thevar1able/ClickHouse
1
0
Fork 0
mirror of https://github.com/ClickHouse/ClickHouse.git synced 2024-09-24 02:30:51 +00:00
Code Issues Packages Projects Releases Wiki Activity

dbms: Server: Use Intel SIMD instructions in order to optimize round/ceil/floor. [#METR-15210]

Browse Source
This commit is contained in:
Alexey Arno 2015-05-19 02:02:19 +03:00
parent dec1c493aa
commit 10bcfe73d1
1 changed files with 31 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

44
dbms/include/DB/Functions/FunctionsRound.h
View File

@ -100,15 +100,19 @@ namespace
struct Rounding32
{
using Data = std::tuple<Float32, Float32, Float32, Float32>;
using Scale = __m128;
static inline void apply(const Data & in, size_t scale, Data & out)
static inline void prepareScale(size_t scale, Scale & mm_scale)
{
Float32 fscale = static_cast<Float32>(scale);
mm_scale = _mm_load1_ps(&fscale);
}
static inline void apply(const Data & in, const Scale & mm_scale, Data & out)
{
Float32 input[4] __attribute__((aligned(16))) = { std::get<0>(in), std::get<1>(in), std::get<2>(in), std::get<3>(in) };
__m128 mm_value = _mm_load_ps(input);
Float32 fscale = static_cast<Float32>(scale);
__m128 mm_scale = _mm_load1_ps(&fscale);
mm_value = _mm_mul_ps(mm_value, mm_scale);
mm_value = _mm_round_ps(mm_value, rounding_mode);
mm_value = _mm_div_ps(mm_value, mm_scale);
@ -123,15 +127,19 @@ namespace
struct Rounding64
{
using Data = std::tuple<Float64, Float64>;
using Scale = __m128d;
static inline void apply(const Data & in, size_t scale, Data & out)
static inline void prepareScale(size_t scale, Scale & mm_scale)
{
Float64 fscale = static_cast<Float64>(scale);
mm_scale = _mm_load1_pd(&fscale);
}
static inline void apply(const Data & in, const Scale & mm_scale, Data & out)
{
Float64 input[2] __attribute__((aligned(16))) = { std::get<0>(in), std::get<1>(in) };
__m128d mm_value = _mm_load_pd(input);
Float64 fscale = static_cast<Float64>(scale);
__m128d mm_scale = _mm_load1_pd(&fscale);
mm_value = _mm_mul_pd(mm_value, mm_scale);
mm_value = _mm_round_pd(mm_value, rounding_mode);
mm_value = _mm_div_pd(mm_value, mm_scale);
@ -164,10 +172,13 @@ namespace
{
using Op = Rounding32<rounding_mode>;
using Data = typename Op::Data;
using Scale = typename Op::Scale;
static inline void apply(const PODArray<Float32> & in, UInt8 precision, typename ColumnVector<Float32>::Container_t & out)
{
size_t scale = PowersTable::values[precision];
Scale scale;
Op::prepareScale(PowersTable::values[precision], scale);
size_t size = in.size();
size_t i;
@ -209,7 +220,9 @@ namespace
return val;
else
{
size_t scale = PowersTable::values[precision];
Scale scale;
Op::prepareScale(PowersTable::values[precision], scale);
Data res;
Op::apply(std::make_tuple(val, 0, 0, 0), scale, res);
return std::get<0>(res);
@ -222,10 +235,13 @@ namespace
{
using Op = Rounding64<rounding_mode>;
using Data = typename Op::Data;
using Scale = typename Op::Scale;
static inline void apply(const PODArray<Float64> & in, UInt8 precision, typename ColumnVector<Float64>::Container_t & out)
{
size_t scale = PowersTable::values[precision];
Scale scale;
Op::prepareScale(PowersTable::values[precision], scale);
size_t size = in.size();
size_t i;
@ -250,7 +266,9 @@ namespace
return val;
else
{
size_t scale = PowersTable::values[precision];
Scale scale;
Op::prepareScale(PowersTable::values[precision], scale);
Data res;
Op::apply(std::make_tuple(val, 0), scale, res);
return std::get<0>(res);
@ -362,7 +380,7 @@ namespace
static IFunction * create(const Context & context) { return new FunctionRounding; }
private:
using PowersOf10 = FillArray<std::numeric_limits<DB::Float64>::digits10 + 1>::result;
using PowersOf10 = FillArray<std::numeric_limits<Float64>::digits10 + 1>::result;
private:
template<typename T>