packages feed

cuda-0.13.0.0: src/Foreign/CUDA/Driver/Device.chs

{-# LANGUAGE BangPatterns             #-}
{-# LANGUAGE CPP                      #-}
{-# LANGUAGE EmptyDataDecls           #-}
{-# LANGUAGE ForeignFunctionInterface #-}
{-# LANGUAGE RecordWildCards          #-}
{-# LANGUAGE TemplateHaskell          #-}
#ifdef USE_EMPTY_CASE
{-# LANGUAGE EmptyCase                #-}
#endif
--------------------------------------------------------------------------------
-- |
-- Module    : Foreign.CUDA.Driver.Device
-- Copyright : [2009..2023] Trevor L. McDonell
-- License   : BSD
--
-- Device management for low-level driver interface
--
--------------------------------------------------------------------------------

module Foreign.CUDA.Driver.Device (

  -- * Device Management
  Device(..),
  DeviceProperties(..), DeviceAttribute(..), Compute(..), ComputeMode(..), InitFlag,
  initialise, capability, device, attribute, count, name, props, uuid, totalMem,

) where

#include "cbits/stubs.h"
{# context lib="cuda" #}

-- Friends
import Foreign.CUDA.Analysis.Device
import Foreign.CUDA.Driver.Error
import Foreign.CUDA.Internal.C2HS

-- System
import Control.Applicative
import Control.Monad                                    ( liftM )
import Data.Bits
import Data.UUID.Types
import Foreign
import Foreign.C
import Prelude


--------------------------------------------------------------------------------
-- Data Types
--------------------------------------------------------------------------------

-- |
-- A CUDA device
--
newtype Device = Device { useDevice :: {# type CUdevice #}}
  deriving (Eq, Show)


-- |
-- Device attributes
--
{# enum CUdevice_attribute as DeviceAttribute
    { underscoreToCase
    , MAX as CU_DEVICE_ATTRIBUTE_MAX }          -- ignore
    with prefix="CU_DEVICE_ATTRIBUTE" deriving (Eq, Show) #}


#if CUDA_VERSION < 5000
{# pointer *CUdevprop as ^ foreign -> CUDevProp nocode #}

--
-- Properties of the compute device (internal helper).
-- Replaced by cuDeviceGetAttribute in CUDA-5.0 and later.
--
data CUDevProp = CUDevProp
  {
    cuMaxThreadsPerBlock :: !Int,               -- Maximum number of threads per block
    cuMaxBlockSize       :: !(Int,Int,Int),     -- Maximum size of each dimension of a block
    cuMaxGridSize        :: !(Int,Int,Int),     -- Maximum size of each dimension of a grid
    cuSharedMemPerBlock  :: !Int64,             -- Shared memory available per block in bytes
    cuTotalConstMem      :: !Int64,             -- Constant memory available on device in bytes
    cuWarpSize           :: !Int,               -- Warp size in threads (SIMD width)
    cuMemPitch           :: !Int64,             -- Maximum pitch in bytes allowed by memory copies
    cuRegsPerBlock       :: !Int,               -- 32-bit registers available per block
    cuClockRate          :: !Int,               -- Clock frequency in kilohertz
    cuTextureAlignment   :: !Int64              -- Alignment requirement for textures
  }
  deriving (Show)


instance Storable CUDevProp where
  sizeOf _    = {#sizeof CUdevprop#}
  alignment _ = alignment (undefined :: Ptr ())

  poke _ _    = error "no instance for Foreign.Storable.poke DeviceProperties"
  peek p      = do
    tb <- fromIntegral `fmap` {#get CUdevprop.maxThreadsPerBlock#} p
    sm <- fromIntegral `fmap` {#get CUdevprop.sharedMemPerBlock#} p
    cm <- fromIntegral `fmap` {#get CUdevprop.totalConstantMemory#} p
    ws <- fromIntegral `fmap` {#get CUdevprop.SIMDWidth#} p
    mp <- fromIntegral `fmap` {#get CUdevprop.memPitch#} p
    rb <- fromIntegral `fmap` {#get CUdevprop.regsPerBlock#} p
    cl <- fromIntegral `fmap` {#get CUdevprop.clockRate#} p
    ta <- fromIntegral `fmap` {#get CUdevprop.textureAlign#} p

    [t1,t2,t3] <- peekArrayWith fromIntegral 3 =<< {#get CUdevprop.maxThreadsDim#} p
    [g1,g2,g3] <- peekArrayWith fromIntegral 3 =<< {#get CUdevprop.maxGridSize#} p

    return CUDevProp
      {
        cuMaxThreadsPerBlock = tb,
        cuMaxBlockSize       = (t1,t2,t3),
        cuMaxGridSize        = (g1,g2,g3),
        cuSharedMemPerBlock  = sm,
        cuTotalConstMem      = cm,
        cuWarpSize           = ws,
        cuMemPitch           = mp,
        cuRegsPerBlock       = rb,
        cuClockRate          = cl,
        cuTextureAlignment   = ta
      }
#endif


-- |
-- Possible option flags for CUDA initialisation. Dummy instance until the API
-- exports actual option values.
--
data InitFlag
instance Enum InitFlag where
#ifdef USE_EMPTY_CASE
  toEnum   x = error ("InitFlag.toEnum: Cannot match " ++ show x)
  fromEnum x = case x of {}
#endif


--------------------------------------------------------------------------------
-- Initialisation
--------------------------------------------------------------------------------

-- |
-- Initialise the CUDA driver API. This must be called before any other
-- driver function.
--
-- <http://docs.nvidia.com/cuda/cuda-driver-api/group__CUDA__INITIALIZE.html#group__CUDA__INITIALIZE_1g0a2f1517e1bd8502c7194c3a8c134bc3>
--
{-# INLINEABLE initialise #-}
initialise :: [InitFlag] -> IO ()
initialise !flags = do
  enable_constructors
  cuInit flags

{-# INLINE enable_constructors #-}
{# fun unsafe enable_constructors { } -> `()' #}

{-# INLINE cuInit #-}
{# fun unsafe cuInit
  { combineBitMasks `[InitFlag]' } -> `()' checkStatus*- #}


--------------------------------------------------------------------------------
-- Device Management
--------------------------------------------------------------------------------

-- |
-- Return the compute compatibility revision supported by the device
--
{-# INLINEABLE capability #-}
capability :: Device -> IO Compute
#if CUDA_VERSION >= 5000
capability !dev =
  Compute <$> attribute dev ComputeCapabilityMajor
          <*> attribute dev ComputeCapabilityMinor
#else
-- Deprecated as of CUDA-5.0
--
capability !dev = uncurry Compute <$> cuDeviceComputeCapability dev

{-# INLINE cuDeviceComputeCapability #-}
{# fun unsafe cuDeviceComputeCapability
  { alloca-   `Int'    peekIntConv*
  , alloca-   `Int'    peekIntConv*
  , useDevice `Device'
  }
  -> `()' checkStatus*- #}
#endif


-- |
-- Return a handle to the compute device at the given ordinal.
--
-- <http://docs.nvidia.com/cuda/cuda-driver-api/group__CUDA__DEVICE.html#group__CUDA__DEVICE_1g8bdd1cc7201304b01357b8034f6587cb>
--
{-# INLINEABLE device #-}
{# fun unsafe cuDeviceGet as device
  { alloca-  `Device' dev*
  ,          `Int'
  }
  -> `()' checkStatus*- #}
  where
    dev = liftM Device . peek


-- |
-- Return the selected attribute for the given device.
--
-- <http://docs.nvidia.com/cuda/cuda-driver-api/group__CUDA__DEVICE.html#group__CUDA__DEVICE_1g9c3e1414f0ad901d3278a4d6645fc266>
--
{-# INLINEABLE attribute #-}
attribute :: Device -> DeviceAttribute -> IO Int
attribute !d !a = cuDeviceGetAttribute a d
  where
    {# fun unsafe cuDeviceGetAttribute
      { alloca-   `Int'             peekIntConv*
      , cFromEnum `DeviceAttribute'
      , useDevice `Device'
      }
      -> `()' checkStatus*- #}


-- |
-- Return the number of device with compute capability > 1.0.
--
-- <http://docs.nvidia.com/cuda/cuda-driver-api/group__CUDA__DEVICE.html#group__CUDA__DEVICE_1g52b5ce05cb8c5fb6831b2c0ff2887c74>
--
{-# INLINEABLE count #-}
{# fun unsafe cuDeviceGetCount as count
  { alloca- `Int' peekIntConv*
  }
  -> `()' checkStatus*- #}


-- |
-- The identifying name of the device.
--
-- <http://docs.nvidia.com/cuda/cuda-driver-api/group__CUDA__DEVICE.html#group__CUDA__DEVICE_1gef75aa30df95446a845f2a7b9fffbb7f>
--
{-# INLINEABLE name #-}
{# fun unsafe cuDeviceGetName as name
  { allocaS-  `String'& peekS*
  , useDevice `Device'
  }
  -> `()' checkStatus*- #}
  where
    len       = 512
    allocaS a = allocaBytes len $ \p -> a (p, fromIntegral len)
    peekS s _ = peekCString s


-- | Returns a UUID for the device.
--
-- Since CUDA-13: If the device is in MIG mode, this function returns its MIG
-- UUID which uniquely identifies the subscribed MIG compute instance.
--
-- Requires CUDA-9.2
--
-- <https://docs.nvidia.com/cuda/cuda-driver-api/group__CUDA__DEVICE.html#group__CUDA__DEVICE_1g987b46b884c101ed5be414ab4d9e60e4>
--
-- @since 0.10.0.0
--
{-# INLINE uuid #-}
uuid :: Device -> IO UUID
#if CUDA_VERSION < 9020
uuid = requireSDK 'uuid 9.2
#else
uuid !dev =
  allocaBytes 16 $ \ptr -> do
    cuDeviceGetUuid ptr dev
    unpack ptr
  where
    {-# INLINE cuDeviceGetUuid #-}
#if CUDA_VERSION < 13000
    {# fun unsafe cuDeviceGetUuid
      {           `Ptr ()'
      , useDevice `Device'
      } -> `()' checkStatus*- #}
#else
    {# fun unsafe cuDeviceGetUuid_v2 as cuDeviceGetUuid
      {           `Ptr ()'
      , useDevice `Device'
      } -> `()' checkStatus*- #}
#endif

    {-# INLINE unpack #-}
    unpack :: Ptr () -> IO UUID
    unpack !p = do
      let !q = castPtr p
      a <- word <$> peekElemOff q  0 <*> peekElemOff q  1 <*> peekElemOff q  2 <*> peekElemOff q  3
      b <- word <$> peekElemOff q  4 <*> peekElemOff q  5 <*> peekElemOff q  6 <*> peekElemOff q  7
      c <- word <$> peekElemOff q  8 <*> peekElemOff q  9 <*> peekElemOff q 10 <*> peekElemOff q 11
      d <- word <$> peekElemOff q 12 <*> peekElemOff q 13 <*> peekElemOff q 14 <*> peekElemOff q 15
      return $! fromWords a b c d

    {-# INLINE word #-}
    word :: Word8 -> Word8 -> Word8 -> Word8 -> Word32
    word !a !b !c !d
      =  (fromIntegral a `shiftL` 24)
     .|. (fromIntegral b `shiftL` 16)
     .|. (fromIntegral c `shiftL`  8)
     .|. (fromIntegral d            )
#endif


-- |
-- Return the properties of the selected device
--
{-# INLINEABLE props #-}
props :: Device -> IO DeviceProperties
props !d = do

#if CUDA_VERSION < 5000
  -- Old versions of the CUDA API used the separate cuDeviceGetProperties
  -- function to probe some properties, and cuDeviceGetAttribute for
  -- others. As of CUDA-5.0, the former was deprecated and its
  -- functionality subsumed by the latter, which we use below.
  --
  p   <- resultIfOk =<< cuDeviceGetProperties d
  let totalConstMem      = cuTotalConstMem p
      sharedMemPerBlock  = cuSharedMemPerBlock p
      regsPerBlock       = cuRegsPerBlock p
      warpSize           = cuWarpSize p
      maxThreadsPerBlock = cuMaxThreadsPerBlock p
      maxBlockSize       = cuMaxBlockSize p
      maxGridSize        = cuMaxGridSize p
      clockRate          = cuClockRate p
      memPitch           = cuMemPitch p
      textureAlignment   = cuTextureAlignment p
#else
  totalConstMem         <- fromIntegral <$> attribute d TotalConstantMemory
  sharedMemPerBlock     <- fromIntegral <$> attribute d SharedMemoryPerBlock
  memPitch              <- fromIntegral <$> attribute d MaxPitch
  textureAlignment      <- fromIntegral <$> attribute d TextureAlignment
  clockRate             <- attribute d ClockRate
  warpSize              <- attribute d WarpSize
  regsPerBlock          <- attribute d RegistersPerBlock
  maxThreadsPerBlock    <- attribute d MaxThreadsPerBlock
  maxBlockSize          <- (,,) <$> attribute d MaxBlockDimX <*> attribute d MaxBlockDimY <*> attribute d MaxBlockDimZ
  maxGridSize           <- (,,) <$> attribute d MaxGridDimX <*> attribute d MaxGridDimY <*> attribute d MaxGridDimZ
#endif

  -- The rest of the properties.
  --
  deviceName                    <- name d
  computeCapability             <- capability d
  totalGlobalMem                <- totalMem d
  multiProcessorCount           <- attribute d MultiprocessorCount
  computeMode                   <- toEnum <$> attribute d ComputeMode
  deviceOverlap                 <- toBool <$> attribute d GpuOverlap
  kernelExecTimeoutEnabled      <- toBool <$> attribute d KernelExecTimeout
  integrated                    <- toBool <$> attribute d Integrated
  canMapHostMemory              <- toBool <$> attribute d CanMapHostMemory
#if CUDA_VERSION >= 3000
  concurrentKernels             <- toBool <$> attribute d ConcurrentKernels
  eccEnabled                    <- toBool <$> attribute d EccEnabled
  maxTextureDim1D               <- attribute d MaximumTexture1dWidth
  maxTextureDim2D               <- (,)  <$> attribute d MaximumTexture2dWidth <*> attribute d MaximumTexture2dHeight
  maxTextureDim3D               <- (,,) <$> attribute d MaximumTexture3dWidth <*> attribute d MaximumTexture3dHeight <*> attribute d MaximumTexture3dDepth
#endif
#if CUDA_VERSION >= 4000
  asyncEngineCount              <- attribute d AsyncEngineCount
  cacheMemL2                    <- attribute d L2CacheSize
  maxThreadsPerMultiProcessor   <- attribute d MaxThreadsPerMultiprocessor
  memBusWidth                   <- attribute d GlobalMemoryBusWidth
  memClockRate                  <- attribute d MemoryClockRate
  pciInfo                       <- PCI <$> attribute d PciBusId <*> attribute d PciDeviceId <*> attribute d PciDomainId
  unifiedAddressing             <- toBool <$> attribute d UnifiedAddressing
  tccDriverEnabled              <- toBool <$> attribute d TccDriver
#endif
#if CUDA_VERSION >= 5050
  streamPriorities              <- toBool <$> attribute d StreamPrioritiesSupported
#endif
#if CUDA_VERSION >= 6000
  globalL1Cache                 <- toBool <$> attribute d GlobalL1CacheSupported
  localL1Cache                  <- toBool <$> attribute d LocalL1CacheSupported
  managedMemory                 <- toBool <$> attribute d ManagedMemory
  multiGPUBoard                 <- toBool <$> attribute d MultiGpuBoard
  multiGPUBoardGroupID          <- attribute d MultiGpuBoardGroupId
#endif
#if CUDA_VERSION >= 8000
  preemption                    <- toBool <$> attribute d ComputePreemptionSupported
  singleToDoublePerfRatio       <- attribute d SingleToDoublePrecisionPerfRatio
#endif
#if CUDA_VERSION >= 9000
  cooperativeLaunch             <- toBool <$> attribute d CooperativeLaunch
  cooperativeLaunchMultiDevice  <- toBool <$> attribute d CooperativeMultiDeviceLaunch
#endif

  return DeviceProperties{..}


#if CUDA_VERSION < 5000
-- Deprecated as of CUDA-5.0
{-# INLINE cuDeviceGetProperties #-}
{# fun unsafe cuDeviceGetProperties
  { alloca-   `CUDevProp' peek*
  , useDevice `Device'          } -> `Status' cToEnum #}
#endif


-- |
-- The total memory available on the device (bytes).
--
-- <http://docs.nvidia.com/cuda/cuda-driver-api/group__CUDA__DEVICE.html#group__CUDA__DEVICE_1gc6a0d6551335a3780f9f3c967a0fde5d>
--
{-# INLINEABLE totalMem #-}
{# fun unsafe cuDeviceTotalMem as totalMem
  { alloca-   `Int64'  peekIntConv*
  , useDevice `Device'
  }
  -> `()' checkStatus*- #}