AVX向量化学习(三)-if判断的处理

AVX-if判断的处理

使用AVX指令集对if判断进行处理

使用到的AVX函数介绍

1.

__m256d _mm256_blendv_pd (__m256d a, __m256d b, __m256d mask)

Description

Blend packed double-precision (64-bit) floating-point elements from a and b using mask, and store the results in dst.

Operation

FOR j := 0 to 3
	i := j*64
	IF mask[i+63]
		dst[i+63:i] := b[i+63:i]
	ELSE
		dst[i+63:i] := a[i+63:i]
	FI
ENDFOR
dst[MAX:256] := 0

2.

__m256d _mm256_cmp_pd (__m256d a, __m256d b, const int imm8)

Description

Compare packed double-precision (64-bit) floating-point elements in a and b based on the comparison operand specified by imm8, and store the results in dst.

Operation

CASE (imm8[4:0]) OF
0: OP := _CMP_EQ_OQ
1: OP := _CMP_LT_OS
2: OP := _CMP_LE_OS
3: OP := _CMP_UNORD_Q 
4: OP := _CMP_NEQ_UQ
5: OP := _CMP_NLT_US
6: OP := _CMP_NLE_US
7: OP := _CMP_ORD_Q
8: OP := _CMP_EQ_UQ
9: OP := _CMP_NGE_US
10: OP := _CMP_NGT_US
11: OP := _CMP_FALSE_OQ
12: OP := _CMP_NEQ_OQ
13: OP := _CMP_GE_OS
14: OP := _CMP_GT_OS
15: OP := _CMP_TRUE_UQ
16: OP := _CMP_EQ_OS
17: OP := _CMP_LT_OQ
18: OP := _CMP_LE_OQ
19: OP := _CMP_UNORD_S
20: OP := _CMP_NEQ_US
21: OP := _CMP_NLT_UQ
22: OP := _CMP_NLE_UQ
23: OP := _CMP_ORD_S
24: OP := _CMP_EQ_US
25: OP := _CMP_NGE_UQ 
26: OP := _CMP_NGT_UQ 
27: OP := _CMP_FALSE_OS 
28: OP := _CMP_NEQ_OS 
29: OP := _CMP_GE_OQ
30: OP := _CMP_GT_OQ
31: OP := _CMP_TRUE_US
ESAC
FOR j := 0 to 3
	i := j*64
	dst[i+63:i] := ( a[i+63:i] OP b[i+63:i] ) ? 0xFFFFFFFFFFFFFFFF : 0
ENDFOR
dst[MAX:256] := 0

程序源代码

#include<stdio.h>
#include <immintrin.h>
int main()
{
	double a[9]={1.1,2.2,3.3,4.4,5.5,6.6,7.7,8.8,2.1};
	double b[9]={2.1,3.2,6.4,8.6,3.7,9.9,5.1,4.2,6.6};
	double d[9]={0};     //记录原始if判断后的值
	double e[9]={0};     //记录AVX-if判断后的值
	
	__m256d v0;
	__m256d v1;
	__m256d v2,v3;
	__m256d v4;
	
	for(int i=0;i<9;i++)
	{
		if(a[i]>b[i])
		{
			d[i] = a[i];
		}
		else
		{
			d[i]=b[i];
		}
	}

	int i=0;
	
	for(;i<9-4;i+=4)
	{	
			v0 = _mm256_loadu_pd(a+i);
			v1 = _mm256_loadu_pd(b+i);
			v2=_mm256_add_pd(v0,v1);
			v3 =_mm256_blendv_pd(v0,v1,_mm256_cmp_pd(v0,v1,_CMP_LE_OQ));
			_mm256_storeu_pd(e+i,v3);
			
	}
    
	for(;i<9;i++)
	{
		if(a[i]>b[i])
		{
			e[i] = a[i];
		}
		else
		{
			e[i]=b[i];
		}
	}
    
	printf("this is d.\n");
		for(int i=0;i<9;i++)
	{
		printf("%lf\n",d[i]);
	}
    
	printf("this is e with AVX.\n");
		for(int i=0;i<9;i++)
	{
		printf("%lf\n",e[i]);
	}
	return 0;
 } 

程序输出

this is d.
2.100000
3.200000
6.400000
8.600000
5.500000
9.900000
7.700000
8.800000
6.600000
this is e with AVX.
2.100000
3.200000
6.400000
8.600000
5.500000
9.900000
7.700000
8.800000
6.600000

相关链接

[https://software.intel.com/sites/landingpage/IntrinsicsGuide/]: “ Intel® Intrinsics Guide”

[simd - How to choose AVX compare predicate variants - Stack Overflow]: “Stack Overflow”

[blendvpd (officedaytime.com)]: “_mm256_Blendv_pd()原理解释”