blas1 routines

[ks-fujii]

Hi,
I'd like to know whether MKL blas1/2 routines are threaded or not.
I think some routines are threaded, but some are not.
Based on my actual measurement, daxpy or ddot are threaded, but dnrm is not.
Do you have something like documents about threaded routine of blas1/2?

Reagards
Fujii

[Gennady_F_Intel]

Hi Fujii,
Please refer to the MKL user guide document, Chapter 6, ""Using the Intel MKL Parallelism"
you can there the list of MKL's treaded routines.
The snip from the userguide related blas1/2 see below:

Level1 and Level2 BLAS functions:
- Level1 BLAS: *axpy, *copy, *swap, ddot/sdot, drot/srot
- Level2 BLAS: *gemv, *trmv, dsyr/ssyr, dsyr2/ssyr2, dsymv/ssymv

Note that these functions are threaded only for:
- Intel 64 architecture
- Intel Core^TM2 Duo and Intel Core^TM i7 processors
********************************************************
Therefore, You are right, daxpy anf ddot are threaded but dnrm is not.
--Gennady

[Gennady_F_Intel]

For additional information (mailnly performance improvements) please refer to the MKL KB article called "Threaded BLAS Level 1 and Level 2 functions are available in Intel MKL 10.2 on Intel Xeon 5500 series (Nehalem) processors"

http://software.intel.com/en-us/articles/threaded-blas-level-1-and-2-on-nehalem/
--Gennady

[ks-fujii]

Quoting - Gennady Fedorov (Intel)

Hi Gennady,

I couldn't find any description about Level1 and 2 threaded routines on the MKL user guide document.
I checked the documents of MKL 10.1 and MKL 10.2-beta.
What is the version you mentioned?

Regards
Fujii

[Gennady_F_Intel]

Quoting - ks-fujii

Hi Fujii, I mentioned both versions.
For example: MKL Users Guide ( for the windows os),
See pp 6-1,6-2
--Gennady

[ks-fujii]

Hi,Gennady,

I checked 6-1,6-2 , but I couldn't.
The section is as follows.

>>>>
Using Intel MKL Parallelism
Intel MKL is threaded in a number of places:
Direct sparse solver
LAPACK
- Linear equations, computational routines:
- factorization: *getrf, *gbtrf, *potrf, *pptrf, *sytrf, *hetrf, *sptrf, *hptrf
- solving: *gbtrs, *gttrs, *pptrs, *pbtrs, *pttrs, *sytrs, *sptrs, *hptrs,*tptrs, *tbtrs
- Orthogonal factorization, computational routines:*geqrf, *ormqr, *unmqr, *ormlq, *unmlq, *ormql, *unmql, *ormrq, *unmrq
- Singular Value Decomposition, computational routines: *gebrd, *bdsqr
- Symmetric Eigenvalue Problems, computational routines:*sytrd, *hetrd, *sptrd, *hptrd, *steqr, *stedc
Note that a number of other LAPACK routines, which are based on threaded LAPACK or
BLAS routines, make effective use of parallelism: *gesv, *posv, *gels, *gesvd,*syev, *heev, etc.
All Level 3 BLAS, Sparse BLAS matrix-vector and matrix-matrix multiply routines forthe compressed sparse row and diagonal formats
VML
All FFTs (except 1D transformations when DFTI_NUMBER_OF_TRANSFORMS=1 and sizes are not power of two).
<<<<

[Gennady_F_Intel]

Opps, Iwas mistaken aboutthe version of user guide - I used MKL v.10.2 Gold version, coming soon.
I am sorry about that.
--Gennady

[ks-fujii]

Quoting - Gennady Fedorov (Intel)

Opps, Iwas mistaken aboutthe version of user guide - I used MKL v.10.2 Gold version, coming soon.
I am sorry about that.
--Gennady

Hi Gennady,

When will it be released?
I'm looking forward to it.

--
Fujii

[Gennady_F_Intel]

Quoting - ks-fujii

Hi Gennady,

When will it be released?
I'm looking forward to it.

--
Fujii

I can't provide you the exact data, but it will happen in the end of June or beginner of July
--Gennady