Re-add more primops for atomic ops on byte arrays

This is the second attempt to add this functionality. The first
attempt was reverted in 950fcae46a82569e7cd1fba1637a23b419e00ecd, due
to register allocator failure on x86. Given how the register
allocator currently works, we don't have enough registers on x86 to
support cmpxchg using complicated addressing modes. Instead we fall
back to a simpler addressing mode on x86.

Adds the following primops:

 * atomicReadIntArray#
 * atomicWriteIntArray#
 * fetchSubIntArray#
 * fetchOrIntArray#
 * fetchXorIntArray#
 * fetchAndIntArray#

Makes these pre-existing out-of-line primops inline:

 * fetchAddIntArray#
 * casIntArray#

1 files changed, 19 insertions, 0 deletions

diff --git a/compiler/cmm/CmmMachOp.hs b/compiler/cmm/CmmMachOp.hs
index c4ec393ad6..d8ce492de1 100644
--- a/compiler/cmm/CmmMachOp.hs
+++ b/compiler/cmm/CmmMachOp.hs
@@ -19,6 +19,9 @@ module CmmMachOp
     -- CallishMachOp
     , CallishMachOp(..), callishMachOpHints
     , pprCallishMachOp
+
+    -- Atomic read-modify-write
+    , AtomicMachOp(..)
    )
 where
 
@@ -547,8 +550,24 @@ data CallishMachOp
 
   | MO_PopCnt Width
   | MO_BSwap Width
+
+  -- Atomic read-modify-write.
+  | MO_AtomicRMW Width AtomicMachOp
+  | MO_AtomicRead Width
+  | MO_AtomicWrite Width
+  | MO_Cmpxchg Width
   deriving (Eq, Show)
 
+-- | The operation to perform atomically.
+data AtomicMachOp =
+      AMO_Add
+    | AMO_Sub
+    | AMO_And
+    | AMO_Nand
+    | AMO_Or
+    | AMO_Xor
+      deriving (Eq, Show)
+
 pprCallishMachOp :: CallishMachOp -> SDoc
 pprCallishMachOp mo = text (show mo)