提交 9c27c13e 编写于 作者: T tensor-tang

follow comments using macro to separate the original implements

上级 84cb542c
...@@ -172,11 +172,10 @@ void SequenceToBatch::sequence2BatchCopy(Matrix &batch, ...@@ -172,11 +172,10 @@ void SequenceToBatch::sequence2BatchCopy(Matrix &batch,
batchData, seqData, idxData, seqWidth, batchCount, seq2batch); batchData, seqData, idxData, seqWidth, batchCount, seq2batch);
} else { } else {
if (seq2batch) { if (seq2batch) {
#ifdef PADDLE_USE_MKLML
const int blockMemSize = 8 * 1024; const int blockMemSize = 8 * 1024;
const int blockSize = blockMemSize / sizeof(real); const int blockSize = blockMemSize / sizeof(real);
#ifdef PADDLE_USE_MKLML
#pragma omp parallel for collapse(2) #pragma omp parallel for collapse(2)
#endif
for (int i = 0; i < batchCount; ++i) { for (int i = 0; i < batchCount; ++i) {
for (int j = 0; j < seqWidth; j += blockSize) { for (int j = 0; j < seqWidth; j += blockSize) {
memcpy(batch.rowBuf(i) + j, memcpy(batch.rowBuf(i) + j,
...@@ -185,6 +184,13 @@ void SequenceToBatch::sequence2BatchCopy(Matrix &batch, ...@@ -185,6 +184,13 @@ void SequenceToBatch::sequence2BatchCopy(Matrix &batch,
: blockMemSize); : blockMemSize);
} }
} }
#else
for (int i = 0; i < batchCount; ++i) {
memcpy(batch.rowBuf(i),
sequence.rowBuf(idxData[i]),
seqWidth * sizeof(real));
}
#endif
} else { } else {
#ifdef PADDLE_USE_MKLML #ifdef PADDLE_USE_MKLML
#pragma omp parallel for #pragma omp parallel for
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册