fdvmDDe (1158418), страница 5

Просмтор этого файла доступен только зарегистрированным пользователям. Но у нас супер быстрая регистрация: достаточно только электронной почты!

Текст из файла (страница 5)

DO label I1 = ...

. . .

DO label In = ...

loop-body

label CONTINUE

is translated into

[ ACROSS-block-1 ]

[ REDUCTION-block-1 ]

* creating parallel loop

ipl = crtpl(n)

[ SHADOW-RENEW-block-1 ]

[ SHADOW-START-block ]

[ SHADOW-WAIT-block ]

* mapping parallel loop (1)

it = mappl(ipl,A,...)

[ SHADOW-RENEW-block-2 ]

[ REDUCTION-block-2 ]

[ REMOTE-ACCESS-block ]

* inquiry of continuation of parallel loop execution

lab1 if(dopl(ipl) .eq. 0) go to lab2

DO label I1 = ...

. . .

DO label In = ...

loop-body

label CONTINUE

go to lab1

* terminating parallel loop

lab2 it = endpl(ipl)

[ ACROSS-block-2 ]

[ REDUCTION-block-3 ]

The function generates and inserts in procedure all the statements preceeding the DO nest. In addition, the initial, end and step value of do-variables in parallel DO-nest are changed. The statements following end statement of parallel loop:

label CONTINUE

are generated by TransFunc( ) when this statement is processing.

If compilation mode is set on debugging, the CALL statements:

call dbegpl(...)

call diter(...)

call dendl(...)

are created and inserted in block (1) before IF statement, before first statement of parallel loop body ,and after the statement

go to lab1

accordingly.

If compilation mode is set on performance analyzing, the CALL statements:

call bploop(...)

call eloop(...)

are created and inserted before the first and after the last statement of the block (1).

The following functions are called to create blocks implementing ACROSS, SHADOW_RENEW, REDUCTION, and REMOTE_ACCESS specifications:

DepList

ShadowList

doIfForReduction

ReductionList

RemoteVariableList

void ParallelLoop_Debug ( SgStatement *stmt )

If compilation mode is set on debugging, the CALL statements:

call dbegpl(...)

call diter(...)

are created and inserted before the DO loop nest and before the first statement of parallel loop body accordingly. This function generates the REDUCTION-block-1 and REDUCTION-block-2 if need be.

If compilation mode is set on performance analyzing, the CALL statements:

call bploop(...)

call eloop(...)

is created and inserted before and after the DO loop nest.

Generates and inserts in procedure the statements:

* creating reduction

dvm000(irv) = crtrgf(reduction-function, red-var,…)

* including reduction in reduction group

dvm000(i) = insred(gref,dvm000(irv),…)

for each reduction in reduction list. The first statement is inserted after stmt1 and the second one after stmt2.

Generates and inserts in procedure the statement:

* including shadow edge in the group

dvm000(i) = inssh(gref,array,...)

for each array in the renewee-list.

The function generates and inserts in procedure the statements for reading remote data in buffer (REMOTE-ACCESS-block).

1) In case of synchronous REMOTE_ACCESS specification the following statements are generated:

{

* creating buffer array

it = crtrbl(array-header,buffer-header,…)

* starting load of buffer array

it = loadrb(buffer-header,0)

* waiting for completion of loading buffer array

it = waitrb(buffer-header)

* correcting coefficient C_NB of buffer array elements addressing,

* where NB is rank of buffer array

buffer-header(NB+2) = buffer-header(NB+1)-

* buffer-header(NB)*buffer-header(NB+3) … -

* buffer-header(3)*buffer-header(2*NB+2)

}... for each remote-access reference

2) In case of asynchronous REMOTE_ACCESS specification (with group RMG) the following statements are generated:

IF (RMG(2) .EQ. 0) THEN

{

* creating buffer array

it = crtrbl(array-header,buffer-header,…)

* correcting coefficient C_NB of buffer array elements addressing

buffer-header(NB+2) = buffer-header(NB+1)-

* buffer-header(NB)*buffer-header(NB+3) … -

* buffer-header(3)*buffer-header(2*NB+2)

* starting load of buffer array

it = loadrb(buffer-header,0)

* waiting for completion of loading buffer array

it = waitrb(buffer-header)

* including buffer array in group RMG

it = insrb(RMG(1),buffer-header)

}... for each remote-access reference

ELSE

IF (RMG(3) .EQ. 1) THEN

* waiting for completion of loading all the buffer arrays of group

it = waitbg(RMG(1))

RMG(3) = 0

ENDIF

ENDIF

5.2Translating input/output statements (module io.cpp)

The compiler module io.cpp involves the functions for translating input/output statements.

In DVM model, input, output and other operations with external files are executed by single processor ( I/O processor ), which is determined by run-time system. I/O of a replicated variable deals with variable copy allocated on I/O processor. I/O of a distributed array deals with buffer array allocated on I/O processor. Inputted data are sent to all other processors owing the variables of input list. When the distributed array is outputted, data are transferred into the buffer from other processors owing elements of the array.

The function analyzes input/output item list. If there are not any distributed array references in the list, this function returns 1, else it returns 0.

Calls ImplicitLoopTest( ) , IOitemTest( ).

The function analyzes item list of implicit loop. If there are not any distributed array references in the list, this function returns 1, else it returns 0.

If the I/O item is not a distributed array reference this function returns 1, else it returns 0.

The function analyzes the control information list of the data transfer statement and assigns the value of control parameters (UNIT,FMT,ERR, and so on) to the elements of array ioc[]. If there are some syntax errors it returns 0, else 1.

The function analyzes the control information list of the BACKSPACE, REWIND and ENDFILE statement and assigns the value of control parameters (UNIT,ERR, and so on) to the elements of array ioc[]. If there are some syntax errors it returns 0, else 1.

The function analyzes the control information list of the OPEN, CLOSE and INQUIRE statement and assigns the value of control parameters (UNIT,ERR, and so on) to the elements of array ioc[]. If there are some syntax errors it returns 0, else 1.

In case of I/O of distributed array the memory is allocated in user program for buffer I/O.

Let A(N₁,N₂,...,N_k) is distributed array of rank k, BUF(L) - vector of the same type as array A (named i000io if A is of type integer, or r000io if A is of type real,...).

The function replaces a statement I/O of A with the sequence of statements according to the following scheme:

s₀ = L

s_i = [s_i-1 / N_i] i  {1 : k}

n = max(i)

s_i 0, i{0 : k}

m = 1 (if n = k)

Характеристики

Список файлов учебной работы