Running duckling binary on ARM based architecture (NVIDIA Xavier, Linux) - unknown RTS option: -N

[IliaGrigoriev]

Dear Facebook Team,

I have the following problem when trying to run the successfully built duckling binary on ARM based architecture on SD Card of NVIDIA Xavier, Ubuntu 18.04, Stack version 1.5.1 aarch64:

stack build
stack exec duckling-example-exe

I get the following output:

duckling-example-exe: unknown RTS option: -N
duckling-example-exe: 
duckling-example-exe: Usage: <prog> <args> [+RTS <rtsopts> | -RTS <args>] ... --RTS <args>
duckling-example-exe: 
duckling-example-exe:    +RTS    Indicates run time system options follow
duckling-example-exe:    -RTS    Indicates program arguments follow
duckling-example-exe:   --RTS    Indicates that ALL subsequent arguments will be given to the
duckling-example-exe:            program (including any of these RTS flags)
duckling-example-exe: 
duckling-example-exe: The following run time system options are available:
duckling-example-exe: 
duckling-example-exe:   -?       Prints this message and exits; the program is not executed
duckling-example-exe:   --info   Print information about the RTS used by this program
duckling-example-exe: 
duckling-example-exe:   -K<size>  Sets the maximum stack size (default: 80% of the heap)
duckling-example-exe:             Egs: -K32k -K512k -K8M
duckling-example-exe:   -ki<size> Sets the initial thread stack size (default 1k)  Egs: -ki4k -ki2m
duckling-example-exe:   -kc<size> Sets the stack chunk size (default 32k)
duckling-example-exe:   -kb<size> Sets the stack chunk buffer size (default 1k)
duckling-example-exe: 
duckling-example-exe:   -A<size> Sets the minimum allocation area size (default 512k) Egs: -A1m -A10k
duckling-example-exe:   -n<size> Allocation area chunk size (0 = disabled, default: 0)
duckling-example-exe:   -O<size> Sets the minimum size of the old generation (default 1M)
duckling-example-exe:   -M<size> Sets the maximum heap size (default unlimited)  Egs: -M256k -M1G
duckling-example-exe:   -H<size> Sets the minimum heap size (default 0M)   Egs: -H24m  -H1G
duckling-example-exe:   -xb<addr> Sets the address from which a suitable start for the heap memory
duckling-example-exe:             will be searched from. This is useful if the default address
duckling-example-exe:             clashes with some third-party library.
duckling-example-exe:   -m<n>    Minimum % of heap which must be available (default 3%)
duckling-example-exe:   -G<n>    Number of generations (default: 2)
duckling-example-exe:   -c<n>    Use in-place compaction instead of copying in the oldest generation
duckling-example-exe:            when live data is at least <n>% of the maximum heap size set with
duckling-example-exe:            -M (default: 30%)
duckling-example-exe:   -c       Use in-place compaction for all oldest generation collections
duckling-example-exe:            (the default is to use copying)
duckling-example-exe:   -w       Use mark-region for the oldest generation (experimental)
duckling-example-exe:   -I<sec>  Perform full GC after <sec> idle time (default: 0.3, 0 == off)
duckling-example-exe: 
duckling-example-exe:   -T         Collect GC statistics (useful for in-program statistics access)
duckling-example-exe:   -t[<file>] One-line GC statistics (if <file> omitted, uses stderr)
duckling-example-exe:   -s[<file>] Summary  GC statistics (if <file> omitted, uses stderr)
duckling-example-exe:   -S[<file>] Detailed GC statistics (if <file> omitted, uses stderr)
duckling-example-exe: 
duckling-example-exe: 
duckling-example-exe:   -Z       Don't squeeze out update frames on stack overflow
duckling-example-exe:   -B       Sound the bell at the start of each garbage collection
duckling-example-exe: 
duckling-example-exe:   -h       Heap residency profile (output file <program>.hp)
duckling-example-exe:   -i<sec>  Time between heap profile samples (seconds, default: 0.1)
duckling-example-exe: 
duckling-example-exe:   -C<secs>  Context-switch interval in seconds.
duckling-example-exe:             0 or no argument means switch as often as possible.
duckling-example-exe:             Default: 0.02 sec.
duckling-example-exe:   -V<secs>  Master tick interval in seconds (0 == disable timer).
duckling-example-exe:             This sets the resolution for -C and the heap profile timer -i,
duckling-example-exe:             and is the frequence of time profile samples.
duckling-example-exe:             Default: 0.01 sec.
duckling-example-exe: 
duckling-example-exe:   --install-signal-handlers=<yes|no>
duckling-example-exe:             Install signal handlers (default: yes)
duckling-example-exe:   -e<n>     Maximum number of outstanding local sparks (default: 4096)
duckling-example-exe:   -xq       The allocation limit given to a thread after it receives
duckling-example-exe:             an AllocationLimitExceeded exception. (default: 100k)
duckling-example-exe: 
duckling-example-exe: RTS options may also be specified using the GHCRTS environment variable.
duckling-example-exe: 
duckling-example-exe: Other RTS options may be available for programs compiled a different way.
duckling-example-exe: The GHC User's Guide has full details.
duckling-example-exe:

I get the same output, when using newer version of stack (2.1.1), the lines inside duckling folder are:

./stack-2.1.1-linux-aarch64/stack build ../

and then:

./stack-2.1.1-linux-aarch64/stack exec duckling-example-exe

What am I doing wrong?

Many thanks in advance.

Best regards, Ilia

[chessai]

Stack isn't the issue. Looks like GHC disables SMP (symmetric multiprocessor) on older versions of ARM: https://gitlab.haskell.org/ghc/ghc/-/blob/master/hadrian/src/Oracles/Flag.hs#L69

Looks like you need a minimum of ARMv7. Can you confirm what ARM version you are on?

[IliaGrigoriev]

Hi. First of all, thank you very much for you reply. Here is the output from cat /proc/cpuinfo:

processor	: 0
model name	: ARMv8 Processor rev 0 (v8l)
BogoMIPS	: 62.50
Features	: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp
CPU implementer	: 0x4e
CPU architecture: 8
CPU variant	: 0x0
CPU part	: 0x004
CPU revision	: 0
MTS version	: 45309758

processor	: 1
model name	: ARMv8 Processor rev 0 (v8l)
BogoMIPS	: 62.50
Features	: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp
CPU implementer	: 0x4e
CPU architecture: 8
CPU variant	: 0x0
CPU part	: 0x004
CPU revision	: 0
MTS version	: 45309758

processor	: 2
model name	: ARMv8 Processor rev 0 (v8l)
BogoMIPS	: 62.50
Features	: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp
CPU implementer	: 0x4e
CPU architecture: 8
CPU variant	: 0x0
CPU part	: 0x004
CPU revision	: 0
MTS version	: 45309758

processor	: 3
model name	: ARMv8 Processor rev 0 (v8l)
BogoMIPS	: 62.50
Features	: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp
CPU implementer	: 0x4e
CPU architecture: 8
CPU variant	: 0x0
CPU part	: 0x004
CPU revision	: 0
MTS version	: 45309758

processor	: 4
model name	: ARMv8 Processor rev 0 (v8l)
BogoMIPS	: 62.50
Features	: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp
CPU implementer	: 0x4e
CPU architecture: 8
CPU variant	: 0x0
CPU part	: 0x004
CPU revision	: 0
MTS version	: 45309758

processor	: 5
model name	: ARMv8 Processor rev 0 (v8l)
BogoMIPS	: 62.50
Features	: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp
CPU implementer	: 0x4e
CPU architecture: 8
CPU variant	: 0x0
CPU part	: 0x004
CPU revision	: 0
MTS version	: 45309758

processor	: 6
model name	: ARMv8 Processor rev 0 (v8l)
BogoMIPS	: 62.50
Features	: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp
CPU implementer	: 0x4e
CPU architecture: 8
CPU variant	: 0x0
CPU part	: 0x004
CPU revision	: 0
MTS version	: 45309758

processor	: 7
model name	: ARMv8 Processor rev 0 (v8l)
BogoMIPS	: 62.50
Features	: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp
CPU implementer	: 0x4e
CPU architecture: 8
CPU variant	: 0x0
CPU part	: 0x004
CPU revision	: 0
MTS version	: 45309758

Best regards, Ilia

[IliaGrigoriev]

./stack-2.1.1-linux-aarch64/stack ghc -- --version
The Glorious Glasgow Haskell Compilation System, version 8.0.2

uname -a
Linux ilia 4.9.140-tegra #1 SMP PREEMPT Mon Dec 9 22:52:02 PST 2019 aarch64 aarch64 aarch64 GNU/Linux

[chessai]

Can you try just removing the -N option from the rts-options?

[chessai]

Bump on this @IliaGrigoriev

[IliaGrigoriev]

Hi @chessai,

sorry, there has been an overload of additional tasks since January this year. I should be able to come back to the task, which required duckling setup on ARM in the mid April this year.

Best wishes, Ilia

[finusdev]

So how can we run Duckling on aarm64 now? Already tried also with docker

[vieenrose]

You could use the following command to remove -N option, as workaround

find . -type f -name "*.cabal" -exec sed -i 's/-with-rtsopts=-N//g' {} +

After building Duckling v0.1.6.1 on Jetson Xavier 8GB, the resulting executable is now runnable on Xavier as well as on Nano.