thanks for explanation, well, how can I change that code, so it will take
count of Threads (or gorutines) instead size of matrix for one gorutine?
Dátum: štvrtok 25. marca 2021, čas: 19:21:02 UTC+1, odosielateľ:
jas...@gmail.com
> Blast from the past so it's hard to be sure, but I think that was how many
> rows or columns to pick for parallel sub matrices to multiply.
>
> On Thu, Mar 25, 2021, 1:17 PM Gabriel Pcklub <gabriel...@gmail.com> wrote:
>
>> Hello, what is Threshold constant in code for? I wanted to try your code,
>> but with Threshold 1 it take all of 16GB RAM I have and crash on sigkill.
>> When I use Threshold 8 it seems ok, but when I use 64, it again take a lot
>> of RAM... I thought it's count of Threads that program will use, while
>> running on CPU. But it does give unexpected or strange results.
>>
>> Dátum: štvrtok 14. apríla 2011, čas: 14:55:09 UTC+2, odosielateľ: Dmitry
>> Vyukov
>>
>>> Hi,
>>>
>>> I want to create a parallel matrix multiplication benchmark for Go. I
>>> used to fork-join style parallelism, so I implemented it as follows:
>>>
>>> type Matrix [][]float64
>>>
>>> func matrix_make(n int) Matrix {
>>> M := make([][]float64, n);
>>> for i := 0; i != n; i += 1 {
>>> M[i] = make([]float64, n)
>>> }
>>> return M
>>> }
>>>
>>> const (Threshold = 8)
>>>
>>> func matmult_pimpl (sync chan int, A Matrix, B Matrix, C Matrix, i0 int,
>>> i1 int, j0 int, j1 int, k0 int, k1 int) {
>>> di := i1 - i0
>>> dj := j1 - j0
>>> dk := k1 - k0
>>> if (di >= dj && di >= dk && di >= Threshold) {
>>> mi := i0 + di / 2
>>> sync0 := make(chan int, 1)
>>> go matmult_pimpl(sync0, A, B, C, i0, mi, j0, j1, k0, k1)
>>> matmult_pimpl(nil, A, B, C, mi, i1, j0, j1, k0, k1)
>>> <- sync0
>>> } else if (dj >= dk && dj >= Threshold) {
>>> mj := j0 + dj / 2
>>> sync0 := make(chan int, 1)
>>> go matmult_pimpl(sync0, A, B, C, i0, i1, j0, mj, k0, k1)
>>> matmult_pimpl(nil, A, B, C, i0, i1, mj, j1, k0, k1)
>>> <- sync0
>>> } else if (dk >= Threshold) {
>>> mk := k0 + dk / 2
>>> matmult_pimpl(nil, A, B, C, i0, i1, j0, j1, k0, mk)
>>> matmult_pimpl(nil, A, B, C, i0, i1, j0, j1, mk, k1)
>>> } else {
>>> for i := i0; i < i1; i += 1 {
>>> for j := j0; j < j1; j += 1 {
>>> for k := k0; k < k1; k += 1 {
>>> C[i][j] += A[i][k] * B[k][j];
>>> }
>>> }
>>> }
>>> }
>>> if sync != nil {
>>> sync <- 0
>>> }
>>> }
>>>
>>> func matmult_parallel(A Matrix, B Matrix) Matrix {
>>> n := len(A)
>>> C := matrix_make(n)
>>> matmult_pimpl(nil, A, B, C, 0, n, 0, n, 0, n)
>>> return C
>>> }
>>>
>>> Is it a good implementation for Go? Is it enough "Go-ish"? What would be
>>> a "canonical" way?
>>> TIA
>>>
>>> --
>> You received this message because you are subscribed to a topic in the
>> Google Groups "golang-nuts" group.
>> To unsubscribe from this topic, visit
>> https://groups.google.com/d/topic/golang-nuts/RVX-BhHcugM/unsubscribe.
>> To unsubscribe from this group and all its topics, send an email to
>> golang-nuts...@googlegroups.com.
>> To view this discussion on the web visit
>> https://groups.google.com/d/msgid/golang-nuts/32e55537-d2d0-4e9a-8425-c85ad6a790a1n%40googlegroups.com
>>
>> <https://groups.google.com/d/msgid/golang-nuts/32e55537-d2d0-4e9a-8425-c85ad6a790a1n%40googlegroups.com?utm_medium=email&utm_source=footer>
>> .
>>
>
--
You received this message because you are subscribed to the Google Groups
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to golang-nuts+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/golang-nuts/8fa9bb79-938a-4f89-a0c9-792776151a6en%40googlegroups.com.
