Re: [PATCH 2/2] MIPS: Loongson64: Wrap -mno-branch-likely with cc-option

From: Nathan Chancellor
Date: Mon Jan 24 2022 - 23:47:40 EST

Next message: Li Chen: "RE: Why does cdns_pcie_ep_set_bar use sz > SZ_2G for is_64bits in pcie-cadence-ep.c?"
Previous message: Lu Baolu: "Re: [PATCH 6/7] iommu: Use right way to retrieve iommu_ops"
In reply to: Nick Desaulniers: "Re: [PATCH 2/2] MIPS: Loongson64: Wrap -mno-branch-likely with cc-option"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Mon, Jan 24, 2022 at 02:49:29PM -0800, Nick Desaulniers wrote:
> On Mon, Jan 24, 2022 at 2:37 PM Nathan Chancellor <nathan@xxxxxxxxxx> wrote:
> >
> > On Mon, Jan 24, 2022 at 12:40:58PM -0800, Nick Desaulniers wrote:
> > > On Thu, Jan 20, 2022 at 1:40 PM Nathan Chancellor <nathan@xxxxxxxxxx> wrote:
> > > >
> > > > This flag is not supported by clang, which results in a warning:
> > > >
> > > > clang-14: warning: argument unused during compilation: '-mno-branch-likely' [-Wunused-command-line-argument]
> > > >
> > > > This breaks cc-option, which adds -Werror to make this warning fatal and
> > > > catch flags that are not supported. Wrap the flag in cc-option so that
> > > > it does not cause cc-option to fail, which can cause randconfigs to be
> > > > really noisy.
> > > >
> > > > Signed-off-by: Nathan Chancellor <nathan@xxxxxxxxxx>
> > >
> > > We should convert more of these tests to Kconfig checks that run once
> > > per configuration, rather than multiple times for a build. IIRC Linus
> > > mentioned this somewhere...yeah, the -Wimplicit-fallthrough patches.
> > > See
> > > dee2b702bcf06 ("kconfig: Add support for -Wimplicit-fallthrough")
> > >
> > > I wonder if we can check ARCH or SUBARCH in Kconfig to limit invoking
> > > the tool under test for certain arch specific command line flags?
> > >
> > > I'll take this patch over such a larger change, but I think towards
> > > the goal of speeding up already configured builds, we eventually want
> > > to be migrating cc-option and ld-option checks to Kconfig.
> > >
> > > Reviewed-by: Nick Desaulniers <ndesaulniers@xxxxxxxxxx>
> >
> > Something like this appears to work, if that is more preferrable?
> >
> > diff --git a/arch/mips/Kconfig b/arch/mips/Kconfig
> > index 058446f01487..a27a3ade810e 100644
> > --- a/arch/mips/Kconfig
> > +++ b/arch/mips/Kconfig
> > @@ -3232,3 +3232,9 @@ endmenu
> > source "arch/mips/kvm/Kconfig"
> >
> > source "arch/mips/vdso/Kconfig"
> > +
> > +config CC_MNO_BRANCH_LIKELY
> > + string
> > + default "-mno-branch-likely"
> > + depends on MACH_LOONGSON64 || MACH_LOONGSON2EF
> > + depends on $(cc-option,-mno-branch-likely)
> > diff --git a/arch/mips/loongson2ef/Platform b/arch/mips/loongson2ef/Platform
> > index 50e659aca543..66ed09581417 100644
> > --- a/arch/mips/loongson2ef/Platform
> > +++ b/arch/mips/loongson2ef/Platform
> > @@ -41,6 +41,6 @@ cflags-y += $(call cc-option,-mno-loongson-mmi)
> > # Loongson Machines' Support
> > #
> >
> > -cflags-$(CONFIG_MACH_LOONGSON2EF) += -I$(srctree)/arch/mips/include/asm/mach-loongson2ef -mno-branch-likely
> > +cflags-$(CONFIG_MACH_LOONGSON2EF) += -I$(srctree)/arch/mips/include/asm/mach-loongson2ef $(CONFIG_CC_MNO_BRANCH_LIKELY)
>
> Does that allow someone to modify the value for CC_MNO_BRANCH_LIKELY
> in menuconfig?

No because it is a hidden symbol; in other words:

config ...
string

rather than

config ...
string "..."

> If so, I'd rather the Makefiles have:
>
> cflags-$(CONFIG_CC_MNO_BRANCH_LIKELY) += -mno-branch-likely
>
> and CONFIG_CC_MNO_BRANCH_LIKELY be a bool rather than an editable
> string. I think that makes the Makefile more readable; you don't have
> to see what CONFIG_CC_MNO_BRANCH_LIKELY expands to in a different
> file.
>
> See also CC_HAS_ASM_GOTO and CC_HAS_ASM_GOTO_OUTPUT in init/Kconfig.

Regardless, your suggestion is simpler. I can send that as v2 if this is
preferrable.

diff --git a/arch/mips/Kconfig b/arch/mips/Kconfig
index 058446f01487..4f83fb5610cb 100644
--- a/arch/mips/Kconfig
+++ b/arch/mips/Kconfig
@@ -3232,3 +3232,7 @@ endmenu
source "arch/mips/kvm/Kconfig"

source "arch/mips/vdso/Kconfig"
+
+config CC_HAS_MNO_BRANCH_LIKELY
+ def_bool y
+ depends on $(cc-option,-mno-branch-likely)
diff --git a/arch/mips/loongson2ef/Platform b/arch/mips/loongson2ef/Platform
index 50e659aca543..eebabf9df6ac 100644
--- a/arch/mips/loongson2ef/Platform
+++ b/arch/mips/loongson2ef/Platform
@@ -41,6 +41,7 @@ cflags-y += $(call cc-option,-mno-loongson-mmi)
# Loongson Machines' Support
#

-cflags-$(CONFIG_MACH_LOONGSON2EF) += -I$(srctree)/arch/mips/include/asm/mach-loongson2ef -mno-branch-likely
+cflags-$(CONFIG_MACH_LOONGSON2EF) += -I$(srctree)/arch/mips/include/asm/mach-loongson2ef
+cflags-$(CONFIG_CC_HAS_MNO_BRANCH_LIKELY) += -mno-branch-likely
load-$(CONFIG_LEMOTE_FULOONG2E) += 0xffffffff80100000
load-$(CONFIG_LEMOTE_MACH2F) += 0xffffffff80200000
diff --git a/arch/mips/loongson64/Platform b/arch/mips/loongson64/Platform
index 3e660d6d3c2b..2fd9dc218a68 100644
--- a/arch/mips/loongson64/Platform
+++ b/arch/mips/loongson64/Platform
@@ -33,5 +33,6 @@ cflags-y += $(call cc-option,-mno-loongson-mmi)
# Loongson Machines' Support
#

-cflags-$(CONFIG_MACH_LOONGSON64) += -I$(srctree)/arch/mips/include/asm/mach-loongson64 -mno-branch-likely
+cflags-$(CONFIG_MACH_LOONGSON64) += -I$(srctree)/arch/mips/include/asm/mach-loongson64
+cflags-$(CONFIG_CC_HAS_MNO_BRANCH_LIKELY) += -mno-branch-likely
load-$(CONFIG_CPU_LOONGSON64) += 0xffffffff80200000

> > load-$(CONFIG_LEMOTE_FULOONG2E) += 0xffffffff80100000
> > load-$(CONFIG_LEMOTE_MACH2F) += 0xffffffff80200000
> > diff --git a/arch/mips/loongson64/Platform b/arch/mips/loongson64/Platform
> > index 3e660d6d3c2b..88fbdfe9ffcc 100644
> > --- a/arch/mips/loongson64/Platform
> > +++ b/arch/mips/loongson64/Platform
> > @@ -33,5 +33,5 @@ cflags-y += $(call cc-option,-mno-loongson-mmi)
> > # Loongson Machines' Support
> > #
> >
> > -cflags-$(CONFIG_MACH_LOONGSON64) += -I$(srctree)/arch/mips/include/asm/mach-loongson64 -mno-branch-likely
> > +cflags-$(CONFIG_MACH_LOONGSON64) += -I$(srctree)/arch/mips/include/asm/mach-loongson64 $(CONFIG_CC_MNO_BRANCH_LIKELY)
> > load-$(CONFIG_CPU_LOONGSON64) += 0xffffffff80200000
> >
> > > > ---
> > > > arch/mips/loongson64/Platform | 3 ++-
> > > > 1 file changed, 2 insertions(+), 1 deletion(-)
> > > >
> > > > diff --git a/arch/mips/loongson64/Platform b/arch/mips/loongson64/Platform
> > > > index 981d3abc150e..acf9edc9b15d 100644
> > > > --- a/arch/mips/loongson64/Platform
> > > > +++ b/arch/mips/loongson64/Platform
> > > > @@ -26,5 +26,6 @@ cflags-y += $(call cc-option,-mno-loongson-mmi)
> > > > # Loongson Machines' Support
> > > > #
> > > >
> > > > -cflags-$(CONFIG_MACH_LOONGSON64) += -I$(srctree)/arch/mips/include/asm/mach-loongson64 -mno-branch-likely
> > > > +cflags-$(CONFIG_MACH_LOONGSON64) += -I$(srctree)/arch/mips/include/asm/mach-loongson64
> > > > +cflags-$(CONFIG_MACH_LOONGSON64) += $(call cc-option,-mno-branch-likely)
> > > > load-$(CONFIG_CPU_LOONGSON64) += 0xffffffff80200000
> > > > --
> > > > 2.34.1
> > > >
> > > >
> > >
> > >
> > > --
> > > Thanks,
> > > ~Nick Desaulniers
> >
>
>
> --
> Thanks,
> ~Nick Desaulniers

Next message: Li Chen: "RE: Why does cdns_pcie_ep_set_bar use sz > SZ_2G for is_64bits in pcie-cadence-ep.c?"
Previous message: Lu Baolu: "Re: [PATCH 6/7] iommu: Use right way to retrieve iommu_ops"
In reply to: Nick Desaulniers: "Re: [PATCH 2/2] MIPS: Loongson64: Wrap -mno-branch-likely with cc-option"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]