Single instruction, multiple data (SIMD) is the concept of having each instruction operate on a small chunk or vector of data elements. CPU vector instruction sets include: x86 SSE and AVX, ARM NEON, and PowerPC AltiVec. To efficiently use SIMD instructions, data needs to be in structure-of-arrays form and should occur in longer streams. Naively "SIMD optimized" code frequently surprises by running slower than the original.
Questions tagged [simd]
2540 questions
48
votes
6 answers
AVX2 what is the most efficient way to pack left based on a mask?
If you have an input array, and an output array, but you only want to write those elements which pass a certain condition, what would be the most efficient way to do this in AVX2?
I've seen in SSE where it was done like…
Froglegs
- 1,095
- 1
- 11
- 21
46
votes
2 answers
Why is np.dot so much faster than np.sum?
Why is np.dot so much faster than np.sum? Following this answer we know that np.sum is slow and has faster alternatives.
For example:
In [20]: A = np.random.rand(1000)
In [21]: B = np.random.rand(1000)
In [22]: %timeit np.sum(A)
3.21 µs ± 270 ns…
Simd
- 19,447
- 42
- 136
- 271
43
votes
1 answer
Difference between MOVDQA and MOVAPS x86 instructions?
I'm looking Intel datasheet: Intel® 64 and IA-32 Architectures
Software Developer’s Manual and I can't find the difference between
MOVDQA: Move Aligned Double Quadword
MOVAPS: Move Aligned Packed Single-Precision
In Intel datasheet I can find…
GJ.
- 10,810
- 2
- 45
- 62
41
votes
2 answers
CPU SIMD vs GPU SIMD?
GPU uses the SIMD paradigm, that is, the same portion of code will be executed in parallel, and applied to various elements of a data set.
However, CPU also uses SIMD, and provide instruction-level parallelism. For example, as far as I know,…
Carmellose
- 4,815
- 10
- 38
- 56
41
votes
8 answers
Why is strcmp not SIMD optimized?
I've tried to compile this program on an x64 computer:
#include
int main(int argc, char* argv[])
{
return ::std::strcmp(argv[0],
"really really really really really really really really really"
"really really really really…
user1095108
- 14,119
- 9
- 58
- 116
40
votes
4 answers
Why vectorizing the loop over 64-bit elements does not have performance improvement over large buffers?
I am investigating the effect of vectorization on the performance of the program. In this regard, I have written following code:
#include
#include
#include
#define LEN 10000000
int main(){
struct timeval…
Pouya
- 1,871
- 3
- 20
- 25
36
votes
5 answers
How to check if compiled code uses SSE and AVX instructions?
I wrote some code to do a bunch of math, and it needs to go fast, so I need it to use SSE and AVX instructions. I'm compiling it using g++ with the flags -O3 and -march=native, so I think it's using SSE and AVX instructions, but I'm not sure. Most…
BadProgrammer99
- 759
- 1
- 5
- 13
34
votes
4 answers
What's missing/sub-optimal in this memcpy implementation?
I've become interested in writing a memcpy() as an educational exercise. I won't write a whole treatise of what I did and didn't think about, but here's
some guy's implementation:
__forceinline // Since Size is usually known,
//…
einpoklum
- 118,144
- 57
- 340
- 684
32
votes
2 answers
Implementation of __builtin_clz
What is the implementation of GCC's (4.6+) __builtin_clz? Does it correspond to some CPU instruction on Intel x86_64 (AVX)?
Cartesius00
- 23,584
- 43
- 124
- 195
32
votes
1 answer
What are the best instruction sequences to generate vector constants on the fly?
"Best" means fewest instructions (or fewest uops, if any instructions decode to more than one uop). Machine-code size in bytes is a tie-breaker for equal insn count.
Constant-generation is by its very nature the start of a fresh dependency chain,…
Peter Cordes
- 328,167
- 45
- 605
- 847
32
votes
3 answers
Intel AVX: 256-bits version of dot product for double precision floating point variables
The Intel Advanced Vector Extensions (AVX) offers no dot product in the 256-bit version (YMM register) for double precision floating point variables. The "Why?" question have been very briefly treated in another forum (here) and on Stack Overflow…
gleeen.gould
- 599
- 1
- 5
- 22
31
votes
5 answers
Why ARM NEON not faster than plain C++?
Here is a C++ code:
#define ARR_SIZE_TEST ( 8 * 1024 * 1024 )
void cpp_tst_add( unsigned* x, unsigned* y )
{
for ( register int i = 0; i < ARR_SIZE_TEST; ++i )
{
x[ i ] = x[ i ] + y[ i ];
}
}
Here is a neon version:
void…
Smalti
- 507
- 1
- 5
- 13
31
votes
1 answer
Crash with icc: can the compiler invent writes where none existed in the abstract machine?
Consider the following simple program:
#include
#include
#include
void replace(char *str, size_t len) {
for (size_t i = 0; i < len; i++) {
if (str[i] == '/') {
str[i] = '_';
}
…
BeeOnRope
- 60,350
- 16
- 207
- 386
31
votes
4 answers
print a __m128i variable
I'm trying to learn to code using intrinsics and below is a code which does addition
compiler used: icc
#include
#include
int main()
{
__m128i a = _mm_set_epi32(1,2,3,4);
__m128i b = _mm_set_epi32(1,2,3,4);
…
arunmoezhi
- 3,082
- 6
- 35
- 54
30
votes
3 answers
How to write portable simd code for complex multiplicative reduction
I want to write fast simd code to compute the multiplicative reduction of a complex array. In standard C this is:
#include
complex float f(complex float x[], int n ) {
complex float p = 1.0;
for (int i = 0; i < n; i++)
p *=…
Simd
- 19,447
- 42
- 136
- 271