How can compiler optimizations affect on a cache-unfriendly algorithm?
I've noticed an interesting behavior in the code of this question which is also comes from Agner Fog in Optimizing software in C++ and it reduces to how data is accessed and stored in the cache (cache associativity). The explanations is clear for me, but then someone pings about volatile...
That is if we add volatile qualifier to the matrix declaration: volatile int mat[MATSIZE][MATSIZE]; the running time for value 512 dramatically decreases: 2144 → 1562 μs.
As we know volatile prevents compilers from caching the value (in a CPU register) and from optimizing away accesses to that value when they seem unnecessary from the POV of a program.
One possible version assumes that the computation process happens only in RAM and no cpu caches is used in the case of volatile. But on the other hand the run-time for value 513 again is less than for 512: 1490 μs...
Unfortunately I cannot confirm that the volatile version is running faster. I did a test run of both the volatile and the non-volatile version, and the time comparison of both can be seen in the chart below. When measuring performance in order for optimizing code it is always important to take several steps (not just one or two, but in the range of thousands) and take the mode (https://en.wikipedia.org/wiki/Mode_(statistics)) of the gathered data as per Alexandrescu's Fastware.
There are various peaks, and deep valleys, but looking at the chart you cannot conclude that the volatile is faster.
Indeed, the code generated by the compilers is different, but not to such an extent, you can check it out on https://godbolt.org/g/ILw3tg
- Simple frame by frame video decoder library
- GCC no longer implements <varargs.h>
- Contents of IO buffer unknown == unsafe?
- Avoiding strcpy overflow destination warning
- Sort program not working, not sure why
- Fast & accurate atan/arctan approximation algorithm
- What's the difference between strtok_r and strtok_s in C?
- How memory address for pointer to arrays is same as an element in 2D array?
- Which is the best way to suppress "unused variable" warning
- How to use ellipsis in c's case statement?
- How can I exclude non-numeric keys? CS50 Caesar Pset2
- Fast ceiling of an integer division in C / C++
- Is there an invalid pthread_t id?
- How to Implement Universal Setter/Getter Functions for Interrupt-Driven Variables in Embedded C?
- How does SIMD (avx) processing work? for example, if I want 10 32 bit floats how do i fit in a 256 bit avx vector?
- FDCAN problems on STM32G4
- How does the call macro enable mutual recursion between functions f and g in this Hanoi Tower implementation?
- Running test on Rocket core CPU - global variable initialized to 0 is unsuccessful, output wrong value instead
- Interacting with C arrays without knowing the size
- Combination of two strings
- carriage return by fgets
- How to use special characters in C?
- Why does 1.0/100.0 == 0.1/10.0 give True?
- Is it correct to compare pointers in C?
- Force free() to return malloc memory back to OS
- How can I print to standard error in C with 'printf'?
- What is the standard behavior of fread in C on Windows?
- How is strtok removing lines it shouldn't have access to?
- Using array as smart point in C
- Assigning string to malloced 2d char array not working as intended