MachOp.hs 20.9 KB
Newer Older
1 2
{-# OPTIONS_GHC -Wno-incomplete-uni-patterns #-}

3
module GHC.Cmm.MachOp
4 5
    ( MachOp(..)
    , pprMachOp, isCommutableMachOp, isAssociativeMachOp
6
    , isComparisonMachOp, maybeIntComparison, machOpResultType
7
    , machOpArgReps, maybeInvertComparison, isFloatComparison
8 9 10 11 12 13

    -- MachOp builders
    , mo_wordAdd, mo_wordSub, mo_wordEq, mo_wordNe,mo_wordMul, mo_wordSQuot
    , mo_wordSRem, mo_wordSNeg, mo_wordUQuot, mo_wordURem
    , mo_wordSGe, mo_wordSLe, mo_wordSGt, mo_wordSLt, mo_wordUGe
    , mo_wordULe, mo_wordUGt, mo_wordULt
14 15
    , mo_wordAnd, mo_wordOr, mo_wordXor, mo_wordNot
    , mo_wordShl, mo_wordSShr, mo_wordUShr
16
    , mo_u_8To32, mo_s_8To32, mo_u_16To32, mo_s_16To32
17 18
    , mo_u_8ToWord, mo_s_8ToWord, mo_u_16ToWord, mo_s_16ToWord
    , mo_u_32ToWord, mo_s_32ToWord
19
    , mo_32To8, mo_32To16, mo_WordTo8, mo_WordTo16, mo_WordTo32, mo_WordTo64
20 21

    -- CallishMachOp
22
    , CallishMachOp(..), callishMachOpHints
23
    , pprCallishMachOp
24
    , machOpMemcpyishAlign
25 26 27

    -- Atomic read-modify-write
    , AtomicMachOp(..)
28 29 30
   )
where

31 32
import GhcPrelude

33
import GHC.Cmm.Type
34
import Outputable
Sylvain Henry's avatar
Sylvain Henry committed
35
import GHC.Driver.Session
36 37 38 39 40 41 42

-----------------------------------------------------------------------------
--              MachOp
-----------------------------------------------------------------------------

{- |
Machine-level primops; ones which we can reasonably delegate to the
Simon Marlow's avatar
Simon Marlow committed
43
native code generators to handle.
44

Simon Marlow's avatar
Simon Marlow committed
45 46 47
Most operations are parameterised by the 'Width' that they operate on.
Some operations have separate signed and unsigned versions, and float
and integer versions.
48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111
-}

data MachOp
  -- Integer operations (insensitive to signed/unsigned)
  = MO_Add Width
  | MO_Sub Width
  | MO_Eq  Width
  | MO_Ne  Width
  | MO_Mul Width                -- low word of multiply

  -- Signed multiply/divide
  | MO_S_MulMayOflo Width       -- nonzero if signed multiply overflows
  | MO_S_Quot Width             -- signed / (same semantics as IntQuotOp)
  | MO_S_Rem  Width             -- signed % (same semantics as IntRemOp)
  | MO_S_Neg  Width             -- unary -

  -- Unsigned multiply/divide
  | MO_U_MulMayOflo Width       -- nonzero if unsigned multiply overflows
  | MO_U_Quot Width             -- unsigned / (same semantics as WordQuotOp)
  | MO_U_Rem  Width             -- unsigned % (same semantics as WordRemOp)

  -- Signed comparisons
  | MO_S_Ge Width
  | MO_S_Le Width
  | MO_S_Gt Width
  | MO_S_Lt Width

  -- Unsigned comparisons
  | MO_U_Ge Width
  | MO_U_Le Width
  | MO_U_Gt Width
  | MO_U_Lt Width

  -- Floating point arithmetic
  | MO_F_Add  Width
  | MO_F_Sub  Width
  | MO_F_Neg  Width             -- unary -
  | MO_F_Mul  Width
  | MO_F_Quot Width

  -- Floating point comparison
  | MO_F_Eq Width
  | MO_F_Ne Width
  | MO_F_Ge Width
  | MO_F_Le Width
  | MO_F_Gt Width
  | MO_F_Lt Width

  -- Bitwise operations.  Not all of these may be supported
  -- at all sizes, and only integral Widths are valid.
  | MO_And   Width
  | MO_Or    Width
  | MO_Xor   Width
  | MO_Not   Width
  | MO_Shl   Width
  | MO_U_Shr Width      -- unsigned shift right
  | MO_S_Shr Width      -- signed shift right

  -- Conversions.  Some of these will be NOPs.
  -- Floating-point conversions use the signed variant.
  | MO_SF_Conv Width Width      -- Signed int -> Float
  | MO_FS_Conv Width Width      -- Float -> Signed int
  | MO_SS_Conv Width Width      -- Signed int -> Signed int
  | MO_UU_Conv Width Width      -- unsigned int -> unsigned int
Michal Terepeta's avatar
Michal Terepeta committed
112 113 114 115 116 117 118 119
  | MO_XX_Conv Width Width      -- int -> int; puts no requirements on the
                                -- contents of upper bits when extending;
                                -- narrowing is simply truncation; the only
                                -- expectation is that we can recover the
                                -- original value by applying the opposite
                                -- MO_XX_Conv, e.g.,
                                --   MO_XX_CONV W64 W8 (MO_XX_CONV W8 W64 x)
                                -- is equivalent to just x.
120
  | MO_FF_Conv Width Width      -- Float -> Float
121 122 123 124

  -- Vector element insertion and extraction operations
  | MO_V_Insert  Length Width   -- Insert scalar into vector
  | MO_V_Extract Length Width   -- Extract scalar from vector
125

126
  -- Integer vector operations
127 128
  | MO_V_Add Length Width
  | MO_V_Sub Length Width
129 130 131 132 133 134
  | MO_V_Mul Length Width

  -- Signed vector multiply/divide
  | MO_VS_Quot Length Width
  | MO_VS_Rem  Length Width
  | MO_VS_Neg  Length Width
135

136 137 138 139
  -- Unsigned vector multiply/divide
  | MO_VU_Quot Length Width
  | MO_VU_Rem  Length Width

140
  -- Floating point vector element insertion and extraction operations
141 142
  | MO_VF_Insert  Length Width   -- Insert scalar into vector
  | MO_VF_Extract Length Width   -- Extract scalar from vector
143

144
  -- Floating point vector operations
145 146
  | MO_VF_Add  Length Width
  | MO_VF_Sub  Length Width
147
  | MO_VF_Neg  Length Width      -- unary negation
148 149
  | MO_VF_Mul  Length Width
  | MO_VF_Quot Length Width
150 151 152

  -- Alignment check (for -falignment-sanitisation)
  | MO_AlignmentCheck Int Width
153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173
  deriving (Eq, Show)

pprMachOp :: MachOp -> SDoc
pprMachOp mo = text (show mo)



-- -----------------------------------------------------------------------------
-- Some common MachReps

-- A 'wordRep' is a machine word on the target architecture
-- Specifically, it is the size of an Int#, Word#, Addr#
-- and the unit of allocation on the stack and the heap
-- Any pointer is also guaranteed to be a wordRep.

mo_wordAdd, mo_wordSub, mo_wordEq, mo_wordNe,mo_wordMul, mo_wordSQuot
    , mo_wordSRem, mo_wordSNeg, mo_wordUQuot, mo_wordURem
    , mo_wordSGe, mo_wordSLe, mo_wordSGt, mo_wordSLt, mo_wordUGe
    , mo_wordULe, mo_wordUGt, mo_wordULt
    , mo_wordAnd, mo_wordOr, mo_wordXor, mo_wordNot, mo_wordShl, mo_wordSShr, mo_wordUShr
    , mo_u_8ToWord, mo_s_8ToWord, mo_u_16ToWord, mo_s_16ToWord, mo_u_32ToWord, mo_s_32ToWord
174 175 176 177 178
    , mo_WordTo8, mo_WordTo16, mo_WordTo32, mo_WordTo64
    :: DynFlags -> MachOp

mo_u_8To32, mo_s_8To32, mo_u_16To32, mo_s_16To32
    , mo_32To8, mo_32To16
179 180
    :: MachOp

181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228
mo_wordAdd      dflags = MO_Add (wordWidth dflags)
mo_wordSub      dflags = MO_Sub (wordWidth dflags)
mo_wordEq       dflags = MO_Eq  (wordWidth dflags)
mo_wordNe       dflags = MO_Ne  (wordWidth dflags)
mo_wordMul      dflags = MO_Mul (wordWidth dflags)
mo_wordSQuot    dflags = MO_S_Quot (wordWidth dflags)
mo_wordSRem     dflags = MO_S_Rem (wordWidth dflags)
mo_wordSNeg     dflags = MO_S_Neg (wordWidth dflags)
mo_wordUQuot    dflags = MO_U_Quot (wordWidth dflags)
mo_wordURem     dflags = MO_U_Rem (wordWidth dflags)

mo_wordSGe      dflags = MO_S_Ge  (wordWidth dflags)
mo_wordSLe      dflags = MO_S_Le  (wordWidth dflags)
mo_wordSGt      dflags = MO_S_Gt  (wordWidth dflags)
mo_wordSLt      dflags = MO_S_Lt  (wordWidth dflags)

mo_wordUGe      dflags = MO_U_Ge  (wordWidth dflags)
mo_wordULe      dflags = MO_U_Le  (wordWidth dflags)
mo_wordUGt      dflags = MO_U_Gt  (wordWidth dflags)
mo_wordULt      dflags = MO_U_Lt  (wordWidth dflags)

mo_wordAnd      dflags = MO_And (wordWidth dflags)
mo_wordOr       dflags = MO_Or  (wordWidth dflags)
mo_wordXor      dflags = MO_Xor (wordWidth dflags)
mo_wordNot      dflags = MO_Not (wordWidth dflags)
mo_wordShl      dflags = MO_Shl (wordWidth dflags)
mo_wordSShr     dflags = MO_S_Shr (wordWidth dflags)
mo_wordUShr     dflags = MO_U_Shr (wordWidth dflags)

mo_u_8To32             = MO_UU_Conv W8 W32
mo_s_8To32             = MO_SS_Conv W8 W32
mo_u_16To32            = MO_UU_Conv W16 W32
mo_s_16To32            = MO_SS_Conv W16 W32

mo_u_8ToWord    dflags = MO_UU_Conv W8  (wordWidth dflags)
mo_s_8ToWord    dflags = MO_SS_Conv W8  (wordWidth dflags)
mo_u_16ToWord   dflags = MO_UU_Conv W16 (wordWidth dflags)
mo_s_16ToWord   dflags = MO_SS_Conv W16 (wordWidth dflags)
mo_s_32ToWord   dflags = MO_SS_Conv W32 (wordWidth dflags)
mo_u_32ToWord   dflags = MO_UU_Conv W32 (wordWidth dflags)

mo_WordTo8      dflags = MO_UU_Conv (wordWidth dflags) W8
mo_WordTo16     dflags = MO_UU_Conv (wordWidth dflags) W16
mo_WordTo32     dflags = MO_UU_Conv (wordWidth dflags) W32
mo_WordTo64     dflags = MO_UU_Conv (wordWidth dflags) W64

mo_32To8               = MO_UU_Conv W32 W8
mo_32To16              = MO_UU_Conv W32 W16
229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252


-- ----------------------------------------------------------------------------
-- isCommutableMachOp

{- |
Returns 'True' if the MachOp has commutable arguments.  This is used
in the platform-independent Cmm optimisations.

If in doubt, return 'False'.  This generates worse code on the
native routes, but is otherwise harmless.
-}
isCommutableMachOp :: MachOp -> Bool
isCommutableMachOp mop =
  case mop of
        MO_Add _                -> True
        MO_Eq _                 -> True
        MO_Ne _                 -> True
        MO_Mul _                -> True
        MO_S_MulMayOflo _       -> True
        MO_U_MulMayOflo _       -> True
        MO_And _                -> True
        MO_Or _                 -> True
        MO_Xor _                -> True
253 254
        MO_F_Add _              -> True
        MO_F_Mul _              -> True
255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276
        _other                  -> False

-- ----------------------------------------------------------------------------
-- isAssociativeMachOp

{- |
Returns 'True' if the MachOp is associative (i.e. @(x+y)+z == x+(y+z)@)
This is used in the platform-independent Cmm optimisations.

If in doubt, return 'False'.  This generates worse code on the
native routes, but is otherwise harmless.
-}
isAssociativeMachOp :: MachOp -> Bool
isAssociativeMachOp mop =
  case mop of
        MO_Add {} -> True       -- NB: does not include
        MO_Mul {} -> True --     floatint point!
        MO_And {} -> True
        MO_Or  {} -> True
        MO_Xor {} -> True
        _other    -> False

277

278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307
-- ----------------------------------------------------------------------------
-- isComparisonMachOp

{- |
Returns 'True' if the MachOp is a comparison.

If in doubt, return False.  This generates worse code on the
native routes, but is otherwise harmless.
-}
isComparisonMachOp :: MachOp -> Bool
isComparisonMachOp mop =
  case mop of
    MO_Eq   _  -> True
    MO_Ne   _  -> True
    MO_S_Ge _  -> True
    MO_S_Le _  -> True
    MO_S_Gt _  -> True
    MO_S_Lt _  -> True
    MO_U_Ge _  -> True
    MO_U_Le _  -> True
    MO_U_Gt _  -> True
    MO_U_Lt _  -> True
    MO_F_Eq {} -> True
    MO_F_Ne {} -> True
    MO_F_Ge {} -> True
    MO_F_Le {} -> True
    MO_F_Gt {} -> True
    MO_F_Lt {} -> True
    _other     -> False

308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326
{- |
Returns @Just w@ if the operation is an integer comparison with width
@w@, or @Nothing@ otherwise.
-}
maybeIntComparison :: MachOp -> Maybe Width
maybeIntComparison mop =
  case mop of
    MO_Eq   w  -> Just w
    MO_Ne   w  -> Just w
    MO_S_Ge w  -> Just w
    MO_S_Le w  -> Just w
    MO_S_Gt w  -> Just w
    MO_S_Lt w  -> Just w
    MO_U_Ge w  -> Just w
    MO_U_Le w  -> Just w
    MO_U_Gt w  -> Just w
    MO_U_Lt w  -> Just w
    _ -> Nothing

327 328 329 330 331 332 333 334 335 336 337
isFloatComparison :: MachOp -> Bool
isFloatComparison mop =
  case mop of
    MO_F_Eq {} -> True
    MO_F_Ne {} -> True
    MO_F_Ge {} -> True
    MO_F_Le {} -> True
    MO_F_Gt {} -> True
    MO_F_Lt {} -> True
    _other     -> False

338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367
-- -----------------------------------------------------------------------------
-- Inverting conditions

-- Sometimes it's useful to be able to invert the sense of a
-- condition.  Not all conditional tests are invertible: in
-- particular, floating point conditionals cannot be inverted, because
-- there exist floating-point values which return False for both senses
-- of a condition (eg. !(NaN > NaN) && !(NaN /<= NaN)).

maybeInvertComparison :: MachOp -> Maybe MachOp
maybeInvertComparison op
  = case op of  -- None of these Just cases include floating point
        MO_Eq r   -> Just (MO_Ne r)
        MO_Ne r   -> Just (MO_Eq r)
        MO_U_Lt r -> Just (MO_U_Ge r)
        MO_U_Gt r -> Just (MO_U_Le r)
        MO_U_Le r -> Just (MO_U_Gt r)
        MO_U_Ge r -> Just (MO_U_Lt r)
        MO_S_Lt r -> Just (MO_S_Ge r)
        MO_S_Gt r -> Just (MO_S_Le r)
        MO_S_Le r -> Just (MO_S_Gt r)
        MO_S_Ge r -> Just (MO_S_Lt r)
        _other    -> Nothing

-- ----------------------------------------------------------------------------
-- machOpResultType

{- |
Returns the MachRep of the result of a MachOp.
-}
368 369
machOpResultType :: DynFlags -> MachOp -> [CmmType] -> CmmType
machOpResultType dflags mop tys =
370 371 372 373 374 375 376 377 378 379 380 381
  case mop of
    MO_Add {}           -> ty1  -- Preserve GC-ptr-hood
    MO_Sub {}           -> ty1  -- of first arg
    MO_Mul    r         -> cmmBits r
    MO_S_MulMayOflo r   -> cmmBits r
    MO_S_Quot r         -> cmmBits r
    MO_S_Rem  r         -> cmmBits r
    MO_S_Neg  r         -> cmmBits r
    MO_U_MulMayOflo r   -> cmmBits r
    MO_U_Quot r         -> cmmBits r
    MO_U_Rem  r         -> cmmBits r

382 383 384 385 386 387
    MO_Eq {}            -> comparisonResultRep dflags
    MO_Ne {}            -> comparisonResultRep dflags
    MO_S_Ge {}          -> comparisonResultRep dflags
    MO_S_Le {}          -> comparisonResultRep dflags
    MO_S_Gt {}          -> comparisonResultRep dflags
    MO_S_Lt {}          -> comparisonResultRep dflags
388

389 390 391 392
    MO_U_Ge {}          -> comparisonResultRep dflags
    MO_U_Le {}          -> comparisonResultRep dflags
    MO_U_Gt {}          -> comparisonResultRep dflags
    MO_U_Lt {}          -> comparisonResultRep dflags
393 394 395 396 397 398

    MO_F_Add r          -> cmmFloat r
    MO_F_Sub r          -> cmmFloat r
    MO_F_Mul r          -> cmmFloat r
    MO_F_Quot r         -> cmmFloat r
    MO_F_Neg r          -> cmmFloat r
399 400 401 402 403 404
    MO_F_Eq  {}         -> comparisonResultRep dflags
    MO_F_Ne  {}         -> comparisonResultRep dflags
    MO_F_Ge  {}         -> comparisonResultRep dflags
    MO_F_Le  {}         -> comparisonResultRep dflags
    MO_F_Gt  {}         -> comparisonResultRep dflags
    MO_F_Lt  {}         -> comparisonResultRep dflags
405 406 407 408 409 410 411 412 413 414 415

    MO_And {}           -> ty1  -- Used for pointer masking
    MO_Or {}            -> ty1
    MO_Xor {}           -> ty1
    MO_Not   r          -> cmmBits r
    MO_Shl   r          -> cmmBits r
    MO_U_Shr r          -> cmmBits r
    MO_S_Shr r          -> cmmBits r

    MO_SS_Conv _ to     -> cmmBits to
    MO_UU_Conv _ to     -> cmmBits to
Michal Terepeta's avatar
Michal Terepeta committed
416
    MO_XX_Conv _ to     -> cmmBits to
417 418 419
    MO_FS_Conv _ to     -> cmmBits to
    MO_SF_Conv _ to     -> cmmFloat to
    MO_FF_Conv _ to     -> cmmFloat to
420

421 422 423 424 425 426 427 428 429 430 431
    MO_V_Insert  l w    -> cmmVec l (cmmBits w)
    MO_V_Extract _ w    -> cmmBits w

    MO_V_Add l w        -> cmmVec l (cmmBits w)
    MO_V_Sub l w        -> cmmVec l (cmmBits w)
    MO_V_Mul l w        -> cmmVec l (cmmBits w)

    MO_VS_Quot l w      -> cmmVec l (cmmBits w)
    MO_VS_Rem  l w      -> cmmVec l (cmmBits w)
    MO_VS_Neg  l w      -> cmmVec l (cmmBits w)

432 433 434
    MO_VU_Quot l w      -> cmmVec l (cmmBits w)
    MO_VU_Rem  l w      -> cmmVec l (cmmBits w)

435 436 437 438 439 440 441 442
    MO_VF_Insert  l w   -> cmmVec l (cmmFloat w)
    MO_VF_Extract _ w   -> cmmFloat w

    MO_VF_Add  l w      -> cmmVec l (cmmFloat w)
    MO_VF_Sub  l w      -> cmmVec l (cmmFloat w)
    MO_VF_Mul  l w      -> cmmVec l (cmmFloat w)
    MO_VF_Quot l w      -> cmmVec l (cmmFloat w)
    MO_VF_Neg  l w      -> cmmVec l (cmmFloat w)
443 444

    MO_AlignmentCheck _ _ -> ty1
445 446 447
  where
    (ty1:_) = tys

448
comparisonResultRep :: DynFlags -> CmmType
449 450 451 452 453 454 455 456 457 458 459
comparisonResultRep = bWord  -- is it?


-- -----------------------------------------------------------------------------
-- machOpArgReps

-- | This function is used for debugging only: we can check whether an
-- application of a MachOp is "type-correct" by checking that the MachReps of
-- its arguments are the same as the MachOp expects.  This is used when
-- linting a CmmExpr.

460 461
machOpArgReps :: DynFlags -> MachOp -> [Width]
machOpArgReps dflags op =
462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501
  case op of
    MO_Add    r         -> [r,r]
    MO_Sub    r         -> [r,r]
    MO_Eq     r         -> [r,r]
    MO_Ne     r         -> [r,r]
    MO_Mul    r         -> [r,r]
    MO_S_MulMayOflo r   -> [r,r]
    MO_S_Quot r         -> [r,r]
    MO_S_Rem  r         -> [r,r]
    MO_S_Neg  r         -> [r]
    MO_U_MulMayOflo r   -> [r,r]
    MO_U_Quot r         -> [r,r]
    MO_U_Rem  r         -> [r,r]

    MO_S_Ge r           -> [r,r]
    MO_S_Le r           -> [r,r]
    MO_S_Gt r           -> [r,r]
    MO_S_Lt r           -> [r,r]

    MO_U_Ge r           -> [r,r]
    MO_U_Le r           -> [r,r]
    MO_U_Gt r           -> [r,r]
    MO_U_Lt r           -> [r,r]

    MO_F_Add r          -> [r,r]
    MO_F_Sub r          -> [r,r]
    MO_F_Mul r          -> [r,r]
    MO_F_Quot r         -> [r,r]
    MO_F_Neg r          -> [r]
    MO_F_Eq  r          -> [r,r]
    MO_F_Ne  r          -> [r,r]
    MO_F_Ge  r          -> [r,r]
    MO_F_Le  r          -> [r,r]
    MO_F_Gt  r          -> [r,r]
    MO_F_Lt  r          -> [r,r]

    MO_And   r          -> [r,r]
    MO_Or    r          -> [r,r]
    MO_Xor   r          -> [r,r]
    MO_Not   r          -> [r]
502 503 504
    MO_Shl   r          -> [r, wordWidth dflags]
    MO_U_Shr r          -> [r, wordWidth dflags]
    MO_S_Shr r          -> [r, wordWidth dflags]
505 506 507

    MO_SS_Conv from _   -> [from]
    MO_UU_Conv from _   -> [from]
Michal Terepeta's avatar
Michal Terepeta committed
508
    MO_XX_Conv from _   -> [from]
509 510 511 512
    MO_SF_Conv from _   -> [from]
    MO_FS_Conv from _   -> [from]
    MO_FF_Conv from _   -> [from]

513 514
    MO_V_Insert  l r    -> [typeWidth (vec l (cmmBits r)),r,wordWidth dflags]
    MO_V_Extract l r    -> [typeWidth (vec l (cmmBits r)),wordWidth dflags]
515

516 517 518 519 520 521 522 523
    MO_V_Add _ r        -> [r,r]
    MO_V_Sub _ r        -> [r,r]
    MO_V_Mul _ r        -> [r,r]

    MO_VS_Quot _ r      -> [r,r]
    MO_VS_Rem  _ r      -> [r,r]
    MO_VS_Neg  _ r      -> [r]

524 525 526
    MO_VU_Quot _ r      -> [r,r]
    MO_VU_Rem  _ r      -> [r,r]

527 528
    MO_VF_Insert  l r   -> [typeWidth (vec l (cmmFloat r)),r,wordWidth dflags]
    MO_VF_Extract l r   -> [typeWidth (vec l (cmmFloat r)),wordWidth dflags]
529

530 531 532 533 534
    MO_VF_Add  _ r      -> [r,r]
    MO_VF_Sub  _ r      -> [r,r]
    MO_VF_Mul  _ r      -> [r,r]
    MO_VF_Quot _ r      -> [r,r]
    MO_VF_Neg  _ r      -> [r]
535

536 537
    MO_AlignmentCheck _ r -> [r]

538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556
-----------------------------------------------------------------------------
-- CallishMachOp
-----------------------------------------------------------------------------

-- CallishMachOps tend to be implemented by foreign calls in some backends,
-- so we separate them out.  In Cmm, these can only occur in a
-- statement position, in contrast to an ordinary MachOp which can occur
-- anywhere in an expression.
data CallishMachOp
  = MO_F64_Pwr
  | MO_F64_Sin
  | MO_F64_Cos
  | MO_F64_Tan
  | MO_F64_Sinh
  | MO_F64_Cosh
  | MO_F64_Tanh
  | MO_F64_Asin
  | MO_F64_Acos
  | MO_F64_Atan
557 558 559
  | MO_F64_Asinh
  | MO_F64_Acosh
  | MO_F64_Atanh
560
  | MO_F64_Log
chessai's avatar
chessai committed
561
  | MO_F64_Log1P
562
  | MO_F64_Exp
chessai's avatar
chessai committed
563
  | MO_F64_ExpM1
564
  | MO_F64_Fabs
565 566 567 568 569 570 571 572 573 574 575
  | MO_F64_Sqrt
  | MO_F32_Pwr
  | MO_F32_Sin
  | MO_F32_Cos
  | MO_F32_Tan
  | MO_F32_Sinh
  | MO_F32_Cosh
  | MO_F32_Tanh
  | MO_F32_Asin
  | MO_F32_Acos
  | MO_F32_Atan
576 577 578
  | MO_F32_Asinh
  | MO_F32_Acosh
  | MO_F32_Atanh
579
  | MO_F32_Log
chessai's avatar
chessai committed
580
  | MO_F32_Log1P
581
  | MO_F32_Exp
chessai's avatar
chessai committed
582
  | MO_F32_ExpM1
583
  | MO_F32_Fabs
584
  | MO_F32_Sqrt
585

586 587
  | MO_UF_Conv Width

Sylvain Henry's avatar
Sylvain Henry committed
588
  | MO_S_Mul2    Width
589
  | MO_S_QuotRem Width
590
  | MO_U_QuotRem Width
Ian Lynagh's avatar
Ian Lynagh committed
591
  | MO_U_QuotRem2 Width
Ian Lynagh's avatar
Ian Lynagh committed
592
  | MO_Add2      Width
Sebastian Graf's avatar
Sebastian Graf committed
593
  | MO_AddWordC  Width
nkaretnikov's avatar
nkaretnikov committed
594
  | MO_SubWordC  Width
595 596
  | MO_AddIntC   Width
  | MO_SubIntC   Width
Ian Lynagh's avatar
Ian Lynagh committed
597
  | MO_U_Mul2    Width
598

599
  | MO_ReadBarrier
600 601
  | MO_WriteBarrier
  | MO_Touch         -- Keep variables live (when using interior pointers)
602

gmainland's avatar
gmainland committed
603
  -- Prefetch
604
  | MO_Prefetch_Data Int -- Prefetch hint. May change program performance but not
gmainland's avatar
gmainland committed
605
                     -- program behavior.
606 607 608 609 610 611
                     -- the Int can be 0-3. Needs to be known at compile time
                     -- to interact with code generation correctly.
                     --  TODO: add support for prefetch WRITES,
                     --  currently only exposes prefetch reads, which
                     -- would the majority of use cases in ghc anyways

gmainland's avatar
gmainland committed
612

613 614 615 616 617 618
  -- These three MachOps are parameterised by the known alignment
  -- of the destination and source (for memcpy/memmove) pointers.
  -- This information may be used for optimisation in backends.
  | MO_Memcpy Int
  | MO_Memset Int
  | MO_Memmove Int
619
  | MO_Memcmp Int
tibbe's avatar
tibbe committed
620 621

  | MO_PopCnt Width
622 623
  | MO_Pdep Width
  | MO_Pext Width
624 625 626
  | MO_Clz Width
  | MO_Ctz Width

627
  | MO_BSwap Width
628
  | MO_BRev Width
629 630 631 632 633 634

  -- Atomic read-modify-write.
  | MO_AtomicRMW Width AtomicMachOp
  | MO_AtomicRead Width
  | MO_AtomicWrite Width
  | MO_Cmpxchg Width
635 636
  deriving (Eq, Show)

637 638 639 640 641 642 643 644 645 646
-- | The operation to perform atomically.
data AtomicMachOp =
      AMO_Add
    | AMO_Sub
    | AMO_And
    | AMO_Nand
    | AMO_Or
    | AMO_Xor
      deriving (Eq, Show)

647 648
pprCallishMachOp :: CallishMachOp -> SDoc
pprCallishMachOp mo = text (show mo)
649

650 651
callishMachOpHints :: CallishMachOp -> ([ForeignHint], [ForeignHint])
callishMachOpHints op = case op of
652 653 654
  MO_Memcpy _  -> ([], [AddrHint,AddrHint,NoHint])
  MO_Memset _  -> ([], [AddrHint,NoHint,NoHint])
  MO_Memmove _ -> ([], [AddrHint,AddrHint,NoHint])
655
  MO_Memcmp _  -> ([], [AddrHint, AddrHint, NoHint])
656
  _            -> ([],[])
657
  -- empty lists indicate NoHint
658 659 660 661 662 663 664

-- | The alignment of a 'memcpy'-ish operation.
machOpMemcpyishAlign :: CallishMachOp -> Maybe Int
machOpMemcpyishAlign op = case op of
  MO_Memcpy  align -> Just align
  MO_Memset  align -> Just align
  MO_Memmove align -> Just align
665
  MO_Memcmp  align -> Just align
666
  _                -> Nothing