- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
I find a strange information. I use Intel(R) Instrinsic Guide produced by Intel.Software.
And I can read that :
Synopsis __m128d _mm_sqrt_sd (__m128d a)
#include "emmintrin.h"
Instruction:
sqrtsd CPUID Feature Flag: SSE2
Description:
Computes the square root of the lower double-precision floating-point value of a.
Operation:
r0 := sqrt(a0)
r1 := a1
Latency & Throughput Information
CPUID(s) Parameters Latency Throughput
0F_03 xmm, xmm 39 39
0F_02 xmm, xmm 38 38
06_2A xmm, xmm 22 22
06_25/2C/1A/1E/1F/2E xmm, xmm 34 30
06_17/1D xmm, xmm 31 25
06_0F xmm, xmm 60 57
06_0E xmm, xmm 58 57
06_0D xmm, xmm 58 57
Ok ! Interesting. But when I read emmintrin.h I saw that :
extern __m128d __ICL_INTRINCC _mm_sqrt_sd(__m128d, __m128d);
How it's possible ? A sqrt scalar for double with two value... ? What is behavior expected ?
Thanks for you answers
Regards
PS : Intel Parallel Studio XE 2013
Intel(R) Guide Instrinsic 2.8
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page