From: Mark Abraham <mark.j.abraham@gmail.com>
Date: Sun, 16 Feb 2014 22:36:56 +0000 (+0100)
Subject: Fix outstanding references to "CPU acceleration"
X-Git-Url: http://biod.pnpi.spb.ru/gitweb/?a=commitdiff_plain;h=cc9ecd5b8a2a495c03f59f3d5dd77cb3327a42bf;p=alexxy%2Fgromacs.git

Fix outstanding references to "CPU acceleration"

Since we now use GMX_SIMD rather than GMX_CPU_ACCELERATION we
should be consistent.

Change-Id: I6a798870e5b89fc92bdbea18680e017101a2f704
---

diff --git a/cmake/gmxSetBuildInformation.cmake b/cmake/gmxSetBuildInformation.cmake
index cb1158d818..42a65bf977 100644
--- a/cmake/gmxSetBuildInformation.cmake
+++ b/cmake/gmxSetBuildInformation.cmake
@@ -1,7 +1,7 @@
 #
 # This file is part of the GROMACS molecular simulation package.
 #
-# Copyright (c) 2012,2013, by the GROMACS development team, led by
+# Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
 # Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
 # and including many others, as listed in the AUTHORS file in the
 # top-level source directory and at http://www.gromacs.org.
@@ -82,7 +82,7 @@ macro(gmx_set_build_information)
     endif(CMAKE_HOST_UNIX)
 
     if(NOT CMAKE_CROSSCOMPILING)
-        # Get CPU acceleration information
+        # Get CPU information, e.g. for deciding what SIMD support exists
         set(_compile_definitions "@GCC_INLINE_ASM_DEFINE@ -I${CMAKE_SOURCE_DIR}/src/gromacs/legacyheaders -DGMX_CPUID_STANDALONE")
         if(GMX_TARGET_X86)
             set(_compile_definitions "${_compile_definitions} -DGMX_TARGET_X86")
diff --git a/manual/install.tex b/manual/install.tex
index 8e49093273..227572a649 100644
--- a/manual/install.tex
+++ b/manual/install.tex
@@ -246,7 +246,7 @@ you should consult your local documentation for details.
 \item   {\tt GMX_DETAILED_PERF_STATS}: when set, print slightly more detailed performance information
         to the {\tt .log} file. The resulting output is the way performance summary is reported in versions
         4.5.x and thus may be useful for anyone using scripts to parse {\tt .log} files or standard output.
-\item   {\tt GMX_DISABLE_CPU_ACCELERATION}: disables CPU architecture-specific SIMD-optimized (SSE2, SSE4, AVX, etc.)
+\item   {\tt GMX_DISABLE_SIMD_KERNELS}: disables architecture-specific SIMD-optimized (SSE2, SSE4.1, AVX, etc.)
         non-bonded kernels thus forcing the use of plain C kernels.
 \item   {\tt GMX_DISABLE_CUDA_TIMING}: timing of asynchronously executed GPU operations can have a
         non-negligible overhead with short step times. Disabling timing can improve performance in these cases.
diff --git a/src/gromacs/mdlib/forcerec.c b/src/gromacs/mdlib/forcerec.c
index 8db994eb5f..d2c1643fea 100644
--- a/src/gromacs/mdlib/forcerec.c
+++ b/src/gromacs/mdlib/forcerec.c
@@ -2272,14 +2272,14 @@ void init_forcerec(FILE              *fp,
         bNoSolvOpt         = TRUE;
     }
 
-    if ( (getenv("GMX_DISABLE_CPU_ACCELERATION") != NULL) || (getenv("GMX_NOOPTIMIZEDKERNELS") != NULL) )
+    if ( (getenv("GMX_DISABLE_SIMD_KERNELS") != NULL) || (getenv("GMX_NOOPTIMIZEDKERNELS") != NULL) )
     {
         fr->use_simd_kernels = FALSE;
         if (fp != NULL)
         {
             fprintf(fp,
-                    "\nFound environment variable GMX_DISABLE_CPU_ACCELERATION.\n"
-                    "Disabling the usage of most SIMD-specific kernel (e.g. SSE2/SSE4/AVX) routines.\n\n");
+                    "\nFound environment variable GMX_DISABLE_SIMD_KERNELS.\n"
+                    "Disabling the usage of any SIMD-specific kernel routines (e.g. SSE2/SSE4.1/AVX).\n\n");
         }
     }
 
diff --git a/src/programs/mdrun/runner.c b/src/programs/mdrun/runner.c
index a6b838578a..153736f9b7 100644
--- a/src/programs/mdrun/runner.c
+++ b/src/programs/mdrun/runner.c
@@ -709,7 +709,7 @@ static void prepare_verlet_scheme(FILE                           *fplog,
         verletbuf_list_setup_t ls;
         real                   rlist_new;
 
-        /* Here we assume CPU acceleration is on. But as currently
+        /* Here we assume SIMD-enabled kernels are being used. But as currently
          * calc_verlet_buffer_size gives the same results for 4x8 and 4x4
          * and 4x2 gives a larger buffer than 4x4, this is ok.
          */
@@ -1538,7 +1538,8 @@ int mdrunner(gmx_hw_opt_t *hw_opt,
         hw_opt->gpu_opt.ncuda_dev_use = 0;
     }
 
-    /* check consistency of CPU acceleration and number of GPUs selected */
+    /* check consistency across ranks of things like SIMD
+     * support and number of GPUs selected */
     gmx_check_hw_runconf_consistency(fplog, hwinfo, cr, hw_opt, bUseGPU);
 
     if (DOMAINDECOMP(cr))