Using share memory method, a part of result data is not matched when I run convolution code repeatily

I worte convolation code in my work. in that code, the share memroy method is used in that code.
Unfortunate, I notes a part of result data is not matched when I run that repeatily.
Code is show as below.


if ((curBlockStartOffset + inputShareMemStartPoint) < curBlockEndOffset)
{
	float accI = 0;
	float accQ = 0;

	for (int j = 0; j < streamCB.d_FirDataCols; j++)
	{
		accI += inputIDataShared[inputShareMemStartPoint + j] * firData3DTablesData[curInputDataFirRowOffset + j];
		accQ += inputQDataShared[inputShareMemStartPoint + j] * firData3DTablesData[curInputDataFirRowOffset + j];
	}

#ifdef GPU_DEBUG
		if (streamCB.processID == 0 && blockIdx.x == 0 && threadIdx.x == 311)
		{
			for (int j = 0; j < streamCB.d_FirDataCols; j++)
			{
				printf("inputIDataShared[%d]:%d, \tinputQDataShared[%d]:%d, \tfirData3DTablesData[%d]:%f\n", inputShareMemStartPoint + j, inputIDataShared[inputShareMemStartPoint + j]
					, inputShareMemStartPoint + j, inputQDataShared[inputShareMemStartPoint + j]
					, curInputDataFirRowOffset + j, firData3DTablesData[curInputDataFirRowOffset + j]);
			}
			printf("StreamId:%d, CurBlockId:%d, CurBlockOutputDataIndex:%d, InputShareMemStartPoint:%d,FirTableIndex:%d, CurInputDataFirRowOffset:%d, AccI:%f\t, AccQ:%f\n"
				, streamCB.processID, blockIdx.x, threadIdx.x, inputShareMemStartPoint, streamCB.d_TimeDelayAdjustTableData[blockIdx.x], curInputDataFirRowOffset, accI, accQ);
	}
#endif

However, I got the different output value both AccIand AccQ even the same variable inputs when I run twice. Below is log details.
First round:


inputIDataShared[385]:361, inputQDataShared[385]:362, firData3DTablesData[1475]:1475.000000
inputIDataShared[386]:362, inputQDataShared[386]:363, firData3DTablesData[1476]:1476.000000
inputIDataShared[387]:363, inputQDataShared[387]:364, firData3DTablesData[1477]:1477.000000
inputIDataShared[388]:364, inputQDataShared[388]:365, firData3DTablesData[1478]:1478.000000
inputIDataShared[389]:365, inputQDataShared[389]:366, firData3DTablesData[1479]:1479.000000
inputIDataShared[390]:366, inputQDataShared[390]:367, firData3DTablesData[1480]:1480.000000
inputIDataShared[391]:367, inputQDataShared[391]:368, firData3DTablesData[1481]:1481.000000
inputIDataShared[392]:368, inputQDataShared[392]:369, firData3DTablesData[1482]:1482.000000
inputIDataShared[393]:369, inputQDataShared[393]:370, firData3DTablesData[1483]:1483.000000
inputIDataShared[394]:370, inputQDataShared[394]:371, firData3DTablesData[1484]:1484.000000
inputIDataShared[395]:371, inputQDataShared[395]:372, firData3DTablesData[1485]:1485.000000
inputIDataShared[396]:372, inputQDataShared[396]:373, firData3DTablesData[1486]:1486.000000
inputIDataShared[397]:373, inputQDataShared[397]:374, firData3DTablesData[1487]:1487.000000
inputIDataShared[398]:374, inputQDataShared[398]:375, firData3DTablesData[1488]:1488.000000
inputIDataShared[399]:375, inputQDataShared[399]:376, firData3DTablesData[1489]:1489.000000
inputIDataShared[400]:376, inputQDataShared[400]:377, firData3DTablesData[1490]:1490.000000
inputIDataShared[401]:377, inputQDataShared[401]:378, firData3DTablesData[1491]:1491.000000
inputIDataShared[402]:378, inputQDataShared[402]:379, firData3DTablesData[1492]:1492.000000
inputIDataShared[403]:379, inputQDataShared[403]:380, firData3DTablesData[1493]:1493.000000
inputIDataShared[404]:380, inputQDataShared[404]:381, firData3DTablesData[1494]:1494.000000
inputIDataShared[405]:381, inputQDataShared[405]:382, firData3DTablesData[1495]:1495.000000
inputIDataShared[406]:382, inputQDataShared[406]:383, firData3DTablesData[1496]:1496.000000
inputIDataShared[407]:383, inputQDataShared[407]:384, firData3DTablesData[1497]:1497.000000
inputIDataShared[408]:384, inputQDataShared[408]:385, firData3DTablesData[1498]:1498.000000
inputIDataShared[409]:385, inputQDataShared[409]:386, firData3DTablesData[1499]:1499.000000
StreamId:0, CurBlockId:0, CurBlockOutputDataIndex:311, InputShareMemStartPoint:385,FirTableIndex:-5, CurInputDataFirRowOffset:1475, AccI:12264637.000000 , AccQ:12297384.000000


Second round:


inputIDataShared[385]:361, inputQDataShared[385]:362, firData3DTablesData[1475]:1475.000000
inputIDataShared[386]:362, inputQDataShared[386]:363, firData3DTablesData[1476]:1476.000000
inputIDataShared[387]:363, inputQDataShared[387]:364, firData3DTablesData[1477]:1477.000000
inputIDataShared[388]:364, inputQDataShared[388]:365, firData3DTablesData[1478]:1478.000000
inputIDataShared[389]:365, inputQDataShared[389]:366, firData3DTablesData[1479]:1479.000000
inputIDataShared[390]:366, inputQDataShared[390]:367, firData3DTablesData[1480]:1480.000000
inputIDataShared[391]:367, inputQDataShared[391]:368, firData3DTablesData[1481]:1481.000000
inputIDataShared[392]:368, inputQDataShared[392]:369, firData3DTablesData[1482]:1482.000000
inputIDataShared[393]:369, inputQDataShared[393]:370, firData3DTablesData[1483]:1483.000000
inputIDataShared[394]:370, inputQDataShared[394]:371, firData3DTablesData[1484]:1484.000000
inputIDataShared[395]:371, inputQDataShared[395]:372, firData3DTablesData[1485]:1485.000000
inputIDataShared[396]:372, inputQDataShared[396]:373, firData3DTablesData[1486]:1486.000000
inputIDataShared[397]:373, inputQDataShared[397]:374, firData3DTablesData[1487]:1487.000000
inputIDataShared[398]:374, inputQDataShared[398]:375, firData3DTablesData[1488]:1488.000000
inputIDataShared[399]:375, inputQDataShared[399]:376, firData3DTablesData[1489]:1489.000000
inputIDataShared[400]:376, inputQDataShared[400]:377, firData3DTablesData[1490]:1490.000000
inputIDataShared[401]:377, inputQDataShared[401]:378, firData3DTablesData[1491]:1491.000000
inputIDataShared[402]:378, inputQDataShared[402]:379, firData3DTablesData[1492]:1492.000000
inputIDataShared[403]:379, inputQDataShared[403]:380, firData3DTablesData[1493]:1493.000000
inputIDataShared[404]:380, inputQDataShared[404]:381, firData3DTablesData[1494]:1494.000000
inputIDataShared[405]:381, inputQDataShared[405]:382, firData3DTablesData[1495]:1495.000000
inputIDataShared[406]:382, inputQDataShared[406]:383, firData3DTablesData[1496]:1496.000000
inputIDataShared[407]:383, inputQDataShared[407]:384, firData3DTablesData[1497]:1497.000000
inputIDataShared[408]:384, inputQDataShared[408]:385, firData3DTablesData[1498]:1498.000000
inputIDataShared[409]:385, inputQDataShared[409]:386, firData3DTablesData[1499]:1499.000000
StreamId:0, CurBlockId:0, CurBlockOutputDataIndex:311, InputShareMemStartPoint:385,FirTableIndex:-5, CurInputDataFirRowOffset:1475, AccI:13867575.000000 , AccQ:13904750.000000


I have been troubled by this issue for several days, Someone can help me to double check it? Thanks