[Git][ghc/ghc][wip/romes/12935] Work around LLVM assembler bug!

[Rodrigo Mesquita (@alt-romes)]

Rodrigo Mesquita pushed to branch wip/romes/12935 at Glasgow Haskell Compiler / GHC


Commits:
f1c12a63 by Rodrigo Mesquita at 2024-05-24T20:43:44+01:00
Work around LLVM assembler bug!

In a really stupid way)

- - - - -


3 changed files:

- compiler/GHC/Cmm/CLabel.hs
- compiler/GHC/Cmm/Dataflow/Label.hs
- compiler/GHC/Cmm/UniqueRenamer.hs


Changes:

=====================================
compiler/GHC/Cmm/CLabel.hs
=====================================
@@ -1896,7 +1896,7 @@ mapInternalNonDetUniques f = \case
   -- ROMES:TODO: what about `RtsApFast NonDetFastString`?
   RtsLabel rtsLblInfo -> pure $ RtsLabel rtsLblInfo
   LocalBlockLabel unique -> LocalBlockLabel <$> f unique
-  fl at ForeignLabel{} -> pure fl 
+  fl at ForeignLabel{} -> pure fl
   AsmTempLabel unique -> AsmTempLabel <$> f unique
   AsmTempDerivedLabel clbl fs -> AsmTempDerivedLabel <$> mapInternalNonDetUniques f clbl <*> pure fs
   StringLitLabel unique -> StringLitLabel <$> f unique


=====================================
compiler/GHC/Cmm/Dataflow/Label.hs
=====================================
@@ -279,6 +279,7 @@ mapToList (LM m) = [(mkHooplLabel k, v) | (k, v) <- M.toList m]
 mapFromList :: [(Label, v)] -> LabelMap v
 mapFromList assocs = LM (M.fromList [(lblToUnique k, v) | (k, v) <- assocs])
 
+{-# INLINE mapFromListWith #-}
 mapFromListWith :: (v -> v -> v) -> [(Label, v)] -> LabelMap v
 mapFromListWith f assocs = LM (M.fromListWith f [(lblToUnique k, v) | (k, v) <- assocs])
 


=====================================
compiler/GHC/Cmm/UniqueRenamer.hs
=====================================
@@ -49,7 +49,13 @@ instance Outputable DetUniqFM where
 type DetRnM = State DetUniqFM
 
 emptyDetUFM :: DetUniqFM
-emptyDetUFM = DetUniqFM { mapping = emptyUFM, supply = 1 }
+emptyDetUFM = DetUniqFM
+  { mapping = emptyUFM
+  -- NB: A lower initial value can get us label `Lsl` which is not parsed
+  -- correctly in older versions of LLVM assembler (llvm-project#80571)
+  -- So we use a x s.t. w64ToBase62 x > "R" > "L" > "r" > "l"
+  , supply = 54
+  }
 
 renameDetUniq :: Unique -> DetRnM Unique
 renameDetUniq uq = do
@@ -61,7 +67,10 @@ renameDetUniq uq = do
           det_uniq = mkUnique tag new_w
       modify' (\DetUniqFM{mapping, supply} ->
         -- Update supply and mapping
-        DetUniqFM{mapping = addToUFM mapping uq det_uniq, supply = supply + 1})
+        DetUniqFM
+          { mapping = addToUFM mapping uq det_uniq
+          , supply = supply + 1
+          })
       return det_uniq
     Just det_uniq ->
       return det_uniq
@@ -138,7 +147,7 @@ instance UniqRenamable CmmLit where
 instance UniqRenamable a {- for 'Body' and on 'RawCmmStatics' -}
   => UniqRenamable (LabelMap a) where
     -- ROMES:TODO: Can a rename of the map have collisions and we lose values? Think harder...
-  uniqRename lm = mapFromList <$> traverse (\(l,x) -> (,) <$> uniqRename l <*> uniqRename x) (mapToList lm)
+  uniqRename lm = mapFromListWith (\_ _ -> error "very bad") <$> traverse (\(l,x) -> (,) <$> uniqRename l <*> uniqRename x) (mapToList lm)
 
 instance UniqRenamable CmmGraph where
   uniqRename (CmmGraph e g) = CmmGraph <$> uniqRename e <*> uniqRename g
@@ -158,7 +167,6 @@ instance UniqRenamable (Block CmmNode n m) where
     BlockCO n bn -> BlockCO <$> uniqRename n <*> uniqRename bn
     BlockCC n1 bn n2 -> BlockCC <$> uniqRename n1 <*> uniqRename bn <*> uniqRename n2
     BlockOC bn n -> BlockOC <$> uniqRename bn <*> uniqRename n
-                                                             
     BNil    -> pure BNil
     BMiddle n -> BMiddle <$> uniqRename n
     BCat    b1 b2 -> BCat <$> uniqRename b1 <*> uniqRename b2



View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/f1c12a63b9fcfffe23ac5466b44d37ce01c9313c

-- 
View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/f1c12a63b9fcfffe23ac5466b44d37ce01c9313c
You're receiving this email because of your account on gitlab.haskell.org.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/ghc-commits/attachments/20240524/cb83ee1f/attachment-0001.html>