Potential bug in double_m operator==

[Matthias Kretz]

On Thursday 12 June 2014 14:37:55 Georgios Bitzes wrote:
> Latest commit (b7f0e8ce59ff866894a8cdb33ce7d53cbfc737bf) fixes the issue
> for AVX and AVX2, but not SSE4.. It's strange that you cannot reproduce -
> I'm configuring Vc with:
> cmake -DCMAKE_INSTALL_PREFIX=<..> -DBUILD_TESTING=OFF <..>

OK, but that doesn't change the code in question. Can you run the mask_sse 
unit test? It's where I added your test case. Because mask_sse does not fail 
for me with SSE4 enabled...

> Compiler version is "Ubuntu/Linaro 4.7.3-2ubuntu4" on a Linux Mint 14 with
> a E3-1280 v3. Also tested on icpc 14.0.2.

I'm using GCC 4.8, but I also don't see a failure on the dashboard.

> By the way, (b == a) produces the correct result for SSE4, too ..

That's a strong indication that its using the testps instruction. But I can't 
find a call for it in sse/*.

-Matthias

-- 
─────────────────────────────────────────────────────────────
 Dipl.-Phys. Matthias Kretz

 Web:   http://compeng.uni-frankfurt.de/?mkretz

 SIMD easy and portable: http://compeng.uni-frankfurt.de/?vc
─────────────────────────────────────────────────────────────