@Raul Thank you for your suggestion. I have pushed a patch to fix force
macro emulation for some avx emulation. Please test again.
On Sat, Oct 28, 2023 at 8:35 PM Raul Miller <rauldmil...@gmail.com> wrote:
> Another workaround occurs to me:
>
> When compiling with -O0, only build the avx2 instance of jconsole.
>
> The /bin/sh test for -O0:
>
> case "$CFLAGS" in *-O0*) true;; *) false;; esac
>
> This could be used as is, as the control statement for a shell if
> statement. Though I'd like to restructure the script/buildga.sh script
> for conciseness and readability if I do this, if that would be ok.
>
> That said, I don't know where your efforts to fix the emulation issue are.
>
> Thanks,
>
> --
> Raul
>
> On Thu, Oct 26, 2023 at 10:53 AM bill lam <bbill....@gmail.com> wrote:
> >
> > I can reproduce the asm error. I suspect the problem is that the
> particular
> > intel asm instruction only supports immediate addressing for i (value
> known
> > during compilation time), but the argument was passed in as an inline
> > function parameter. The workaround is to replace the inline function with
> > macro call. I will work on it over the weekend.
> >
> > All files under the script are meant for the github action yaml. They are
> > intended for github runners, Developers should run test/tsu.ijs directly
> > after building binaries (using shell scripts under make2 on *nix or
> makevs
> > etc on windows) on their local machine.
> >
> > On Thu, Oct 26, 2023 at 10:16 PM Raul Miller <rauldmil...@gmail.com>
> wrote:
> >
> > > Well.. I think a part of the answer is .github/workflows/jsource.yml
> > >
> > > There's lots of things going on, and some of them are difficult to
> > > understand.
> > >
> > > For my purposes, I don't mind supplying linux on the command line. If
> > > I had enough machines, and wanted to automate based on uname, I guess
> > > I'd update the script with tr '[A-Z]' '[a-z]' or create a wrapper
> > > script which did that. (Probably a wrapper script, patterned after the
> > > github workflow yaml.)
> > >
> > > I guess the fundamental problems here are:
> > >
> > > (1) different platforms require entirely different build chains, and
> > >
> > > (2) our developer community is relatively isolated, and
> > >
> > > (3) it takes energy and time to implement working changes
> > >
> > > I guess maybe I should submit a patch for testga.sh to fix what seem
> > > to be obvious problems (the unquoted $1 and the local context for
> > > tests). I was hoping that bill could say a few words about the state
> > > of and/or plan for the script, though.
> > >
> > > Thanks,
> > >
> > > --
> > > Raul
> > >
> > > On Thu, Oct 26, 2023 at 9:36 AM 'Viktor Grigorov' via Source
> > > <sou...@jsoftware.com> wrote:
> > > >
> > > > On "9.5.0-beta6" the script has some issues. There are common to most
> > > the dash script I've seen in the zip and the main repo. Likely not
> what you
> > > are discussing, but worth mentioning, since you are trying to have a
> glibc
> > > unix release.
> > > >
> > > > I get
> > > >
> > > > script/testga.sh: line 16: [: =: unary operator expected
> > > > script/testga.sh: line 18: [: =: unary operator expected
> > > > script/testga.sh: line 22: [: =: unary operator expected
> > > > script/testga.sh: line 24: [: =: unary operator expected
> > > > script/testga.sh: line 26: [: =: unary operator expected
> > > > argument is linux|darwin|raspberry|openbsd|freebsd|wasm
> > > >
> > > > when running sans argument. $1 isn't doubly quoted on those lines.
> > > >
> > > > Running with an argument of linux---why lowercase when uname returns
> it
> > > capitalized..?---I get that j64 cannot be accessed. This is because it
> is
> > > suffixed with avx or avx2 or avx512. Symlinking my version to j64/ and
> > > running again, it cannot stat 'mpir/linux/x86_64/libgmpd.so'. I have
> only
> > > one such named file in mpir/linux/i386/. I don't know whether I can
> test
> > > whether I have extended integers, since I haven't got the library
> working
> > > yet, but any x prepended to an integer doesn't cause an issue. Whatever
> > > libgmpd.so does in addition to libgmp.so is beyond me currently.
> > > >
> > > > Could provide more info, if desired.
> > > >
> > > > Oct 26, 2023, 15:46 by rauldmil...@gmail.com:
> > > >
> > > > > I understand the error message in g13x.ijs
> > > > >
> > > > > I guess the question is: is script/testga.sh used anywhere (it sure
> > > > > looks like it's intended to be used broadly)? If so, why don't
> other
> > > > > people get this error?
> > > > >
> > > > > I haven't changed anything about the testing process.
> > > > >
> > > > > (Meanwhile - if anyone else stumbles over this - it looks like
> there's
> > > > > a comment in avx-intremu.h explaining the error and what to do
> about
> > > > > it.)
> > > > >
> > > > > Thanks,
> > > > >
> > > > > --
> > > > > Raul
> > > > >
> > > > >
> > > > > On Thu, Oct 26, 2023 at 5:29 AM Henry Rich <henryhr...@gmail.com>
> > > wrote:
> > > > >
> > > > >>
> > > > >> I have nothing on the asm problem.
> > > > >>
> > > > >> On the other, the error is just what it looks like: you are
> publicly
> > > > >> assigning a name that was previously assigned privately. The
> scripts
> > > are
> > > > >> sometimes loaded under an explicit definition, sometimes not, so
> =. is
> > > > >> unpredictable.
> > > > >>
> > > > >> hhr
> > > > >>
> > > > >> On Thu, Oct 26, 2023, 1:34 AM Raul Miller <rauldmil...@gmail.com>
> > > wrote:
> > > > >>
> > > > >> > I'm testing out a change to the support for extended integers,
> and
> > > > >> > I've run across what seems like an unrelated issue in g13x.ijs,
> when
> > > > >> > running script/testga.sh linux:
> > > > >> >
> > > > >> > NB. 13!:13
> > > > >> > -------------------------------------------------------------
> > > > >> >
> > > > >> > mean=: sum % #
> > > > >> > sum =: [: +/ ".@('t=:13!:13 $0'&[) ] ]
> > > > >> >
> > > > >> > 13!:0 ]1
> > > > >> > 1: mean x=: ?4 5$100
> > > > >> > |domain error: sum
> > > > >> > |public assignment to a name with a private value
> > > > >> > | t =:13!:13$0
> > > > >> >
> > > > >> > Not sure what to do about this...
> > > > >> >
> > > > >> > Meanwhile, running test/tsu.ijs manually shows me an error with
> my
> > > > >> > changes. I'd like to use a debugger to inspect what's happening
> to
> > > > >> > cause that error. However, when I attempt to build for
> debugging, I
> > > > >> > get a compiler error which I do not understand:
> > > > >> >
> > > > >> > CFLAGS="-ggdb -O0" script/buildga.sh linux
> > > > >> > ...
> > > > >> > clang -c -o viavx6.o ../../../../jsrc/viavx6.c -fPIC
> > > > >> > -fvisibility=hidden -fno-strict-aliasing -fwrapv -Werror
> -Wextra
> > > > >> > -Wno-unknown-warning-option -Wsign-compare
> > > > >> > -Wtautological-constant-out-of-range-compare -Wuninitialized
> > > > >> > -Wno-char-subscripts -Wno-consumed
> > > > >> > -Wno-delete-non-abstract-non-virtual-dtor
> > > > >> > -Wno-deprecated-non-prototype -Wno-empty-body
> > > > >> > -Wno-implicit-float-conversion
> -Wno-implicit-int-float-conversion
> > > > >> > -Wno-incompatible-function-pointer-types -Wno-int-conversion
> > > > >> > -Wno-int-in-bool-context -Wno-missing-braces
> > > > >> > -Wno-missing-field-initializers -Wno-null-pointer-arithmetic
> > > > >> > -Wno-null-pointer-subtraction -Wno-parentheses
> -Wno-pass-failed
> > > > >> > -Wno-pointer-sign -Wno-pointer-to-int-cast
> > > -Wno-shift-count-overflow
> > > > >> > -Wno-sometimes-uninitialized -Wno-string-plus-int
> > > > >> > -Wno-unknown-pragmas -Wno-unsequenced
> -Wno-unused-but-set-variable
> > > > >> > -Wno-unused-function -Wno-unused-parameter -Wno-unused-value
> > > > >> > -Wno-unused-variable -ggdb -O0 -DPYXES=1 -DSLEEF=1
> > > > >> > -I../../../../mpir/include -DEMU_AVX2=1 -msse3
> > > > >> > In file included from ../../../../jsrc/viavx4.c:6:
> > > > >> > In file included from ../../../../jsrc/j.h:119:
> > > > >> > ../../../../jsrc/avxintrin-emu.h:1112:19: error: invalid
> operand for
> > > > >> > inline asm constraint 'i'
> > > > >> > __asm__ ( "cmppd %[pred_], %[m2_], %[res_]" : [res_]
> "+x"
> > > > >> > (res) : [m2_] "x" (m2), [pred_] "i" (predicate) );
> > > > >> > ^
> > > > >> > ../../../../jsrc/avxintrin-emu.h:1112:19: error: invalid
> operand for
> > > > >> > inline asm constraint 'i'
> > > > >> > ../../../../jsrc/avxintrin-emu.h:1112:19: error: invalid
> operand for
> > > > >> > inline asm constraint 'i'
> > > > >> > ../../../../jsrc/avxintrin-emu.h:1112:19: error: invalid
> operand for
> > > > >> > inline asm constraint 'i'
> > > > >> > ../../../../jsrc/avxintrin-emu.h:1112:19: error: invalid
> operand for
> > > > >> > inline asm constraint 'i'
> > > > >> > ../../../../jsrc/avxintrin-emu.h:1112:19: error: invalid
> operand for
> > > > >> > inline asm constraint 'i'
> > > > >> > ...
> > > > >> >
> > > > >> > If anyone has clues for me, that would be great.
> > > > >> >
> > > > >> > Thanks,
> > > > >> >
> > > > >> > --
> > > > >> > Raul
> > > > >> >
> > > ----------------------------------------------------------------------
> > > > >> > For information about J forums see
> > > http://www.jsoftware.com/forums.htm
> > > > >> >
> > > > >>
> ----------------------------------------------------------------------
> > > > >> For information about J forums see
> > > http://www.jsoftware.com/forums.htm
> > > > >>
> > > > >
> ----------------------------------------------------------------------
> > > > > For information about J forums see
> http://www.jsoftware.com/forums.htm
> > > > >
> > > >
> > > >
> ----------------------------------------------------------------------
> > > > For information about J forums see
> http://www.jsoftware.com/forums.htm
> > > ----------------------------------------------------------------------
> > > For information about J forums see http://www.jsoftware.com/forums.htm
> > >
> > ----------------------------------------------------------------------
> > For information about J forums see http://www.jsoftware.com/forums.htm
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
