topic is that 32 or 64 bit? in Intel® oneAPI Math Kernel Library

11.3 Update 1 integer divide by zero when using MKL_DIRECT

AndrewC — Mon, 18 Jan 2016 17:35:31 GMT

I am using the "direct" calls set by MKL_DIRECT and am moving from MKL 11.2 to MKL 11.3 Update 1

I am finding my code is throwing a "integer divide by zero" exception at this call when using 11.3 Update 1. The "calling" code is well tested and worked fine in MKL 11.2.

If I "turn off" MKL_DIRECT, the code again works as expected.

	else if (!AisN && BisN)
		mkl_dc_gemm_xx_mnk_pst(m, n, k, alpha, A, lda, B, ldb, beta, C, ldc, MKL_DC_MOV, MKL_DC_MOV, MKL_DC_MT, MKL_DC_MN);

Hello,

Murat_G_Intel — Mon, 18 Jan 2016 22:34:48 GMT

Hello,

Could you please send the *GEMM call in your code that causes this exception? Also, which compiler version are you using?

Thank you,

Efe

I am using Intel Compiler 16

AndrewC — Mon, 18 Jan 2016 23:47:53 GMT

I am using Intel Compiler 16.0 Update 1 on Windows x64 ( Visual Studio 2015)

The call is a simple "dgemm'" call as below with m=4,n=4. An "integer divide by 0" exception must be a good clue as to possible issue,

This is a pretty simple regression case for me..

Used to work in MKL 11.2 with MKL_DIRECT.
Fails in 11.3 Update 1 MKL_DIRECT.
Works fine in 11.3 Update 1 when MKL_DIRECT is undefined.

dgemm( &transa, &transb, &m, &n,&k,&alpha, a, &lda, b, &ldb,&beta, c, &ldc );

is that 32 or 64 bit?

Gennady_F_Intel — Tue, 19 Jan 2016 03:06:29 GMT

is that 32 or 64 bit?

I tried MKL 11.3.1 direct

Murat_G_Intel — Tue, 19 Jan 2016 06:05:49 GMT

I tried MKL 11.3.1 direct call on a Linux system (should try on windows next). Although I do not see "integer divide by 0" exception, I do see incorrect results with the 16.0.1 compiler. I get incorrect results with both MKL 11.2.3 and MKL 11.3.1. If I switch to an older compiler (15.0.3), the errors for both MKL builds disappear.

Vasci, if this is easy on your side, could you try MKL 11.3.1 with an 15.0.* compiler to see if you still see an error with MKL 11.3.1? If you confirm that MKL 11.3.1 works fine with an 15.0 compiler, then we can confirm that this is a compiler related issue.

Thank you,

Efe

Unfortunately I am stuck with

AndrewC — Tue, 19 Jan 2016 14:31:47 GMT

Unfortunately I am stuck with Intel Compiler 16.0 Update 1. I am transitioning to Visual Studio 2015 Update 1, and Intel Compiler 15.0 is language incompatible with the VS 2015 Update 1 STL headers as the STL use some C++ features unsupported by IC 15.0. And IC 15.0 is not going to be updated anymore.

I will try to create a reproducer - at the moment the failure is happening inside a regression test..

@Gennady this is 64-bit

Hi Vasci,

Murat_G_Intel — Tue, 19 Jan 2016 18:31:21 GMT

Hi Vasci,

No problem, I thought this may be hard as well. After further investigation, we suspect that this may be related to the -openmp (Linux), /Qopenmp (Windows) compiler flag. If I link with libiomp5md.lib explicitly, I do not see the errors. Similarly, If I link with sequential library where we do not need to pass the openmp flags, I do not see any errors.

Thank you,

Efe

the problem is escalated. We

Gennady_F_Intel — Wed, 20 Jan 2016 06:10:09 GMT

the problem is escalated. We will keep you updated when the fix of the problem will be available.

We have a patched MKL direct

Murat_G_Intel — Sat, 23 Jan 2016 00:01:45 GMT

We have a patched MKL direct call header file that will work fine with the Intel 16.0 compiler. Please let us know if you are interested in trying this modified header file so that we can send it to you.

Yes, I am interested in this

AndrewC — Thu, 11 Feb 2016 22:28:53 GMT

Yes, I am interested in this patch. I replied to another e-mail re this, and it went to a black hole.

Has this been issue been

AndrewC — Tue, 03 May 2016 19:31:51 GMT

Has this been issue been corrected in the just released Parallel Studio 2016 Update 3?

Yes, it should be fixed in

Murat_G_Intel — Tue, 03 May 2016 20:33:51 GMT

Yes, it should be fixed in Update 3. It was fixed in MKL 11.3.3 release.

OK, great. The fix did not

AndrewC — Tue, 03 May 2016 20:45:34 GMT

OK, great. The fix did not seem to be mentioned in the release notes for Update 3.

Andrew, yes, the problem has

Gennady_F_Intel — Wed, 04 May 2016 05:27:27 GMT

Andrew, yes, the problem has been fixed into the latest update 3 ( MKL v. 11.3 ). I will update the corresponding bug fix list article asap.

btw Andrew, have you already

Gennady_F_Intel — Wed, 04 May 2016 05:46:03 GMT

btw Andrew, have you already tried how it works with MKL 11.3 update 3?

Not yet, just downloaded the

AndrewC — Wed, 04 May 2016 14:41:00 GMT

Not yet, just downloaded the full XE 2016 package and will try it out as soon as feasible.