pcgeadd.c File Reference

#include "common.h"

Include dependency graph for pcgeadd.c:

Go to the source code of this file.

Macros
#define	A(m, n)   BLKADDR(A, PLASMA_Complex32_t, m, n)
#define	B(m, n)   BLKADDR(B, PLASMA_Complex32_t, m, n)

Functions
void	plasma_pcgeadd (plasma_context_t *plasma)
void	plasma_pcgeadd_quark (PLASMA_Complex32_t alpha, PLASMA_desc A, PLASMA_desc B, PLASMA_sequence *sequence, PLASMA_request *request)

---

Detailed Description

PLASMA auxiliary routines PLASMA is a software package provided by Univ. of Tennessee, Univ. of California Berkeley and Univ. of Colorado Denver

Version:

2.4.5

Author:

Emmanuel Agullo

Mathieu Faverge

Date:

2010-11-15 c Tue Nov 22 14:35:40 2011

Definition in file pcgeadd.c.

---

Macro Definition Documentation

#define A	(		m,
			n
	)		BLKADDR(A, PLASMA_Complex32_t, m, n)

Definition at line 18 of file pcgeadd.c.

#define B	(		m,
			n
	)		BLKADDR(B, PLASMA_Complex32_t, m, n)

Definition at line 19 of file pcgeadd.c.

---

Function Documentation

void plasma_pcgeadd

(

plasma_context_t *

plasma

)

Declarations of parallel functions (static scheduling) - alphabetical order

Definition at line 23 of file pcgeadd.c.

References A, B, BLKLDD, CORE_cgeadd(), plasma_desc_t::m, plasma_desc_t::mb, plasma_desc_t::mt, plasma_desc_t::n, plasma_desc_t::nb, plasma_desc_t::nt, PLASMA_RANK, PLASMA_SIZE, PLASMA_SUCCESS, plasma_unpack_args_5, and plasma_sequence_t::status.

{
    PLASMA_Complex32_t alpha;
    PLASMA_desc A;
    PLASMA_desc B;
    PLASMA_sequence *sequence;
    PLASMA_request *request;

    int X, Y;
    int m, n;
    int next_m;
    int next_n;
    int ldam, ldbm;

    plasma_unpack_args_5(alpha, A, B, sequence, request);
    if (sequence->status != PLASMA_SUCCESS)
        return;

    n = 0;
    m = PLASMA_RANK;
    while (m >= A.mt && n < A.nt) {
        n++;
        m = m-A.mt;
    }

    while (n < A.nt) {
        next_m = m;
        next_n = n;

        next_m += PLASMA_SIZE;
        while (next_m >= A.mt && next_n < A.nt) {
            next_n++;
            next_m = next_m-A.mt;
        }

        X = m == A.mt-1 ? A.m-A.mb*m : A.nb;
        Y = n == A.nt-1 ? A.n-A.nb*n : A.nb;
        ldam = BLKLDD(A, m);
        ldbm = BLKLDD(B, m);
        CORE_cgeadd(X, Y, alpha, A(m, n), ldam, B(m, n), ldbm);

        m = next_m;
        n = next_n;
    }
}

Here is the call graph for this function:

void plasma_pcgeadd_quark	(	PLASMA_Complex32_t	alpha,
		PLASMA_desc	A,
		PLASMA_desc	B,
		PLASMA_sequence *	sequence,
		PLASMA_request *	request
	)

Declarations of parallel functions (dynamic scheduling) - alphabetical order

Definition at line 72 of file pcgeadd.c.

References B, BLKLDD, plasma_desc_t::m, plasma_desc_t::mb, plasma_desc_t::mt, plasma_desc_t::n, plasma_desc_t::nb, plasma_desc_t::nt, plasma_context_self(), PLASMA_SUCCESS, plasma_context_struct::quark, QUARK_CORE_cgeadd(), plasma_sequence_t::quark_sequence, QUARK_Task_Flag_Set(), Quark_Task_Flags_Initializer, plasma_sequence_t::status, and TASK_SEQUENCE.

{
    plasma_context_t *plasma;
    Quark_Task_Flags task_flags = Quark_Task_Flags_Initializer;

    int X, Y;
    int m, n;
    int ldam, ldbm;

    plasma = plasma_context_self();
    if (sequence->status != PLASMA_SUCCESS)
        return;
    QUARK_Task_Flag_Set(&task_flags, TASK_SEQUENCE, (intptr_t)sequence->quark_sequence);

    for (m = 0; m < A.mt; m++) {
        X = m == A.mt-1 ? A.m-m*A.mb : A.mb;
        ldam = BLKLDD(A, m);
        ldbm = BLKLDD(B, m);

        for (n = 0; n < A.nt; n++) {
            Y = n == A.nt-1 ? A.n-n*A.nb : A.nb;
            QUARK_CORE_cgeadd(
                plasma->quark, &task_flags,
                X, Y, A.mb,
                alpha, A(m, n), ldam,
                       B(m, n), ldbm);
        }
    }
}

Here is the call graph for this function: