You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.

getarch.c 43 kB

rebase? (#1) * With the Intel compiler on Linux, prefer ifort for the final link step icc has known problems with mixed-language builds that ifort can handle just fine. Fixes #1956 * Rename operands to put lda on the input/output constraint list * Fix wrong constraints in inline assembly for #2009 * Fix inline assembly constraints rework indices to allow marking argument lda4 as input and output. For #2009 * Fix inline assembly constraints rework indices to allow marking argument lda as input and output. * Fix inline assembly constraints * Fix inline assembly constraints * Fix inline assembly constraints in Bulldozer TRSM kernels rework indices to allow marking i,as and bs as both input and output (marked operand n1 as well for simplicity). For #2009 * Correct range_n limiting same bug as seen in #1388, somehow missed in corresponding PR #1389 * Allow multithreading TRMV again revert workaround introduced for issue #1332 as the actual cause appears to be my incorrect fix from #1262 (see #1388) * Fix error introduced during cleanup * Reduce list of kernels in the dynamic arch build to make compilation complete reliably within the 1h limit again * init * move fix to right place * Fix missing -c option in AVX512 test * Fix AVX512 test always returning false due to missing compiler option * Make x86_32 imply NO_AVX2, NO_AVX512 in addition to NO_AVX fixes #2033 * Keep xcode8.3 for osx BINARY=32 build as xcode10 deprecated i386 * Make sure that AVX512 is disabled in 32bit builds for #2033 * Improve handling of NO_STATIC and NO_SHARED to avoid surprises from defining either as zero. Fixes #2035 by addressing some concerns from #1422 * init * address warning introed with #1814 et al * Restore locking optimizations for OpenMP case restore another accidentally dropped part of #1468 that was missed in #2004 to address performance regression reported in #1461 * HiSilicon tsv110 CPUs optimization branch add HiSilicon tsv110 CPUs optimization branch * add TARGET support for HiSilicon tsv110 CPUs * add TARGET support for HiSilicon tsv110 CPUs * add TARGET support for HiSilicon tsv110 CPUs * Fix module definition conflicts between LAPACK and ReLAPACK for #2043 * Do not compile in AVX512 check if AVX support is disabled xgetbv is function depends on NO_AVX being undefined - we could change that too, but that combo is unlikely to work anyway * ctest.c : add __POWERPC__ for PowerMac * Fix crash in sgemm SSE/nano kernel on x86_64 Fix bug #2047. Signed-off-by: Celelibi <celelibi@gmail.com> * param.h : enable defines for PPC970 on DarwinOS fixes: gemm.c: In function 'sgemm_': ../common_param.h:981:18: error: 'SGEMM_DEFAULT_P' undeclared (first use in this function) #define SGEMM_P SGEMM_DEFAULT_P ^ * common_power.h: force DCBT_ARG 0 on PPC970 Darwin without this, we see ../kernel/power/gemv_n.S:427:Parameter syntax error and many more similar entries that relates to this assembly command dcbt 8, r24, r18 this change makes the DCBT_ARG = 0 and openblas builds through to completion on PowerMac 970 Tests pass * Make TARGET=GENERIC compatible with DYNAMIC_ARCH=1 for issue #2048 * make DYNAMIC_ARCH=1 package work on TSV110. * make DYNAMIC_ARCH=1 package work on TSV110 * Add Intel Denverton for #2048 * Add Intel Denverton * Change 64-bit detection as explained in #2056 * Trivial typo fix as suggested in #2022 * Disable the AVX512 DGEMM kernel (again) Due to as yet unresolved errors seen in #1955 and #2029 * Use POSIX getenv on Cygwin The Windows-native GetEnvironmentVariable cannot be relied on, as Cygwin does not always copy environment variables set through Cygwin to the Windows environment block, particularly after fork(). * Fix for #2063: The DllMain used in Cygwin did not run the thread memory pool cleanup upon THREAD_DETACH which is needed when compiled with USE_TLS=1. * Also call CloseHandle on each thread, as well as on the event so as to not leak thread handles. * AIX asm syntax changes needed for shared object creation * power9 makefile. dgemm based on power8 kernel with following changes : 32x unrolled 16x4 kernel and 8x4 kernel using (lxv stxv butterfly rank1 update). improvement from 17 to 22-23gflops. dtrmm cases were added into dgemm itself * Expose CBLAS interfaces for I?MIN and I?MAX * Build CBLAS interfaces for I?MIN and I?MAX * Add declarations for ?sum and cblas_?sum * Add interface for ?sum (derived from ?asum) * Add ?sum * Add implementations of ssum/dsum and csum/zsum as trivial copies of asum/zsasum with the fabs calls replaced by fmov to preserve code structure * Add ARM implementations of ?sum (trivial copies of the respective ?asum with the fabs calls removed) * Add ARM64 implementations of ?sum as trivial copies of the respective ?asum kernels with the fabs calls removed * Add ia64 implementation of ?sum as trivial copy of asum with the fabs calls removed * Add MIPS implementation of ?sum as trivial copy of ?asum with the fabs calls removed * Add MIPS64 implementation of ?sum as trivial copy of ?asum with the fabs replaced by mov to preserve code structure * Add POWER implementation of ?sum as trivial copy of ?asum with the fabs replaced by fmr to preserve code structure * Add SPARC implementation of ?sum as trivial copy of ?asum with the fabs replaced by fmov to preserve code structure * Add x86 implementation of ?sum as trivial copy of ?asum with the fabs calls removed * Add x86_64 implementation of ?sum as trivial copy of ?asum with the fabs calls removed * Add ZARCH implementation of ?sum as trivial copies of the respective ?asum kernels with the ABS and vflpsb calls removed * Detect 32bit environment on 64bit ARM hardware for #2056, using same approach as #2058 * Add cmake defaults for ?sum kernels * Add ?sum * Add ?sum definitions for generic kernel * Add declarations for ?sum * Add -lm and disable EXPRECISION support on *BSD fixes #2075 * Add in runtime CPU detection for POWER. * snprintf define consolidated to common.h * Support INTERFACE64=1 * Add support for INTERFACE64 and fix XERBLA calls 1. Replaced all instances of "int" with "blasint" 2. Added string length as "hidden" third parameter in calls to fortran XERBLA * Correct length of name string in xerbla call * Avoid out-of-bounds accesses in LAPACK EIG tests see https://github.com/Reference-LAPACK/lapack/issues/333 * Correct INFO=4 condition * Disable reallocation of work array in xSYTRF as it appears to cause memory management problems (seen in the LAPACK tests) * Disable repeated recursion on Ab_BR in ReLAPACK xGBTRF due to crashes in LAPACK tests * sgemm/strmm * Update Changelog with changes from 0.3.6 * Increment version to 0.3.7.dev * Increment version to 0.3.7.dev * Misc. typo fixes Found via `codespell -q 3 -w -L ith,als,dum,nd,amin,nto,wis,ba -S ./relapack,./kernel,./lapack-netlib` * Correct argument of CPU_ISSET for glibc <2.5 fixes #2104 * conflict resolve * Revert reference/ fixes * Revert Changelog.txt typos * Disable the SkyLakeX DGEMMITCOPY kernel as well as a stopgap measure for https://github.com/numpy/numpy/issues/13401 as mentioned in #1955 * Disable DGEMMINCOPY as well for now #1955 * init * Fix errors in cpu enumeration with glibc 2.6 for #2114 * Change two http links to https Closes #2109 * remove redundant code #2113 * Set up CI with Azure Pipelines [skip ci] * TST: add native POWER8 to CI * add native POWER8 testing to Travis CI matrix with ppc64le os entry * Update link to IBM MASS library, update cpu support status * first try migrating one of the arm builds from travis * fix tabbing in azure commands * Update azure-pipelines.yml take out offending lines (although stolen from https://github.com/conda-forge/opencv-feedstock azure-pipelines fiie) * Update azure-pipelines.yml * Update azure-pipelines.yml * Update azure-pipelines.yml * Update azure-pipelines.yml * DOC: Add Azure CI status badge * Add ARMV6 build to azure CI setup (#2122) using aytekinar's Alpine image and docker script from the Travis setup [skip ci] * TST: Azure manylinux1 & clean-up * remove some of the steps & comments from the original Azure yml template * modify the trigger section to use develop since OpenBLAS primarily uses this branch; use the same batching behavior as downstream projects NumPy/ SciPy * remove Travis emulated ARMv6 gcc build because this now happens in Azure * use documented Ubuntu vmImage name for Azure and add in a manylinux1 test run to the matrix [skip appveyor] * Add NO_AFFINITY to available options on Linux, and set it to ON to match the gmake default. Fixes second part of #2114 * Replace ISMIN and ISAMIN kernels on all x86_64 platforms (#2125) * Mark iamax_sse.S as unsuitable for MIN due to issue #2116 * Use iamax.S rather than iamax_sse.S for ISMIN/ISAMIN on all x86_64 as workaround for #2116 * Move ARMv8 gcc build from Travis to Azure * Move ARMv8 gcc build from Travis to Azure * Update .travis.yml * Test drone CI * install make * remove sudo * Install gcc * Install perl * Install gfortran and add a clang job * gfortran->gcc-gfortran * Switch to ubuntu and parallel jobs * apt update * Fix typo * update yes * no need of gcc in clang build * Add a cmake build as well * Add cmake builds and print options * build without lapack on cmake * parallel build * See if ubuntu 19.04 fixes the ICE * Remove qemu armv8 builds * arm32 build * Fix typo * TST: add SkylakeX AVX512 CI test * adapt the C-level reproducer code for some recent SkylakeX AVX512 kernel issues, provided by Isuru Fernando and modified by Martin Kroeker, for usage in the utest suite * add an Intel SDE SkylakeX emulation utest run to the Azure CI matrix; a custom Docker build was required because Ubuntu image provided by Azure does not support AVX512VL instructions * Add option USE_LOCKING for single-threaded build with locking support for calling from concurrent threads * Add option USE_LOCKING for single-threaded build with locking support * Add option USE_LOCKING for SMP-like locking in USE_THREAD=0 builds * Add option USE_LOCKING but keep default settings intact * Remove unrelated change * Do not try ancient PGI hacks with recent versions of that compiler should fix #2139 * Build and run utests in any case, they do their own checks for fortran availability * Add softfp support in min/max kernels fix for #1912 * Revert "Add softfp support in min/max kernels" * Separate implementations of AMAX and IAMAX on arm As noted in #1912 and comment on #1942, the combined implementation happens to "do the right thing" on hardfp, but cannot return both value and index on softfp where they would have to share the return register * Ensure correct output for DAMAX with softfp * Use generic kernels for complex (I)AMAX to support softfp * improved zgemm power9 based on power8 * upload thread safety test folder * hook up c++ thread safety test (main Makefile) * add c++ thread test option to Makefile.rule * Document NO_AVX512 for #2151 * sgemm pipeline improved, zgemm rewritten without inner packs, ABI lxvx v20 fixed with vs52 * Fix detection of AVX512 capable compilers in getarch 21eda8b5 introduced a check in getarch.c to test if the compiler is capable of AVX512. This check currently fails, since the used __AVX2__ macro is only defined if getarch itself was compiled with AVX2/AVX512 support. Make sure this is the case by building getarch with -march=native on x86_64. It is only supposed to run on the build host anyway. * c_check: Unlink correct file * power9 zgemm ztrmm optimized * conflict resolve * Add gfortran workaround for ABI violations in LAPACKE for #2154 (see gcc bug 90329) * Add gfortran workaround for ABI violations for #2154 (see gcc bug 90329) * Add gfortran workaround for potential ABI violation for #2154 * Update fc.cmake * Remove any inadvertent use of -march=native from DYNAMIC_ARCH builds from #2143, -march=native precludes use of more specific options like -march=skylake-avx512 in individual kernels, and defeats the purpose of dynamic arch anyway. * Avoid unintentional activation of TLS code via USE_TLS=0 fixes #2149 * Do not force gcc options on non-gcc compilers fixes compile failure with pgi 18.10 as reported on OpenBLAS-users * Update Makefile.x86_64 * Zero ecx with a mov instruction PGI assembler does not like the initialization in the constraints. * Fix mov syntax * new sgemm 8x16 * Update dtrmm_kernel_16x4_power8.S * PGI compiler does not like -march=native * Fix build on FreeBSD/powerpc64. Signed-off-by: Piotr Kubaj <pkubaj@anongoth.pl> * Fix build for PPC970 on FreeBSD pt. 1 FreeBSD needs DCBT_ARG=0 as well. * Fix build for PPC970 on FreeBSD pt.2 FreeBSD needs those macros too. * cgemm/ctrmm power9 * Utest needs CBLAS but not necessarily FORTRAN * Add mingw builds to Appveyor config * Add getarch flags to disable AVX on x86 (and other small fixes to match Makefile behaviour) * Make disabling DYNAMIC_ARCH on unsupported systems work needs to be unset in the cache for the change to have any effect * Mingw32 needs leading underscore on object names (also copy BUNDERSCORE settings for FORTRAN from the corresponding Makefile)
7 years ago
15 years ago
11 years ago
Simplifying ARMv8 build parameters ARMv8 builds were a bit mixed up, with ThunderX2 code in ARMv8 mode (which is not right because TX2 is ARMv8.1) as well as requiring a few redundancies in the defines, making it harder to maintain and understand what core has what. A few other minor issues were also fixed. Tests were made on the following cores: A53, A57, A72, Falkor, ThunderX, ThunderX2, and XGene. Tests were: OpenBLAS/test, OpenBLAS/benchmark, BLAS-Tester. A summary: * Removed TX2 code from ARMv8 build, to make sure it is compatible with all ARMv8 cores, not just v8.1. Also, the TX2 code has actually harmed performance on big cores. * Commoned up ARMv8 architectures' defines in params.h, to make sure that all will benefit from ARMv8 settings, in addition to their own. * Adding a few more cores, using ARMv8's include strategy, to benefit from compiler optimisations using mtune. Also updated cache information from the manuals, making sure we set good conservative values by default. Removed Vulcan, as it's an alias to TX2. * Auto-detecting most of those cores, but also updating the forced compilation in getarch.c, to make sure the parameters are the same whether compiled natively or forced arch. Benefits: * ARMv8 build is now guaranteed to work on all ARMv8 cores * Improved performance for ARMv8 builds on some cores (A72, Falkor, ThunderX1 and 2: up to 11%) over current develop * Improved performance for *all* cores comparing to develop branch before TX2's patch (9% ~ 36%) * ThunderX1 builds are 14% faster than ARMv8 on TX1, 9% faster than current develop's branch and 8% faster than deveop before tx2 patches Issues: * Regression from current develop branch for A53 (-12%) and A57 (-3%) with ARMv8 builds, but still faster than before TX2's commit (+15% and +24% respectively). This can be improved with a simplification of TX2's code, to be done in future patches. At least the code is guaranteed to be ARMv8.0 now. Comments: * CortexA57 builds are unchanged on A57 hardware from develop's branch, which makes sense, as it's untouched. * CortexA72 builds improve over A57 on A72 hardware, even if they're using the same includes due to new compiler tunning in the makefile.
7 years ago
Simplifying ARMv8 build parameters ARMv8 builds were a bit mixed up, with ThunderX2 code in ARMv8 mode (which is not right because TX2 is ARMv8.1) as well as requiring a few redundancies in the defines, making it harder to maintain and understand what core has what. A few other minor issues were also fixed. Tests were made on the following cores: A53, A57, A72, Falkor, ThunderX, ThunderX2, and XGene. Tests were: OpenBLAS/test, OpenBLAS/benchmark, BLAS-Tester. A summary: * Removed TX2 code from ARMv8 build, to make sure it is compatible with all ARMv8 cores, not just v8.1. Also, the TX2 code has actually harmed performance on big cores. * Commoned up ARMv8 architectures' defines in params.h, to make sure that all will benefit from ARMv8 settings, in addition to their own. * Adding a few more cores, using ARMv8's include strategy, to benefit from compiler optimisations using mtune. Also updated cache information from the manuals, making sure we set good conservative values by default. Removed Vulcan, as it's an alias to TX2. * Auto-detecting most of those cores, but also updating the forced compilation in getarch.c, to make sure the parameters are the same whether compiled natively or forced arch. Benefits: * ARMv8 build is now guaranteed to work on all ARMv8 cores * Improved performance for ARMv8 builds on some cores (A72, Falkor, ThunderX1 and 2: up to 11%) over current develop * Improved performance for *all* cores comparing to develop branch before TX2's patch (9% ~ 36%) * ThunderX1 builds are 14% faster than ARMv8 on TX1, 9% faster than current develop's branch and 8% faster than deveop before tx2 patches Issues: * Regression from current develop branch for A53 (-12%) and A57 (-3%) with ARMv8 builds, but still faster than before TX2's commit (+15% and +24% respectively). This can be improved with a simplification of TX2's code, to be done in future patches. At least the code is guaranteed to be ARMv8.0 now. Comments: * CortexA57 builds are unchanged on A57 hardware from develop's branch, which makes sense, as it's untouched. * CortexA72 builds improve over A57 on A72 hardware, even if they're using the same includes due to new compiler tunning in the makefile.
7 years ago
Simplifying ARMv8 build parameters ARMv8 builds were a bit mixed up, with ThunderX2 code in ARMv8 mode (which is not right because TX2 is ARMv8.1) as well as requiring a few redundancies in the defines, making it harder to maintain and understand what core has what. A few other minor issues were also fixed. Tests were made on the following cores: A53, A57, A72, Falkor, ThunderX, ThunderX2, and XGene. Tests were: OpenBLAS/test, OpenBLAS/benchmark, BLAS-Tester. A summary: * Removed TX2 code from ARMv8 build, to make sure it is compatible with all ARMv8 cores, not just v8.1. Also, the TX2 code has actually harmed performance on big cores. * Commoned up ARMv8 architectures' defines in params.h, to make sure that all will benefit from ARMv8 settings, in addition to their own. * Adding a few more cores, using ARMv8's include strategy, to benefit from compiler optimisations using mtune. Also updated cache information from the manuals, making sure we set good conservative values by default. Removed Vulcan, as it's an alias to TX2. * Auto-detecting most of those cores, but also updating the forced compilation in getarch.c, to make sure the parameters are the same whether compiled natively or forced arch. Benefits: * ARMv8 build is now guaranteed to work on all ARMv8 cores * Improved performance for ARMv8 builds on some cores (A72, Falkor, ThunderX1 and 2: up to 11%) over current develop * Improved performance for *all* cores comparing to develop branch before TX2's patch (9% ~ 36%) * ThunderX1 builds are 14% faster than ARMv8 on TX1, 9% faster than current develop's branch and 8% faster than deveop before tx2 patches Issues: * Regression from current develop branch for A53 (-12%) and A57 (-3%) with ARMv8 builds, but still faster than before TX2's commit (+15% and +24% respectively). This can be improved with a simplification of TX2's code, to be done in future patches. At least the code is guaranteed to be ARMv8.0 now. Comments: * CortexA57 builds are unchanged on A57 hardware from develop's branch, which makes sense, as it's untouched. * CortexA72 builds improve over A57 on A72 hardware, even if they're using the same includes due to new compiler tunning in the makefile.
7 years ago
Simplifying ARMv8 build parameters ARMv8 builds were a bit mixed up, with ThunderX2 code in ARMv8 mode (which is not right because TX2 is ARMv8.1) as well as requiring a few redundancies in the defines, making it harder to maintain and understand what core has what. A few other minor issues were also fixed. Tests were made on the following cores: A53, A57, A72, Falkor, ThunderX, ThunderX2, and XGene. Tests were: OpenBLAS/test, OpenBLAS/benchmark, BLAS-Tester. A summary: * Removed TX2 code from ARMv8 build, to make sure it is compatible with all ARMv8 cores, not just v8.1. Also, the TX2 code has actually harmed performance on big cores. * Commoned up ARMv8 architectures' defines in params.h, to make sure that all will benefit from ARMv8 settings, in addition to their own. * Adding a few more cores, using ARMv8's include strategy, to benefit from compiler optimisations using mtune. Also updated cache information from the manuals, making sure we set good conservative values by default. Removed Vulcan, as it's an alias to TX2. * Auto-detecting most of those cores, but also updating the forced compilation in getarch.c, to make sure the parameters are the same whether compiled natively or forced arch. Benefits: * ARMv8 build is now guaranteed to work on all ARMv8 cores * Improved performance for ARMv8 builds on some cores (A72, Falkor, ThunderX1 and 2: up to 11%) over current develop * Improved performance for *all* cores comparing to develop branch before TX2's patch (9% ~ 36%) * ThunderX1 builds are 14% faster than ARMv8 on TX1, 9% faster than current develop's branch and 8% faster than deveop before tx2 patches Issues: * Regression from current develop branch for A53 (-12%) and A57 (-3%) with ARMv8 builds, but still faster than before TX2's commit (+15% and +24% respectively). This can be improved with a simplification of TX2's code, to be done in future patches. At least the code is guaranteed to be ARMv8.0 now. Comments: * CortexA57 builds are unchanged on A57 hardware from develop's branch, which makes sense, as it's untouched. * CortexA72 builds improve over A57 on A72 hardware, even if they're using the same includes due to new compiler tunning in the makefile.
7 years ago
Simplifying ARMv8 build parameters ARMv8 builds were a bit mixed up, with ThunderX2 code in ARMv8 mode (which is not right because TX2 is ARMv8.1) as well as requiring a few redundancies in the defines, making it harder to maintain and understand what core has what. A few other minor issues were also fixed. Tests were made on the following cores: A53, A57, A72, Falkor, ThunderX, ThunderX2, and XGene. Tests were: OpenBLAS/test, OpenBLAS/benchmark, BLAS-Tester. A summary: * Removed TX2 code from ARMv8 build, to make sure it is compatible with all ARMv8 cores, not just v8.1. Also, the TX2 code has actually harmed performance on big cores. * Commoned up ARMv8 architectures' defines in params.h, to make sure that all will benefit from ARMv8 settings, in addition to their own. * Adding a few more cores, using ARMv8's include strategy, to benefit from compiler optimisations using mtune. Also updated cache information from the manuals, making sure we set good conservative values by default. Removed Vulcan, as it's an alias to TX2. * Auto-detecting most of those cores, but also updating the forced compilation in getarch.c, to make sure the parameters are the same whether compiled natively or forced arch. Benefits: * ARMv8 build is now guaranteed to work on all ARMv8 cores * Improved performance for ARMv8 builds on some cores (A72, Falkor, ThunderX1 and 2: up to 11%) over current develop * Improved performance for *all* cores comparing to develop branch before TX2's patch (9% ~ 36%) * ThunderX1 builds are 14% faster than ARMv8 on TX1, 9% faster than current develop's branch and 8% faster than deveop before tx2 patches Issues: * Regression from current develop branch for A53 (-12%) and A57 (-3%) with ARMv8 builds, but still faster than before TX2's commit (+15% and +24% respectively). This can be improved with a simplification of TX2's code, to be done in future patches. At least the code is guaranteed to be ARMv8.0 now. Comments: * CortexA57 builds are unchanged on A57 hardware from develop's branch, which makes sense, as it's untouched. * CortexA72 builds improve over A57 on A72 hardware, even if they're using the same includes due to new compiler tunning in the makefile.
7 years ago
Simplifying ARMv8 build parameters ARMv8 builds were a bit mixed up, with ThunderX2 code in ARMv8 mode (which is not right because TX2 is ARMv8.1) as well as requiring a few redundancies in the defines, making it harder to maintain and understand what core has what. A few other minor issues were also fixed. Tests were made on the following cores: A53, A57, A72, Falkor, ThunderX, ThunderX2, and XGene. Tests were: OpenBLAS/test, OpenBLAS/benchmark, BLAS-Tester. A summary: * Removed TX2 code from ARMv8 build, to make sure it is compatible with all ARMv8 cores, not just v8.1. Also, the TX2 code has actually harmed performance on big cores. * Commoned up ARMv8 architectures' defines in params.h, to make sure that all will benefit from ARMv8 settings, in addition to their own. * Adding a few more cores, using ARMv8's include strategy, to benefit from compiler optimisations using mtune. Also updated cache information from the manuals, making sure we set good conservative values by default. Removed Vulcan, as it's an alias to TX2. * Auto-detecting most of those cores, but also updating the forced compilation in getarch.c, to make sure the parameters are the same whether compiled natively or forced arch. Benefits: * ARMv8 build is now guaranteed to work on all ARMv8 cores * Improved performance for ARMv8 builds on some cores (A72, Falkor, ThunderX1 and 2: up to 11%) over current develop * Improved performance for *all* cores comparing to develop branch before TX2's patch (9% ~ 36%) * ThunderX1 builds are 14% faster than ARMv8 on TX1, 9% faster than current develop's branch and 8% faster than deveop before tx2 patches Issues: * Regression from current develop branch for A53 (-12%) and A57 (-3%) with ARMv8 builds, but still faster than before TX2's commit (+15% and +24% respectively). This can be improved with a simplification of TX2's code, to be done in future patches. At least the code is guaranteed to be ARMv8.0 now. Comments: * CortexA57 builds are unchanged on A57 hardware from develop's branch, which makes sense, as it's untouched. * CortexA72 builds improve over A57 on A72 hardware, even if they're using the same includes due to new compiler tunning in the makefile.
7 years ago
Simplifying ARMv8 build parameters ARMv8 builds were a bit mixed up, with ThunderX2 code in ARMv8 mode (which is not right because TX2 is ARMv8.1) as well as requiring a few redundancies in the defines, making it harder to maintain and understand what core has what. A few other minor issues were also fixed. Tests were made on the following cores: A53, A57, A72, Falkor, ThunderX, ThunderX2, and XGene. Tests were: OpenBLAS/test, OpenBLAS/benchmark, BLAS-Tester. A summary: * Removed TX2 code from ARMv8 build, to make sure it is compatible with all ARMv8 cores, not just v8.1. Also, the TX2 code has actually harmed performance on big cores. * Commoned up ARMv8 architectures' defines in params.h, to make sure that all will benefit from ARMv8 settings, in addition to their own. * Adding a few more cores, using ARMv8's include strategy, to benefit from compiler optimisations using mtune. Also updated cache information from the manuals, making sure we set good conservative values by default. Removed Vulcan, as it's an alias to TX2. * Auto-detecting most of those cores, but also updating the forced compilation in getarch.c, to make sure the parameters are the same whether compiled natively or forced arch. Benefits: * ARMv8 build is now guaranteed to work on all ARMv8 cores * Improved performance for ARMv8 builds on some cores (A72, Falkor, ThunderX1 and 2: up to 11%) over current develop * Improved performance for *all* cores comparing to develop branch before TX2's patch (9% ~ 36%) * ThunderX1 builds are 14% faster than ARMv8 on TX1, 9% faster than current develop's branch and 8% faster than deveop before tx2 patches Issues: * Regression from current develop branch for A53 (-12%) and A57 (-3%) with ARMv8 builds, but still faster than before TX2's commit (+15% and +24% respectively). This can be improved with a simplification of TX2's code, to be done in future patches. At least the code is guaranteed to be ARMv8.0 now. Comments: * CortexA57 builds are unchanged on A57 hardware from develop's branch, which makes sense, as it's untouched. * CortexA72 builds improve over A57 on A72 hardware, even if they're using the same includes due to new compiler tunning in the makefile.
7 years ago
Simplifying ARMv8 build parameters ARMv8 builds were a bit mixed up, with ThunderX2 code in ARMv8 mode (which is not right because TX2 is ARMv8.1) as well as requiring a few redundancies in the defines, making it harder to maintain and understand what core has what. A few other minor issues were also fixed. Tests were made on the following cores: A53, A57, A72, Falkor, ThunderX, ThunderX2, and XGene. Tests were: OpenBLAS/test, OpenBLAS/benchmark, BLAS-Tester. A summary: * Removed TX2 code from ARMv8 build, to make sure it is compatible with all ARMv8 cores, not just v8.1. Also, the TX2 code has actually harmed performance on big cores. * Commoned up ARMv8 architectures' defines in params.h, to make sure that all will benefit from ARMv8 settings, in addition to their own. * Adding a few more cores, using ARMv8's include strategy, to benefit from compiler optimisations using mtune. Also updated cache information from the manuals, making sure we set good conservative values by default. Removed Vulcan, as it's an alias to TX2. * Auto-detecting most of those cores, but also updating the forced compilation in getarch.c, to make sure the parameters are the same whether compiled natively or forced arch. Benefits: * ARMv8 build is now guaranteed to work on all ARMv8 cores * Improved performance for ARMv8 builds on some cores (A72, Falkor, ThunderX1 and 2: up to 11%) over current develop * Improved performance for *all* cores comparing to develop branch before TX2's patch (9% ~ 36%) * ThunderX1 builds are 14% faster than ARMv8 on TX1, 9% faster than current develop's branch and 8% faster than deveop before tx2 patches Issues: * Regression from current develop branch for A53 (-12%) and A57 (-3%) with ARMv8 builds, but still faster than before TX2's commit (+15% and +24% respectively). This can be improved with a simplification of TX2's code, to be done in future patches. At least the code is guaranteed to be ARMv8.0 now. Comments: * CortexA57 builds are unchanged on A57 hardware from develop's branch, which makes sense, as it's untouched. * CortexA72 builds improve over A57 on A72 hardware, even if they're using the same includes due to new compiler tunning in the makefile.
7 years ago
Simplifying ARMv8 build parameters ARMv8 builds were a bit mixed up, with ThunderX2 code in ARMv8 mode (which is not right because TX2 is ARMv8.1) as well as requiring a few redundancies in the defines, making it harder to maintain and understand what core has what. A few other minor issues were also fixed. Tests were made on the following cores: A53, A57, A72, Falkor, ThunderX, ThunderX2, and XGene. Tests were: OpenBLAS/test, OpenBLAS/benchmark, BLAS-Tester. A summary: * Removed TX2 code from ARMv8 build, to make sure it is compatible with all ARMv8 cores, not just v8.1. Also, the TX2 code has actually harmed performance on big cores. * Commoned up ARMv8 architectures' defines in params.h, to make sure that all will benefit from ARMv8 settings, in addition to their own. * Adding a few more cores, using ARMv8's include strategy, to benefit from compiler optimisations using mtune. Also updated cache information from the manuals, making sure we set good conservative values by default. Removed Vulcan, as it's an alias to TX2. * Auto-detecting most of those cores, but also updating the forced compilation in getarch.c, to make sure the parameters are the same whether compiled natively or forced arch. Benefits: * ARMv8 build is now guaranteed to work on all ARMv8 cores * Improved performance for ARMv8 builds on some cores (A72, Falkor, ThunderX1 and 2: up to 11%) over current develop * Improved performance for *all* cores comparing to develop branch before TX2's patch (9% ~ 36%) * ThunderX1 builds are 14% faster than ARMv8 on TX1, 9% faster than current develop's branch and 8% faster than deveop before tx2 patches Issues: * Regression from current develop branch for A53 (-12%) and A57 (-3%) with ARMv8 builds, but still faster than before TX2's commit (+15% and +24% respectively). This can be improved with a simplification of TX2's code, to be done in future patches. At least the code is guaranteed to be ARMv8.0 now. Comments: * CortexA57 builds are unchanged on A57 hardware from develop's branch, which makes sense, as it's untouched. * CortexA72 builds improve over A57 on A72 hardware, even if they're using the same includes due to new compiler tunning in the makefile.
7 years ago
Simplifying ARMv8 build parameters ARMv8 builds were a bit mixed up, with ThunderX2 code in ARMv8 mode (which is not right because TX2 is ARMv8.1) as well as requiring a few redundancies in the defines, making it harder to maintain and understand what core has what. A few other minor issues were also fixed. Tests were made on the following cores: A53, A57, A72, Falkor, ThunderX, ThunderX2, and XGene. Tests were: OpenBLAS/test, OpenBLAS/benchmark, BLAS-Tester. A summary: * Removed TX2 code from ARMv8 build, to make sure it is compatible with all ARMv8 cores, not just v8.1. Also, the TX2 code has actually harmed performance on big cores. * Commoned up ARMv8 architectures' defines in params.h, to make sure that all will benefit from ARMv8 settings, in addition to their own. * Adding a few more cores, using ARMv8's include strategy, to benefit from compiler optimisations using mtune. Also updated cache information from the manuals, making sure we set good conservative values by default. Removed Vulcan, as it's an alias to TX2. * Auto-detecting most of those cores, but also updating the forced compilation in getarch.c, to make sure the parameters are the same whether compiled natively or forced arch. Benefits: * ARMv8 build is now guaranteed to work on all ARMv8 cores * Improved performance for ARMv8 builds on some cores (A72, Falkor, ThunderX1 and 2: up to 11%) over current develop * Improved performance for *all* cores comparing to develop branch before TX2's patch (9% ~ 36%) * ThunderX1 builds are 14% faster than ARMv8 on TX1, 9% faster than current develop's branch and 8% faster than deveop before tx2 patches Issues: * Regression from current develop branch for A53 (-12%) and A57 (-3%) with ARMv8 builds, but still faster than before TX2's commit (+15% and +24% respectively). This can be improved with a simplification of TX2's code, to be done in future patches. At least the code is guaranteed to be ARMv8.0 now. Comments: * CortexA57 builds are unchanged on A57 hardware from develop's branch, which makes sense, as it's untouched. * CortexA72 builds improve over A57 on A72 hardware, even if they're using the same includes due to new compiler tunning in the makefile.
7 years ago
rebase? (#1) * With the Intel compiler on Linux, prefer ifort for the final link step icc has known problems with mixed-language builds that ifort can handle just fine. Fixes #1956 * Rename operands to put lda on the input/output constraint list * Fix wrong constraints in inline assembly for #2009 * Fix inline assembly constraints rework indices to allow marking argument lda4 as input and output. For #2009 * Fix inline assembly constraints rework indices to allow marking argument lda as input and output. * Fix inline assembly constraints * Fix inline assembly constraints * Fix inline assembly constraints in Bulldozer TRSM kernels rework indices to allow marking i,as and bs as both input and output (marked operand n1 as well for simplicity). For #2009 * Correct range_n limiting same bug as seen in #1388, somehow missed in corresponding PR #1389 * Allow multithreading TRMV again revert workaround introduced for issue #1332 as the actual cause appears to be my incorrect fix from #1262 (see #1388) * Fix error introduced during cleanup * Reduce list of kernels in the dynamic arch build to make compilation complete reliably within the 1h limit again * init * move fix to right place * Fix missing -c option in AVX512 test * Fix AVX512 test always returning false due to missing compiler option * Make x86_32 imply NO_AVX2, NO_AVX512 in addition to NO_AVX fixes #2033 * Keep xcode8.3 for osx BINARY=32 build as xcode10 deprecated i386 * Make sure that AVX512 is disabled in 32bit builds for #2033 * Improve handling of NO_STATIC and NO_SHARED to avoid surprises from defining either as zero. Fixes #2035 by addressing some concerns from #1422 * init * address warning introed with #1814 et al * Restore locking optimizations for OpenMP case restore another accidentally dropped part of #1468 that was missed in #2004 to address performance regression reported in #1461 * HiSilicon tsv110 CPUs optimization branch add HiSilicon tsv110 CPUs optimization branch * add TARGET support for HiSilicon tsv110 CPUs * add TARGET support for HiSilicon tsv110 CPUs * add TARGET support for HiSilicon tsv110 CPUs * Fix module definition conflicts between LAPACK and ReLAPACK for #2043 * Do not compile in AVX512 check if AVX support is disabled xgetbv is function depends on NO_AVX being undefined - we could change that too, but that combo is unlikely to work anyway * ctest.c : add __POWERPC__ for PowerMac * Fix crash in sgemm SSE/nano kernel on x86_64 Fix bug #2047. Signed-off-by: Celelibi <celelibi@gmail.com> * param.h : enable defines for PPC970 on DarwinOS fixes: gemm.c: In function 'sgemm_': ../common_param.h:981:18: error: 'SGEMM_DEFAULT_P' undeclared (first use in this function) #define SGEMM_P SGEMM_DEFAULT_P ^ * common_power.h: force DCBT_ARG 0 on PPC970 Darwin without this, we see ../kernel/power/gemv_n.S:427:Parameter syntax error and many more similar entries that relates to this assembly command dcbt 8, r24, r18 this change makes the DCBT_ARG = 0 and openblas builds through to completion on PowerMac 970 Tests pass * Make TARGET=GENERIC compatible with DYNAMIC_ARCH=1 for issue #2048 * make DYNAMIC_ARCH=1 package work on TSV110. * make DYNAMIC_ARCH=1 package work on TSV110 * Add Intel Denverton for #2048 * Add Intel Denverton * Change 64-bit detection as explained in #2056 * Trivial typo fix as suggested in #2022 * Disable the AVX512 DGEMM kernel (again) Due to as yet unresolved errors seen in #1955 and #2029 * Use POSIX getenv on Cygwin The Windows-native GetEnvironmentVariable cannot be relied on, as Cygwin does not always copy environment variables set through Cygwin to the Windows environment block, particularly after fork(). * Fix for #2063: The DllMain used in Cygwin did not run the thread memory pool cleanup upon THREAD_DETACH which is needed when compiled with USE_TLS=1. * Also call CloseHandle on each thread, as well as on the event so as to not leak thread handles. * AIX asm syntax changes needed for shared object creation * power9 makefile. dgemm based on power8 kernel with following changes : 32x unrolled 16x4 kernel and 8x4 kernel using (lxv stxv butterfly rank1 update). improvement from 17 to 22-23gflops. dtrmm cases were added into dgemm itself * Expose CBLAS interfaces for I?MIN and I?MAX * Build CBLAS interfaces for I?MIN and I?MAX * Add declarations for ?sum and cblas_?sum * Add interface for ?sum (derived from ?asum) * Add ?sum * Add implementations of ssum/dsum and csum/zsum as trivial copies of asum/zsasum with the fabs calls replaced by fmov to preserve code structure * Add ARM implementations of ?sum (trivial copies of the respective ?asum with the fabs calls removed) * Add ARM64 implementations of ?sum as trivial copies of the respective ?asum kernels with the fabs calls removed * Add ia64 implementation of ?sum as trivial copy of asum with the fabs calls removed * Add MIPS implementation of ?sum as trivial copy of ?asum with the fabs calls removed * Add MIPS64 implementation of ?sum as trivial copy of ?asum with the fabs replaced by mov to preserve code structure * Add POWER implementation of ?sum as trivial copy of ?asum with the fabs replaced by fmr to preserve code structure * Add SPARC implementation of ?sum as trivial copy of ?asum with the fabs replaced by fmov to preserve code structure * Add x86 implementation of ?sum as trivial copy of ?asum with the fabs calls removed * Add x86_64 implementation of ?sum as trivial copy of ?asum with the fabs calls removed * Add ZARCH implementation of ?sum as trivial copies of the respective ?asum kernels with the ABS and vflpsb calls removed * Detect 32bit environment on 64bit ARM hardware for #2056, using same approach as #2058 * Add cmake defaults for ?sum kernels * Add ?sum * Add ?sum definitions for generic kernel * Add declarations for ?sum * Add -lm and disable EXPRECISION support on *BSD fixes #2075 * Add in runtime CPU detection for POWER. * snprintf define consolidated to common.h * Support INTERFACE64=1 * Add support for INTERFACE64 and fix XERBLA calls 1. Replaced all instances of "int" with "blasint" 2. Added string length as "hidden" third parameter in calls to fortran XERBLA * Correct length of name string in xerbla call * Avoid out-of-bounds accesses in LAPACK EIG tests see https://github.com/Reference-LAPACK/lapack/issues/333 * Correct INFO=4 condition * Disable reallocation of work array in xSYTRF as it appears to cause memory management problems (seen in the LAPACK tests) * Disable repeated recursion on Ab_BR in ReLAPACK xGBTRF due to crashes in LAPACK tests * sgemm/strmm * Update Changelog with changes from 0.3.6 * Increment version to 0.3.7.dev * Increment version to 0.3.7.dev * Misc. typo fixes Found via `codespell -q 3 -w -L ith,als,dum,nd,amin,nto,wis,ba -S ./relapack,./kernel,./lapack-netlib` * Correct argument of CPU_ISSET for glibc <2.5 fixes #2104 * conflict resolve * Revert reference/ fixes * Revert Changelog.txt typos * Disable the SkyLakeX DGEMMITCOPY kernel as well as a stopgap measure for https://github.com/numpy/numpy/issues/13401 as mentioned in #1955 * Disable DGEMMINCOPY as well for now #1955 * init * Fix errors in cpu enumeration with glibc 2.6 for #2114 * Change two http links to https Closes #2109 * remove redundant code #2113 * Set up CI with Azure Pipelines [skip ci] * TST: add native POWER8 to CI * add native POWER8 testing to Travis CI matrix with ppc64le os entry * Update link to IBM MASS library, update cpu support status * first try migrating one of the arm builds from travis * fix tabbing in azure commands * Update azure-pipelines.yml take out offending lines (although stolen from https://github.com/conda-forge/opencv-feedstock azure-pipelines fiie) * Update azure-pipelines.yml * Update azure-pipelines.yml * Update azure-pipelines.yml * Update azure-pipelines.yml * DOC: Add Azure CI status badge * Add ARMV6 build to azure CI setup (#2122) using aytekinar's Alpine image and docker script from the Travis setup [skip ci] * TST: Azure manylinux1 & clean-up * remove some of the steps & comments from the original Azure yml template * modify the trigger section to use develop since OpenBLAS primarily uses this branch; use the same batching behavior as downstream projects NumPy/ SciPy * remove Travis emulated ARMv6 gcc build because this now happens in Azure * use documented Ubuntu vmImage name for Azure and add in a manylinux1 test run to the matrix [skip appveyor] * Add NO_AFFINITY to available options on Linux, and set it to ON to match the gmake default. Fixes second part of #2114 * Replace ISMIN and ISAMIN kernels on all x86_64 platforms (#2125) * Mark iamax_sse.S as unsuitable for MIN due to issue #2116 * Use iamax.S rather than iamax_sse.S for ISMIN/ISAMIN on all x86_64 as workaround for #2116 * Move ARMv8 gcc build from Travis to Azure * Move ARMv8 gcc build from Travis to Azure * Update .travis.yml * Test drone CI * install make * remove sudo * Install gcc * Install perl * Install gfortran and add a clang job * gfortran->gcc-gfortran * Switch to ubuntu and parallel jobs * apt update * Fix typo * update yes * no need of gcc in clang build * Add a cmake build as well * Add cmake builds and print options * build without lapack on cmake * parallel build * See if ubuntu 19.04 fixes the ICE * Remove qemu armv8 builds * arm32 build * Fix typo * TST: add SkylakeX AVX512 CI test * adapt the C-level reproducer code for some recent SkylakeX AVX512 kernel issues, provided by Isuru Fernando and modified by Martin Kroeker, for usage in the utest suite * add an Intel SDE SkylakeX emulation utest run to the Azure CI matrix; a custom Docker build was required because Ubuntu image provided by Azure does not support AVX512VL instructions * Add option USE_LOCKING for single-threaded build with locking support for calling from concurrent threads * Add option USE_LOCKING for single-threaded build with locking support * Add option USE_LOCKING for SMP-like locking in USE_THREAD=0 builds * Add option USE_LOCKING but keep default settings intact * Remove unrelated change * Do not try ancient PGI hacks with recent versions of that compiler should fix #2139 * Build and run utests in any case, they do their own checks for fortran availability * Add softfp support in min/max kernels fix for #1912 * Revert "Add softfp support in min/max kernels" * Separate implementations of AMAX and IAMAX on arm As noted in #1912 and comment on #1942, the combined implementation happens to "do the right thing" on hardfp, but cannot return both value and index on softfp where they would have to share the return register * Ensure correct output for DAMAX with softfp * Use generic kernels for complex (I)AMAX to support softfp * improved zgemm power9 based on power8 * upload thread safety test folder * hook up c++ thread safety test (main Makefile) * add c++ thread test option to Makefile.rule * Document NO_AVX512 for #2151 * sgemm pipeline improved, zgemm rewritten without inner packs, ABI lxvx v20 fixed with vs52 * Fix detection of AVX512 capable compilers in getarch 21eda8b5 introduced a check in getarch.c to test if the compiler is capable of AVX512. This check currently fails, since the used __AVX2__ macro is only defined if getarch itself was compiled with AVX2/AVX512 support. Make sure this is the case by building getarch with -march=native on x86_64. It is only supposed to run on the build host anyway. * c_check: Unlink correct file * power9 zgemm ztrmm optimized * conflict resolve * Add gfortran workaround for ABI violations in LAPACKE for #2154 (see gcc bug 90329) * Add gfortran workaround for ABI violations for #2154 (see gcc bug 90329) * Add gfortran workaround for potential ABI violation for #2154 * Update fc.cmake * Remove any inadvertent use of -march=native from DYNAMIC_ARCH builds from #2143, -march=native precludes use of more specific options like -march=skylake-avx512 in individual kernels, and defeats the purpose of dynamic arch anyway. * Avoid unintentional activation of TLS code via USE_TLS=0 fixes #2149 * Do not force gcc options on non-gcc compilers fixes compile failure with pgi 18.10 as reported on OpenBLAS-users * Update Makefile.x86_64 * Zero ecx with a mov instruction PGI assembler does not like the initialization in the constraints. * Fix mov syntax * new sgemm 8x16 * Update dtrmm_kernel_16x4_power8.S * PGI compiler does not like -march=native * Fix build on FreeBSD/powerpc64. Signed-off-by: Piotr Kubaj <pkubaj@anongoth.pl> * Fix build for PPC970 on FreeBSD pt. 1 FreeBSD needs DCBT_ARG=0 as well. * Fix build for PPC970 on FreeBSD pt.2 FreeBSD needs those macros too. * cgemm/ctrmm power9 * Utest needs CBLAS but not necessarily FORTRAN * Add mingw builds to Appveyor config * Add getarch flags to disable AVX on x86 (and other small fixes to match Makefile behaviour) * Make disabling DYNAMIC_ARCH on unsupported systems work needs to be unset in the cache for the change to have any effect * Mingw32 needs leading underscore on object names (also copy BUNDERSCORE settings for FORTRAN from the corresponding Makefile)
7 years ago
rebase? (#1) * With the Intel compiler on Linux, prefer ifort for the final link step icc has known problems with mixed-language builds that ifort can handle just fine. Fixes #1956 * Rename operands to put lda on the input/output constraint list * Fix wrong constraints in inline assembly for #2009 * Fix inline assembly constraints rework indices to allow marking argument lda4 as input and output. For #2009 * Fix inline assembly constraints rework indices to allow marking argument lda as input and output. * Fix inline assembly constraints * Fix inline assembly constraints * Fix inline assembly constraints in Bulldozer TRSM kernels rework indices to allow marking i,as and bs as both input and output (marked operand n1 as well for simplicity). For #2009 * Correct range_n limiting same bug as seen in #1388, somehow missed in corresponding PR #1389 * Allow multithreading TRMV again revert workaround introduced for issue #1332 as the actual cause appears to be my incorrect fix from #1262 (see #1388) * Fix error introduced during cleanup * Reduce list of kernels in the dynamic arch build to make compilation complete reliably within the 1h limit again * init * move fix to right place * Fix missing -c option in AVX512 test * Fix AVX512 test always returning false due to missing compiler option * Make x86_32 imply NO_AVX2, NO_AVX512 in addition to NO_AVX fixes #2033 * Keep xcode8.3 for osx BINARY=32 build as xcode10 deprecated i386 * Make sure that AVX512 is disabled in 32bit builds for #2033 * Improve handling of NO_STATIC and NO_SHARED to avoid surprises from defining either as zero. Fixes #2035 by addressing some concerns from #1422 * init * address warning introed with #1814 et al * Restore locking optimizations for OpenMP case restore another accidentally dropped part of #1468 that was missed in #2004 to address performance regression reported in #1461 * HiSilicon tsv110 CPUs optimization branch add HiSilicon tsv110 CPUs optimization branch * add TARGET support for HiSilicon tsv110 CPUs * add TARGET support for HiSilicon tsv110 CPUs * add TARGET support for HiSilicon tsv110 CPUs * Fix module definition conflicts between LAPACK and ReLAPACK for #2043 * Do not compile in AVX512 check if AVX support is disabled xgetbv is function depends on NO_AVX being undefined - we could change that too, but that combo is unlikely to work anyway * ctest.c : add __POWERPC__ for PowerMac * Fix crash in sgemm SSE/nano kernel on x86_64 Fix bug #2047. Signed-off-by: Celelibi <celelibi@gmail.com> * param.h : enable defines for PPC970 on DarwinOS fixes: gemm.c: In function 'sgemm_': ../common_param.h:981:18: error: 'SGEMM_DEFAULT_P' undeclared (first use in this function) #define SGEMM_P SGEMM_DEFAULT_P ^ * common_power.h: force DCBT_ARG 0 on PPC970 Darwin without this, we see ../kernel/power/gemv_n.S:427:Parameter syntax error and many more similar entries that relates to this assembly command dcbt 8, r24, r18 this change makes the DCBT_ARG = 0 and openblas builds through to completion on PowerMac 970 Tests pass * Make TARGET=GENERIC compatible with DYNAMIC_ARCH=1 for issue #2048 * make DYNAMIC_ARCH=1 package work on TSV110. * make DYNAMIC_ARCH=1 package work on TSV110 * Add Intel Denverton for #2048 * Add Intel Denverton * Change 64-bit detection as explained in #2056 * Trivial typo fix as suggested in #2022 * Disable the AVX512 DGEMM kernel (again) Due to as yet unresolved errors seen in #1955 and #2029 * Use POSIX getenv on Cygwin The Windows-native GetEnvironmentVariable cannot be relied on, as Cygwin does not always copy environment variables set through Cygwin to the Windows environment block, particularly after fork(). * Fix for #2063: The DllMain used in Cygwin did not run the thread memory pool cleanup upon THREAD_DETACH which is needed when compiled with USE_TLS=1. * Also call CloseHandle on each thread, as well as on the event so as to not leak thread handles. * AIX asm syntax changes needed for shared object creation * power9 makefile. dgemm based on power8 kernel with following changes : 32x unrolled 16x4 kernel and 8x4 kernel using (lxv stxv butterfly rank1 update). improvement from 17 to 22-23gflops. dtrmm cases were added into dgemm itself * Expose CBLAS interfaces for I?MIN and I?MAX * Build CBLAS interfaces for I?MIN and I?MAX * Add declarations for ?sum and cblas_?sum * Add interface for ?sum (derived from ?asum) * Add ?sum * Add implementations of ssum/dsum and csum/zsum as trivial copies of asum/zsasum with the fabs calls replaced by fmov to preserve code structure * Add ARM implementations of ?sum (trivial copies of the respective ?asum with the fabs calls removed) * Add ARM64 implementations of ?sum as trivial copies of the respective ?asum kernels with the fabs calls removed * Add ia64 implementation of ?sum as trivial copy of asum with the fabs calls removed * Add MIPS implementation of ?sum as trivial copy of ?asum with the fabs calls removed * Add MIPS64 implementation of ?sum as trivial copy of ?asum with the fabs replaced by mov to preserve code structure * Add POWER implementation of ?sum as trivial copy of ?asum with the fabs replaced by fmr to preserve code structure * Add SPARC implementation of ?sum as trivial copy of ?asum with the fabs replaced by fmov to preserve code structure * Add x86 implementation of ?sum as trivial copy of ?asum with the fabs calls removed * Add x86_64 implementation of ?sum as trivial copy of ?asum with the fabs calls removed * Add ZARCH implementation of ?sum as trivial copies of the respective ?asum kernels with the ABS and vflpsb calls removed * Detect 32bit environment on 64bit ARM hardware for #2056, using same approach as #2058 * Add cmake defaults for ?sum kernels * Add ?sum * Add ?sum definitions for generic kernel * Add declarations for ?sum * Add -lm and disable EXPRECISION support on *BSD fixes #2075 * Add in runtime CPU detection for POWER. * snprintf define consolidated to common.h * Support INTERFACE64=1 * Add support for INTERFACE64 and fix XERBLA calls 1. Replaced all instances of "int" with "blasint" 2. Added string length as "hidden" third parameter in calls to fortran XERBLA * Correct length of name string in xerbla call * Avoid out-of-bounds accesses in LAPACK EIG tests see https://github.com/Reference-LAPACK/lapack/issues/333 * Correct INFO=4 condition * Disable reallocation of work array in xSYTRF as it appears to cause memory management problems (seen in the LAPACK tests) * Disable repeated recursion on Ab_BR in ReLAPACK xGBTRF due to crashes in LAPACK tests * sgemm/strmm * Update Changelog with changes from 0.3.6 * Increment version to 0.3.7.dev * Increment version to 0.3.7.dev * Misc. typo fixes Found via `codespell -q 3 -w -L ith,als,dum,nd,amin,nto,wis,ba -S ./relapack,./kernel,./lapack-netlib` * Correct argument of CPU_ISSET for glibc <2.5 fixes #2104 * conflict resolve * Revert reference/ fixes * Revert Changelog.txt typos * Disable the SkyLakeX DGEMMITCOPY kernel as well as a stopgap measure for https://github.com/numpy/numpy/issues/13401 as mentioned in #1955 * Disable DGEMMINCOPY as well for now #1955 * init * Fix errors in cpu enumeration with glibc 2.6 for #2114 * Change two http links to https Closes #2109 * remove redundant code #2113 * Set up CI with Azure Pipelines [skip ci] * TST: add native POWER8 to CI * add native POWER8 testing to Travis CI matrix with ppc64le os entry * Update link to IBM MASS library, update cpu support status * first try migrating one of the arm builds from travis * fix tabbing in azure commands * Update azure-pipelines.yml take out offending lines (although stolen from https://github.com/conda-forge/opencv-feedstock azure-pipelines fiie) * Update azure-pipelines.yml * Update azure-pipelines.yml * Update azure-pipelines.yml * Update azure-pipelines.yml * DOC: Add Azure CI status badge * Add ARMV6 build to azure CI setup (#2122) using aytekinar's Alpine image and docker script from the Travis setup [skip ci] * TST: Azure manylinux1 & clean-up * remove some of the steps & comments from the original Azure yml template * modify the trigger section to use develop since OpenBLAS primarily uses this branch; use the same batching behavior as downstream projects NumPy/ SciPy * remove Travis emulated ARMv6 gcc build because this now happens in Azure * use documented Ubuntu vmImage name for Azure and add in a manylinux1 test run to the matrix [skip appveyor] * Add NO_AFFINITY to available options on Linux, and set it to ON to match the gmake default. Fixes second part of #2114 * Replace ISMIN and ISAMIN kernels on all x86_64 platforms (#2125) * Mark iamax_sse.S as unsuitable for MIN due to issue #2116 * Use iamax.S rather than iamax_sse.S for ISMIN/ISAMIN on all x86_64 as workaround for #2116 * Move ARMv8 gcc build from Travis to Azure * Move ARMv8 gcc build from Travis to Azure * Update .travis.yml * Test drone CI * install make * remove sudo * Install gcc * Install perl * Install gfortran and add a clang job * gfortran->gcc-gfortran * Switch to ubuntu and parallel jobs * apt update * Fix typo * update yes * no need of gcc in clang build * Add a cmake build as well * Add cmake builds and print options * build without lapack on cmake * parallel build * See if ubuntu 19.04 fixes the ICE * Remove qemu armv8 builds * arm32 build * Fix typo * TST: add SkylakeX AVX512 CI test * adapt the C-level reproducer code for some recent SkylakeX AVX512 kernel issues, provided by Isuru Fernando and modified by Martin Kroeker, for usage in the utest suite * add an Intel SDE SkylakeX emulation utest run to the Azure CI matrix; a custom Docker build was required because Ubuntu image provided by Azure does not support AVX512VL instructions * Add option USE_LOCKING for single-threaded build with locking support for calling from concurrent threads * Add option USE_LOCKING for single-threaded build with locking support * Add option USE_LOCKING for SMP-like locking in USE_THREAD=0 builds * Add option USE_LOCKING but keep default settings intact * Remove unrelated change * Do not try ancient PGI hacks with recent versions of that compiler should fix #2139 * Build and run utests in any case, they do their own checks for fortran availability * Add softfp support in min/max kernels fix for #1912 * Revert "Add softfp support in min/max kernels" * Separate implementations of AMAX and IAMAX on arm As noted in #1912 and comment on #1942, the combined implementation happens to "do the right thing" on hardfp, but cannot return both value and index on softfp where they would have to share the return register * Ensure correct output for DAMAX with softfp * Use generic kernels for complex (I)AMAX to support softfp * improved zgemm power9 based on power8 * upload thread safety test folder * hook up c++ thread safety test (main Makefile) * add c++ thread test option to Makefile.rule * Document NO_AVX512 for #2151 * sgemm pipeline improved, zgemm rewritten without inner packs, ABI lxvx v20 fixed with vs52 * Fix detection of AVX512 capable compilers in getarch 21eda8b5 introduced a check in getarch.c to test if the compiler is capable of AVX512. This check currently fails, since the used __AVX2__ macro is only defined if getarch itself was compiled with AVX2/AVX512 support. Make sure this is the case by building getarch with -march=native on x86_64. It is only supposed to run on the build host anyway. * c_check: Unlink correct file * power9 zgemm ztrmm optimized * conflict resolve * Add gfortran workaround for ABI violations in LAPACKE for #2154 (see gcc bug 90329) * Add gfortran workaround for ABI violations for #2154 (see gcc bug 90329) * Add gfortran workaround for potential ABI violation for #2154 * Update fc.cmake * Remove any inadvertent use of -march=native from DYNAMIC_ARCH builds from #2143, -march=native precludes use of more specific options like -march=skylake-avx512 in individual kernels, and defeats the purpose of dynamic arch anyway. * Avoid unintentional activation of TLS code via USE_TLS=0 fixes #2149 * Do not force gcc options on non-gcc compilers fixes compile failure with pgi 18.10 as reported on OpenBLAS-users * Update Makefile.x86_64 * Zero ecx with a mov instruction PGI assembler does not like the initialization in the constraints. * Fix mov syntax * new sgemm 8x16 * Update dtrmm_kernel_16x4_power8.S * PGI compiler does not like -march=native * Fix build on FreeBSD/powerpc64. Signed-off-by: Piotr Kubaj <pkubaj@anongoth.pl> * Fix build for PPC970 on FreeBSD pt. 1 FreeBSD needs DCBT_ARG=0 as well. * Fix build for PPC970 on FreeBSD pt.2 FreeBSD needs those macros too. * cgemm/ctrmm power9 * Utest needs CBLAS but not necessarily FORTRAN * Add mingw builds to Appveyor config * Add getarch flags to disable AVX on x86 (and other small fixes to match Makefile behaviour) * Make disabling DYNAMIC_ARCH on unsupported systems work needs to be unset in the cache for the change to have any effect * Mingw32 needs leading underscore on object names (also copy BUNDERSCORE settings for FORTRAN from the corresponding Makefile)
7 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368
  1. /*****************************************************************************
  2. Copyright (c) 2011-2014, The OpenBLAS Project
  3. All rights reserved.
  4. Redistribution and use in source and binary forms, with or without
  5. modification, are permitted provided that the following conditions are
  6. met:
  7. 1. Redistributions of source code must retain the above copyright
  8. notice, this list of conditions and the following disclaimer.
  9. 2. Redistributions in binary form must reproduce the above copyright
  10. notice, this list of conditions and the following disclaimer in
  11. the documentation and/or other materials provided with the
  12. distribution.
  13. 3. Neither the name of the OpenBLAS project nor the names of
  14. its contributors may be used to endorse or promote products
  15. derived from this software without specific prior written
  16. permission.
  17. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  18. AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19. IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20. ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  21. LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  22. DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
  23. SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
  24. CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
  25. OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
  26. USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  27. **********************************************************************************/
  28. /*********************************************************************/
  29. /* Copyright 2009, 2010 The University of Texas at Austin. */
  30. /* All rights reserved. */
  31. /* */
  32. /* Redistribution and use in source and binary forms, with or */
  33. /* without modification, are permitted provided that the following */
  34. /* conditions are met: */
  35. /* */
  36. /* 1. Redistributions of source code must retain the above */
  37. /* copyright notice, this list of conditions and the following */
  38. /* disclaimer. */
  39. /* */
  40. /* 2. Redistributions in binary form must reproduce the above */
  41. /* copyright notice, this list of conditions and the following */
  42. /* disclaimer in the documentation and/or other materials */
  43. /* provided with the distribution. */
  44. /* */
  45. /* THIS SOFTWARE IS PROVIDED BY THE UNIVERSITY OF TEXAS AT */
  46. /* AUSTIN ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, */
  47. /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */
  48. /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */
  49. /* DISCLAIMED. IN NO EVENT SHALL THE UNIVERSITY OF TEXAS AT */
  50. /* AUSTIN OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, */
  51. /* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES */
  52. /* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE */
  53. /* GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR */
  54. /* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF */
  55. /* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT */
  56. /* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT */
  57. /* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */
  58. /* POSSIBILITY OF SUCH DAMAGE. */
  59. /* */
  60. /* The views and conclusions contained in the software and */
  61. /* documentation are those of the authors and should not be */
  62. /* interpreted as representing official policies, either expressed */
  63. /* or implied, of The University of Texas at Austin. */
  64. /*********************************************************************/
  65. #if defined(__WIN32__) || defined(__WIN64__) || defined(__CYGWIN32__) || defined(__CYGWIN64__) || defined(_WIN32) || defined(_WIN64)
  66. #define OS_WINDOWS
  67. #endif
  68. #if defined(__i386__) || defined(__x86_64__) || defined(_M_IX86) || defined(_M_X64)
  69. #define INTEL_AMD
  70. #endif
  71. #include <stdio.h>
  72. #include <string.h>
  73. #ifdef OS_WINDOWS
  74. #include <windows.h>
  75. #endif
  76. #if defined(__FreeBSD__) || defined(__OpenBSD__) || defined(__DragonFly__) || defined(__APPLE__)
  77. #include <sys/types.h>
  78. #include <sys/sysctl.h>
  79. #endif
  80. #if defined(linux) || defined(__sun__)
  81. #include <sys/sysinfo.h>
  82. #include <unistd.h>
  83. #endif
  84. #if (( defined(__GNUC__) && __GNUC__ > 6 && defined(__AVX2__)) || (defined(__clang__) && __clang_major__ >= 6))
  85. #else
  86. #define NO_AVX512
  87. #endif
  88. /* #define FORCE_P2 */
  89. /* #define FORCE_KATMAI */
  90. /* #define FORCE_COPPERMINE */
  91. /* #define FORCE_NORTHWOOD */
  92. /* #define FORCE_PRESCOTT */
  93. /* #define FORCE_BANIAS */
  94. /* #define FORCE_YONAH */
  95. /* #define FORCE_CORE2 */
  96. /* #define FORCE_PENRYN */
  97. /* #define FORCE_DUNNINGTON */
  98. /* #define FORCE_NEHALEM */
  99. /* #define FORCE_SANDYBRIDGE */
  100. /* #define FORCE_ATOM */
  101. /* #define FORCE_ATHLON */
  102. /* #define FORCE_OPTERON */
  103. /* #define FORCE_OPTERON_SSE3 */
  104. /* #define FORCE_BARCELONA */
  105. /* #define FORCE_SHANGHAI */
  106. /* #define FORCE_ISTANBUL */
  107. /* #define FORCE_BOBCAT */
  108. /* #define FORCE_BULLDOZER */
  109. /* #define FORCE_PILEDRIVER */
  110. /* #define FORCE_SSE_GENERIC */
  111. /* #define FORCE_VIAC3 */
  112. /* #define FORCE_NANO */
  113. /* #define FORCE_POWER3 */
  114. /* #define FORCE_POWER4 */
  115. /* #define FORCE_POWER5 */
  116. /* #define FORCE_POWER6 */
  117. /* #define FORCE_POWER7 */
  118. /* #define FORCE_POWER8 */
  119. /* #define FORCE_PPCG4 */
  120. /* #define FORCE_PPC970 */
  121. /* #define FORCE_PPC970MP */
  122. /* #define FORCE_PPC440 */
  123. /* #define FORCE_PPC440FP2 */
  124. /* #define FORCE_CELL */
  125. /* #define FORCE_SICORTEX */
  126. /* #define FORCE_LOONGSON3A */
  127. /* #define FORCE_LOONGSON3B */
  128. /* #define FORCE_I6400 */
  129. /* #define FORCE_P6600 */
  130. /* #define FORCE_P5600 */
  131. /* #define FORCE_I6500 */
  132. /* #define FORCE_ITANIUM2 */
  133. /* #define FORCE_SPARC */
  134. /* #define FORCE_SPARCV7 */
  135. /* #define FORCE_ZARCH_GENERIC */
  136. /* #define FORCE_Z13 */
  137. /* #define FORCE_GENERIC */
  138. #ifdef FORCE_P2
  139. #define FORCE
  140. #define FORCE_INTEL
  141. #define ARCHITECTURE "X86"
  142. #define SUBARCHITECTURE "PENTIUM2"
  143. #define ARCHCONFIG "-DPENTIUM2 " \
  144. "-DL1_DATA_SIZE=16384 -DL1_DATA_LINESIZE=32 " \
  145. "-DL2_SIZE=512488 -DL2_LINESIZE=32 " \
  146. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  147. "-DHAVE_CMOV -DHAVE_MMX"
  148. #define LIBNAME "p2"
  149. #define CORENAME "P5"
  150. #endif
  151. #ifdef FORCE_KATMAI
  152. #define FORCE
  153. #define FORCE_INTEL
  154. #define ARCHITECTURE "X86"
  155. #define SUBARCHITECTURE "PENTIUM3"
  156. #define ARCHCONFIG "-DPENTIUM3 " \
  157. "-DL1_DATA_SIZE=16384 -DL1_DATA_LINESIZE=32 " \
  158. "-DL2_SIZE=524288 -DL2_LINESIZE=32 " \
  159. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  160. "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE "
  161. #define LIBNAME "katmai"
  162. #define CORENAME "KATMAI"
  163. #endif
  164. #ifdef FORCE_COPPERMINE
  165. #define FORCE
  166. #define FORCE_INTEL
  167. #define ARCHITECTURE "X86"
  168. #define SUBARCHITECTURE "PENTIUM3"
  169. #define ARCHCONFIG "-DPENTIUM3 " \
  170. "-DL1_DATA_SIZE=16384 -DL1_DATA_LINESIZE=32 " \
  171. "-DL2_SIZE=262144 -DL2_LINESIZE=32 " \
  172. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  173. "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE "
  174. #define LIBNAME "coppermine"
  175. #define CORENAME "COPPERMINE"
  176. #endif
  177. #ifdef FORCE_NORTHWOOD
  178. #define FORCE
  179. #define FORCE_INTEL
  180. #define ARCHITECTURE "X86"
  181. #define SUBARCHITECTURE "PENTIUM4"
  182. #define ARCHCONFIG "-DPENTIUM4 " \
  183. "-DL1_DATA_SIZE=8192 -DL1_DATA_LINESIZE=64 " \
  184. "-DL2_SIZE=524288 -DL2_LINESIZE=64 " \
  185. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 " \
  186. "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 "
  187. #define LIBNAME "northwood"
  188. #define CORENAME "NORTHWOOD"
  189. #endif
  190. #ifdef FORCE_PRESCOTT
  191. #define FORCE
  192. #define FORCE_INTEL
  193. #define ARCHITECTURE "X86"
  194. #define SUBARCHITECTURE "PENTIUM4"
  195. #define ARCHCONFIG "-DPENTIUM4 " \
  196. "-DL1_DATA_SIZE=16384 -DL1_DATA_LINESIZE=64 " \
  197. "-DL2_SIZE=1048576 -DL2_LINESIZE=64 " \
  198. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 " \
  199. "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3"
  200. #define LIBNAME "prescott"
  201. #define CORENAME "PRESCOTT"
  202. #endif
  203. #ifdef FORCE_BANIAS
  204. #define FORCE
  205. #define FORCE_INTEL
  206. #define ARCHITECTURE "X86"
  207. #define SUBARCHITECTURE "BANIAS"
  208. #define ARCHCONFIG "-DPENTIUMM " \
  209. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 " \
  210. "-DL2_SIZE=1048576 -DL2_LINESIZE=64 " \
  211. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  212. "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 "
  213. #define LIBNAME "banias"
  214. #define CORENAME "BANIAS"
  215. #endif
  216. #ifdef FORCE_YONAH
  217. #define FORCE
  218. #define FORCE_INTEL
  219. #define ARCHITECTURE "X86"
  220. #define SUBARCHITECTURE "YONAH"
  221. #define ARCHCONFIG "-DPENTIUMM " \
  222. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 " \
  223. "-DL2_SIZE=1048576 -DL2_LINESIZE=64 " \
  224. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  225. "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 "
  226. #define LIBNAME "yonah"
  227. #define CORENAME "YONAH"
  228. #endif
  229. #ifdef FORCE_CORE2
  230. #define FORCE
  231. #define FORCE_INTEL
  232. #define ARCHITECTURE "X86"
  233. #define SUBARCHITECTURE "CONRORE"
  234. #define ARCHCONFIG "-DCORE2 " \
  235. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 " \
  236. "-DL2_SIZE=1048576 -DL2_LINESIZE=64 " \
  237. "-DDTB_DEFAULT_ENTRIES=256 -DDTB_SIZE=4096 " \
  238. "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSSE3"
  239. #define LIBNAME "core2"
  240. #define CORENAME "CORE2"
  241. #endif
  242. #ifdef FORCE_PENRYN
  243. #define FORCE
  244. #define FORCE_INTEL
  245. #define ARCHITECTURE "X86"
  246. #define SUBARCHITECTURE "PENRYN"
  247. #define ARCHCONFIG "-DPENRYN " \
  248. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 " \
  249. "-DL2_SIZE=1048576 -DL2_LINESIZE=64 " \
  250. "-DDTB_DEFAULT_ENTRIES=256 -DDTB_SIZE=4096 " \
  251. "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSSE3 -DHAVE_SSE4_1"
  252. #define LIBNAME "penryn"
  253. #define CORENAME "PENRYN"
  254. #endif
  255. #ifdef FORCE_DUNNINGTON
  256. #define FORCE
  257. #define FORCE_INTEL
  258. #define ARCHITECTURE "X86"
  259. #define SUBARCHITECTURE "DUNNINGTON"
  260. #define ARCHCONFIG "-DDUNNINGTON " \
  261. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 " \
  262. "-DL2_SIZE=1048576 -DL2_LINESIZE=64 " \
  263. "-DL3_SIZE=16777216 -DL3_LINESIZE=64 " \
  264. "-DDTB_DEFAULT_ENTRIES=256 -DDTB_SIZE=4096 " \
  265. "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSSE3 -DHAVE_SSE4_1"
  266. #define LIBNAME "dunnington"
  267. #define CORENAME "DUNNINGTON"
  268. #endif
  269. #ifdef FORCE_NEHALEM
  270. #define FORCE
  271. #define FORCE_INTEL
  272. #define ARCHITECTURE "X86"
  273. #define SUBARCHITECTURE "NEHALEM"
  274. #define ARCHCONFIG "-DNEHALEM " \
  275. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 " \
  276. "-DL2_SIZE=262144 -DL2_LINESIZE=64 " \
  277. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  278. "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSSE3 -DHAVE_SSE4_1 -DHAVE_SSE4_2"
  279. #define LIBNAME "nehalem"
  280. #define CORENAME "NEHALEM"
  281. #endif
  282. #ifdef FORCE_SANDYBRIDGE
  283. #define FORCE
  284. #define FORCE_INTEL
  285. #define ARCHITECTURE "X86"
  286. #define SUBARCHITECTURE "SANDYBRIDGE"
  287. #define ARCHCONFIG "-DSANDYBRIDGE " \
  288. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 " \
  289. "-DL2_SIZE=262144 -DL2_LINESIZE=64 " \
  290. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  291. "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSSE3 -DHAVE_SSE4_1 -DHAVE_SSE4_2 -DHAVE_AVX"
  292. #define LIBNAME "sandybridge"
  293. #define CORENAME "SANDYBRIDGE"
  294. #endif
  295. #ifdef FORCE_HASWELL
  296. #define FORCE
  297. #define FORCE_INTEL
  298. #define ARCHITECTURE "X86"
  299. #define SUBARCHITECTURE "HASWELL"
  300. #define ARCHCONFIG "-DHASWELL " \
  301. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 " \
  302. "-DL2_SIZE=262144 -DL2_LINESIZE=64 " \
  303. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  304. "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSSE3 -DHAVE_SSE4_1 -DHAVE_SSE4_2 -DHAVE_AVX " \
  305. "-DFMA3"
  306. #define LIBNAME "haswell"
  307. #define CORENAME "HASWELL"
  308. #endif
  309. #ifdef FORCE_SKYLAKEX
  310. #ifdef NO_AVX512
  311. #define FORCE
  312. #define FORCE_INTEL
  313. #define ARCHITECTURE "X86"
  314. #define SUBARCHITECTURE "HASWELL"
  315. #define ARCHCONFIG "-DHASWELL " \
  316. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 " \
  317. "-DL2_SIZE=262144 -DL2_LINESIZE=64 " \
  318. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  319. "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSSE3 -DHAVE_SSE4_1 -DHAVE_SSE4_2 -DHAVE_AVX " \
  320. "-DFMA3"
  321. #define LIBNAME "haswell"
  322. #define CORENAME "HASWELL"
  323. #else
  324. #define FORCE
  325. #define FORCE_INTEL
  326. #define ARCHITECTURE "X86"
  327. #define SUBARCHITECTURE "SKYLAKEX"
  328. #define ARCHCONFIG "-DSKYLAKEX " \
  329. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 " \
  330. "-DL2_SIZE=262144 -DL2_LINESIZE=64 " \
  331. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  332. "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSSE3 -DHAVE_SSE4_1 -DHAVE_SSE4_2 -DHAVE_AVX " \
  333. "-DFMA3 -DHAVE_AVX512VL -march=skylake-avx512"
  334. #define LIBNAME "skylakex"
  335. #define CORENAME "SKYLAKEX"
  336. #endif
  337. #endif
  338. #ifdef FORCE_ATOM
  339. #define FORCE
  340. #define FORCE_INTEL
  341. #define ARCHITECTURE "X86"
  342. #define SUBARCHITECTURE "ATOM"
  343. #define ARCHCONFIG "-DATOM " \
  344. "-DL1_DATA_SIZE=24576 -DL1_DATA_LINESIZE=64 " \
  345. "-DL2_SIZE=524288 -DL2_LINESIZE=64 " \
  346. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=4 " \
  347. "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSSE3"
  348. #define LIBNAME "atom"
  349. #define CORENAME "ATOM"
  350. #endif
  351. #ifdef FORCE_ATHLON
  352. #define FORCE
  353. #define FORCE_INTEL
  354. #define ARCHITECTURE "X86"
  355. #define SUBARCHITECTURE "ATHLON"
  356. #define ARCHCONFIG "-DATHLON " \
  357. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=64 " \
  358. "-DL2_SIZE=1048576 -DL2_LINESIZE=64 " \
  359. "-DDTB_DEFAULT_ENTRIES=32 -DDTB_SIZE=4096 -DHAVE_3DNOW " \
  360. "-DHAVE_3DNOWEX -DHAVE_MMX -DHAVE_SSE "
  361. #define LIBNAME "athlon"
  362. #define CORENAME "ATHLON"
  363. #endif
  364. #ifdef FORCE_OPTERON
  365. #define FORCE
  366. #define FORCE_INTEL
  367. #define ARCHITECTURE "X86"
  368. #define SUBARCHITECTURE "OPTERON"
  369. #define ARCHCONFIG "-DOPTERON " \
  370. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=64 " \
  371. "-DL2_SIZE=1048576 -DL2_LINESIZE=64 " \
  372. "-DDTB_DEFAULT_ENTRIES=32 -DDTB_SIZE=4096 -DHAVE_3DNOW " \
  373. "-DHAVE_3DNOWEX -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 "
  374. #define LIBNAME "opteron"
  375. #define CORENAME "OPTERON"
  376. #endif
  377. #ifdef FORCE_OPTERON_SSE3
  378. #define FORCE
  379. #define FORCE_INTEL
  380. #define ARCHITECTURE "X86"
  381. #define SUBARCHITECTURE "OPTERON"
  382. #define ARCHCONFIG "-DOPTERON " \
  383. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=64 " \
  384. "-DL2_SIZE=1048576 -DL2_LINESIZE=64 " \
  385. "-DDTB_DEFAULT_ENTRIES=32 -DDTB_SIZE=4096 -DHAVE_3DNOW " \
  386. "-DHAVE_3DNOWEX -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3"
  387. #define LIBNAME "opteron"
  388. #define CORENAME "OPTERON"
  389. #endif
  390. #if defined(FORCE_BARCELONA) || defined(FORCE_SHANGHAI) || defined(FORCE_ISTANBUL)
  391. #define FORCE
  392. #define FORCE_INTEL
  393. #define ARCHITECTURE "X86"
  394. #define SUBARCHITECTURE "BARCELONA"
  395. #define ARCHCONFIG "-DBARCELONA " \
  396. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=64 " \
  397. "-DL2_SIZE=524288 -DL2_LINESIZE=64 -DL3_SIZE=2097152 " \
  398. "-DDTB_DEFAULT_ENTRIES=48 -DDTB_SIZE=4096 " \
  399. "-DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 " \
  400. "-DHAVE_SSE4A -DHAVE_MISALIGNSSE -DHAVE_128BITFPU -DHAVE_FASTMOVU"
  401. #define LIBNAME "barcelona"
  402. #define CORENAME "BARCELONA"
  403. #endif
  404. #if defined(FORCE_BOBCAT)
  405. #define FORCE
  406. #define FORCE_INTEL
  407. #define ARCHITECTURE "X86"
  408. #define SUBARCHITECTURE "BOBCAT"
  409. #define ARCHCONFIG "-DBOBCAT " \
  410. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 " \
  411. "-DL2_SIZE=524288 -DL2_LINESIZE=64 " \
  412. "-DDTB_DEFAULT_ENTRIES=40 -DDTB_SIZE=4096 " \
  413. "-DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSSE3 " \
  414. "-DHAVE_SSE4A -DHAVE_MISALIGNSSE -DHAVE_CFLUSH -DHAVE_CMOV"
  415. #define LIBNAME "bobcat"
  416. #define CORENAME "BOBCAT"
  417. #endif
  418. #if defined (FORCE_BULLDOZER)
  419. #define FORCE
  420. #define FORCE_INTEL
  421. #define ARCHITECTURE "X86"
  422. #define SUBARCHITECTURE "BULLDOZER"
  423. #define ARCHCONFIG "-DBULLDOZER " \
  424. "-DL1_DATA_SIZE=49152 -DL1_DATA_LINESIZE=64 " \
  425. "-DL2_SIZE=1024000 -DL2_LINESIZE=64 -DL3_SIZE=16777216 " \
  426. "-DDTB_DEFAULT_ENTRIES=32 -DDTB_SIZE=4096 " \
  427. "-DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 " \
  428. "-DHAVE_SSE4A -DHAVE_MISALIGNSSE -DHAVE_128BITFPU -DHAVE_FASTMOVU " \
  429. "-DHAVE_AVX -DHAVE_FMA4"
  430. #define LIBNAME "bulldozer"
  431. #define CORENAME "BULLDOZER"
  432. #endif
  433. #if defined (FORCE_PILEDRIVER)
  434. #define FORCE
  435. #define FORCE_INTEL
  436. #define ARCHITECTURE "X86"
  437. #define SUBARCHITECTURE "PILEDRIVER"
  438. #define ARCHCONFIG "-DPILEDRIVER " \
  439. "-DL1_DATA_SIZE=16384 -DL1_DATA_LINESIZE=64 " \
  440. "-DL2_SIZE=2097152 -DL2_LINESIZE=64 -DL3_SIZE=12582912 " \
  441. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  442. "-DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSE4_1 -DHAVE_SSE4_2 " \
  443. "-DHAVE_SSE4A -DHAVE_MISALIGNSSE -DHAVE_128BITFPU -DHAVE_FASTMOVU -DHAVE_CFLUSH " \
  444. "-DHAVE_AVX -DHAVE_FMA4 -DHAVE_FMA3"
  445. #define LIBNAME "piledriver"
  446. #define CORENAME "PILEDRIVER"
  447. #endif
  448. #if defined (FORCE_STEAMROLLER)
  449. #define FORCE
  450. #define FORCE_INTEL
  451. #define ARCHITECTURE "X86"
  452. #define SUBARCHITECTURE "STEAMROLLER"
  453. #define ARCHCONFIG "-DSTEAMROLLER " \
  454. "-DL1_DATA_SIZE=16384 -DL1_DATA_LINESIZE=64 " \
  455. "-DL2_SIZE=2097152 -DL2_LINESIZE=64 -DL3_SIZE=12582912 " \
  456. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  457. "-DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSE4_1 -DHAVE_SSE4_2 " \
  458. "-DHAVE_SSE4A -DHAVE_MISALIGNSSE -DHAVE_128BITFPU -DHAVE_FASTMOVU -DHAVE_CFLUSH " \
  459. "-DHAVE_AVX -DHAVE_FMA4 -DHAVE_FMA3"
  460. #define LIBNAME "steamroller"
  461. #define CORENAME "STEAMROLLER"
  462. #endif
  463. #if defined (FORCE_EXCAVATOR)
  464. #define FORCE
  465. #define FORCE_INTEL
  466. #define ARCHITECTURE "X86"
  467. #define SUBARCHITECTURE "EXCAVATOR"
  468. #define ARCHCONFIG "-DEXCAVATOR " \
  469. "-DL1_DATA_SIZE=16384 -DL1_DATA_LINESIZE=64 " \
  470. "-DL2_SIZE=2097152 -DL2_LINESIZE=64 -DL3_SIZE=12582912 " \
  471. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  472. "-DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSE4_1 -DHAVE_SSE4_2 " \
  473. "-DHAVE_SSE4A -DHAVE_MISALIGNSSE -DHAVE_128BITFPU -DHAVE_FASTMOVU -DHAVE_CFLUSH " \
  474. "-DHAVE_AVX -DHAVE_FMA4 -DHAVE_FMA3"
  475. #define LIBNAME "excavator"
  476. #define CORENAME "EXCAVATOR"
  477. #endif
  478. #if defined (FORCE_ZEN)
  479. #define FORCE
  480. #define FORCE_INTEL
  481. #define ARCHITECTURE "X86"
  482. #define SUBARCHITECTURE "ZEN"
  483. #define ARCHCONFIG "-DZEN " \
  484. "-DL1_CODE_SIZE=32768 -DL1_CODE_LINESIZE=64 -DL1_CODE_ASSOCIATIVE=8 " \
  485. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 -DL2_CODE_ASSOCIATIVE=8 " \
  486. "-DL2_SIZE=524288 -DL2_LINESIZE=64 -DL2_ASSOCIATIVE=8 " \
  487. "-DL3_SIZE=16777216 -DL3_LINESIZE=64 -DL3_ASSOCIATIVE=8 " \
  488. "-DITB_DEFAULT_ENTRIES=64 -DITB_SIZE=4096 " \
  489. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  490. "-DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSE4_1 -DHAVE_SSE4_2 " \
  491. "-DHAVE_SSE4A -DHAVE_MISALIGNSSE -DHAVE_128BITFPU -DHAVE_FASTMOVU -DHAVE_CFLUSH " \
  492. "-DHAVE_AVX -DHAVE_FMA3 -DFMA3"
  493. #define LIBNAME "zen"
  494. #define CORENAME "ZEN"
  495. #endif
  496. #ifdef FORCE_SSE_GENERIC
  497. #define FORCE
  498. #define FORCE_INTEL
  499. #define ARCHITECTURE "X86"
  500. #define SUBARCHITECTURE "GENERIC"
  501. #define ARCHCONFIG "-DGENERIC " \
  502. "-DL1_DATA_SIZE=16384 -DL1_DATA_LINESIZE=64 " \
  503. "-DL2_SIZE=524288 -DL2_LINESIZE=64 " \
  504. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 " \
  505. "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2"
  506. #define LIBNAME "generic"
  507. #define CORENAME "GENERIC"
  508. #endif
  509. #ifdef FORCE_VIAC3
  510. #define FORCE
  511. #define FORCE_INTEL
  512. #define ARCHITECTURE "X86"
  513. #define SUBARCHITECTURE "VIAC3"
  514. #define ARCHCONFIG "-DVIAC3 " \
  515. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=32 " \
  516. "-DL2_SIZE=65536 -DL2_LINESIZE=32 " \
  517. "-DDTB_DEFAULT_ENTRIES=128 -DDTB_SIZE=4096 " \
  518. "-DHAVE_MMX -DHAVE_SSE "
  519. #define LIBNAME "viac3"
  520. #define CORENAME "VIAC3"
  521. #endif
  522. #ifdef FORCE_NANO
  523. #define FORCE
  524. #define FORCE_INTEL
  525. #define ARCHITECTURE "X86"
  526. #define SUBARCHITECTURE "NANO"
  527. #define ARCHCONFIG "-DNANO " \
  528. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=64 " \
  529. "-DL2_SIZE=1048576 -DL2_LINESIZE=64 " \
  530. "-DDTB_DEFAULT_ENTRIES=128 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 " \
  531. "-DHAVE_CMOV -DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSSE3"
  532. #define LIBNAME "nano"
  533. #define CORENAME "NANO"
  534. #endif
  535. #ifdef FORCE_POWER3
  536. #define FORCE
  537. #define ARCHITECTURE "POWER"
  538. #define SUBARCHITECTURE "POWER3"
  539. #define SUBDIRNAME "power"
  540. #define ARCHCONFIG "-DPOWER3 " \
  541. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=128 " \
  542. "-DL2_SIZE=2097152 -DL2_LINESIZE=128 " \
  543. "-DDTB_DEFAULT_ENTRIES=256 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
  544. #define LIBNAME "power3"
  545. #define CORENAME "POWER3"
  546. #endif
  547. #ifdef FORCE_POWER4
  548. #define FORCE
  549. #define ARCHITECTURE "POWER"
  550. #define SUBARCHITECTURE "POWER4"
  551. #define SUBDIRNAME "power"
  552. #define ARCHCONFIG "-DPOWER4 " \
  553. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=128 " \
  554. "-DL2_SIZE=1509949 -DL2_LINESIZE=128 " \
  555. "-DDTB_DEFAULT_ENTRIES=128 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=6 "
  556. #define LIBNAME "power4"
  557. #define CORENAME "POWER4"
  558. #endif
  559. #ifdef FORCE_POWER5
  560. #define FORCE
  561. #define ARCHITECTURE "POWER"
  562. #define SUBARCHITECTURE "POWER5"
  563. #define SUBDIRNAME "power"
  564. #define ARCHCONFIG "-DPOWER5 " \
  565. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=128 " \
  566. "-DL2_SIZE=1509949 -DL2_LINESIZE=128 " \
  567. "-DDTB_DEFAULT_ENTRIES=128 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=6 "
  568. #define LIBNAME "power5"
  569. #define CORENAME "POWER5"
  570. #endif
  571. #if defined(FORCE_POWER6) || defined(FORCE_POWER7)
  572. #define FORCE
  573. #define ARCHITECTURE "POWER"
  574. #define SUBARCHITECTURE "POWER6"
  575. #define SUBDIRNAME "power"
  576. #define ARCHCONFIG "-DPOWER6 " \
  577. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=128 " \
  578. "-DL2_SIZE=4194304 -DL2_LINESIZE=128 " \
  579. "-DDTB_DEFAULT_ENTRIES=128 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
  580. #define LIBNAME "power6"
  581. #define CORENAME "POWER6"
  582. #endif
  583. #if defined(FORCE_POWER8)
  584. #define FORCE
  585. #define ARCHITECTURE "POWER"
  586. #define SUBARCHITECTURE "POWER8"
  587. #define SUBDIRNAME "power"
  588. #define ARCHCONFIG "-DPOWER8 " \
  589. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=128 " \
  590. "-DL2_SIZE=4194304 -DL2_LINESIZE=128 " \
  591. "-DDTB_DEFAULT_ENTRIES=128 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
  592. #define LIBNAME "power8"
  593. #define CORENAME "POWER8"
  594. #endif
  595. #if defined(FORCE_POWER9)
  596. #define FORCE
  597. #define ARCHITECTURE "POWER"
  598. #define SUBARCHITECTURE "POWER9"
  599. #define SUBDIRNAME "power"
  600. #define ARCHCONFIG "-DPOWER9 " \
  601. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=128 " \
  602. "-DL2_SIZE=4194304 -DL2_LINESIZE=128 " \
  603. "-DDTB_DEFAULT_ENTRIES=128 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
  604. #define LIBNAME "power9"
  605. #define CORENAME "POWER9"
  606. #endif
  607. #ifdef FORCE_PPCG4
  608. #define FORCE
  609. #define ARCHITECTURE "POWER"
  610. #define SUBARCHITECTURE "PPCG4"
  611. #define SUBDIRNAME "power"
  612. #define ARCHCONFIG "-DPPCG4 " \
  613. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=32 " \
  614. "-DL2_SIZE=262144 -DL2_LINESIZE=32 " \
  615. "-DDTB_DEFAULT_ENTRIES=128 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
  616. #define LIBNAME "ppcg4"
  617. #define CORENAME "PPCG4"
  618. #endif
  619. #ifdef FORCE_PPC970
  620. #define FORCE
  621. #define ARCHITECTURE "POWER"
  622. #define SUBARCHITECTURE "PPC970"
  623. #define SUBDIRNAME "power"
  624. #define ARCHCONFIG "-DPPC970 " \
  625. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=128 " \
  626. "-DL2_SIZE=512488 -DL2_LINESIZE=128 " \
  627. "-DDTB_DEFAULT_ENTRIES=128 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
  628. #define LIBNAME "ppc970"
  629. #define CORENAME "PPC970"
  630. #endif
  631. #ifdef FORCE_PPC970MP
  632. #define FORCE
  633. #define ARCHITECTURE "POWER"
  634. #define SUBARCHITECTURE "PPC970"
  635. #define SUBDIRNAME "power"
  636. #define ARCHCONFIG "-DPPC970 " \
  637. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=128 " \
  638. "-DL2_SIZE=1024976 -DL2_LINESIZE=128 " \
  639. "-DDTB_DEFAULT_ENTRIES=128 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
  640. #define LIBNAME "ppc970mp"
  641. #define CORENAME "PPC970"
  642. #endif
  643. #ifdef FORCE_PPC440
  644. #define FORCE
  645. #define ARCHITECTURE "POWER"
  646. #define SUBARCHITECTURE "PPC440"
  647. #define SUBDIRNAME "power"
  648. #define ARCHCONFIG "-DPPC440 " \
  649. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=32 " \
  650. "-DL2_SIZE=16384 -DL2_LINESIZE=128 " \
  651. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=16 "
  652. #define LIBNAME "ppc440"
  653. #define CORENAME "PPC440"
  654. #endif
  655. #ifdef FORCE_PPC440FP2
  656. #define FORCE
  657. #define ARCHITECTURE "POWER"
  658. #define SUBARCHITECTURE "PPC440FP2"
  659. #define SUBDIRNAME "power"
  660. #define ARCHCONFIG "-DPPC440FP2 " \
  661. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=32 " \
  662. "-DL2_SIZE=16384 -DL2_LINESIZE=128 " \
  663. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=16 "
  664. #define LIBNAME "ppc440FP2"
  665. #define CORENAME "PPC440FP2"
  666. #endif
  667. #ifdef FORCE_CELL
  668. #define FORCE
  669. #define ARCHITECTURE "POWER"
  670. #define SUBARCHITECTURE "CELL"
  671. #define SUBDIRNAME "power"
  672. #define ARCHCONFIG "-DCELL " \
  673. "-DL1_DATA_SIZE=262144 -DL1_DATA_LINESIZE=128 " \
  674. "-DL2_SIZE=512488 -DL2_LINESIZE=128 " \
  675. "-DDTB_DEFAULT_ENTRIES=128 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
  676. #define LIBNAME "cell"
  677. #define CORENAME "CELL"
  678. #endif
  679. #ifdef FORCE_SICORTEX
  680. #define FORCE
  681. #define ARCHITECTURE "MIPS"
  682. #define SUBARCHITECTURE "SICORTEX"
  683. #define SUBDIRNAME "mips"
  684. #define ARCHCONFIG "-DSICORTEX " \
  685. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=32 " \
  686. "-DL2_SIZE=512488 -DL2_LINESIZE=32 " \
  687. "-DDTB_DEFAULT_ENTRIES=32 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
  688. #define LIBNAME "mips"
  689. #define CORENAME "sicortex"
  690. #endif
  691. #ifdef FORCE_LOONGSON3A
  692. #define FORCE
  693. #define ARCHITECTURE "MIPS"
  694. #define SUBARCHITECTURE "LOONGSON3A"
  695. #define SUBDIRNAME "mips64"
  696. #define ARCHCONFIG "-DLOONGSON3A " \
  697. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=32 " \
  698. "-DL2_SIZE=512488 -DL2_LINESIZE=32 " \
  699. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=4 "
  700. #define LIBNAME "loongson3a"
  701. #define CORENAME "LOONGSON3A"
  702. #else
  703. #endif
  704. #ifdef FORCE_LOONGSON3B
  705. #define FORCE
  706. #define ARCHITECTURE "MIPS"
  707. #define SUBARCHITECTURE "LOONGSON3B"
  708. #define SUBDIRNAME "mips64"
  709. #define ARCHCONFIG "-DLOONGSON3B " \
  710. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=32 " \
  711. "-DL2_SIZE=512488 -DL2_LINESIZE=32 " \
  712. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=4 "
  713. #define LIBNAME "loongson3b"
  714. #define CORENAME "LOONGSON3B"
  715. #else
  716. #endif
  717. #ifdef FORCE_I6400
  718. #define FORCE
  719. #define ARCHITECTURE "MIPS"
  720. #define SUBARCHITECTURE "I6400"
  721. #define SUBDIRNAME "mips64"
  722. #define ARCHCONFIG "-DI6400 " \
  723. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=32 " \
  724. "-DL2_SIZE=1048576 -DL2_LINESIZE=32 " \
  725. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
  726. #define LIBNAME "i6400"
  727. #define CORENAME "I6400"
  728. #else
  729. #endif
  730. #ifdef FORCE_P6600
  731. #define FORCE
  732. #define ARCHITECTURE "MIPS"
  733. #define SUBARCHITECTURE "P6600"
  734. #define SUBDIRNAME "mips64"
  735. #define ARCHCONFIG "-DP6600 " \
  736. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=32 " \
  737. "-DL2_SIZE=1048576 -DL2_LINESIZE=32 " \
  738. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
  739. #define LIBNAME "p6600"
  740. #define CORENAME "P6600"
  741. #else
  742. #endif
  743. #ifdef FORCE_P5600
  744. #define FORCE
  745. #define ARCHITECTURE "MIPS"
  746. #define SUBARCHITECTURE "P5600"
  747. #define SUBDIRNAME "mips"
  748. #define ARCHCONFIG "-DP5600 " \
  749. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=32 " \
  750. "-DL2_SIZE=1048576 -DL2_LINESIZE=32 " \
  751. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
  752. #define LIBNAME "p5600"
  753. #define CORENAME "P5600"
  754. #else
  755. #endif
  756. #ifdef FORCE_I6500
  757. #define FORCE
  758. #define ARCHITECTURE "MIPS"
  759. #define SUBARCHITECTURE "I6500"
  760. #define SUBDIRNAME "mips64"
  761. #define ARCHCONFIG "-DI6500 " \
  762. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=32 " \
  763. "-DL2_SIZE=1048576 -DL2_LINESIZE=32 " \
  764. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
  765. #define LIBNAME "i6500"
  766. #define CORENAME "I6500"
  767. #else
  768. #endif
  769. #ifdef FORCE_ITANIUM2
  770. #define FORCE
  771. #define ARCHITECTURE "IA64"
  772. #define SUBARCHITECTURE "ITANIUM2"
  773. #define SUBDIRNAME "ia64"
  774. #define ARCHCONFIG "-DITANIUM2 " \
  775. "-DL1_DATA_SIZE=262144 -DL1_DATA_LINESIZE=128 " \
  776. "-DL2_SIZE=1572864 -DL2_LINESIZE=128 -DDTB_SIZE=16384 -DDTB_DEFAULT_ENTRIES=128 "
  777. #define LIBNAME "itanium2"
  778. #define CORENAME "itanium2"
  779. #endif
  780. #ifdef FORCE_SPARC
  781. #define FORCE
  782. #define ARCHITECTURE "SPARC"
  783. #define SUBARCHITECTURE "SPARC"
  784. #define SUBDIRNAME "sparc"
  785. #define ARCHCONFIG "-DSPARC -DV9 " \
  786. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=64 " \
  787. "-DL2_SIZE=1572864 -DL2_LINESIZE=64 -DDTB_SIZE=8192 -DDTB_DEFAULT_ENTRIES=64 "
  788. #define LIBNAME "sparc"
  789. #define CORENAME "sparc"
  790. #endif
  791. #ifdef FORCE_SPARCV7
  792. #define FORCE
  793. #define ARCHITECTURE "SPARC"
  794. #define SUBARCHITECTURE "SPARC"
  795. #define SUBDIRNAME "sparc"
  796. #define ARCHCONFIG "-DSPARC -DV7 " \
  797. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=64 " \
  798. "-DL2_SIZE=1572864 -DL2_LINESIZE=64 -DDTB_SIZE=8192 -DDTB_DEFAULT_ENTRIES=64 "
  799. #define LIBNAME "sparcv7"
  800. #define CORENAME "sparcv7"
  801. #endif
  802. #ifdef FORCE_GENERIC
  803. #define FORCE
  804. #define ARCHITECTURE "GENERIC"
  805. #define SUBARCHITECTURE "GENERIC"
  806. #define SUBDIRNAME "generic"
  807. #define ARCHCONFIG "-DGENERIC " \
  808. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=128 " \
  809. "-DL2_SIZE=512488 -DL2_LINESIZE=128 " \
  810. "-DDTB_DEFAULT_ENTRIES=128 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=8 "
  811. #define LIBNAME "generic"
  812. #define CORENAME "generic"
  813. #endif
  814. #ifdef FORCE_ARMV7
  815. #define FORCE
  816. #define ARCHITECTURE "ARM"
  817. #define SUBARCHITECTURE "ARMV7"
  818. #define SUBDIRNAME "arm"
  819. #define ARCHCONFIG "-DARMV7 " \
  820. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=32 " \
  821. "-DL2_SIZE=512488 -DL2_LINESIZE=32 " \
  822. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=4 " \
  823. "-DHAVE_VFPV3 -DHAVE_VFP"
  824. #define LIBNAME "armv7"
  825. #define CORENAME "ARMV7"
  826. #else
  827. #endif
  828. #ifdef FORCE_CORTEXA9
  829. #define FORCE
  830. #define ARCHITECTURE "ARM"
  831. #define SUBARCHITECTURE "CORTEXA9"
  832. #define SUBDIRNAME "arm"
  833. #define ARCHCONFIG "-DCORTEXA9 -DARMV7 " \
  834. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=32 " \
  835. "-DL2_SIZE=1048576 -DL2_LINESIZE=32 " \
  836. "-DDTB_DEFAULT_ENTRIES=128 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=4 " \
  837. "-DHAVE_VFPV3 -DHAVE_VFP -DHAVE_NEON"
  838. #define LIBNAME "cortexa9"
  839. #define CORENAME "CORTEXA9"
  840. #else
  841. #endif
  842. #ifdef FORCE_CORTEXA15
  843. #define FORCE
  844. #define ARCHITECTURE "ARM"
  845. #define SUBARCHITECTURE "CORTEXA15"
  846. #define SUBDIRNAME "arm"
  847. #define ARCHCONFIG "-DCORTEXA15 -DARMV7 " \
  848. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=32 " \
  849. "-DL2_SIZE=1048576 -DL2_LINESIZE=32 " \
  850. "-DDTB_DEFAULT_ENTRIES=128 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=4 " \
  851. "-DHAVE_VFPV3 -DHAVE_VFP -DHAVE_NEON"
  852. #define LIBNAME "cortexa15"
  853. #define CORENAME "CORTEXA15"
  854. #else
  855. #endif
  856. #ifdef FORCE_ARMV6
  857. #define FORCE
  858. #define ARCHITECTURE "ARM"
  859. #define SUBARCHITECTURE "ARMV6"
  860. #define SUBDIRNAME "arm"
  861. #define ARCHCONFIG "-DARMV6 " \
  862. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=32 " \
  863. "-DL2_SIZE=512488 -DL2_LINESIZE=32 " \
  864. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=4 " \
  865. "-DHAVE_VFP"
  866. #define LIBNAME "armv6"
  867. #define CORENAME "ARMV6"
  868. #else
  869. #endif
  870. #ifdef FORCE_ARMV5
  871. #define FORCE
  872. #define ARCHITECTURE "ARM"
  873. #define SUBARCHITECTURE "ARMV5"
  874. #define SUBDIRNAME "arm"
  875. #define ARCHCONFIG "-DARMV5 " \
  876. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=32 " \
  877. "-DL2_SIZE=512488 -DL2_LINESIZE=32 " \
  878. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=4 "
  879. #define LIBNAME "armv5"
  880. #define CORENAME "ARMV5"
  881. #else
  882. #endif
  883. #ifdef FORCE_ARMV8
  884. #define FORCE
  885. #define ARCHITECTURE "ARM64"
  886. #define SUBARCHITECTURE "ARMV8"
  887. #define SUBDIRNAME "arm64"
  888. #define ARCHCONFIG "-DARMV8 " \
  889. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 " \
  890. "-DL2_SIZE=262144 -DL2_LINESIZE=64 " \
  891. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=32 " \
  892. "-DHAVE_VFPV4 -DHAVE_VFPV3 -DHAVE_VFP -DHAVE_NEON -DARMV8"
  893. #define LIBNAME "armv8"
  894. #define CORENAME "ARMV8"
  895. #endif
  896. #ifdef FORCE_CORTEXA53
  897. #define FORCE
  898. #define ARCHITECTURE "ARM64"
  899. #define SUBARCHITECTURE "CORTEXA53"
  900. #define SUBDIRNAME "arm64"
  901. #define ARCHCONFIG "-DCORTEXA53 " \
  902. "-DL1_CODE_SIZE=32768 -DL1_CODE_LINESIZE=64 -DL1_CODE_ASSOCIATIVE=3 " \
  903. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 -DL1_DATA_ASSOCIATIVE=2 " \
  904. "-DL2_SIZE=262144 -DL2_LINESIZE=64 -DL2_ASSOCIATIVE=16 " \
  905. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  906. "-DHAVE_VFPV4 -DHAVE_VFPV3 -DHAVE_VFP -DHAVE_NEON -DARMV8"
  907. #define LIBNAME "cortexa53"
  908. #define CORENAME "CORTEXA53"
  909. #else
  910. #endif
  911. #ifdef FORCE_CORTEXA57
  912. #define FORCE
  913. #define ARCHITECTURE "ARM64"
  914. #define SUBARCHITECTURE "CORTEXA57"
  915. #define SUBDIRNAME "arm64"
  916. #define ARCHCONFIG "-DCORTEXA57 " \
  917. "-DL1_CODE_SIZE=49152 -DL1_CODE_LINESIZE=64 -DL1_CODE_ASSOCIATIVE=3 " \
  918. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 -DL1_DATA_ASSOCIATIVE=2 " \
  919. "-DL2_SIZE=2097152 -DL2_LINESIZE=64 -DL2_ASSOCIATIVE=16 " \
  920. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  921. "-DHAVE_VFPV4 -DHAVE_VFPV3 -DHAVE_VFP -DHAVE_NEON -DARMV8"
  922. #define LIBNAME "cortexa57"
  923. #define CORENAME "CORTEXA57"
  924. #else
  925. #endif
  926. #ifdef FORCE_CORTEXA72
  927. #define FORCE
  928. #define ARCHITECTURE "ARM64"
  929. #define SUBARCHITECTURE "CORTEXA72"
  930. #define SUBDIRNAME "arm64"
  931. #define ARCHCONFIG "-DCORTEXA72 " \
  932. "-DL1_CODE_SIZE=49152 -DL1_CODE_LINESIZE=64 -DL1_CODE_ASSOCIATIVE=3 " \
  933. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 -DL1_DATA_ASSOCIATIVE=2 " \
  934. "-DL2_SIZE=2097152 -DL2_LINESIZE=64 -DL2_ASSOCIATIVE=16 " \
  935. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  936. "-DHAVE_VFPV4 -DHAVE_VFPV3 -DHAVE_VFP -DHAVE_NEON -DARMV8"
  937. #define LIBNAME "cortexa72"
  938. #define CORENAME "CORTEXA72"
  939. #else
  940. #endif
  941. #ifdef FORCE_CORTEXA73
  942. #define FORCE
  943. #define ARCHITECTURE "ARM64"
  944. #define SUBARCHITECTURE "CORTEXA73"
  945. #define SUBDIRNAME "arm64"
  946. #define ARCHCONFIG "-DCORTEXA73 " \
  947. "-DL1_CODE_SIZE=49152 -DL1_CODE_LINESIZE=64 -DL1_CODE_ASSOCIATIVE=3 " \
  948. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 -DL1_DATA_ASSOCIATIVE=2 " \
  949. "-DL2_SIZE=2097152 -DL2_LINESIZE=64 -DL2_ASSOCIATIVE=16 " \
  950. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  951. "-DHAVE_VFPV4 -DHAVE_VFPV3 -DHAVE_VFP -DHAVE_NEON -DARMV8"
  952. #define LIBNAME "cortexa73"
  953. #define CORENAME "CORTEXA73"
  954. #else
  955. #endif
  956. #ifdef FORCE_FALKOR
  957. #define FORCE
  958. #define ARCHITECTURE "ARM64"
  959. #define SUBARCHITECTURE "FALKOR"
  960. #define SUBDIRNAME "arm64"
  961. #define ARCHCONFIG "-DFALKOR " \
  962. "-DL1_CODE_SIZE=49152 -DL1_CODE_LINESIZE=64 -DL1_CODE_ASSOCIATIVE=3 " \
  963. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 -DL1_DATA_ASSOCIATIVE=2 " \
  964. "-DL2_SIZE=2097152 -DL2_LINESIZE=64 -DL2_ASSOCIATIVE=16 " \
  965. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  966. "-DHAVE_VFPV4 -DHAVE_VFPV3 -DHAVE_VFP -DHAVE_NEON -DARMV8"
  967. #define LIBNAME "falkor"
  968. #define CORENAME "FALKOR"
  969. #else
  970. #endif
  971. #ifdef FORCE_THUNDERX
  972. #define FORCE
  973. #define ARCHITECTURE "ARM64"
  974. #define SUBARCHITECTURE "THUNDERX"
  975. #define SUBDIRNAME "arm64"
  976. #define ARCHCONFIG "-DTHUNDERX " \
  977. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=128 " \
  978. "-DL2_SIZE=16777216 -DL2_LINESIZE=128 -DL2_ASSOCIATIVE=16 " \
  979. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  980. "-DHAVE_VFPV4 -DHAVE_VFPV3 -DHAVE_VFP -DHAVE_NEON -DARMV8"
  981. #define LIBNAME "thunderx"
  982. #define CORENAME "THUNDERX"
  983. #else
  984. #endif
  985. #ifdef FORCE_THUNDERX2T99
  986. #define ARMV8
  987. #define FORCE
  988. #define ARCHITECTURE "ARM64"
  989. #define SUBARCHITECTURE "THUNDERX2T99"
  990. #define SUBDIRNAME "arm64"
  991. #define ARCHCONFIG "-DTHUNDERX2T99 " \
  992. "-DL1_CODE_SIZE=32768 -DL1_CODE_LINESIZE=64 -DL1_CODE_ASSOCIATIVE=8 " \
  993. "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=64 -DL1_DATA_ASSOCIATIVE=8 " \
  994. "-DL2_SIZE=262144 -DL2_LINESIZE=64 -DL2_ASSOCIATIVE=8 " \
  995. "-DL3_SIZE=33554432 -DL3_LINESIZE=64 -DL3_ASSOCIATIVE=32 " \
  996. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  997. "-DHAVE_VFPV4 -DHAVE_VFPV3 -DHAVE_VFP -DHAVE_NEON -DARMV8"
  998. #define LIBNAME "thunderx2t99"
  999. #define CORENAME "THUNDERX2T99"
  1000. #else
  1001. #endif
  1002. #ifdef FORCE_TSV110
  1003. #define FORCE
  1004. #define ARCHITECTURE "ARM64"
  1005. #define SUBARCHITECTURE "TSV110"
  1006. #define SUBDIRNAME "arm64"
  1007. #define ARCHCONFIG "-DTSV110 " \
  1008. "-DL1_CODE_SIZE=65536 -DL1_CODE_LINESIZE=64 -DL1_CODE_ASSOCIATIVE=4 " \
  1009. "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=64 -DL1_DATA_ASSOCIATIVE=4 " \
  1010. "-DL2_SIZE=524288 -DL2_LINESIZE=64 -DL2_ASSOCIATIVE=8 " \
  1011. "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
  1012. "-DHAVE_VFPV4 -DHAVE_VFPV3 -DHAVE_VFP -DHAVE_NEON -DARMV8"
  1013. #define LIBNAME "tsv110"
  1014. #define CORENAME "TSV110"
  1015. #else
  1016. #endif
  1017. #ifdef FORCE_ZARCH_GENERIC
  1018. #define FORCE
  1019. #define ARCHITECTURE "ZARCH"
  1020. #define SUBARCHITECTURE "ZARCH_GENERIC"
  1021. #define ARCHCONFIG "-DZARCH_GENERIC " \
  1022. "-DDTB_DEFAULT_ENTRIES=64"
  1023. #define LIBNAME "zarch_generic"
  1024. #define CORENAME "ZARCH_GENERIC"
  1025. #endif
  1026. #ifdef FORCE_Z13
  1027. #define FORCE
  1028. #define ARCHITECTURE "ZARCH"
  1029. #define SUBARCHITECTURE "Z13"
  1030. #define ARCHCONFIG "-DZ13 " \
  1031. "-DDTB_DEFAULT_ENTRIES=64"
  1032. #define LIBNAME "z13"
  1033. #define CORENAME "Z13"
  1034. #endif
  1035. #ifdef FORCE_Z14
  1036. #define FORCE
  1037. #define ARCHITECTURE "ZARCH"
  1038. #define SUBARCHITECTURE "Z14"
  1039. #define ARCHCONFIG "-DZ14 " \
  1040. "-DDTB_DEFAULT_ENTRIES=64"
  1041. #define LIBNAME "z14"
  1042. #define CORENAME "Z14"
  1043. #endif
  1044. #ifndef FORCE
  1045. #ifdef USER_TARGET
  1046. #error "The TARGET specified on the command line or in Makefile.rule is not supported. Please choose a target from TargetList.txt"
  1047. #endif
  1048. #if defined(__powerpc__) || defined(__powerpc) || defined(powerpc) || \
  1049. defined(__PPC__) || defined(PPC) || defined(_POWER) || defined(__POWERPC__)
  1050. #ifndef POWER
  1051. #define POWER
  1052. #endif
  1053. #define OPENBLAS_SUPPORTED
  1054. #endif
  1055. #if defined(__zarch__) || defined(__s390x__)
  1056. #define ZARCH
  1057. #include "cpuid_zarch.c"
  1058. #define OPENBLAS_SUPPORTED
  1059. #endif
  1060. #ifdef INTEL_AMD
  1061. #include "cpuid_x86.c"
  1062. #define OPENBLAS_SUPPORTED
  1063. #endif
  1064. #ifdef __ia64__
  1065. #include "cpuid_ia64.c"
  1066. #define OPENBLAS_SUPPORTED
  1067. #endif
  1068. #ifdef __alpha
  1069. #include "cpuid_alpha.c"
  1070. #define OPENBLAS_SUPPORTED
  1071. #endif
  1072. #ifdef POWER
  1073. #include "cpuid_power.c"
  1074. #define OPENBLAS_SUPPORTED
  1075. #endif
  1076. #ifdef sparc
  1077. #include "cpuid_sparc.c"
  1078. #define OPENBLAS_SUPPORTED
  1079. #endif
  1080. #ifdef __mips__
  1081. #ifdef __mips64
  1082. #include "cpuid_mips64.c"
  1083. #else
  1084. #include "cpuid_mips.c"
  1085. #endif
  1086. #define OPENBLAS_SUPPORTED
  1087. #endif
  1088. #ifdef __arm__
  1089. #include "cpuid_arm.c"
  1090. #define OPENBLAS_SUPPORTED
  1091. #endif
  1092. #ifdef __aarch64__
  1093. #include "cpuid_arm64.c"
  1094. #define OPENBLAS_SUPPORTED
  1095. #endif
  1096. #ifndef OPENBLAS_SUPPORTED
  1097. #error "This arch/CPU is not supported by OpenBLAS."
  1098. #endif
  1099. #else
  1100. #endif
  1101. static int get_num_cores(void) {
  1102. #ifdef OS_WINDOWS
  1103. SYSTEM_INFO sysinfo;
  1104. #elif defined(__FreeBSD__) || defined(__OpenBSD__) || defined(__DragonFly__) || defined(__APPLE__)
  1105. int m[2], count;
  1106. size_t len;
  1107. #endif
  1108. #if defined(linux) || defined(__sun__)
  1109. //returns the number of processors which are currently online
  1110. return sysconf(_SC_NPROCESSORS_CONF);
  1111. #elif defined(OS_WINDOWS)
  1112. GetSystemInfo(&sysinfo);
  1113. return sysinfo.dwNumberOfProcessors;
  1114. #elif defined(__FreeBSD__) || defined(__OpenBSD__) || defined(__DragonFly__) || defined(__APPLE__)
  1115. m[0] = CTL_HW;
  1116. m[1] = HW_NCPU;
  1117. len = sizeof(int);
  1118. sysctl(m, 2, &count, &len, NULL, 0);
  1119. return count;
  1120. #else
  1121. return 2;
  1122. #endif
  1123. }
  1124. int main(int argc, char *argv[]){
  1125. #ifdef FORCE
  1126. char buffer[8192], *p, *q;
  1127. int length;
  1128. #endif
  1129. if (argc == 1) return 0;
  1130. switch (argv[1][0]) {
  1131. case '0' : /* for Makefile */
  1132. #ifdef FORCE
  1133. printf("CORE=%s\n", CORENAME);
  1134. #else
  1135. #if defined(INTEL_AMD) || defined(POWER) || defined(__mips__) || defined(__arm__) || defined(__aarch64__) || defined(ZARCH) || defined(sparc)
  1136. printf("CORE=%s\n", get_corename());
  1137. #endif
  1138. #endif
  1139. #ifdef FORCE
  1140. printf("LIBCORE=%s\n", LIBNAME);
  1141. #else
  1142. printf("LIBCORE=");
  1143. get_libname();
  1144. printf("\n");
  1145. #endif
  1146. printf("NUM_CORES=%d\n", get_num_cores());
  1147. #if defined(__arm__) && !defined(FORCE)
  1148. get_features();
  1149. #endif
  1150. #ifdef INTEL_AMD
  1151. #ifndef FORCE
  1152. get_sse();
  1153. #else
  1154. sprintf(buffer, "%s", ARCHCONFIG);
  1155. p = &buffer[0];
  1156. while (*p) {
  1157. if ((*p == '-') && (*(p + 1) == 'D')) {
  1158. p += 2;
  1159. while ((*p != ' ') && (*p != '\0')) {
  1160. if (*p == '=') {
  1161. printf("=");
  1162. p ++;
  1163. while ((*p != ' ') && (*p != '\0')) {
  1164. printf("%c", *p);
  1165. p ++;
  1166. }
  1167. } else {
  1168. printf("%c", *p);
  1169. p ++;
  1170. if ((*p == ' ') || (*p =='\0')) printf("=1");
  1171. }
  1172. }
  1173. printf("\n");
  1174. } else p ++;
  1175. }
  1176. #endif
  1177. #endif
  1178. #ifdef MAKE_NB_JOBS
  1179. #if MAKE_NB_JOBS > 0
  1180. printf("MAKE += -j %d\n", MAKE_NB_JOBS);
  1181. #else
  1182. // Let make use parent -j argument or -j1 if there
  1183. // is no make parent
  1184. #endif
  1185. #elif NO_PARALLEL_MAKE==1
  1186. printf("MAKE += -j 1\n");
  1187. #else
  1188. printf("MAKE += -j %d\n", get_num_cores());
  1189. #endif
  1190. break;
  1191. case '1' : /* For config.h */
  1192. #ifdef FORCE
  1193. sprintf(buffer, "%s -DCORE_%s\n", ARCHCONFIG, CORENAME);
  1194. p = &buffer[0];
  1195. while (*p) {
  1196. if ((*p == '-') && (*(p + 1) == 'D')) {
  1197. p += 2;
  1198. printf("#define ");
  1199. while ((*p != ' ') && (*p != '\0')) {
  1200. if (*p == '=') {
  1201. printf(" ");
  1202. p ++;
  1203. while ((*p != ' ') && (*p != '\0')) {
  1204. printf("%c", *p);
  1205. p ++;
  1206. }
  1207. } else {
  1208. if (*p != '\n')
  1209. printf("%c", *p);
  1210. p ++;
  1211. }
  1212. }
  1213. printf("\n");
  1214. } else p ++;
  1215. }
  1216. #else
  1217. get_cpuconfig();
  1218. #endif
  1219. #ifdef FORCE
  1220. printf("#define CHAR_CORENAME \"%s\"\n", CORENAME);
  1221. #else
  1222. #if defined(INTEL_AMD) || defined(POWER) || defined(__mips__) || defined(__arm__) || defined(__aarch64__) || defined(ZARCH) || defined(sparc)
  1223. printf("#define CHAR_CORENAME \"%s\"\n", get_corename());
  1224. #endif
  1225. #endif
  1226. break;
  1227. case '2' : /* SMP */
  1228. if (get_num_cores() > 1) printf("SMP=1\n");
  1229. break;
  1230. }
  1231. fflush(stdout);
  1232. return 0;
  1233. }