Compiler Usage Guidelines for 64-Bit Operating Systems on AMD64 ...

More documents

Recommendations

Info

<strong>Compiler</strong> <strong>Usage</strong> <strong>Guidelines</strong> <strong>for</strong> AMD<strong>64</strong> Plat<strong>for</strong>ms The GCC 4.0 and later version compilers can per<strong>for</strong>m loop vectorization by using the -ftree-vectorize flag. 3.2.4 Other Switches 32035 Rev. 3.22 November 2007 In addition to the switches mentioned in Table 3 on page 23, the following list of switches may also improve the program per<strong>for</strong>mance. It is worth experimenting with these switches. -march=k8. For the FSF GCC 4.2.0 SuSE 4.2.0and Red Hat 4.2.0 compilers, using this switch may give you a per<strong>for</strong>mance advantage in some cases. -march=amdfam10. For applications to be executed on AMD Family 10h processor-based plat<strong>for</strong>ms, this switch results in better per<strong>for</strong>mance. Note: The amdfam10 option is not available on all GCC compiler releases. See your compiler documentation <strong>for</strong> further in<strong>for</strong>mation. Profile Guided Optimization. The <strong>64</strong>-bit GCC compiler also allows profile guided optimization. Table 4 shows the profile guided optimization switches <strong>for</strong> the different GCC compilers. Table 4. Profile Guided Optimization <strong>for</strong> <strong>64</strong>-<strong>Bit</strong> GCC <strong>Compiler</strong>s <strong>for</strong> Linux ® <strong>Compiler</strong> Version Optimization Switches SuSE GCC 4.2.0 Step 1.Compile the program with -fprofile-arcs. and Red Hat gcc-ssa Step 2.Run the executable produced in Step 1. Running the (<strong>for</strong> C/C++ and Fortran) executable generates several files with profile in<strong>for</strong>mation (*.da). Step 3.Recompile the program with -fbranch-probabilities. FSF GCC 4.2.0 and Step 1.Compile the program with -fprofile-generate. Red Hat GCC 4.2.0 Step 2.Run the executable produced in Step 1. Running the executable generates several files with profile in<strong>for</strong>mation (*.da). Step 3.Recompile the program with -fprofile-use. -Bsymbolic. Sarting from GCC 4.1, gcc compiler no longer requires the -Bsymbolic switch. GCC 4.1 and later versions offer -combine -fwhole -program, which should be used together, but require that makefiles be changed to use a single command to compile and link all files of an application, slowing down builds. So it should only be used <strong>for</strong> non-debug builds. Un<strong>for</strong>tunately, these options may fail compiling some files. -minline-all-stringops. When using the GCC 3.4 compiler on Red Hat Enterprise Linux 4, experiment with the switch -minline-all-stringops. This switch is not recommended <strong>for</strong> GCC 3.4 on SuSE Linux Enterprise Server. Linking with ACML. The AMD Core Math Library (ACML) includes BLAS, LAPACK and FFT routines that are optimized <strong>for</strong> AMD Athlon <strong>64</strong>, AMD Opteron, and AMD Family 10h processors. If the program uses these routines, using ACML in place of generic C/Fortran 24 Per<strong>for</strong>mance-Centric <strong>Compiler</strong> Switches Chapter 3
32035 Rev. 3.22 November 2007 <strong>Compiler</strong> <strong>Usage</strong> <strong>Guidelines</strong> <strong>for</strong> AMD<strong>64</strong> Plat<strong>for</strong>ms implementation may greatly improve the per<strong>for</strong>mance. For additional details on how to install this library and use it, see http://developer.amd.com/assets/acml_userguide.pdf. -fno-rtti. This switch disables generation of in<strong>for</strong>mation about every class, with virtual functions, <strong>for</strong> use by the C++ runtime type identification features (dynamic_cast and typeid). If the user does not use those parts of the language, some space can be conserved by using this switch. Generate 32-<strong>Bit</strong> Binaries with -m32. The user can also use <strong>64</strong>-bit GCC compilers to generate 32-bit binaries by using the -m32 switch. This can improve per<strong>for</strong>mance if the program has lots of variables of the type long and/or pointers. As these data-types are 32-bit in x86, this switch will reduce the memory footprint of the program. Also, the user should use the recommended switches <strong>for</strong> the 32-bit GCC compiler (section 3.8.4 on page 32) when -m32 is used. Users can obtain more details on these switches by trying info gcc on their Linux systems. 3.3 Intel <strong>Compiler</strong>s (<strong>64</strong>-<strong>Bit</strong>) <strong>for</strong> Linux ® Intel provides <strong>64</strong>-bit compilers <strong>for</strong> Linux that can be used <strong>for</strong> AMD<strong>64</strong> systems. The current version (as of Augus 2007) is 10.0. All options described in this section apply to this version. 3.3.1 Invocation Commands The following commands invoke specific compilers: icpc invokes the C++ compiler. icc invokes the C compiler. i<strong>for</strong>t invokes the Fortran compiler. 3.3.2 Generic Per<strong>for</strong>mance Switches The switches -xW -ipo -O3 -static are generally recommended. 3.3.3 Other Switches In addition to the generic per<strong>for</strong>mance switches, it is worth experimenting with the following switches. Profile Guided Optimization. Intel compilers allow profile guided optimization. Use the following steps <strong>for</strong> profile guided optimization with Intel compilers. 1. Compile the program with the -prof_gen switch. The -ipo or -ip switch is ignored by the compiler if used with -prof_gen. 2. Run the executable produced in Step 1. Running this executable generates several files with profile in<strong>for</strong>mation (*.dyn and *.dpi). 3. Recompile the program with the -prof_use switch. It is recommended to also use the -ipo switch in this stage. Chapter 3 Per<strong>for</strong>mance-Centric <strong>Compiler</strong> Switches 25
Page 1 and 2: Compiler U
Page 3 and 4: 32035 Rev. 3.22 November 2007 Conte
Page 5 and 6: 32035 Rev. 3.22 November 2007 <stro
Page 9 and 10: 32035 Rev. 3.22 November 2007 Table
Page 11 and 12: 32035 Rev. 3.22 November 2007 Revis
Page 13 and 14: 32035 Rev. 3.22 November 2007 Chapt
Page 23: 32035 Rev. 3.22 November 2007 <stro
Page 29 and 30: 32035 Rev. 3.22 November 2007 3.7 S
Page 31 and 32: 32035 Rev. 3.22 November 2007 Table
Page 39 and 40: 32035 Rev. 3.22 November 2007 3.13.
Page 45 and 46: 32035 Rev. 3.22 November 2007 4.2.2
Page 53 and 54: 32035 Rev. 3.22 November 2007 4.12.
Page 57 and 58: 32035 Rev. 3.22 November 2007 5.1.2

Compiler Usage Guidelines for 64-Bit Operating Systems on AMD64 ...

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?