[Git][ghc/ghc][wip/T16742] PrelRules: Don't break let/app invariant in shiftRule

Ben Gamari gitlab at gitlab.haskell.org
Mon Jun 10 13:32:45 UTC 2019



Ben Gamari pushed to branch wip/T16742 at Glasgow Haskell Compiler / GHC


Commits:
87b21352 by Ben Gamari at 2019-06-10T13:32:36Z
PrelRules: Don't break let/app invariant in shiftRule

Previously shiftRule would rewrite as invalid shift like
```
let x = I# (uncheckedIShiftL# n 80)
in ...
```
to
```
let x = I# (error "invalid shift")
in ...
```
However, this breaks the let/app invariant as `error` is not
okay-for-speculation. There isn't an easy way to avoid this so let's not
try. Instead we just take advantage of the undefined nature of invalid
shifts and return zero.

- - - - -


1 changed file:

- compiler/prelude/PrelRules.hs


Changes:

=====================================
compiler/prelude/PrelRules.hs
=====================================
@@ -475,8 +475,7 @@ shiftRule shift_op
              -> return e1
              -- See Note [Guarding against silly shifts]
              | shift_len < 0 || shift_len > wordSizeInBits dflags
-             -> return $ mkRuntimeErrorApp rUNTIME_ERROR_ID wordPrimTy
-                           ("Bad shift length " ++ show shift_len)
+             -> return $ mkLitNumberWrap dflags nt 0 (exprType e1)
 
            -- Do the shift at type Integer, but shift length is Int
            Lit (LitNumber nt x t)
@@ -701,7 +700,23 @@ can't constant fold it, but if it gets to the assember we get
      Error: operand type mismatch for `shl'
 
 So the best thing to do is to rewrite the shift with a call to error,
-when the second arg is stupid.
+when the second arg is large. However, in general we cannot do this; consider
+this case
+
+    let x = I# (uncheckedIShiftL# n 80)
+    in ...
+
+Here x contains an invalid shift and consequently we would like to rewrite it
+as follows:
+
+    let x = I# (error "invalid shift)
+    in ...
+
+This was originally done in the fix to #16449 but this breaks the let/app
+invariant (see Note [CoreSyn let/app invariant] in CoreSyn) as noted in #16742.
+Consequently we instead take advantage of the fact that large shifts are
+undefined behavior and transform the invalid shift into an "obviously
+incorrect" value.
 
 There are two cases:
 



View it on GitLab: https://gitlab.haskell.org/ghc/ghc/commit/87b213520d22de4d0bca831c0ac75f8a8468ca59

-- 
View it on GitLab: https://gitlab.haskell.org/ghc/ghc/commit/87b213520d22de4d0bca831c0ac75f8a8468ca59
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/20190610/c352ec4d/attachment-0001.html>


More information about the ghc-commits mailing list