This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH PR62178]Improve candidate selecting in IVOPT, 2nd try.


Bin.Cheng wrote:
> do we have some compilation time benchmarks for GCC?

I'm using the llvm test-suite to see compile time differences:

$ git clone http://llvm.org/git/test-suite.git /path/to/test-suite
$ /path/to/test-suite/configure --without-llvmsrc --without-llvmobj --with-externals=/path/to/spec
$ make -k TEST=simple TARGET_LLVMGCC=/path/to/gcc TARGET_CXX=/path/to/g++ TARGET_CC=/path/to/gcc TARGET_LLVMGXX=/path/to/g++ CC_UNDER_TEST_IS_GCC=1 TARGET_FLAGS=  USE_REFERENCE_OUTPUT=1 CC_UNDER_TEST_TARGET_IS_AARCH64=1 OPTFLAGS="-O3" LLC_OPTFLAGS="-O3" ENABLE_OPTIMIZED=1 ARCH=AArch64 ENABLE_HASHED_PROGRAM_OUTPUT=1 DISABLE_JIT=1 report report.simple.csv
$ head -1 report.simple.csv
Program,CC,CC_Time,CC_Real_Time,Exec,Exec_Time,Exec_Real_Time
$ awk -F, '{print $1, $3 }' < report.simple.csv

Here is how to get benchmark code size:
$ make -k TEST=codesize TARGET_LLVMGCC=/path/to/gcc TARGET_CXX=/path/to/g++ TARGET_CC=/path/to/gcc TARGET_LLVMGXX=/path/to/g++ TARGET_FLAGS= USE_REFERENCE_OUTPUT=1 CC_UNDER_TEST_TARGET_IS_AARCH64=1 CC_UNDER_TEST_IS_CLANG=1 OPTFLAGS="-O3" LLC_OPTFLAGS="-O3" ENABLE_OPTIMIZED=1 ARCH=AArch64 ENABLE_HASHED_PROGRAM_OUTPUT=1 DISABLE_JIT=1 2>/dev/null | grep ^size: > test.codesize.txt


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]