[Git][ghc/ghc][wip/andreask/bound_cmm_folding] Cmm constant folding: Narrow results to operations bitwidth.

[Andreas Klebinger (@AndreasK)]

Andreas Klebinger pushed to branch wip/andreask/bound_cmm_folding at Glasgow Haskell Compiler / GHC


Commits:
eae742da by Andreas Klebinger at 2024-07-01T15:37:32+02:00
Cmm constant folding: Narrow results to operations bitwidth.

When constant folding operations on literals ensure the result
is still within bounds by explicitly narrowing the results.

Not doing so results in a lot of spurious assembler warnings
especially when testing primops.

- - - - -


3 changed files:

- compiler/GHC/Cmm/Opt.hs
- + testsuite/tests/cmm/opt/T24556.cmm
- testsuite/tests/cmm/opt/all.T


Changes:

=====================================
compiler/GHC/Cmm/Opt.hs
=====================================
@@ -50,6 +50,7 @@ constantFoldExprOpt e = wrapRecExpOpt f e
           CmmMachOp op' args' -> fromMaybe (CmmMachOp op' args') <$> cmmMachOpFoldOptM cfg op' args'
           e -> pure e
     f (CmmRegOff r 0) = pure (CmmReg r)
+    f (CmmLit (CmmInt x rep)) = pure (CmmLit $ CmmInt (narrowU rep x) rep)
     f e = pure e
 
 constantFoldExpr :: Platform -> CmmExpr -> CmmExpr
@@ -82,7 +83,7 @@ cmmMachOpFoldM
 
 cmmMachOpFoldM _ op [CmmLit (CmmInt x rep)]
   = Just $! case op of
-      MO_S_Neg _ -> CmmLit (CmmInt (-x) rep)
+      MO_S_Neg _ -> CmmLit (CmmInt (narrowS rep (-x)) rep)
       MO_Not _   -> CmmLit (CmmInt (complement x) rep)
 
         -- these are interesting: we must first narrow to the
@@ -159,9 +160,9 @@ cmmMachOpFoldM platform mop [CmmLit (CmmInt x xrep), CmmLit (CmmInt y _)]
         MO_S_Lt _ -> Just $! CmmLit (CmmInt (if x_s <  y_s then 1 else 0) (wordWidth platform))
         MO_S_Le _ -> Just $! CmmLit (CmmInt (if x_s <= y_s then 1 else 0) (wordWidth platform))
 
-        MO_Add r -> Just $! CmmLit (CmmInt (x + y) r)
-        MO_Sub r -> Just $! CmmLit (CmmInt (x - y) r)
-        MO_Mul r -> Just $! CmmLit (CmmInt (x * y) r)
+        MO_Add r -> Just $! CmmLit (CmmInt (narrowU r $ x + y) r)
+        MO_Sub r -> Just $! CmmLit (CmmInt (narrowS r $ x - y) r)
+        MO_Mul r -> Just $! CmmLit (CmmInt (narrowU r $ x * y) r)
         MO_U_Quot r | y /= 0 -> Just $! CmmLit (CmmInt (x_u `quot` y_u) r)
         MO_U_Rem  r | y /= 0 -> Just $! CmmLit (CmmInt (x_u `rem`  y_u) r)
         MO_S_Quot r | y /= 0 -> Just $! CmmLit (CmmInt (x_s `quot` y_s) r)
@@ -171,7 +172,7 @@ cmmMachOpFoldM platform mop [CmmLit (CmmInt x xrep), CmmLit (CmmInt y _)]
         MO_Or    r -> Just $! CmmLit (CmmInt (x .|. y) r)
         MO_Xor   r -> Just $! CmmLit (CmmInt (x `xor` y) r)
 
-        MO_Shl   r -> Just $! CmmLit (CmmInt (x   `shiftL` fromIntegral y) r)
+        MO_Shl   r -> Just $! CmmLit (CmmInt (narrowU r $ x   `shiftL` fromIntegral y) r)
         MO_U_Shr r -> Just $! CmmLit (CmmInt (x_u `shiftR` fromIntegral y) r)
         MO_S_Shr r -> Just $! CmmLit (CmmInt (x_s `shiftR` fromIntegral y) r)
 


=====================================
testsuite/tests/cmm/opt/T24556.cmm
=====================================
@@ -0,0 +1,12 @@
+#include "Cmm.h"
+
+func(bits64 buffer) {
+  I8[buffer]   = %lobits8(255 + 45);
+  I8[buffer+1] = %lobits8(310 - 10);
+  I8[buffer+2] = %lobits8(30  * 10);
+  I8[buffer+3] = %lobits8(150 << 1);
+  // This one comes from test-primops
+  I64[buffer+4] = %zx64(((1 :: bits16) & ((1 :: bits16) & (((516 :: bits16) * (154 :: bits16)) + bits16[buffer + (0 :: bits64)]))));
+  return(1);
+}
+


=====================================
testsuite/tests/cmm/opt/all.T
=====================================
@@ -3,3 +3,8 @@
 test('T15188', cmm_src, makefile_test, [])
 test('T18141', normal, compile, [''])
 test('T20142', normal, compile, [''])
+
+# Cmm opt should not produce oversized literals in the assembly output.
+# We check this by telling the assembler to exit on warnings.
+test('T24556', [only_ways('optasm'), cmm_src], compile, ['-O -opta -Xassembler -opta --fatal-warnings'])
+



View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/eae742da78e538205ce335ab9f521afcdb390d78

-- 
View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/eae742da78e538205ce335ab9f521afcdb390d78
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/20240701/99419326/attachment-0001.html>