RASO && RASO_parallel avx64_256_512

HEM5_avxOR.h

@@ -15,6 +15,7 @@
 #define mmfor(i,a,b) for(int i=(a);i>=(b);--i)
 
 // 动动模式 用avx2 做256位的或运算
+// 用于和 HEM5 对比
 class HEM5_avxOR
 {
 private:

README.md

@@ -130,9 +130,9 @@
 | 13   | 综合性能实验                                           | $\phi_s$= 1 ~ 10, $\phi_e$ = 1 ~ 40, w=0~1, d=100, alpha=1, attrSubsetNum=25 | 1081                                       |
 | 14   | 构建时间、删除时间实验(即实验 6)                       | $\phi_s$=5~30, d=$\phi_e$=30, w=0.8                          | 1082-1087                                  |
 | 15   | 内存实验(即实验 5)                                     | n = 300**k** ~ 9**m**                                        | 1029-1034                                  |
-| 16   | SIMD实验 on HEMDD (匹配时间和或运算时间下降比例-2个图) | $\psi_e$=d=[16, 32, 64, 128, 256, 512], blockSize=256,512    | 1355-1360, 1361-1366，1393-1401, 1409-1417 |
+| 16   | SIMD实验 on HEMDD (匹配时间和或运算时间下降比例-2个图) | $\psi_e$=d=[16, 32, 64, 128, 256, 512], BlockSize=256,512    | 1355-1360, 1361-1366，1393-1401, 1409-1417 |
 | 17   | pHEMDD 并行实验（加速比和效率-两图）                   | $\psi_e=d=$ 512, pD = [1, 2, 4, 8, 16, 32, 64]               | 1402-1408                                  |
-| 18   | pHEMDD-avxOR 组合实验（和HEMDD、pHEMDD 比较）          | $\psi_e=d=$ 512, pD = [1, 2, 4, 8, 16, 32, 64], blockSize=512 | 1402-1408                                  |
+| 18   | pHEMDD-avxOR 组合实验（和HEMDD、pHEMDD 比较）          | $\psi_e=d=$ 512, pD = [1, 2, 4, 8, 16, 32, 64], BlockSize=512 | 1402-1408                                  |
 | 19   | pHEMDD-RAS-avxOR 高负载高维度实验（和HEMDD-RAS比较）   | n = $2^{27}$, d=2048, attrSubset=32, $\psi_e$=64, pD = [1, 2, 4, 8, 16, 32, 64], $b_e$ = 1 | 1386-1392                                  |
 |20| 事件值实验| 0-0.1M, 0.1M-0.2M,...,0.9M-1M    | 随机宽度: 1658-1667，固定w0.4: 1668-1677  |
 

Util.h

@@ -129,7 +129,7 @@ class Util {
 
 
 // 256
-#if blockSize == 256
+#if BlockSize == 256
  void Util::bitsetOr(bitset<subs> &b1, const bitset<subs> &b2) { // b1=b1|b2;
  	__m256i b1_256, b2_256;
  	long long int *begin1 = reinterpret_cast<long long int *> (&b1);
@@ -149,7 +149,7 @@ class Util {
  		begin2 += ptrIncrement;
  	}
  }
-#elif blockSize==512
+#elif BlockSize==512
 // 512
 void Util::bitsetOr(bitset<subs> &b1, const bitset<subs> &b2) { // b1=b1|b2;
 	__m512i b1_512, b2_512;

constant.cpp

@@ -38,10 +38,10 @@ int be = 4;                 // bit exponent. In each dimension, there are 2 * 2^
 int be2 = 50;                // When be=-1, be2 is adopted. In each dimension, there are 2 * be2 bitsets.
 const int lvls = 2;         // HEMSC number of levels
 int gs = 5;                 // HEMSR Group Size 一组有多少个维度
-// const int blockSize = 512;  // 每次对256位做逻辑运算
+// const int BlockSize = 512;  // 每次对256位做逻辑运算
 const int ptrSize = 64;     // 每次只能加载__int64 64位
-const int blockNum = (subs + blockSize - 1) / blockSize; //  把做 subs 位的逻辑运算 简化为做 blockNum 次逻辑运算
-const int ptrIncrement = blockSize / ptrSize;       //  需要加载的次数
+const int blockNum = (subs + BlockSize - 1) / BlockSize; //  把做 subs 位的逻辑运算 简化为做 blockNum 次逻辑运算
+const int ptrIncrement = BlockSize / ptrSize;       //  需要加载的次数
 const __m256i mask = _mm256_set1_epi32(0x80000000);
 
 // Tama

constant.h

@@ -14,7 +14,7 @@ extern const bool verifyID; // whether to verify Insertion and Deletion function
 extern const int interval;	// ÿƥ����ô����¼����һ����־��Ϣ
 extern const double subp;	// subp�����Ķ��ĺ��¼���ν�ʶ�����ǰcons��ά����
 extern double realMatchNum;
-const int subs = 1000000;	 // 1048576;//134217728;  // Number of subscriptions.
+const int subs = 1048576;    // 1048576;//134217728;  // Number of subscriptions.
 extern const int32_t pubs;	 // Number of publications.
 extern int32_t atts;		 // Total number of attributes, i.e. dimensions.
 extern int attrGroup;		 // The number of attribute group in #atts attributes.
@@ -90,8 +90,8 @@ extern int be;		   // ÿ��ά���� 2 * 2^bits��bits����
 extern int be2;		   // be=-1ʱÿ��ά���� 2 * be2 ��bits����
 extern const int lvls; // HEMSC ����
 extern int gs;		   // HEMSR һ���ж��ٸ�ά��
-// extern const int blockSize;
-#define blockSize 256
+// extern const int BlockSize;
+#define BlockSize 256
 extern const int ptrSize;
 extern const int blockNum; // ���������
 extern const int ptrIncrement;

run_models.cpp

@@ -74,7 +74,7 @@ void run_rein(const intervalGenerator& gen, unordered_map<int, bool> deleteNo)
 					 + " ms AvgMarkTime= " + to_string(rein.markTime / pubs / 1000000)
 					 + " ms AvgBitTime= " + to_string(rein.bitTime / pubs / 1000000)
 					 + " ms pD= " + to_string(parallelDegree)
-					 + " avx= " + to_string(blockSize)
+					 + " avx= " + to_string(BlockSize)
 					 + " numBuk= " + Util::Int2String(rein.numBucket)
 					 + " numSub= " + Util::Int2String(subs)
 					 + " subSize= " + Util::Int2String(cons)
@@ -2763,7 +2763,7 @@ void run_HEM5(const intervalGenerator& gen, unordered_map<int, bool> deleteNo)
 					 + " ms AvgMarkTime= " + to_string(hem5.markTime / pubs / 1000000)
 					 + " ms OrTime= " + to_string(hem5.orTime / pubs / 1000000)
 					 + " ms AvgBitTime= " + to_string(hem5.bitTime / pubs / 1000000)
-					 + " ms avx= " + to_string(blockSize)
+					 + " ms avx= " + to_string(BlockSize)
 					 + " numBuk= " + Util::Int2String(hem5.numBucket)
 					 + " numSub= " + Util::Int2String(subs)
 					 + " subSize= " + Util::Int2String(cons)
@@ -2886,7 +2886,7 @@ void run_HEM5_avxOR(const intervalGenerator& gen, unordered_map<int, bool> delet
 					 + " ms AvgMarkTime= " + to_string(hem5_avxor.markTime / pubs / 1000000)
 					 + " ms OrTime= " + to_string(hem5_avxor.orTime / pubs / 1000000)
 					 + " ms AvgBitTime= " + to_string(hem5_avxor.bitTime / pubs / 1000000)
-					 + " ms avx= " + to_string(blockSize)
+					 + " ms avx= " + to_string(BlockSize)
 					 + " numBuk= " + Util::Int2String(hem5_avxor.numBucket)
 					 + " numSub= " + Util::Int2String(subs)
 					 + " subSize= " + Util::Int2String(cons)
@@ -3116,7 +3116,7 @@ void run_HEM5_avxOR_parallel(const intervalGenerator& gen, unordered_map<int, bo
 					 + " ms MergeTime= " + to_string(hem5_avxOR_p.mergeTime / pubs / 1000000)
 					 + " ms AvgBitTime= " + to_string(hem5_avxOR_p.bitTime / pubs / 1000000)
 					 + " ms pD= " + to_string(parallelDegree)
-					 + " avx= " + to_string(blockSize)
+					 + " avx= " + to_string(BlockSize)
 					 + " numBuk= " + Util::Int2String(hem5_avxOR_p.numBucket)
 					 + " numSub= " + Util::Int2String(subs)
 					 + " subSize= " + Util::Int2String(cons)
@@ -3423,7 +3423,7 @@ void run_HEM5_RAS_avxOR_parallel(const intervalGenerator& gen, unordered_map<int
 		matchTimeList.push_back((double)eventTime / 1000000);
 		matchSubList.push_back(matchSubs);
 		if (i % interval == 0)
-			cout << "HEM5DD_RAS_avxOR_Parallel" + to_string(blockSize) + "-Parallel: Event " << i << " is matched.\n";
+			cout << "HEM5DD_RAS_avxOR_Parallel" + to_string(BlockSize) + "-Parallel: Event " << i << " is matched.\n";
 	}
 
 	if (display)
@@ -3452,7 +3452,7 @@ void run_HEM5_RAS_avxOR_parallel(const intervalGenerator& gen, unordered_map<int
 					 + " ms MergeTime= " + to_string(hem5_ras_a_p.mergeTime / pubs / 1000000)
 					 + " ms AvgBitTime= " + to_string(hem5_ras_a_p.bitTime / pubs / 1000000)
 					 + " ms pD= " + to_string(parallelDegree)
-					 + " avx= " + to_string(blockSize)
+					 + " avx= " + to_string(BlockSize)
 					 + " numBuk= " + Util::Int2String(hem5_ras_a_p.numBucket)
 					 + " numSub= " + Util::Int2String(subs)
 					 + " subSize= " + Util::Int2String(cons)