Skip to contentSkip to conversation

arch/arm: introduce big.LITTLE configurations

PATCH Superseded Yann E. MORIN 4 participants 14 messages Jul 8, 2017 – Jul 19, 2017

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Yann E. MORIN

author
#0 lore raw

v2 cover arch/arm: introduce big.LITTLE configurations

Hello All!

This series adds four new CPU variants to the arm choice, to enable optimising for big.LITTLE CPUs.

We first fix the current values in the first patch, then introduce the actual new variants, and then make use of them in the sole package that needs to know of them.

This has been only slightly tested by simply building the toolchains for the six possibilities:
- the four variants in arm mode (32-bit),
- the two armv8 variants in aarch64 mode (64-bit).

The variants were introduced in various versions of gcc, so we protect the internal gcc as well as the external toolchains accordingly.

To be noted: openblas also has a cortex-a15 optimisation, so maybe that optimisation can also be used for the cortex-a15.cortex-a7 variant as well. But I have absolutely no clue what openblas is doing with that anyway, so I left it alone for someone else to deal with...

Regards,
Yann E. MORIN.

The following changes since commit 644c024edeb7e20bd96ce4426aeca2cbcd74d9b9

syslinux: drop patch 0003, not needed after gnu-efi update (2017-07-04 18:45:26 +0200)

are available in the git repository at:

git://git.buildroot.org/~ymorin/git/buildroot.git

for you to fetch changes up to d1065a08c9b2a83d4611a6566f068528a5b1ca3c

package/kvm-unit-test: available for big.LITTLE arm cores too (2017-07-08 15:51:23 +0200)

----------------------------------------------------------------
Yann E. MORIN (4):
arch/arm: fix -mcpu default values for AArch64 package/kvm-unit-tests: introduce _ARCH_SUPPORTS arch/arm: add big.LITTLE cpu variants package/kvm-unit-test: available for big.LITTLE arm cores too

 arch/Config.in.arm                                 | 50 ++++++++++++++++++----
 package/gcc/Config.in.host                         |  5 ++-
 package/kvm-unit-tests/Config.in                   | 19 ++++----
 .../Config.in                                      |  2 +
 .../toolchain-external-codesourcery-arm/Config.in  |  4 ++
 5 files changed, 62 insertions(+), 18 deletions(-)
Signature · 6 lines
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'
author
#1 lore raw

PATCH v2 1/4 arch/arm: fix -mcpu default values for AArch64

We have to specify the -mcpu value, even in 64-bit mode.

For AArch64, +fp and +simd are the default, so they are totally useless.

Signed-off-by "Yann E. MORIN" <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Baruch Siach <[email protected]>

---
Changes v1 -> v2:
- squash the two changes into one patch (Thomas)
- remove useless parentheses (Baruch)
- remove superfluous indentation

Notes: If anything, it is the +nofp or +nosimd that should be specified. This can be done in a future patch, though.

---
 arch/Config.in.arm | 12 +++---------
 1 file changed, 3 insertions(+), 9 deletions(-)
arch/Config.in.arm +3 -9
index b1817ed568..3cafe21700 100644
@@ -534,15 +534,9 @@ config BR2_GCC_TARGET_CPU
 	default "strongarm"	if BR2_strongarm
 	default "xscale"	if BR2_xscale
 	default "iwmmxt"	if BR2_iwmmxt
-	default "cortex-a53"		if (BR2_cortex_a53 && !BR2_ARCH_IS_64)
-	default "cortex-a53+fp"		if (BR2_cortex_a53 && BR2_ARCH_IS_64 && BR2_ARM_FPU_FP_ARMV8)
-	default "cortex-a53+fp+simd"	if (BR2_cortex_a53 && BR2_ARCH_IS_64 && BR2_ARM_FPU_NEON_FP_ARMV8)
-	default "cortex-a57"		if (BR2_cortex_a57 && !BR2_ARCH_IS_64)
-	default "cortex-a57+fp"		if (BR2_cortex_a57 && BR2_ARCH_IS_64 && BR2_ARM_FPU_FP_ARMV8)
-	default "cortex-a57+fp+simd"	if (BR2_cortex_a57 && BR2_ARCH_IS_64 && BR2_ARM_FPU_NEON_FP_ARMV8)
-	default "cortex-a72"		if (BR2_cortex_a72 && !BR2_ARCH_IS_64)
-	default "cortex-a72+fp"		if (BR2_cortex_a72 && BR2_ARCH_IS_64 && BR2_ARM_FPU_FP_ARMV8)
-	default "cortex-a72+fp+simd"	if (BR2_cortex_a72 && BR2_ARCH_IS_64 && BR2_ARM_FPU_NEON_FP_ARMV8)
+	default "cortex-a53"	if BR2_cortex_a53
+	default "cortex-a57"	if BR2_cortex_a57
+	default "cortex-a72"	if BR2_cortex_a72
 
 config BR2_GCC_TARGET_ABI
 	default "aapcs-linux"	if BR2_arm || BR2_armeb
Signature · 1 line
2.11.0
#2 lore raw

PATCH v2 1/4 arch/arm: fix -mcpu default values for AArch64

Quoted 3 lines · Yann E. MORIN

We have to specify the -mcpu value, even in 64-bit mode.

For AArch64, +fp and +simd are the default, so they are totally useless.

However, we typically want to be explicit about things, so when the +nofp and
+nosimd are added we want these lines to return as well. Wouldn't it be better to just add the noXX variants in this patch?

That said, this patch actually fixes the !FP case so

Reviewed-by Arnout Vandecappelle (Essensium/Mind) <[email protected]>

Regards,
Arnout

Quoted 41 lines
Signed-off-by "Yann E. MORIN" <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Baruch Siach <[email protected]>

---
Changes v1 -> v2:
- squash the two changes into one patch (Thomas)
- remove useless parentheses (Baruch)
- remove superfluous indentation

Notes: If anything, it is the +nofp or +nosimd that should be specified. This can be done in a future patch, though.

---
 arch/Config.in.arm | 12 +++---------
 1 file changed, 3 insertions(+), 9 deletions(-)
arch/Config.in.arm +3 -9
index b1817ed568..3cafe21700 100644
@@ -534,15 +534,9 @@ config BR2_GCC_TARGET_CPU
 	default "strongarm"	if BR2_strongarm
 	default "xscale"	if BR2_xscale
 	default "iwmmxt"	if BR2_iwmmxt
-	default "cortex-a53"		if (BR2_cortex_a53 && !BR2_ARCH_IS_64)
-	default "cortex-a53+fp"		if (BR2_cortex_a53 && BR2_ARCH_IS_64 && BR2_ARM_FPU_FP_ARMV8)
-	default "cortex-a53+fp+simd"	if (BR2_cortex_a53 && BR2_ARCH_IS_64 && BR2_ARM_FPU_NEON_FP_ARMV8)
-	default "cortex-a57"		if (BR2_cortex_a57 && !BR2_ARCH_IS_64)
-	default "cortex-a57+fp"		if (BR2_cortex_a57 && BR2_ARCH_IS_64 && BR2_ARM_FPU_FP_ARMV8)
-	default "cortex-a57+fp+simd"	if (BR2_cortex_a57 && BR2_ARCH_IS_64 && BR2_ARM_FPU_NEON_FP_ARMV8)
-	default "cortex-a72"		if (BR2_cortex_a72 && !BR2_ARCH_IS_64)
-	default "cortex-a72+fp"		if (BR2_cortex_a72 && BR2_ARCH_IS_64 && BR2_ARM_FPU_FP_ARMV8)
-	default "cortex-a72+fp+simd"	if (BR2_cortex_a72 && BR2_ARCH_IS_64 && BR2_ARM_FPU_NEON_FP_ARMV8)
+	default "cortex-a53"	if BR2_cortex_a53
+	default "cortex-a57"	if BR2_cortex_a57
+	default "cortex-a72"	if BR2_cortex_a72
 
 config BR2_GCC_TARGET_ABI
 	default "aapcs-linux"	if BR2_arm || BR2_armeb
Signature · 6 lines
Arnout Vandecappelle                          arnout at mind be
Senior Embedded Software Architect            +32-16-286500
Essensium/Mind                                http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint:  7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
author
#3 lore raw

PATCH v2 1/4 arch/arm: fix -mcpu default values for AArch64

Arnout, All,

On 2017-07-08 17:33 +0200, Arnout Vandecappelle spake thusly:

Quoted 8 lines · Yann E. MORIN
Quoted 3 lines

We have to specify the -mcpu value, even in 64-bit mode.

For AArch64, +fp and +simd are the default, so they are totally useless.

However, we typically want to be explicit about things, so when the +nofp and
+nosimd are added we want these lines to return as well. Wouldn't it be better to just add the noXX variants in this patch?

When we talked about that with Thomas, he said he preferred we drop them altogether.

Also, adding the no variants is not that easy, because it involves all of 64-bit, vfpv8 and neon.

However, we do not have an option for noen on 64-bit for now. We can't re-use the existing 32-bit neon option either, because some packages expect that neon is the 32-bit variant.

So, it is a little bit more involved.

Since the no variants were not really working anyway for now, there is yet no regression yet.

I also just decided to stay with just adding the big.LITTLE variants rather than send another many-patches series like I'm used to.

Quoted 3 lines

That said, this patch actually fixes the !FP case so

Reviewed-by Arnout Vandecappelle (Essensium/Mind) <[email protected]>

Thanks!

Regards,
Yann E. MORIN.

Quoted 52 lines

Regards,
Arnout

Quoted 41 lines
Signed-off-by "Yann E. MORIN" <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Baruch Siach <[email protected]>

---
Changes v1 -> v2:
- squash the two changes into one patch (Thomas)
- remove useless parentheses (Baruch)
- remove superfluous indentation

Notes: If anything, it is the +nofp or +nosimd that should be specified. This can be done in a future patch, though.

---
 arch/Config.in.arm | 12 +++---------
 1 file changed, 3 insertions(+), 9 deletions(-)
arch/Config.in.arm +3 -9
index b1817ed568..3cafe21700 100644
@@ -534,15 +534,9 @@ config BR2_GCC_TARGET_CPU
 	default "strongarm"	if BR2_strongarm
 	default "xscale"	if BR2_xscale
 	default "iwmmxt"	if BR2_iwmmxt
-	default "cortex-a53"		if (BR2_cortex_a53 && !BR2_ARCH_IS_64)
-	default "cortex-a53+fp"		if (BR2_cortex_a53 && BR2_ARCH_IS_64 && BR2_ARM_FPU_FP_ARMV8)
-	default "cortex-a53+fp+simd"	if (BR2_cortex_a53 && BR2_ARCH_IS_64 && BR2_ARM_FPU_NEON_FP_ARMV8)
-	default "cortex-a57"		if (BR2_cortex_a57 && !BR2_ARCH_IS_64)
-	default "cortex-a57+fp"		if (BR2_cortex_a57 && BR2_ARCH_IS_64 && BR2_ARM_FPU_FP_ARMV8)
-	default "cortex-a57+fp+simd"	if (BR2_cortex_a57 && BR2_ARCH_IS_64 && BR2_ARM_FPU_NEON_FP_ARMV8)
-	default "cortex-a72"		if (BR2_cortex_a72 && !BR2_ARCH_IS_64)
-	default "cortex-a72+fp"		if (BR2_cortex_a72 && BR2_ARCH_IS_64 && BR2_ARM_FPU_FP_ARMV8)
-	default "cortex-a72+fp+simd"	if (BR2_cortex_a72 && BR2_ARCH_IS_64 && BR2_ARM_FPU_NEON_FP_ARMV8)
+	default "cortex-a53"	if BR2_cortex_a53
+	default "cortex-a57"	if BR2_cortex_a57
+	default "cortex-a72"	if BR2_cortex_a72
 
 config BR2_GCC_TARGET_ABI
 	default "aapcs-linux"	if BR2_arm || BR2_armeb
Signature · 6 lines
Arnout Vandecappelle                          arnout at mind be
Senior Embedded Software Architect            +32-16-286500
Essensium/Mind                                http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint:  7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
Signature · 6 lines
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'
#4 lore raw

PATCH v2 1/4 arch/arm: fix -mcpu default values for AArch64

In reply to Yann E. MORIN
Quoted 9 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

"Yann" == Yann E MORIN <[email protected]> writes:

We have to specify the -mcpu value, even in 64-bit mode.
For AArch64, +fp and +simd are the default, so they are totally useless.

Signed-off-by "Yann E. MORIN" <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Baruch Siach <[email protected]>
Quoted 5 lines

---
Changes v1 -> v2:
- squash the two changes into one patch (Thomas)
- remove useless parentheses (Baruch)
- remove superfluous indentation

Committed to 2017.02.x and 2017.05.x, thanks.

Signature · 1 line
Bye, Peter Korsgaard
author
#5 lore raw

PATCH v2 2/4 package/kvm-unit-tests: introduce _ARCH_SUPPORTS

In reply to Yann E. MORIN

Move all architecture options to their own symbol, so that it is easier to add more variants in the future.

Signed-off-by "Yann E. MORIN" <[email protected]>
Cc Cyril Bur <[email protected]>
---
 package/kvm-unit-tests/Config.in | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)
package/kvm-unit-tests/Config.in +10 -8
index 7eab0c25d6..0283e97dfc 100644
@@ -1,15 +1,17 @@
-config BR2_PACKAGE_KVM_UNIT_TESTS
-	bool "kvm-unit-tests"
+config BR2_PACKAGE_KVM_UNIT_TESTS_ARCH_SUPPORTS
+	bool
+	default y if BR2_cortex_a7 || BR2_cortex_a12 || \
+		BR2_cortex_a15 || BR2_cortex_a17
+	default y if BR2_powerpc64 || BR2_powerpc64le
 	# on i386 and x86-64, __builtin_reachable is used, so we need
 	# gcc 4.5 at least. on i386, we use the target gcc, while on
 	# x86-64 we use the host gcc (see .mk file for details)
 	# On ARM, it uses virtualization extensions
-	depends on BR2_cortex_a7 || BR2_cortex_a12 || \
-		BR2_cortex_a15 || BR2_cortex_a17 || \
-		(BR2_i386 && BR2_TOOLCHAIN_GCC_AT_LEAST_4_5) || \
-		BR2_powerpc64 || \
-		BR2_powerpc64le || \
-		(BR2_x86_64 && BR2_HOST_GCC_AT_LEAST_4_5)
+	default y if (BR2_i386 || BR2_x86_64) && BR2_TOOLCHAIN_GCC_AT_LEAST_4_5
+
+config BR2_PACKAGE_KVM_UNIT_TESTS
+	bool "kvm-unit-tests"
+	depends on BR2_PACKAGE_KVM_UNIT_TESTS_ARCH_SUPPORTS
 	select BR2_HOSTARCH_NEEDS_IA32_COMPILER if BR2_x86_64=y
 	help
 	  kvm-unit-tests is a project as old as KVM. As its name
Signature · 1 line
2.11.0
#6 lore raw

PATCH v2 2/4 package/kvm-unit-tests: introduce _ARCH_SUPPORTS

Quoted 22 lines · Yann E. MORIN

Move all architecture options to their own symbol, so that it is easier to add more variants in the future.

Signed-off-by "Yann E. MORIN" <[email protected]>
Cc Cyril Bur <[email protected]>
---
 package/kvm-unit-tests/Config.in | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)
package/kvm-unit-tests/Config.in +5 -2
index 7eab0c25d6..0283e97dfc 100644
@@ -1,15 +1,17 @@
-config BR2_PACKAGE_KVM_UNIT_TESTS
-	bool "kvm-unit-tests"
+config BR2_PACKAGE_KVM_UNIT_TESTS_ARCH_SUPPORTS
+	bool
+	default y if BR2_cortex_a7 || BR2_cortex_a12 || \
+		BR2_cortex_a15 || BR2_cortex_a17
+	default y if BR2_powerpc64 || BR2_powerpc64le
 	# on i386 and x86-64, __builtin_reachable is used, so we need

package/kvm-unit-tests/Config.in line 10

Quoted 10 lines
package/kvm-unit-tests/Config.in +1 -6
 	# gcc 4.5 at least. on i386, we use the target gcc, while on
 	# x86-64 we use the host gcc (see .mk file for details)
 	# On ARM, it uses virtualization extensions
-	depends on BR2_cortex_a7 || BR2_cortex_a12 || \
-		BR2_cortex_a15 || BR2_cortex_a17 || \
-		(BR2_i386 && BR2_TOOLCHAIN_GCC_AT_LEAST_4_5) || \
-		BR2_powerpc64 || \
-		BR2_powerpc64le || \
-		(BR2_x86_64 && BR2_HOST_GCC_AT_LEAST_4_5)
+	default y if (BR2_i386 || BR2_x86_64) && BR2_TOOLCHAIN_GCC_AT_LEAST_4_5

GCC is *not* an arch dependency, so this should just be

	default y if BR2_i386
	default y if BR2_powerpc64 || BR2_powerpc64le
	default y if BR2_x86_64

package/kvm-unit-tests/Config.in line 17

Quoted 7 lines
package/kvm-unit-tests/Config.in +4 -0
+
+config BR2_PACKAGE_KVM_UNIT_TESTS
+	bool "kvm-unit-tests"
+	depends on BR2_PACKAGE_KVM_UNIT_TESTS_ARCH_SUPPORTS
 	select BR2_HOSTARCH_NEEDS_IA32_COMPILER if BR2_x86_64=y
 	help
 	  kvm-unit-tests is a project as old as KVM. As its name

And here

	depends on !((BR2_i386 || BR2_x86_64) || BR2_TOOLCHAIN_GCC_AT_LEAST_4_5

with the big comment above it.

And also the comment is missing at the moment - which is probably why you
didn't notice that it's not an architecture dependency.

Regards,
Arnout

Signature · 6 lines
Arnout Vandecappelle                          arnout at mind be
Senior Embedded Software Architect            +32-16-286500
Essensium/Mind                                http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint:  7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
author
#7 lore raw

PATCH v2 2/4 package/kvm-unit-tests: introduce _ARCH_SUPPORTS

Arnout, All,

On 2017-07-08 17:41 +0200, Arnout Vandecappelle spake thusly:

Quoted 27 lines · Yann E. MORIN
Quoted 26 lines

Move all architecture options to their own symbol, so that it is easier to add more variants in the future.

Signed-off-by "Yann E. MORIN" <[email protected]>
Cc Cyril Bur <[email protected]>
---
 package/kvm-unit-tests/Config.in | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)
package/kvm-unit-tests/Config.in +5 -3
index 7eab0c25d6..0283e97dfc 100644
@@ -1,15 +1,17 @@
-config BR2_PACKAGE_KVM_UNIT_TESTS
-	bool "kvm-unit-tests"
+config BR2_PACKAGE_KVM_UNIT_TESTS_ARCH_SUPPORTS
+	bool
+	default y if BR2_cortex_a7 || BR2_cortex_a12 || \
+		BR2_cortex_a15 || BR2_cortex_a17
+	default y if BR2_powerpc64 || BR2_powerpc64le
 	# on i386 and x86-64, __builtin_reachable is used, so we need
 	# gcc 4.5 at least. on i386, we use the target gcc, while on
 	# x86-64 we use the host gcc (see .mk file for details)
 	# On ARM, it uses virtualization extensions
-	depends on BR2_cortex_a7 || BR2_cortex_a12 || \
Quoted 8 lines

package/kvm-unit-tests/Config.in line 10

Quoted 6 lines
package/kvm-unit-tests/Config.in +1 -5
-		BR2_cortex_a15 || BR2_cortex_a17 || \
-		(BR2_i386 && BR2_TOOLCHAIN_GCC_AT_LEAST_4_5) || \
-		BR2_powerpc64 || \
-		BR2_powerpc64le || \
-		(BR2_x86_64 && BR2_HOST_GCC_AT_LEAST_4_5)
+	default y if (BR2_i386 || BR2_x86_64) && BR2_TOOLCHAIN_GCC_AT_LEAST_4_5

GCC is *not* an arch dependency, so this should just be

Yet we already use similar construct in other packages, like host-go:

    config BR2_PACKAGE_HOST_GO_ARCH_SUPPORTS
        depends on (BR2_arm && BR2_TOOLCHAIN_SUPPORTS_PIE) || [...]

or like kodi:

     config BR2_PACKAGE_KODI_ARCH_SUPPORTS
        default y if (BR2_arm || (BR2_mipsel && BR2_TOOLCHAIN_USES_GLIBC) || [...]

or with libsigsegv:

config BR2_PACKAGE_LIBSIGSEGV_ARCH_SUPPORTS
default y if BR2_TOOLCHAIN_USES_GLIBC
default y if BR2_TOOLCHAIN_USES_MUSL
default y if BR2_TOOLCHAIN_USES_UCLIBC && [...archs...]

or libunwind, openal (interesting case), openblas, protobuf.

Arguaby, none of the above have a dependency on gcc, but some are on the C library, and others are on toolchain features...

Quoted 14 lines
	default y if BR2_i386
	default y if BR2_powerpc64 || BR2_powerpc64le
	default y if BR2_x86_64

package/kvm-unit-tests/Config.in line 17

Quoted 7 lines
package/kvm-unit-tests/Config.in +4 -0
+
+config BR2_PACKAGE_KVM_UNIT_TESTS
+	bool "kvm-unit-tests"
+	depends on BR2_PACKAGE_KVM_UNIT_TESTS_ARCH_SUPPORTS
 	select BR2_HOSTARCH_NEEDS_IA32_COMPILER if BR2_x86_64=y
 	help
 	  kvm-unit-tests is a project as old as KVM. As its name

And here

Quoted 6 lines
	depends on !((BR2_i386 || BR2_x86_64) || BR2_TOOLCHAIN_GCC_AT_LEAST_4_5

with the big comment above it.

And also the comment is missing at the moment - which is probably why you
didn't notice that it's not an architecture dependency.

OK, will add.

Thanks!

Regards,
Yann E. MORIN.

Quoted 11 lines

Regards,
Arnout

Signature · 6 lines
Arnout Vandecappelle                          arnout at mind be
Senior Embedded Software Architect            +32-16-286500
Essensium/Mind                                http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint:  7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
Signature · 6 lines
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'
#8 lore raw

PATCH v2 2/4 package/kvm-unit-tests: introduce _ARCH_SUPPORTS

Hello,

Quoted 19 lines · Yann E. MORIN
Quoted 1 line

GCC is *not* an arch dependency, so this should just be

Yet we already use similar construct in other packages, like host-go:

    config BR2_PACKAGE_HOST_GO_ARCH_SUPPORTS
        depends on (BR2_arm && BR2_TOOLCHAIN_SUPPORTS_PIE) || [...]

or like kodi:

     config BR2_PACKAGE_KODI_ARCH_SUPPORTS
        default y if (BR2_arm || (BR2_mipsel && BR2_TOOLCHAIN_USES_GLIBC) || [...]

or with libsigsegv:

config BR2_PACKAGE_LIBSIGSEGV_ARCH_SUPPORTS
default y if BR2_TOOLCHAIN_USES_GLIBC
default y if BR2_TOOLCHAIN_USES_MUSL
default y if BR2_TOOLCHAIN_USES_UCLIBC && [...archs...]

Quoted 4 lines

or libunwind, openal (interesting case), openblas, protobuf.

Arguaby, none of the above have a dependency on gcc, but some are on the C library, and others are on toolchain features...

Yes, but see what I replied to Carlos about adding a BR2_PACKAGE_LINUX_PAM_ARCH_SUPPORTS option, which would have encoded non-architecture dependencies.

Thomas

Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com
author
#9 lore raw

PATCH v2 2/4 package/kvm-unit-tests: introduce _ARCH_SUPPORTS

Thomas, All,

On 2017-07-09 18:03 +0200, Thomas Petazzoni spake thusly:

Quoted 3 lines · Yann E. MORIN
Quoted 2 lines
Quoted 1 line

GCC is *not* an arch dependency, so this should just be

Yet we already use similar construct in other packages, like [...]

[--SNIP--]

Quoted 6 lines
Quoted 2 lines

Arguaby, none of the above have a dependency on gcc, but some are on the C library, and others are on toolchain features...

Yes, but see what I replied to Carlos about adding a BR2_PACKAGE_LINUX_PAM_ARCH_SUPPORTS option, which would have encoded non-architecture dependencies.

And it looks like I agree as I addressed Arnout's comment in v3. ;-)

I was just pointing out that we are not entirely consistent with this rule, as some packages have such non-arch dependencies in their _ARCH_SUPPORT symbols.

I'll try to find some time to fix them. :-)

Regards,
Yann E. MORIN.

Signature · 6 lines
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'
author
#10 lore raw

PATCH v2 3/4 arch/arm: add big.LITTLE cpu variants

In reply to Yann E. MORIN

The big.LITTLE configurations can be optimised for by gcc, and a few users wonder what they should choose when they have such CPUs.

Add new entries for those big.LITTLE configurations.

Signed-off-by "Yann E. MORIN" <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Thomas De Schampheleire <[email protected]>
Cc Baruch Siach <[email protected]>

---
Changes v1 -> v2:
- remove superfluous parentheses (Baruch)

---
 arch/Config.in.arm                                 | 38 ++++++++++++++++++++++
 package/gcc/Config.in.host                         |  5 ++-
 .../Config.in                                      |  2 ++
 .../toolchain-external-codesourcery-arm/Config.in  |  4 +++
 4 files changed, 48 insertions(+), 1 deletion(-)
arch/Config.in.arm +38 -0
index 3cafe21700..09916df7ad 100644
@@ -182,6 +182,15 @@ config BR2_cortex_a15
 	select BR2_ARM_CPU_ARMV7A
 	select BR2_ARCH_HAS_MMU_OPTIONAL
 	depends on !BR2_ARCH_IS_64
+config BR2_cortex_a15_a7
+	bool "cortex-A15/A7 big.LITTLE"
+	select BR2_ARM_CPU_HAS_ARM
+	select BR2_ARM_CPU_HAS_NEON
+	select BR2_ARM_CPU_HAS_VFPV4
+	select BR2_ARM_CPU_HAS_THUMB2
+	select BR2_ARM_CPU_ARMV7A
+	select BR2_ARCH_HAS_MMU_OPTIONAL
+	depends on !BR2_ARCH_IS_64
 config BR2_cortex_a17
 	bool "cortex-A17"
 	select BR2_ARM_CPU_HAS_ARM

@@ -191,6 +200,15 @@ config BR2_cortex_a17
 	select BR2_ARM_CPU_ARMV7A
 	select BR2_ARCH_HAS_MMU_OPTIONAL
 	depends on !BR2_ARCH_IS_64
+config BR2_cortex_a17_a7
+	bool "cortex-A17/A7 big.LITTLE"
+	select BR2_ARM_CPU_HAS_ARM
+	select BR2_ARM_CPU_HAS_NEON
+	select BR2_ARM_CPU_HAS_VFPV4
+	select BR2_ARM_CPU_HAS_THUMB2
+	select BR2_ARM_CPU_ARMV7A
+	select BR2_ARCH_HAS_MMU_OPTIONAL
+	depends on !BR2_ARCH_IS_64
 config BR2_cortex_a53
 	bool "cortex-A53"
 	select BR2_ARM_CPU_HAS_ARM if !BR2_ARCH_IS_64

@@ -207,6 +225,14 @@ config BR2_cortex_a57
 	select BR2_ARM_CPU_HAS_FP_ARMV8
 	select BR2_ARM_CPU_ARMV8
 	select BR2_ARCH_HAS_MMU_OPTIONAL
+config BR2_cortex_a57_a53
+	bool "cortex-A57/A53 big.LITTLE"
+	select BR2_ARM_CPU_HAS_ARM if !BR2_ARCH_IS_64
+	select BR2_ARM_CPU_HAS_NEON if !BR2_ARCH_IS_64
+	select BR2_ARM_CPU_HAS_THUMB2 if !BR2_ARCH_IS_64
+	select BR2_ARM_CPU_HAS_FP_ARMV8
+	select BR2_ARM_CPU_ARMV8
+	select BR2_ARCH_HAS_MMU_OPTIONAL
 config BR2_cortex_a72
 	bool "cortex-A72"
 	select BR2_ARM_CPU_HAS_ARM if !BR2_ARCH_IS_64

@@ -215,6 +241,14 @@ config BR2_cortex_a72
 	select BR2_ARM_CPU_HAS_FP_ARMV8
 	select BR2_ARM_CPU_ARMV8
 	select BR2_ARCH_HAS_MMU_OPTIONAL
+config BR2_cortex_a72_a53
+	bool "cortex-A72/A53 big.LITTLE"
+	select BR2_ARM_CPU_HAS_ARM if !BR2_ARCH_IS_64
+	select BR2_ARM_CPU_HAS_NEON if !BR2_ARCH_IS_64
+	select BR2_ARM_CPU_HAS_THUMB2 if !BR2_ARCH_IS_64
+	select BR2_ARM_CPU_HAS_FP_ARMV8
+	select BR2_ARM_CPU_ARMV8
+	select BR2_ARCH_HAS_MMU_OPTIONAL
 config BR2_cortex_m3
 	bool "cortex-M3"
 	select BR2_ARM_CPU_HAS_THUMB2

@@ -526,7 +560,9 @@ config BR2_GCC_TARGET_CPU
 	default "cortex-a9"	if BR2_cortex_a9
 	default "cortex-a12"	if BR2_cortex_a12
 	default "cortex-a15"	if BR2_cortex_a15
+	default "cortex-a15.cortex-a7"	if BR2_cortex_a15_a7
 	default "cortex-a17"	if BR2_cortex_a17
+	default "cortex-a17.cortex-a7"	if BR2_cortex_a17_a7
 	default "cortex-m3"	if BR2_cortex_m3
 	default "cortex-m4"	if BR2_cortex_m4
 	default "fa526"		if BR2_fa526

@@ -536,7 +572,9 @@ config BR2_GCC_TARGET_CPU
 	default "iwmmxt"	if BR2_iwmmxt
 	default "cortex-a53"	if BR2_cortex_a53
 	default "cortex-a57"	if BR2_cortex_a57
+	default "cortex-a57.cortex-a53"	if BR2_cortex_a57_a53
 	default "cortex-a72"	if BR2_cortex_a72
+	default "cortex-a72.cortex-a53"	if BR2_cortex_a72_a53
 
 config BR2_GCC_TARGET_ABI
 	default "aapcs-linux"	if BR2_arm || BR2_armeb
package/gcc/Config.in.host +4 -1
index 5dcaa03ff0..c072b78ae5 100644
@@ -25,7 +25,8 @@ config BR2_GCC_VERSION_4_9_X
 	# Broken or unsupported architectures
 	depends on !BR2_arc && !BR2_bfin && !BR2_or1k
 	# Broken or unsupported ARM cores
-	depends on !BR2_cortex_a17 && !BR2_cortex_a72
+	depends on !BR2_cortex_a17 && !BR2_cortex_a17_a7
+	depends on !BR2_cortex_a72 && !BR2_cortex_a72_a53
 	# Unsupported MIPS cores
 	depends on !BR2_mips_interaptiv
 	# Unsupported for MIPS R5

@@ -45,6 +46,8 @@ config BR2_GCC_VERSION_5_X
 	bool "gcc 5.x"
 	# Broken or unsupported architectures
 	depends on !BR2_arc && !BR2_bfin && !BR2_or1k
+	# Broken or unsupported ARM cores
+	depends on !BR2_cortex_a57_a53 && !BR2_cortex_a72_a53
 	# musl ppc64 unsupported
 	depends on !(BR2_TOOLCHAIN_USES_MUSL && (BR2_powerpc64 || BR2_powerpc64le))
 	# Unsupported MIPS cores
toolchain/toolchain-external/toolchain-external-codesourcery-aarch64/Config.in +2 -0
index 2fbb218ecc..66a032e9ac 100644
@@ -1,6 +1,8 @@
 config BR2_TOOLCHAIN_EXTERNAL_CODESOURCERY_AARCH64
 	bool "CodeSourcery AArch64 2014.11"
 	depends on BR2_aarch64
+	# a57/a53 and a72/a53 appeared in gcc-6 or were broken before
+	depends on !BR2_cortex_a57_a53 && !BR2_cortex_a72_a53
 	depends on BR2_HOSTARCH = "x86_64" || BR2_HOSTARCH = "x86"
 	depends on !BR2_STATIC_LIBS
 	select BR2_TOOLCHAIN_EXTERNAL_GLIBC
toolchain/toolchain-external/toolchain-external-codesourcery-arm/Config.in +4 -0
index 6331873dbd..e1a7891007 100644
@@ -1,6 +1,10 @@
 config BR2_TOOLCHAIN_EXTERNAL_CODESOURCERY_ARM
 	bool "Sourcery CodeBench ARM 2014.05"
 	depends on BR2_arm
+	# a15/a7 appeared in gcc-4.9, a17/a7 in gcc-5, a57/a53 and a72/a53
+	# in gcc-6, or they each were broken earlier than that.
+	depends on !BR2_cortex_a15_a7 && !BR2_cortex_a17_a7
+	depends on !BR2_cortex_a57_53 && !BR2_cortex_a72_53
 	depends on BR2_HOSTARCH = "x86_64" || BR2_HOSTARCH = "x86"
 	depends on BR2_ARM_EABI
 	# Unsupported ARM cores
Signature · 1 line
2.11.0
#11 lore raw

PATCH v2 3/4 arch/arm: add big.LITTLE cpu variants

Quoted 9 lines · Yann E. MORIN

The big.LITTLE configurations can be optimised for by gcc, and a few users wonder what they should choose when they have such CPUs.

Add new entries for those big.LITTLE configurations.

Signed-off-by "Yann E. MORIN" <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Thomas De Schampheleire <[email protected]>
Cc Baruch Siach <[email protected]>

[snip]

arch/Config.in.arm config BR2_cortex_a15 · line 192

Quoted 8 lines
arch/Config.in.arm +8 -0
+config BR2_cortex_a15_a7
+	bool "cortex-A15/A7 big.LITTLE"
+	select BR2_ARM_CPU_HAS_ARM
+	select BR2_ARM_CPU_HAS_NEON
+	select BR2_ARM_CPU_HAS_VFPV4
+	select BR2_ARM_CPU_HAS_THUMB2
+	select BR2_ARM_CPU_ARMV7A
+	select BR2_ARCH_HAS_MMU_OPTIONAL

Not for this patch, but: is this actually correct? Looking at the kernel's
arch/arm/Kconfig, I see that all the platforms that can be selected when MMU is not selected are <= ARMv6. Well, I'm not entirely sure, it's a bit difficult to map stuff like CPU_XSC3 on an ARM version.

Quoted 1 line
(quoted diff) +1 -0
+	depends on !BR2_ARCH_IS_64

[snip]

Quoted 15 lines
package/gcc/Config.in.host +2 -1
index 5dcaa03ff0..c072b78ae5 100644
@@ -25,7 +25,8 @@ config BR2_GCC_VERSION_4_9_X
 	# Broken or unsupported architectures
 	depends on !BR2_arc && !BR2_bfin && !BR2_or1k
 	# Broken or unsupported ARM cores
-	depends on !BR2_cortex_a17 && !BR2_cortex_a72
+	depends on !BR2_cortex_a17 && !BR2_cortex_a17_a7
+	depends on !BR2_cortex_a72 && !BR2_cortex_a72_a53
 	# Unsupported MIPS cores
 	depends on !BR2_mips_interaptiv
 	# Unsupported for MIPS R5

package/gcc/Config.in.host config BR2_GCC_VERSION_5_X · line 50

Quoted 5 lines
package/gcc/Config.in.host +2 -0
@@ -45,6 +46,8 @@ config BR2_GCC_VERSION_5_X
 	bool "gcc 5.x"
 	# Broken or unsupported architectures
 	depends on !BR2_arc && !BR2_bfin && !BR2_or1k
+	# Broken or unsupported ARM cores
+	depends on !BR2_cortex_a57_a53 && !BR2_cortex_a72_a53

According to [1], a72.a53 was added in GCC 5. And a57.a53 was already added in
GCC 4.9 - is it broken in GCC 5?

toolchain/toolchain-external/toolchain-external-codesourcery-aarch64/Config.in line 4

Quoted 12 lines
package/gcc/Config.in.host +0 -0
 	# musl ppc64 unsupported
 	depends on !(BR2_TOOLCHAIN_USES_MUSL && (BR2_powerpc64 || BR2_powerpc64le))
 	# Unsupported MIPS cores
toolchain/toolchain-external/toolchain-external-codesourcery-aarch64/Config.in +1 -0
index 2fbb218ecc..66a032e9ac 100644
@@ -1,6 +1,8 @@
 config BR2_TOOLCHAIN_EXTERNAL_CODESOURCERY_AARCH64
 	bool "CodeSourcery AArch64 2014.11"
 	depends on BR2_aarch64
+	# a57/a53 and a72/a53 appeared in gcc-6 or were broken before

Ah, here you mention it. Better raise that to the commit message. Do you also
have a reference for it?

Regards,
Arnout

toolchain/toolchain-external/toolchain-external-codesourcery-aarch64/Config.in line 5

Quoted 1 line
toolchain/toolchain-external/toolchain-external-codesourcery-aarch64/Config.in +1 -0
+	depends on !BR2_cortex_a57_a53 && !BR2_cortex_a72_a53

[snip]

[1] https://gcc.gnu.org/gcc-5/changes.html

Signature · 6 lines
Arnout Vandecappelle                          arnout at mind be
Senior Embedded Software Architect            +32-16-286500
Essensium/Mind                                http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint:  7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
author
#12 lore raw

PATCH v2 3/4 arch/arm: add big.LITTLE cpu variants

Arnout, All,

On 2017-07-08 18:34 +0200, Arnout Vandecappelle spake thusly:

Quoted 11 lines · Yann E. MORIN
Quoted 9 lines

The big.LITTLE configurations can be optimised for by gcc, and a few users wonder what they should choose when they have such CPUs.

Add new entries for those big.LITTLE configurations.

Signed-off-by "Yann E. MORIN" <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Thomas De Schampheleire <[email protected]>
Cc Baruch Siach <[email protected]>

[snip]

Quoted 13 lines

arch/Config.in.arm config BR2_cortex_a15 · line 192

Quoted 8 lines
arch/Config.in.arm +8 -0
+config BR2_cortex_a15_a7
+	bool "cortex-A15/A7 big.LITTLE"
+	select BR2_ARM_CPU_HAS_ARM
+	select BR2_ARM_CPU_HAS_NEON
+	select BR2_ARM_CPU_HAS_VFPV4
+	select BR2_ARM_CPU_HAS_THUMB2
+	select BR2_ARM_CPU_ARMV7A
+	select BR2_ARCH_HAS_MMU_OPTIONAL

Not for this patch, but: is this actually correct? Looking at the kernel's
arch/arm/Kconfig, I see that all the platforms that can be selected when MMU is not selected are <= ARMv6. Well, I'm not entirely sure, it's a bit difficult to map stuff like CPU_XSC3 on an ARM version.

I just duplicated for a15-a7 what we had for a15 and a7, which fortunately was exactly identical. So a15-a7 can be noMMU in Buildroot.

Does that make sense? Probably not, but this should be fixed by another patch, then.

Quoted 17 lines
Quoted 1 line
(quoted diff) +1 -0
+	depends on !BR2_ARCH_IS_64

[snip]

Quoted 15 lines
package/gcc/Config.in.host +2 -1
index 5dcaa03ff0..c072b78ae5 100644
@@ -25,7 +25,8 @@ config BR2_GCC_VERSION_4_9_X
 	# Broken or unsupported architectures
 	depends on !BR2_arc && !BR2_bfin && !BR2_or1k
 	# Broken or unsupported ARM cores
-	depends on !BR2_cortex_a17 && !BR2_cortex_a72
+	depends on !BR2_cortex_a17 && !BR2_cortex_a17_a7
+	depends on !BR2_cortex_a72 && !BR2_cortex_a72_a53
 	# Unsupported MIPS cores
 	depends on !BR2_mips_interaptiv
 	# Unsupported for MIPS R5

@@ -45,6 +46,8 @@ config BR2_GCC_VERSION_5_X
Quoted 8 lines

package/gcc/Config.in.host config BR2_GCC_VERSION_5_X · line 50

Quoted 5 lines
package/gcc/Config.in.host +2 -0
 	bool "gcc 5.x"
 	# Broken or unsupported architectures
 	depends on !BR2_arc && !BR2_bfin && !BR2_or1k
+	# Broken or unsupported ARM cores
+	depends on !BR2_cortex_a57_a53 && !BR2_cortex_a72_a53

According to [1], a72.a53 was added in GCC 5. And a57.a53 was already added in
GCC 4.9 - is it broken in GCC 5?

I indeed tried various versions of gcc, starting with gcc-4.9.

The depednencies I added here are based on brokenness of versions earlier than that.

Quoted 7 lines
Quoted 7 lines
package/gcc/Config.in.host +0 -0
 	# musl ppc64 unsupported
 	depends on !(BR2_TOOLCHAIN_USES_MUSL && (BR2_powerpc64 || BR2_powerpc64le))
 	# Unsupported MIPS cores
toolchain/toolchain-external/toolchain-external-codesourcery-aarch64/Config.in +0 -0
index 2fbb218ecc..66a032e9ac 100644
Quoted 8 lines

toolchain/toolchain-external/toolchain-external-codesourcery-aarch64/Config.in line 4

Quoted 5 lines
toolchain/toolchain-external/toolchain-external-codesourcery-aarch64/Config.in +1 -0
@@ -1,6 +1,8 @@
 config BR2_TOOLCHAIN_EXTERNAL_CODESOURCERY_AARCH64
 	bool "CodeSourcery AArch64 2014.11"
 	depends on BR2_aarch64
+	# a57/a53 and a72/a53 appeared in gcc-6 or were broken before

Ah, here you mention it. Better raise that to the commit message. Do you also
have a reference for it?

No, except that gcc did not build: the brokenness manifested itself with gcc's configure script refusing the big-LITTLE variants.

Regards,
Yann E. MORIN.

Quoted 16 lines

Regards,
Arnout

toolchain/toolchain-external/toolchain-external-codesourcery-aarch64/Config.in line 5

Quoted 1 line
toolchain/toolchain-external/toolchain-external-codesourcery-aarch64/Config.in +1 -0
+	depends on !BR2_cortex_a57_a53 && !BR2_cortex_a72_a53

[snip]

[1] https://gcc.gnu.org/gcc-5/changes.html

Signature · 6 lines
Arnout Vandecappelle                          arnout at mind be
Senior Embedded Software Architect            +32-16-286500
Essensium/Mind                                http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint:  7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
Signature · 6 lines
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'
author
#13 lore raw

PATCH v2 4/4 package/kvm-unit-test: available for big.LITTLE arm cores too

In reply to Yann E. MORIN

Since it supports each variants individually, it should also supports the big.LITTLE configurations.

Signed-off-by "Yann E. MORIN" <[email protected]>
Cc Cyril Bur <[email protected]>
Cc Thomas Petazzoni <[email protected]>
---
 package/kvm-unit-tests/Config.in | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
package/kvm-unit-tests/Config.in +2 -1
index 0283e97dfc..ec1826b2da 100644
@@ -1,7 +1,8 @@
 config BR2_PACKAGE_KVM_UNIT_TESTS_ARCH_SUPPORTS
 	bool
 	default y if BR2_cortex_a7 || BR2_cortex_a12 || \
-		BR2_cortex_a15 || BR2_cortex_a17
+		BR2_cortex_a15 || BR2_cortex_a17 || \
+		BR2_cortex_a15_a7 || BR2_cortex_a17_a7
 	default y if BR2_powerpc64 || BR2_powerpc64le
 	# on i386 and x86-64, __builtin_reachable is used, so we need
 	# gcc 4.5 at least. on i386, we use the target gcc, while on
Signature · 1 line
2.11.0