GitLab

Fixes building with -Werror (i.e. validate) and GHC < 6.6

After my talk, I got the idea of 'shallow rewriting' for the
dataflow framework.  Here it is implemented, along with
some related ideas late making Graph and not LGraph primary.

The only bad thing is that the whole bit is stitched together
out of ill-fitting pieces, kind of like Frankenstein's monster.
A new ZipDataflow will rise out of the ashes.

- ZipCfg: add mkBlockId :: Unique -> BlockId
- MkZipCfg: change sequence --> catAGrpahs
- MkZipCfgCmm: add mkCmmIfThen

Not fully validated, but I don't think anything will break

See Note [Float coercions (unlifted)] in Simplify

Test is simpl018

Fix exponential-time behaviour with type synonyms; rename -XPartiallyAppliedTypeSynonyms to -XLiberalTypeSynonyms

Fixes exponential behaviour present in GHC 6.6!

I renamed the flag because the old (not very old) name wasn't
describing what it does.

- This patch re-instates the policy of 6.6.1 to zonk the given constraints
  in the simplifier loop.

MERGE TO STABLE

  MERGE TO STABLE

Ian Lynagh <igloo@earth.li>**20070916005328] {

If iterative coalescing isn't turned on, then do a single aggressive
coalescing pass for the first build/color cycle and then back off to 
conservative coalescing for subseqent passes.

Aggressive coalescing is a cheap way to eliminate lots of the reg-reg
moves, but it can make the graph less colorable - if we turn it on 
for every pass then allocation for code with a large amount of register
pressure (ie SHA1) doesn't converge in a sensible number of cycles.

Coalescences must be applied to the unsimplified graph in the same order 
they were found by the iterative coalescing algorithm - otherwise
the vreg rewrite mapping will be wrong and badness will ensue.

When -dasm-lint is turned on the register conflict graph is checked for 
internal consistency after each build/color pass. Make sure that all 
edges point to valid nodes, that nodes are colored differently to their
neighbours, and if the graph is supposed to be colored, that all nodes
actually have a color.

To help with debugging / nicer -ddump-asm-regalloc-stages

On further testing it turns out that using Chaitin's spill cost formula
of counting the number of uses of a var and then dividing by the degree
of the node isn't actually a win. Perhaps because this isn't counting 
the number of actual spills inserted in the output code.

This would be worth revisiting if other work manages to increase the 
register pressure in the native code.

Track what slots each basic block reloads from. When cleaning spill 
instructions we can use this information to decide whether the slot 
spilled to will ever be read from on this path.

This lets us gain coverage on programs that use fork, like xmonad.

* To be merged to STABLE *

* To be merged to STABLE. *

This change allows you to write
	f :: (Eq a) => (Ord b) => a -> b -> b
Previously you could only have a forall and context after '->'
but not after '=>' which is strange and inconsistent.

Making the parser a bit more generous didn't change the number
of shift/reduce conflicts.

tc236 tests.