diff options
| author |  Eugene Zhulenev <ezhulenev@google.com> | 2018-12-03 17:10:40 -0800 |
|---|---|---|
| committer | Eugene Zhulenev <ezhulenev@google.com> | 2018-12-03 17:10:40 -0800 |
| commit | 0bb15bb6d6a445edb9341c4970d854882347b6d7 (patch) | |
| tree | d3b5384fda538f24e2245dbe09906b9965e414dd /Eigen/src/Core/util/ConfigureVectorization.h | |
| parent | fd0fbfa9b5301e5339c34846c76835cf347ef4cb (diff) | |
Update checks in ConfigureVectorization.h
Diffstat (limited to 'Eigen/src/Core/util/ConfigureVectorization.h')
| -rw-r--r-- | Eigen/src/Core/util/ConfigureVectorization.h | 18 |
1 files changed, 8 insertions, 10 deletions
diff --git a/Eigen/src/Core/util/ConfigureVectorization.h b/Eigen/src/Core/util/ConfigureVectorization.h index c482a0b14..c70ad894e 100644 --- a/Eigen/src/Core/util/ConfigureVectorization.h +++ b/Eigen/src/Core/util/ConfigureVectorization.h @@ -49,7 +49,13 @@ // If the user explicitly disable vectorization, then we also disable alignment #if defined(EIGEN_DONT_VECTORIZE) - #define EIGEN_IDEAL_MAX_ALIGN_BYTES 0 + #if defined(EIGEN_GPUCC) + // GPU code is always vectorized and requires memory alignment for + // statically allocated buffers. + #define EIGEN_IDEAL_MAX_ALIGN_BYTES 16 + #elif + #define EIGEN_IDEAL_MAX_ALIGN_BYTES 0 + #endif #elif defined(__AVX512F__) // 64 bytes static alignment is preferred only if really required #define EIGEN_IDEAL_MAX_ALIGN_BYTES 64 @@ -176,14 +182,6 @@ //---------------------------------------------------------------------- -// If we are compiling for GPU we should also disable vectorization because -// all the packet functions are not marked as __device__ functions. -#ifdef EIGEN_GPUCC -#ifndef EIGEN_DONT_VECTORIZE - #define EIGEN_DONT_VECTORIZE - #endif -#endif - // if alignment is disabled, then disable vectorization. Note: EIGEN_MAX_ALIGN_BYTES is the proper check, it takes into // account both the user's will (EIGEN_MAX_ALIGN_BYTES,EIGEN_DONT_ALIGN) and our own platform checks #if EIGEN_MAX_ALIGN_BYTES==0 @@ -210,7 +208,7 @@ #endif -#ifndef EIGEN_DONT_VECTORIZE +#if !(defined(EIGEN_DONT_VECTORIZE) || defined(EIGEN_GPUCC)) #if defined (EIGEN_SSE2_ON_NON_MSVC_BUT_NOT_OLD_GCC) || defined(EIGEN_SSE2_ON_MSVC_2008_OR_LATER) |