How does GPGPU OpenCL implement exact string search?

How does GPGPU OpenCL implement exact string search?

1. Acceleration Method

(1) store a small amount of constant data, such as the mode string length and text length, in the private memory of the thread.

(2) Save the mode string in the local memory of the GPU, and accelerate the thread's access to the mode string.

(3) Save the text to be searched in global memory, use as many threads as possible to access global memory, and reduce the average thread access time.

(4) A section of the thread operation text in each work-group. Multiple work-groups process large texts in parallel.

　　2. Synchronization

(1) CLK_LOCAL_MEM_FENCE and CLK_GLOBAL_MEM_FENCE are used in the work-group.

(2) The atomic operation on _ global int Is used globally to ensure that each thread writes the result to the correct position in the global memory. The operations supported by the device can be obtained by querying the extension of the device. For example, the kernel function supports atomic operations and printf operations:

　　3. Code instance, large text exact mode string SEARCH

3.1 core function (string_search_kernel.cl ):

Int compare (_ global const uchar * text, _ local const uchar * pattern, uint length ){

For (uint l = 0; l

If (text [l]! = Pattern [l])

Return 0;

}

Return 1;

}

_ Kernel void

StringSearch (

_ Global uchar * text, // Input Text

Const uint textLength, // Length of the text

_ Global const uchar * pattern, // Pattern string

Const uint patternLength, // Pattern length

Const uint maxSearchLength, // Maximum search positions for each work-group

_ Global int * resultCount, // Result counts (global)

_ Global int * resultBuffer, // Save the match result

_ Local uchar * localPattern) // local buffer for the search pattern

{

Int localIdx = get_local_id (0 );

Int localSize = get_local_size (0 );

Int groupIdx = get_group_id (0 );

Uint lastSearchIdx = textLength-patternLength + 1;

Uint beginSearchIdx = groupIdx * maxSearchLength;

Uint endSearchIdx = beginSearchIdx + maxSearchLength;

If (beginSearchIdx> lastSearchIdx)

Return;

If (endSearchIdx> lastSearchIdx)

EndSearchIdx = lastSearchIdx;

For (int idx = localIdx; idx <patternLength; idx + = localSize)

LocalPattern [idx] = pattern [idx];

Barrier (CLK_LOCAL_MEM_FENCE );

For (uint stringPos = beginSearchIdx + localIdx; stringPos

If (compare (text + stringPos, localPattern, patternLength) = 1 ){

Int count = atomic_inc (resultCount );

ResultBuffer [count] = stringPos;

// Printf ("% d", stringPos );

}

Barrier (CLK_LOCAL_MEM_FENCE );

}

}