armv7: how to enable NEON/VFP with Clang-5.0.1 compiler
Hello, I am trying to build direwolf software-TNC(modem) to benchmark my pcDuino's performance. With my small patch, I got working binary. But it does not contains VFP/NEON instruction with "-O3 -ffast-math -mfpu=neon" option. To use VFP/NEON instruction, what option do I specify to armv7 clang? (direwolf is at https://github.com/wb2osz/direwolf and patch for OpenBSD is http://ik1-342-31132.vs.sakura.ne.jp/~uaa/gomitext/2018/20180503/20180503.diff ) Regards, -- SASANO Takayoshi (JG1UAA)
Re: armv7: how to enable NEON/VFP with Clang-5.0.1 compiler
> Date: Thu, 03 May 2018 16:11:24 +0900 > From: SASANO Takayoshi > > Hello, > > I am trying to build direwolf software-TNC(modem) to benchmark my pcDuino's > performance. > > With my small patch, I got working binary. But it does not contains > VFP/NEON instruction with "-O3 -ffast-math -mfpu=neon" option. > > To use VFP/NEON instruction, what option do I specify to armv7 clang? You'll also need "-mfloat-abi=softfp". Note that this is soon going to be the default in -current. At some point in the future we'll switch the default to "-mfloat-abi=hardfp" which is inompatible with "-mfloat-abi=softfp". It is probably a bad idea to push changes that explicitly set -mfloat-abi upstream. > (direwolf is at https://github.com/wb2osz/direwolf and patch for OpenBSD is > http://ik1-342-31132.vs.sakura.ne.jp/~uaa/gomitext/2018/20180503/20180503.diff > )
Re: armv7: how to enable NEON/VFP with Clang-5.0.1 compiler
Hi,
>> To use VFP/NEON instruction, what option do I specify to armv7 clang?
>
> You'll also need "-mfloat-abi=softfp". Note that this is soon going
> to be the default in -current. At some point in the future we'll
> switch the default to "-mfloat-abi=hardfp" which is inompatible with
> "-mfloat-abi=softfp". It is probably a bad idea to push changes that
> explicitly set -mfloat-abi upstream.
Adding -mfloat-abi=softfp option solved the problem. Thanks.
I tried with that option before sending this question to ML
but the disassembled code of objdump-2.17 confused me.
1c44: ed900a2afldss0, [r0, #168]
1c48: e08f0001add r0, pc, r1
1c4c: ed901a02fldss2, [r0, #8]
1c50: eef80bc0cdp 11, 15, cr0, cr8, cr0, {6}
1c54: eef81bc1cdp 11, 15, cr1, cr8, cr1, {6}
1c58: eec10ba0cdp 11, 12, cr0, cr1, cr0, {5}
Same binary with objdump-2.27.
1c44: ed900a2avldrs0, [r0, #168] ; 0xa8
1c48: e08f0001add r0, pc, r1
1c4c: ed901a02vldrs2, [r0, #8]
1c50: eef80bc0vcvt.f64.s32d16, s0
1c54: eef81bc1vcvt.f64.s32d17, s2
1c58: eec10ba0vdiv.f64d16, d17, d16
I thought the first letter of NEON/VFP instruction is "v",
so I misunderstood that these instruction is not available...
Best regards,
--
SASANO Takayoshi (JG1UAA)
Re: gtk libool slow on arm
On Wed, May 02, 2018 at 05:36:33PM -0700, s_g...@telus.net wrote: > Thank you for the advice. I will try the patch, but I want to let the > current php build complete, which may take a few more days. > > Did you get the bme code I sent you? > Yep, I just forgot about it, and now i did "over do" a bit with the necessary cleanup/refactoring.. oops, but no more ugly unions nor "forbidden hacks":) diff below does still miss a few things like BUGS from the man page, as i consider current hardcoded configuration, while useful for most cases, as a bug that should be fixed before ie. spi, HZ=1000, etc.. It could be used for even indoor navigation, so I think the sensor has plenty more potential, than what bme(4) does now offer, but i don't see why you couldn't offer it as-is, if you find it working after properly testing this version, on tech@, good luck:) -Artturi diff --git share/man/man4/Makefile share/man/man4/Makefile index 8004d8b575d..3ac02b89788 100644 --- share/man/man4/Makefile +++ share/man/man4/Makefile @@ -15,7 +15,7 @@ MAN= aac.4 ac97.4 acphy.4 acrtc.4 \ auacer.4 audio.4 aue.4 auglx.4 auich.4 auixp.4 autri.4 auvia.4 \ axe.4 axen.4 axppmic.4 azalia.4 \ bce.4 bcmaux.4 bcmdog.4 bcmrng.4 bcmtemp.4 berkwdt.4 bge.4 \ - bgw.4 bio.4 bktr.4 bmtphy.4 bnx.4 \ + bgw.4 bio.4 bktr.4 bme.4 bmtphy.4 bnx.4 \ boca.4 bpf.4 brgphy.4 bridge.4 brswphy.4 bwfm.4 bwi.4 bytgpio.4 \ cac.4 cas.4 cardbus.4 carp.4 ccp.4 cd.4 cdce.4 cfxga.4 \ ch.4 chvgpio.4 ciphy.4 ciss.4 clcs.4 clct.4 cmpci.4 \ diff --git share/man/man4/bme.4 share/man/man4/bme.4 new file mode 100644 index 000..bb48513d9dc --- /dev/null +++ share/man/man4/bme.4 @@ -0,0 +1,44 @@ +.\"$OpenBSD: tsl.4,v 1.10 2015/10/02 09:29:02 sobrado Exp $ +.\" +.\" Copyright (c) 2005 Mark Kettenis +.\" Copyright (c) 2018 Artturi Alm +.\" +.\" Permission to use, copy, modify, and distribute this software for any +.\" purpose with or without fee is hereby granted, provided that the above +.\" copyright notice and this permission notice appear in all copies. +.\" +.\" THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES +.\" WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF +.\" MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR +.\" ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES +.\" WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN +.\" ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF +.\" OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. +.\" +.Dd $Mdocdate$ +.Dt BME 4 +.Os +.Sh NAME +.Nm bme +.Nd Bosch BME280 Environmental sensor +.Sh SYNOPSIS +.Cd "bme* at iic?" +.Sh DESCRIPTION +The +.Nm +driver provides support for the Bosch BME280 Environmental sensor. +These sensors contain combined digital humidity, pressure and temperature +sensing capabilities. +These values are made available through the +.Xr sysctl 8 +interface. +.Sh SEE ALSO +.Xr iic 4 , +.Xr intro 4 , +.Xr sensorsd 8 , +.Xr sysctl 8 +.Sh HISTORY +The +.Nm +driver first appeared in +.Ox 6.4 . diff --git sys/arch/arm64/conf/GENERIC sys/arch/arm64/conf/GENERIC index d7b7dca12bf..37050606674 100644 --- sys/arch/arm64/conf/GENERIC +++ sys/arch/arm64/conf/GENERIC @@ -296,6 +296,7 @@ ch* at scsibus? uk*at scsibus? # I2C devices +bme* at iic? # BME280 environmental sensor islrtc*at iic? # ISL1208 RTC rkpmic*at iic? # RK808 PMIC diff --git sys/arch/armv7/conf/GENERIC sys/arch/armv7/conf/GENERIC index 052f73bf71f..fae08d5c8cd 100644 --- sys/arch/armv7/conf/GENERIC +++ sys/arch/armv7/conf/GENERIC @@ -178,7 +178,9 @@ iic*at rkiic? rktemp*at fdt? dwdog* at fdt? +# I2C devices axppmic* at iic? # axp209 pmic +bme* at iic? # BME280 environmental sensor crosec*at iic? wskbd* at crosec? mux 1 pcfrtc*at iic? diff --git sys/dev/fdt/bme.c sys/dev/fdt/bme.c new file mode 100644 index 000..04a61dc7d19 --- /dev/null +++ sys/dev/fdt/bme.c @@ -0,0 +1,547 @@ +/* $OpenBSD: tsl2560.c,v 1.7 2008/04/18 01:17:51 deraadt Exp $ */ + +/* + * Copyright (c) 2005 Mark Kettenis + * Copyright (c) 2017 Artturi Alm + * + * Permission to use, copy, modify, and distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR
Re: gtk libool slow on arm
Thank you for the updates. I will build and test as soon as I get finished with the php build. My system is still flailing away at gtk with each ld session taking about 20 minutes. Could be days yet. -Original Message- From: owner-...@openbsd.org On Behalf Of Artturi Alm Sent: May 3, 2018 5:21 PM To: Stephen Graf Cc: arm@openbsd.org Subject: Re: gtk libool slow on arm On Wed, May 02, 2018 at 05:36:33PM -0700, s_g...@telus.net wrote: > Thank you for the advice. I will try the patch, but I want to let the > current php build complete, which may take a few more days. > > Did you get the bme code I sent you? > Yep, I just forgot about it, and now i did "over do" a bit with the necessary cleanup/refactoring.. oops, but no more ugly unions nor "forbidden hacks":) diff below does still miss a few things like BUGS from the man page, as i consider current hardcoded configuration, while useful for most cases, as a bug that should be fixed before ie. spi, HZ=1000, etc.. It could be used for even indoor navigation, so I think the sensor has plenty more potential, than what bme(4) does now offer, but i don't see why you couldn't offer it as-is, if you find it working after properly testing this version, on tech@, good luck:) -Artturi diff --git share/man/man4/Makefile share/man/man4/Makefile index 8004d8b575d..3ac02b89788 100644 --- share/man/man4/Makefile +++ share/man/man4/Makefile @@ -15,7 +15,7 @@ MAN= aac.4 ac97.4 acphy.4 acrtc.4 \ auacer.4 audio.4 aue.4 auglx.4 auich.4 auixp.4 autri.4 auvia.4 \ axe.4 axen.4 axppmic.4 azalia.4 \ bce.4 bcmaux.4 bcmdog.4 bcmrng.4 bcmtemp.4 berkwdt.4 bge.4 \ - bgw.4 bio.4 bktr.4 bmtphy.4 bnx.4 \ + bgw.4 bio.4 bktr.4 bme.4 bmtphy.4 bnx.4 \ boca.4 bpf.4 brgphy.4 bridge.4 brswphy.4 bwfm.4 bwi.4 bytgpio.4 \ cac.4 cas.4 cardbus.4 carp.4 ccp.4 cd.4 cdce.4 cfxga.4 \ ch.4 chvgpio.4 ciphy.4 ciss.4 clcs.4 clct.4 cmpci.4 \ diff --git share/man/man4/bme.4 share/man/man4/bme.4 new file mode 100644 index 000..bb48513d9dc --- /dev/null +++ share/man/man4/bme.4 @@ -0,0 +1,44 @@ +.\"$OpenBSD: tsl.4,v 1.10 2015/10/02 09:29:02 sobrado Exp $ +.\" +.\" Copyright (c) 2005 Mark Kettenis .\" +Copyright (c) 2018 Artturi Alm .\" +.\" Permission to use, copy, modify, and distribute this software for +any .\" purpose with or without fee is hereby granted, provided that +the above .\" copyright notice and this permission notice appear in all copies. +.\" +.\" THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL +WARRANTIES .\" WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED +WARRANTIES OF .\" MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE +AUTHOR BE LIABLE FOR .\" ANY SPECIAL, DIRECT, INDIRECT, OR +CONSEQUENTIAL DAMAGES OR ANY DAMAGES .\" WHATSOEVER RESULTING FROM LOSS +OF USE, DATA OR PROFITS, WHETHER IN AN .\" ACTION OF CONTRACT, +NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF .\" OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. +.\" +.Dd $Mdocdate$ +.Dt BME 4 +.Os +.Sh NAME +.Nm bme +.Nd Bosch BME280 Environmental sensor +.Sh SYNOPSIS +.Cd "bme* at iic?" +.Sh DESCRIPTION +The +.Nm +driver provides support for the Bosch BME280 Environmental sensor. +These sensors contain combined digital humidity, pressure and +temperature sensing capabilities. +These values are made available through the .Xr sysctl 8 interface. +.Sh SEE ALSO +.Xr iic 4 , +.Xr intro 4 , +.Xr sensorsd 8 , +.Xr sysctl 8 +.Sh HISTORY +The +.Nm +driver first appeared in +.Ox 6.4 . diff --git sys/arch/arm64/conf/GENERIC sys/arch/arm64/conf/GENERIC index d7b7dca12bf..37050606674 100644 --- sys/arch/arm64/conf/GENERIC +++ sys/arch/arm64/conf/GENERIC @@ -296,6 +296,7 @@ ch* at scsibus? uk*at scsibus? # I2C devices +bme* at iic? # BME280 environmental sensor islrtc*at iic? # ISL1208 RTC rkpmic*at iic? # RK808 PMIC diff --git sys/arch/armv7/conf/GENERIC sys/arch/armv7/conf/GENERIC index 052f73bf71f..fae08d5c8cd 100644 --- sys/arch/armv7/conf/GENERIC +++ sys/arch/armv7/conf/GENERIC @@ -178,7 +178,9 @@ iic*at rkiic? rktemp*at fdt? dwdog* at fdt? +# I2C devices axppmic* at iic? # axp209 pmic +bme* at iic? # BME280 environmental sensor crosec*at iic? wskbd* at crosec? mux 1 pcfrtc*at iic? diff --git sys/dev/fdt/bme.c sys/dev/fdt/bme.c new file mode 100644 index 000..04a61dc7d19 --- /dev/null +++ sys/dev/fdt/bme.c @@ -0,0 +1,547 @@ +/* $OpenBSD: tsl2560.c,v 1.7 2008/04/18 01:17:51 deraadt Exp $ */ + +/* + * Copyright (c) 2005 Mark Kettenis + * Copyright (c) 2017 Artturi Alm + * + * Permission to use, copy, modify, and distribute this software for +any + * purpose with or without fee is hereby granted, provided that the +above + * copy
