GitLab

* Vectorisation avoidance is now the default
* Types and values from unvectorised modules are permitted in scalar code
* Simplified the VECTORISE pragmas (see http://hackage.haskell.org/trac/ghc/wiki/DataParallel/VectPragma for the spec)
* Vectorisation information is now included in the annotated Core AST

Keep the warning off for template-haskell and bytestring for the moment.

Move the following primop rules from GHC.Base to PrelRules:

"narrow32Int#"  forall x#. narrow32Int#   x# = x#
"narrow32Word#" forall x#. narrow32Word#   x# = x#

"int2Word2Int"  forall x#. int2Word# (word2Int# x#) = x#
"word2Int2Word" forall x#. word2Int# (int2Word# x#) = x#

All the flags that 'ways' imply are now dynamic

Ported various rules for numeric types from GHC.Base. Added new rules
for bitwise operations, shifts and word comparisons.

Also, temporarely disable that warning for validate builds, until we
finish fixing them all.

This patch makes
  fiExpr (AnnLam ...)
  noFloatIntoExpr (AnnLam ...)
line up with each other.  Previously they behaved inconsistently
with bad results.

The details are in Trac #7088.    There's a regression test
in simplCore/should_compile/T7088

The main thing is that we now keep unsolved Derived constraints in the
wc_flats of a WantedConstraints, rather than discarding them each time.
This actually fixes a poential (admittedly obscure) bug, when we currently
discard a superclass constraint, and may never re-generate it, and may
thereby miss a functional dependency.

Instead, reportErrors filters out Derived constraints that we don't want
to report.

The other changes are all small refactorings following our walk-through.

This warning was suggested by Trac #6082, where we had
a library RULE that failed to fire because its function
was inlined too soon.

       _ = e

Thse are used in a few libraries, either to add type constraints
via a signature, or to mention some variables that are only
otherwise mentioned in one #ifdef branch

See Note [Calculate unfolding guidance on the non-occ-anal'd expression]
This makes a big difference to residency (530M vs over 800M when compiling
Cabal).

This fixes the majority of the regression in Trac #6104

I suspect I have done the wrong thing; I hope someone can improve.

Fixes Trac #7085

parts of WantedConstraints, and other small refactorings.

We don't need them any more, now that we have DynFlags inside SDoc