • at -O2 and above, the generated code uses a more elaborate inline expansion.


Benchmarking your code repeatedly shows a substantial variation from one run to another, but increasing the number of iterations shows that:




  • the -O1 code is much slower than the C library implementation: 32240 vs 3090

  • the -O2 code is faster than the -O1 but still substantially slower than the C ibrary code: 8570 vs 3090.


This behavior is specific to gcc and the glibc. The same test on OS/X with clang and Apple's Libc does not show a significant difference, which is not a surprise as Godbolt shows that clang generates a call to the C library strlen at all optimisation levels.



This could be considered a bug in gcc/glibc but more extensive benchmarking might show that the overhead of calling strlen has a more important impact than the lack of performance of the inline code for small strings. The strings on which you benchmark are uncommonly large, so focusing the benchmark on ultra-long strings might not give meaningful results.



I updated the benchmark for smaller strings and it shows similar performance for string lengths varying from 0 to 100 at -O0 and -O2 but still a much worse performance at -O1, 3 times slower.



Here is the updated code:



#include <stdlib.h>
#include <string.h>
#include <time.h>

void benchmark(int repeat, int minlen, int maxlen) {
char *s = malloc(maxlen + 1);
memset(s, 'A', minlen);
long long bytes = 0, calls = 0;
clock_t clk = clock();
for (int n = 0; n < repeat; n++) {
for (int i = minlen; i < maxlen; ++i) {
bytes += i + 1;
calls += 1;
s[i] = '';
s[strlen(s)] = 'A';
}
}
clk = clock() - clk;
free(s);
double avglen = (minlen + maxlen - 1) / 2.0;
double ns = (double)clk * 1e9 / CLOCKS_PER_SEC;
printf("average length %7.0f -> avg time: %7.3f ns/byte, %7.3f ns/calln",
avglen, ns / bytes, ns / calls);
}

int main() {
benchmark(10000000, 0, 1);
benchmark(1000000, 0, 10);
benchmark(1000000, 5, 15);
benchmark(100000, 0, 100);
benchmark(100000, 50, 150);
benchmark(10000, 0, 1000);
benchmark(10000, 500, 1500);
benchmark(1000, 0, 10000);
benchmark(1000, 5000, 15000);
benchmark(100, 1000000 - 50, 1000000 + 50);
return 0;
}


Here is the output:




chqrlie> gcc -std=c99 -O0 benchstrlen.c && ./a.out
average length 0 -> avg time: 14.000 ns/byte, 14.000 ns/call
average length 4 -> avg time: 2.364 ns/byte, 13.000 ns/call
average length 10 -> avg time: 1.238 ns/byte, 13.000 ns/call
average length 50 -> avg time: 0.317 ns/byte, 16.000 ns/call
average length 100 -> avg time: 0.169 ns/byte, 17.000 ns/call
average length 500 -> avg time: 0.074 ns/byte, 37.000 ns/call
average length 1000 -> avg time: 0.068 ns/byte, 68.000 ns/call
average length 5000 -> avg time: 0.064 ns/byte, 318.000 ns/call
average length 10000 -> avg time: 0.062 ns/byte, 622.000 ns/call
average length 1000000 -> avg time: 0.062 ns/byte, 62000.000 ns/call
chqrlie> gcc -std=c99 -O1 benchstrlen.c && ./a.out
average length 0 -> avg time: 20.000 ns/byte, 20.000 ns/call
average length 4 -> avg time: 3.818 ns/byte, 21.000 ns/call
average length 10 -> avg time: 2.190 ns/byte, 23.000 ns/call
average length 50 -> avg time: 0.990 ns/byte, 50.000 ns/call
average length 100 -> avg time: 0.816 ns/byte, 82.000 ns/call
average length 500 -> avg time: 0.679 ns/byte, 340.000 ns/call
average length 1000 -> avg time: 0.664 ns/byte, 664.000 ns/call
average length 5000 -> avg time: 0.651 ns/byte, 3254.000 ns/call
average length 10000 -> avg time: 0.649 ns/byte, 6491.000 ns/call
average length 1000000 -> avg time: 0.648 ns/byte, 648000.000 ns/call
chqrlie> gcc -std=c99 -O2 benchstrlen.c && ./a.out
average length 0 -> avg time: 10.000 ns/byte, 10.000 ns/call
average length 4 -> avg time: 2.000 ns/byte, 11.000 ns/call
average length 10 -> avg time: 1.048 ns/byte, 11.000 ns/call
average length 50 -> avg time: 0.337 ns/byte, 17.000 ns/call
average length 100 -> avg time: 0.299 ns/byte, 30.000 ns/call
average length 500 -> avg time: 0.202 ns/byte, 101.000 ns/call
average length 1000 -> avg time: 0.188 ns/byte, 188.000 ns/call
average length 5000 -> avg time: 0.174 ns/byte, 868.000 ns/call
average length 10000 -> avg time: 0.172 ns/byte, 1716.000 ns/call
average length 1000000 -> avg time: 0.172 ns/byte, 172000.000 ns/call






share|improve this answer














share|improve this answer



share|improve this answer








edited 20 mins ago

























answered 1 hour ago









chqrliechqrlie

62.9k848107




62.9k848107













  • Wouldn't it still be better for the inlined version to use the same optimizations as the library strlen, giving the best of both worlds?

    – Daniel H
    1 hour ago






  • 1





    It would, but the hand optimized version in the C library might be larger and more complicated to inline. I have not looked into this recently, but there used to be a mix of complex platform specific macros in <string.h> and hard coded optimisations in the gcc code generator. Definitely still room for improvement on intel targets.

    – chqrlie
    1 hour ago











  • Does it change if you use -march=native -mtune=native?

    – Deduplicator
    10 mins ago











  • Note that the GNU C library function for strlen() is likely optimised for extremely large strings (that no sane programmer will care about) at the expense of performance for small strings (that are extremely common); and the optimisations done by the library version should never be done. The problem here is that the OP's code doesn't keep track of the string's length itself (e.g. with an int len; variable) and should not have used strlen() at all, making the code so bad for performance that "optimised for something no sane programmer would care about" actually helped.

    – Brendan
    7 mins ago





















  • Wouldn't it still be better for the inlined version to use the same optimizations as the library strlen, giving the best of both worlds?

    – Daniel H
    1 hour ago






  • 1





    It would, but the hand optimized version in the C library might be larger and more complicated to inline. I have not looked into this recently, but there used to be a mix of complex platform specific macros in <string.h> and hard coded optimisations in the gcc code generator. Definitely still room for improvement on intel targets.

    – chqrlie
    1 hour ago











  • Does it change if you use -march=native -mtune=native?

    – Deduplicator
    10 mins ago











  • Note that the GNU C library function for strlen() is likely optimised for extremely large strings (that no sane programmer will care about) at the expense of performance for small strings (that are extremely common); and the optimisations done by the library version should never be done. The problem here is that the OP's code doesn't keep track of the string's length itself (e.g. with an int len; variable) and should not have used strlen() at all, making the code so bad for performance that "optimised for something no sane programmer would care about" actually helped.

    – Brendan
    7 mins ago



















Wouldn't it still be better for the inlined version to use the same optimizations as the library strlen, giving the best of both worlds?

– Daniel H
1 hour ago





Wouldn't it still be better for the inlined version to use the same optimizations as the library strlen, giving the best of both worlds?

– Daniel H
1 hour ago




1




1





It would, but the hand optimized version in the C library might be larger and more complicated to inline. I have not looked into this recently, but there used to be a mix of complex platform specific macros in <string.h> and hard coded optimisations in the gcc code generator. Definitely still room for improvement on intel targets.

– chqrlie
1 hour ago





It would, but the hand optimized version in the C library might be larger and more complicated to inline. I have not looked into this recently, but there used to be a mix of complex platform specific macros in <string.h> and hard coded optimisations in the gcc code generator. Definitely still room for improvement on intel targets.

– chqrlie
1 hour ago













Does it change if you use -march=native -mtune=native?

– Deduplicator
10 mins ago





Does it change if you use -march=native -mtune=native?

– Deduplicator
10 mins ago













Note that the GNU C library function for strlen() is likely optimised for extremely large strings (that no sane programmer will care about) at the expense of performance for small strings (that are extremely common); and the optimisations done by the library version should never be done. The problem here is that the OP's code doesn't keep track of the string's length itself (e.g. with an int len; variable) and should not have used strlen() at all, making the code so bad for performance that "optimised for something no sane programmer would care about" actually helped.

– Brendan
7 mins ago







Note that the GNU C library function for strlen() is likely optimised for extremely large strings (that no sane programmer will care about) at the expense of performance for small strings (that are extremely common); and the optimisations done by the library version should never be done. The problem here is that the OP's code doesn't keep track of the string's length itself (e.g. with an int len; variable) and should not have used strlen() at all, making the code so bad for performance that "optimised for something no sane programmer would care about" actually helped.

– Brendan
7 mins ago






















draft saved

draft discarded




















































Thanks for contributing an answer to Stack Overflow!


  • Please be sure to answer the question. Provide details and share your research!

But avoid



  • Asking for help, clarification, or responding to other answers.

  • Making statements based on opinion; back them up with references or personal experience.


To learn more, see our tips on writing great answers.




draft saved


draft discarded














StackExchange.ready(
function () {
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fstackoverflow.com%2fquestions%2f55563598%2fwhy-is-this-code-6-5x-slower-with-optimizations-enabled%23new-answer', 'question_page');
}
);

Post as a guest















Required, but never shown





















































Required, but never shown














Required, but never shown












Required, but never shown







Required, but never shown

































Required, but never shown














Required, but never shown












Required, but never shown







Required, but never shown







Popular posts from this blog

Gersau Kjelder | Navigasjonsmeny46°59′0″N 8°31′0″E46°59′0″N...

What is the “three and three hundred thousand syndrome”?Who wrote the book Arena?What five creatures were...

Are all UTXOs locked by an address spent in a transaction?UTXO all sent to change address?Signing...