Hi Ard, Sorry for my late response. Thank you help to sync the improvement to other GCC X64 toolchains, especial for the MS VA intrinsics one. I once mentioned I could sync them as below, and I'm glad you already done it. :)
http://thread.gmane.org/gmane.comp.bios.edk2.devel/13751/focus=14015 Steven Shi Intel\SSG\STO\UEFI Firmware Tel: +86 021-61166522 iNet: 821-6522 > -----Original Message----- > From: Ard Biesheuvel [mailto:ard.biesheu...@linaro.org] > Sent: Saturday, July 16, 2016 6:16 AM > To: edk2-devel@lists.01.org; ler...@redhat.com; af...@apple.com; Gao, > Liming <liming....@intel.com>; Shi, Steven <steven....@intel.com>; Zhu, > Yonghong <yonghong....@intel.com>; Kinney, Michael D > <michael.d.kin...@intel.com>; Justen, Jordan L <jordan.l.jus...@intel.com> > Cc: br...@cran.org.uk; pbonz...@redhat.com; Ard Biesheuvel > <ard.biesheu...@linaro.org> > Subject: [PATCH v2 0/5] MdePkg BaseTools: GCC optimization for X64 > > This is v2 of the series to enable compiler optimization under GCC for builds > targetting X64. It includes a patch from Steven Shi, taken from his series to > introduce support for GCC 5.x and Clang 3.8/X64 to EDK2. > > The first patch fixes the issue that __builtin_unreachable() is not > implemented by GCC 4.4 or earlier. > > Patch #2 is Steven's patch to switch to the flavor of VA_LIST that is > explicitly modeled after the MS implementation. This by itself is an > improvement, since the open coded implementation that performs > arithmetic > on the address of explicit arguments to obtain the variadic arguments is > fragile and difficult to maintain, and should be best avoided. > > Patch #3 enables -Os optimization for X64/RELEASE and X86/DEBUG all the > way > back to GCC44. This aligns X64 with IA32, which is already built using -Os > for both DEBUG and RELEASE. > > Note that the reported breakage regarding the use of -Os [0] is likely caused > by the poor man's __builtin_ms_va_list (NO_BUILTIN_VA_FUNCS) that we > used for > GCC/X64 before Steven's patch above switches us to a properly defined > builtin > type. (The poor man's implementation uses pointer arithmetic involving the > address of the last named argument to calculate pointers to the variadic > arguments and dereferences them to obtain the argument values. I'm quite > surprised it works as well as it does at -O0, to be honest) > > I tested this change with both Ovmf and EmulatorPkg built in various ways > and with various versions, with the caveat that I did not always use a > matching > binutils (i.e., of the same era). Since the issues this series deal with are > all code generation issues, I think this is reasonable, but more testing would > be appreciated. (v2: this still applies, I have tested DEBUG and RELEASE > builds > using GCC44, GCC46, GCC47 and GCC48, but all on the same Ubuntu 14.04 > system, > which means they all use the same ld version 2.24.90) > > [0] http://thread.gmane.org/gmane.comp.bios.tianocore.devel/10963 > > Patch #4 applies the visibility 'hidden' GCC pragma globally. Please refer to > the commit log for the motivation. > > Patch #5 switches GCC/X64 to the PIC small code model, which results in > smaller > code. > > Changes in v2: > - Switch to -Os optimization, both for DEBUG and RELEASE builds. This turns > out > to work fine (as far as I could establish) and gives the best results in terms > of code size. The following numbers were produced with the following > invocation > > build -a X64 -t GCC44 -t GCC46 -t GCC47 -t GCC48 -p > OvmfPkg/OvmfPkgX64.dsc \ > -D SECURE_BOOT_ENABLE -b DEBUG -b RELEASE > > DEBUG_GCC44 > > SECFV [14%Full] 212992 total, 31632 used, 181360 free > FVMAIN_COMPACT [74%Full] 1753088 total, 1300904 used, 452184 free > DXEFV [50%Full] 10485760 total, 5272464 used, 5213296 free > PEIFV [25%Full] 917504 total, 231496 used, 686008 free > > DEBUG_GCC46 > > SECFV [14%Full] 212992 total, 31728 used, 181264 free > FVMAIN_COMPACT [73%Full] 1753088 total, 1291672 used, 461416 free > DXEFV [50%Full] 10485760 total, 5263888 used, 5221872 free > PEIFV [25%Full] 917504 total, 231656 used, 685848 free > > DEBUG_GCC47 > > SECFV [14%Full] 212992 total, 31440 used, 181552 free > FVMAIN_COMPACT [73%Full] 1753088 total, 1290584 used, 462504 free > DXEFV [49%Full] 10485760 total, 5234224 used, 5251536 free > PEIFV [25%Full] 917504 total, 230152 used, 687352 free > > DEBUG_GCC48 > > SECFV [14%Full] 212992 total, 31280 used, 181712 free > FVMAIN_COMPACT [73%Full] 1753088 total, 1296208 used, 456880 free > DXEFV [49%Full] 10485760 total, 5229424 used, 5256336 free > PEIFV [25%Full] 917504 total, 229512 used, 687992 free > > RELEASE_GCC44 > > SECFV [6%Full] 212992 total, 14320 used, 198672 free > FVMAIN_COMPACT [56%Full] 1753088 total, 997184 used, 755904 free > DXEFV [33%Full] 10485760 total, 3508336 used, 6977424 free > PEIFV [11%Full] 917504 total, 109896 used, 807608 free > > RELEASE_GCC46 > > SECFV [6%Full] 212992 total, 13936 used, 199056 free > FVMAIN_COMPACT [56%Full] 1753088 total, 987632 used, 765456 free > DXEFV [33%Full] 10485760 total, 3468304 used, 7017456 free > PEIFV [11%Full] 917504 total, 107784 used, 809720 free > > RELEASE_GCC47 > > SECFV [6%Full] 212992 total, 13808 used, 199184 free > FVMAIN_COMPACT [56%Full] 1753088 total, 988072 used, 765016 free > DXEFV [32%Full] 10485760 total, 3449936 used, 7035824 free > PEIFV [11%Full] 917504 total, 106984 used, 810520 free > > RELEASE_GCC48 > > SECFV [6%Full] 212992 total, 13744 used, 199248 free > FVMAIN_COMPACT [56%Full] 1753088 total, 992800 used, 760288 free > DXEFV [32%Full] 10485760 total, 3453456 used, 7032304 free > PEIFV [11%Full] 917504 total, 107464 used, 810040 free > > For comparison, the baseline RELEASE_GCC48 (i.e., using -0O and the large > code > model) is > > SECFV [11%Full] 212992 total, 25392 used, 187600 free > FVMAIN_COMPACT [61%Full] 1753088 total, 1076488 used, 676600 free > DXEFV [54%Full] 10485760 total, 5743440 used, 4742320 free > PEIFV [21%Full] 917504 total, 200328 used, 717176 free > > Ard Biesheuvel (4): > MdePkg: avoid __builtin_unreachable() on GCC v4.4 > BaseTools/tools_def: enable Os optimization for GCC X64 builds > MdePkg X64: force 'hidden' visibility when building with -fpic > BaseTools/tools_def: switch GCC/X64 to the PIC small model > > Shi, Steven (1): > MdePkg: Enable new MS VA intrinsics for GNUC x86 64bits build > > BaseTools/Conf/tools_def.template | 2 +- > MdePkg/Include/Base.h | 29 +++++++++++++++++++- > MdePkg/Include/X64/ProcessorBind.h | 10 +++++++ > 3 files changed, 39 insertions(+), 2 deletions(-) > > -- > 1.9.1 _______________________________________________ edk2-devel mailing list edk2-devel@lists.01.org https://lists.01.org/mailman/listinfo/edk2-devel
