Skip to contentSkip to conversation

arch-armv4: Allow -march=armv4

PATCH In review Khem Raj 2 participants 3 messages Jun 10, 2021 – Jun 17, 2021

lore
Share & tools
lore

Khem Raj

author
#0 lore raw

PATCH 1/1 arch-armv4: Allow -march=armv4

Even though it is deprecated in GCC 6 [1] it has not yet been removed from gcc upstream. We do have active machines in OE ecosystem which use armv4 ( SA11xx ) e.g. collie in meta-handheld so until upstream gcc takes next step to remove them
lets support armv4 again, we are still carrying the relevant gcc patch to support v4 BX fix.

[1] https://gcc.gnu.org/gcc-6/changes.html#arm

Cc Andrea Adami <[email protected]>
---
 meta/conf/machine/include/arm/arch-armv4.inc | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
meta/conf/machine/include/arm/arch-armv4.inc +1 -1
index fac2bdf952..b71739c20a 100644
@@ -2,7 +2,7 @@ DEFAULTTUNE ?= "armv4"
 
 TUNEVALID[arm] = "Enable ARM instruction set"
 TUNEVALID[armv4] = "Enable instructions for ARMv4"
-TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'armv4', ' -march=armv4t', '', d)}"
+TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'armv4', ' -march=armv4${ARMPKGSFX_THUMB}', '', d)}"
 # enable --fix-v4bx when we have armv4 in TUNE_FEATURES, but then disable it when we have also armv5 or thumb
 # maybe we should extend bb.utils.contains to support check for any checkvalues in value, now it does 
 # checkvalues.issubset(val) which cannot be used for negative test of foo neither bar in value
Signature · 1 line
2.32.0
#1 lore raw

Re: [OE-core] [PATCH] arch-armv4: Allow -march=armv4 --> remove qemuarmv5.conf ?

Quoted 8 lines · Khem Raj

Even though it is deprecated in GCC 6 [1] it has not yet been removed from gcc upstream. We do have active machines in OE ecosystem which use armv4 ( SA11xx ) e.g. collie in meta-handheld so until upstream gcc takes next step to remove them
lets support armv4 again, we are still carrying the relevant gcc patch to support v4 BX fix.

[1] https://gcc.gnu.org/gcc-6/changes.html#arm

Huh, I was going to send an email about removing support for qemuarmv5. Do we follow gcc or drop older arches before that?

Our current list of qemu machines on master is:

meta/conf/machine/qemuarm64.conf
meta/conf/machine/qemuarm.conf
meta/conf/machine/qemuarmv5.conf
meta/conf/machine/qemumips64.conf
meta/conf/machine/qemumips.conf
meta/conf/machine/qemuppc64.conf
meta/conf/machine/qemuppc.conf
meta/conf/machine/qemuriscv32.conf
meta/conf/machine/qemuriscv64.conf
meta/conf/machine/qemux86-64.conf
meta/conf/machine/qemux86.conf

qemuarmv5 was added in:

commit 6fc70eb4f3494bee2be10ee24fe3ea1c8b5ff988

Author: Jon Mason <[email protected]>
Date:   Tue Mar 5 17:32:19 2019

qemuarm: Swap for an arm7ve (A15) configuration

Add new QEMU BSP for a Arm Cortex-A15 system and use this as qemuarm, moving the old armv5te Versatile PB based machine to qemuarmv5.

The new machine uses the QEMU virt machine type, which should be faster to emulate and updates the qemuarm support to a modern architecture.

../Randy

Quoted 25 lines
Cc Andrea Adami <[email protected]>
---
  meta/conf/machine/include/arm/arch-armv4.inc | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
meta/conf/machine/include/arm/arch-armv4.inc +1 -1
index fac2bdf952..b71739c20a 100644
@@ -2,7 +2,7 @@ DEFAULTTUNE ?= "armv4"
 
  TUNEVALID[arm] = "Enable ARM instruction set"
  TUNEVALID[armv4] = "Enable instructions for ARMv4"
-TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'armv4', ' -march=armv4t', '', d)}"
+TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'armv4', ' -march=armv4${ARMPKGSFX_THUMB}', '', d)}"
  # enable --fix-v4bx when we have armv4 in TUNE_FEATURES, but then disable it when we have also armv5 or thumb
  # maybe we should extend bb.utils.contains to support check for any checkvalues in value, now it does
  # checkvalues.issubset(val) which cannot be used for negative test of foo neither bar in value
Signature · 2 lines
# Randy MacLeod
# Wind River Linux
author
#2 lore raw

Re: [OE-core] [PATCH] arch-armv4: Allow -march=armv4 --> remove qemuarmv5.conf ?

Quoted 8 lines · Randy MacLeod
Quoted 7 lines · Khem Raj

Even though it is deprecated in GCC 6 [1] it has not yet been removed from gcc upstream. We do have active machines in OE ecosystem which use armv4 ( SA11xx ) e.g. collie in meta-handheld so until upstream gcc takes next step to remove them
lets support armv4 again, we are still carrying the relevant gcc patch to support v4 BX fix.

Quoted 6 lines
Quoted 1 line

[1] https://gcc.gnu.org/gcc-6/changes.html#arm

Huh, I was going to send an email about removing support for qemuarmv5. Do we follow gcc or drop older arches before that?

We use armv5te for default tune for that
Machine which is still supported and so is armv4t
Armv4 is a special case where I kept it even after it’s deprecated since there are users in OE community who are actively using it with master so we will stretch it to the point where gcc removes it and then perhaps remove it so no immediate action is needed

As far as armv5 ( without t ) is concerned we can remove it

Quoted 73 lines

Our current list of qemu machines on master is:

meta/conf/machine/qemuarm64.conf
meta/conf/machine/qemuarm.conf
meta/conf/machine/qemuarmv5.conf
meta/conf/machine/qemumips64.conf
meta/conf/machine/qemumips.conf
meta/conf/machine/qemuppc64.conf
meta/conf/machine/qemuppc.conf
meta/conf/machine/qemuriscv32.conf
meta/conf/machine/qemuriscv64.conf
meta/conf/machine/qemux86-64.conf
meta/conf/machine/qemux86.conf

qemuarmv5 was added in:

commit 6fc70eb4f3494bee2be10ee24fe3ea1c8b5ff988

Author: Jon Mason <[email protected]>
Date:   Tue Mar 5 17:32:19 2019

qemuarm: Swap for an arm7ve (A15) configuration

Add new QEMU BSP for a Arm Cortex-A15 system and use this as qemuarm, moving the old armv5te Versatile PB based machine to qemuarmv5.

The new machine uses the QEMU virt machine type, which should be faster to emulate and updates the qemuarm support to a modern architecture.

../Randy

Review

Quoted 8 lines
Cc Andrea Adami <[email protected]>
---
  meta/conf/machine/include/arm/arch-armv4.inc | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
(quoted diff) +0 -0

b/meta/conf/machine/include/arm/arch-armv4.inc

meta/conf/machine/include/arm/arch-armv4.inc DEFAULTTUNE ?= "armv4" · line 5

Quoted 8 lines

index fac2bdf952..b71739c20a 100644

meta/conf/machine/include/arm/arch-armv4.inc +0 -1
@@ -2,7 +2,7 @@ DEFAULTTUNE ?= "armv4"
 
  TUNEVALID[arm] = "Enable ARM instruction set"
  TUNEVALID[armv4] = "Enable instructions for ARMv4"
-TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'armv4', '

-march=armv4t', '', d)}"

meta/conf/machine/include/arm/arch-armv4.inc DEFAULTTUNE ?= "armv4" · line 5

Quoted 1 line
meta/conf/machine/include/arm/arch-armv4.inc +1 -0
+TUNE_CCARGS .= "${@bb.utils.contains('TUNE_FEATURES', 'armv4', '

-march=armv4${ARMPKGSFX_THUMB}', '', d)}"

meta/conf/machine/include/arm/arch-armv4.inc DEFAULTTUNE ?= "armv4" · line 6

Quoted 1 line
meta/conf/machine/include/arm/arch-armv4.inc +0 -0
  # enable --fix-v4bx when we have armv4 in TUNE_FEATURES, but then

disable it when we have also armv5 or thumb

meta/conf/machine/include/arm/arch-armv4.inc DEFAULTTUNE ?= "armv4" · line 7

Quoted 1 line
meta/conf/machine/include/arm/arch-armv4.inc +0 -0
  # maybe we should extend bb.utils.contains to support check for any

checkvalues in value, now it does

meta/conf/machine/include/arm/arch-armv4.inc DEFAULTTUNE ?= "armv4" · line 8

Quoted 1 line
meta/conf/machine/include/arm/arch-armv4.inc +0 -0
  # checkvalues.issubset(val) which cannot be used for negative test of

foo neither bar in value

Quoted 5 lines

--
# Randy MacLeod
# Wind River Linux