install numba on Fedora (not with conda)

[Neal Becker]

I attempted to install numba via pip3.

I used alternatives to point llvm-config to llvm-3.8 (because numba didn't
seem to want to build with llvm-3.7).
alternatives --display llvm-config

$ llvm-config - status is manual.
link currently points to /usr/bin/llvm-config-64
/usr/bin/llvm-config-64-3.7 - priority 64
/usr/bin/llvm-config-64 - priority 64
Current `best' version is /usr/bin/llvm-config-64-3.7.

$ llvm-config --version
3.8.0

Install went OK, but when I tested:
In [100]: find_runs(df)
---------------------------------------------------------------------------
KeyError Traceback (most recent call last)
/home/nbecker/.local/lib/python3.5/site-packages/numba/bytecode.py in
_fetch_opcode(self)
250 try:
--> 251 info = BYTECODE_TABLE[opcode]
252 except KeyError:

KeyError: 121

During handling of the above exception, another exception occurred:

NotImplementedError Traceback (most recent call last)
<ipython-input-103-e18a660e554b> in <module>()
----> 1 find_runs(df)

/home/nbecker/.local/lib/python3.5/site-packages/numba/dispatcher.py in
_compile_for_args(self, *args, **kws)
284 argtypes.append(self.typeof_pyval(a))
285 try:
--> 286 return self.compile(tuple(argtypes))
287 except errors.TypingError as e:
288 # Intercept typing error that may be due to an argument

/home/nbecker/.local/lib/python3.5/site-packages/numba/dispatcher.py in
compile(self, sig)
530
531 self._cache_misses[sig] += 1
--> 532 cres = self._compiler.compile(args, return_type)
533 self.add_overload(cres)
534 self._cache.save_overload(sig, cres)

/home/nbecker/.local/lib/python3.5/site-packages/numba/dispatcher.py in
compile(self, args, return_type)
79 impl,
80 args=args,
return_type=return_type,
---> 81 flags=flags,
locals=self.locals)
82 # Check typing error if object mode is used
83 if cres.typing_error is not None and not
flags.enable_pyobject:

/home/nbecker/.local/lib/python3.5/site-packages/numba/compiler.py in
compile_extra(typingctx, targetctx, func, args, return_type, flags, locals,
library)
682 pipeline = Pipeline(typingctx, targetctx, library,
683 args, return_type, flags, locals)
--> 684 return pipeline.compile_extra(func)
685
686

/home/nbecker/.local/lib/python3.5/site-packages/numba/compiler.py in
compile_extra(self, func)
341 return self.cr
342 else:
--> 343 raise e
344
345 self.bc = bc

/home/nbecker/.local/lib/python3.5/site-packages/numba/compiler.py in
compile_extra(self, func)
335
336 try:
--> 337 bc = self.extract_bytecode(self.func_id)
338 except BaseException as e:
339 if self.status.can_giveup:

/home/nbecker/.local/lib/python3.5/site-packages/numba/compiler.py in
extract_bytecode(self, func_id)
325 Extract bytecode from function
326 """
--> 327 bc = bytecode.ByteCode(func_id)
328 if config.DUMP_BYTECODE:
329 print(bc.dump())

/home/nbecker/.local/lib/python3.5/site-packages/numba/bytecode.py in
__init__(self, func_id)
292
293 # A map of {offset: ByteCodeInst}
--> 294 table = OrderedDict(ByteCodeIter(code))
295 self._compute_lineno(table, code)
296

/home/nbecker/.local/lib/python3.5/site-packages/numba/bytecode.py in
next(self)
261
262 def next(self):
--> 263 offset, opcode, arg = self._fetch_opcode()
264 if opcode == EXTENDED_ARG:
265 hi_arg = arg

/home/nbecker/.local/lib/python3.5/site-packages/numba/bytecode.py in
_fetch_opcode(self)
253 ts = "offset=%d opcode=0x%x opname=%s"
254 tv = offset, opcode, dis.opname[opcode]
--> 255 raise NotImplementedError(ts % tv)
256 if info.argsize:
257 arg = self.read_arg(info.argsize)

NotImplementedError: offset=32 opcode=0x79 opname=SETUP_EXCEPT

[Siu Kwan Lam]

The exception indicates the use of an unsupported operation SETUP_EXCEPT.  This is likely due to the use of try..except blocks.  It seems the installation went fine but numba error out on the user code.

--
You received this message because you are subscribed to the Google Groups "Numba Public Discussion - Public" group.
To unsubscribe from this group and stop receiving emails from it, send an email to numba-users...@continuum.io.
To post to this group, send email to numba...@continuum.io.
To view this discussion on the web visit https://groups.google.com/a/continuum.io/d/msgid/numba-users/o2eear%24uql%241%40blaine.gmane.org.
For more options, visit https://groups.google.com/a/continuum.io/d/optout.

--

Siu Kwan Lam

Software Engineer

Continuum Analytics