Re: [dpdk-dev] [PATCH v4 01/10] build: add an option to enable LTO build

[Bruce Richardson <bruce.richardson@intel.com>]

On Tue, Oct 22, 2019 at 01:54:03PM +0200, Andrzej Ostruszka wrote:
> This patch adds an option to enable link time optimization.  In addition
> to LTO option itself (-flto) fat-lto-objects are being used.  This is
> because during the build pmdinfogen scans the generated ELF objects to
> find this_pmd_name* symbol in symbol table.  Without fat-lto-objects gcc
> produces ELF only with extra symbols for internal use during linking.
> 
> Signed-off-by: Andrzej Ostruszka <aostruszka@marvell.com>
> ---
>  .travis.yml                                  |  7 ++++
>  config/common_base                           |  5 +++
>  config/meson.build                           | 13 +++++++
>  doc/guides/prog_guide/lto.rst                | 36 ++++++++++++++++++++
>  doc/guides/rel_notes/release_19_11.rst       |  8 +++++
>  lib/librte_distributor/rte_distributor.c     | 18 +++++-----
>  lib/librte_distributor/rte_distributor_v20.c | 18 +++++-----
>  lib/librte_lpm/rte_lpm.c                     | 28 +++++++--------
>  lib/librte_lpm/rte_lpm6.c                    | 16 ++++-----
>  lib/librte_timer/rte_timer.c                 | 20 +++++------
>  mk/toolchain/gcc/rte.toolchain-compat.mk     |  4 +++
>  mk/toolchain/gcc/rte.vars.mk                 | 12 +++++++
>  mk/toolchain/icc/rte.vars.mk                 |  8 +++++
>  13 files changed, 143 insertions(+), 50 deletions(-)
>  create mode 100644 doc/guides/prog_guide/lto.rst
> 
> diff --git a/.travis.yml b/.travis.yml
> index 781f9f666..70d221852 100644
> --- a/.travis.yml
> +++ b/.travis.yml
> @@ -31,6 +31,7 @@ env:
>    - DEF_LIB="static" OPTS="-Denable_kmods=false"
>    - DEF_LIB="shared" OPTS="-Denable_kmods=false"
>    - DEF_LIB="shared" RUN_TESTS=1 BUILD_DOCS=1
> +  - DEF_LIB="shared" OPTS="-Db_lto=true"
>  
>  matrix:
>    include:
> @@ -100,6 +101,12 @@ matrix:
>        apt:
>          packages:
>            - *extra_packages
> +  - env: DEF_LIB="shared" OPTS="-Db_lto=true" EXTRA_PACKAGES=1
> +    compiler: gcc
> +    addons:
> +      apt:
> +        packages:
> +          - *extra_packages
>  
>  
>  script: ./.ci/${TRAVIS_OS_NAME}-build.sh
> diff --git a/config/common_base b/config/common_base
> index 8ef75c203..73a55fdec 100644
> --- a/config/common_base
> +++ b/config/common_base
> @@ -49,6 +49,11 @@ CONFIG_RTE_FORCE_INTRINSICS=n
>  #
>  CONFIG_RTE_ARCH_STRICT_ALIGN=n
>  
> +#
> +# Enable link time optimization
> +#
> +CONFIG_RTE_ENABLE_LTO=n
> +
>  #
>  # Compile to share library
>  #
> diff --git a/config/meson.build b/config/meson.build
> index 2bafea530..1a5093118 100644
> --- a/config/meson.build
> +++ b/config/meson.build
> @@ -196,3 +196,16 @@ add_project_arguments('-D_GNU_SOURCE', language: 'c')
>  if is_freebsd
>  	add_project_arguments('-D__BSD_VISIBLE', language: 'c')
>  endif
> +
> +if get_option('b_lto')
> +	if cc.has_argument('-ffat-lto-objects')
> +		add_project_arguments('-ffat-lto-objects', language: 'c')
> +	else
> +		error('compiler does not support fat LTO objects - please turn LTO off')
> +	endif
> +	# workaround for gcc bug 81440
> +	if cc.get_id() == 'gcc' and cc.version().version_compare('<8.0')
> +		add_project_arguments('-Wno-lto-type-mismatch', language: 'c')
> +		add_project_link_arguments('-Wno-lto-type-mismatch', language: 'c')
> +	endif
> +endif

This LGTM

> diff --git a/doc/guides/prog_guide/lto.rst b/doc/guides/prog_guide/lto.rst
> new file mode 100644
> index 000000000..b6daabc86
> --- /dev/null
<snip>
> diff --git a/mk/toolchain/gcc/rte.toolchain-compat.mk b/mk/toolchain/gcc/rte.toolchain-compat.mk
> index ea40a11c0..ad4fad83c 100644
> --- a/mk/toolchain/gcc/rte.toolchain-compat.mk
> +++ b/mk/toolchain/gcc/rte.toolchain-compat.mk
> @@ -88,6 +88,10 @@ else
>  		MACHINE_CFLAGS := $(filter-out -march% -mtune% -msse%,$(MACHINE_CFLAGS))
>  	endif
>  
> +	ifeq ($(shell test $(GCC_VERSION) -lt 45 && echo 1), 1)
> +		CONFIG_RTE_ENABLE_LTO=n
> +	endif
> +

I don't think we support GCC versions that far back any more, but this is
fairly harmless, so probably ok to keep.

Acked-by: Bruce Richardson <bruce.richarson@intel.com>