Hacked up code to test run ratios of SSB sequences.
diff --git a/storage/PackedRowStoreTupleStorageSubBlock.cpp b/storage/PackedRowStoreTupleStorageSubBlock.cpp
index 4cb6d02..bf086da 100644
--- a/storage/PackedRowStoreTupleStorageSubBlock.cpp
+++ b/storage/PackedRowStoreTupleStorageSubBlock.cpp
@@ -143,6 +143,8 @@
+ header_->num_tuples * relation_.getFixedByteLength();
const unsigned num_nullable_attrs = relation_.numNullableAttributes();
+ printf("(rr:%f)\n", accessor->getTupleIdSequenceVirtual()->getInternalBitVector().runRatio());
+
InvokeOnAnyValueAccessor(
accessor,
[&](auto *accessor) -> void { // NOLINT(build/c++11)
@@ -214,6 +216,8 @@
+ header_->num_tuples * relation_.getFixedByteLength();
const unsigned num_nullable_attrs = relation_.numNullableAttributes();
+ printf("(rr:%f)\n", accessor->getTupleIdSequenceVirtual()->getInternalBitVector().runRatio());
+
InvokeOnAnyValueAccessor(
accessor,
[&](auto *accessor) -> void { // NOLINT(build/c++11)
diff --git a/storage/SplitRowStoreTupleStorageSubBlock.cpp b/storage/SplitRowStoreTupleStorageSubBlock.cpp
index 6c70d0f..f3af282 100644
--- a/storage/SplitRowStoreTupleStorageSubBlock.cpp
+++ b/storage/SplitRowStoreTupleStorageSubBlock.cpp
@@ -195,6 +195,8 @@
const tuple_id original_num_tuples = header_->num_tuples;
tuple_id pos = 0;
+ printf("(rr:%f)\n", accessor->getTupleIdSequenceVirtual()->getInternalBitVector().runRatio());
+
InvokeOnAnyValueAccessor(
accessor,
[&](auto *accessor) -> void { // NOLINT(build/c++11)
@@ -391,6 +393,8 @@
const tuple_id original_num_tuples = header_->num_tuples;
tuple_id pos = 0;
+ printf("(rr:%f)\n", accessor->getTupleIdSequenceVirtual()->getInternalBitVector().runRatio());
+
InvokeOnAnyValueAccessor(
accessor,
[&](auto *accessor) -> void { // NOLINT(build/c++11)
diff --git a/utility/BitVector.hpp b/utility/BitVector.hpp
index a6197a5..0897ea6 100644
--- a/utility/BitVector.hpp
+++ b/utility/BitVector.hpp
@@ -749,6 +749,62 @@
return num_bits_;
}
+ // Hack to determine the ratio of runs (sequence of 1 or more tuples) to
+ // the total number of tuples. For example, a sequence of 100 selected tuples,
+ // all of the tuples being in a sequence will have a ratio of 1/100 = .01.
+ // A sequence like 10101010101010 will have a run ratio of 1.
+ const double runRatio() const {
+ double onesCount = 0;
+ double runsCount = 0;
+ bool inRun = false;
+
+ //printf("\n");
+
+ // Iterate through each bit.
+ for (std::size_t i = 0; i < size(); ++i) {
+ if (getBit(i)) {
+ ++onesCount;
+ inRun = true;
+ } else {
+ if (inRun) {
+ inRun = false;
+ runsCount++;
+ }
+ }
+ }
+
+ if (inRun) {
+ runsCount++;
+ }
+
+
+ // for (std::size_t position = 0; position < data_array_size_; ++position) {
+ // std::size_t word = data_array_[position];
+
+ // // Iterate through each bit in the word.
+ // for (std::size_t bit = 0; bit < sizeof(std::size_t) * 8; ++bit) {
+ // printf ("%lu", word & 0x1U);
+ // if (word & 0x1U) {
+ // ++onesCount;
+ // inRun = true;
+ // } else {
+ // if (inRun) {
+ // inRun = false;
+ // runsCount++;
+ // }
+ // }
+ // word >>= 1;
+ // }
+ // printf("\n");
+ // }
+
+
+
+ // printf("\n");
+
+ return runsCount / onesCount;
+ }
+
/**
* @brief Find the last 0-bit (strictly before the specified position) in
* this BitVector.
diff --git a/utility/tests/BitVector_unittest.cpp b/utility/tests/BitVector_unittest.cpp
index 52b475a..e6601f0 100644
--- a/utility/tests/BitVector_unittest.cpp
+++ b/utility/tests/BitVector_unittest.cpp
@@ -576,6 +576,36 @@
TestFixture::runUnionWithTest(200);
}
+TYPED_TEST(BitVectorTest, RunRatioTest) {
+ std::unique_ptr<TypeParam> bit_vector(this->createBitVector(TestFixture::kBiggerBitSize));
+ bit_vector->clear();
+
+ // Make a pattern which should have a ratio of 1: "010 ... 101".
+ for (size_t i = 0; i < TestFixture::kBiggerBitSize; ++i) {
+ bit_vector->setBit(i, i%2);
+ }
+
+ EXPECT_NEAR(1.0, bit_vector->runRatio(), 0.0001);
+
+ printf("run ratio: %f\n", bit_vector->runRatio());
+
+ bit_vector->clear();
+ // Make a pattern which should have a ratio of 0.5: "110 ... 011".
+ for (size_t i = 0; i < TestFixture::kBiggerBitSize; ++i) {
+ bit_vector->setBit(i, i%3 != 0);
+ }
+
+ EXPECT_NEAR(0.5, bit_vector->runRatio(), 0.0001);
+ printf("run ratio: %f\n", bit_vector->runRatio());
+
+ // Make a word-edge case: [00000011] [110000000] where the RR == 1
+ bit_vector->clear();
+ bit_vector->setBitRange((sizeof(std::size_t) * 8) - 4, 8, true);
+
+ EXPECT_NEAR(0.125, bit_vector->runRatio(), 0.0001);
+ printf("run ratio: %f\n", bit_vector->runRatio());
+}
+
TYPED_TEST(BitVectorTest, SetBitRangeTest) {
std::unique_ptr<TypeParam> small_bit_vector(this->createBitVector(TestFixture::kSmallBitSize));
small_bit_vector->clear();
