Consider the following attempt to abstract out an unwieldly use of Annotate into an easy-to-use Pin type:
import Clash.Prelude
import Clash.Signal
import Clash.Annotations.SynthesisAttributes
import Clash.Annotations.TopEntity
import Clash.Annotations.TH
createDomain vSystem{vName="Bank2C", vPeriod=20000, vResetKind=Asynchronous, vResetPolarity=ActiveLow}
type Pin t x y = t `Annotate` 'StringAttr "chip_pin" x
`Annotate` 'StringAttr "altera_attribute" ("-name IO_STANDARD \"" `AppendSymbol` y `AppendSymbol` "\"")
top :: HiddenClockResetEnable dom => Signal dom (BitVector 4)
top = s where s = register 0 (s + 1)
topEntity
:: "CLK_50_B2C" ::: Pin (Clock Bank2C) "PIN_AW38" "1.2 V"
-> "CPU_RESET_n" ::: Signal Bank2C Bool
-> "LED" ::: Signal Bank2C (BitVector 4)
topEntity clk rstnButton = exposeClockResetEnable top clk rst en where
en = enableGen
rst = unsafeFromLowPolarity rstnButton
rstSync = resetSynchronizer clk rst en
makeTopEntityWithName 'topEntity "shell"Using the tip of the 1.4 branch, I get:
$ clash --verilog DE10Pro
GHC: Parsing and optimising modules took: 1.584s
GHC: Loading external modules from interface files took: 0.000s
GHC: Parsing annotations took: 0.000s
Clash: Parsing and compiling primitives took 0.147s
<no location info>: error:
Clash error call:
Could not unpack given type to string: AppendSymbol (AppendSymbol "-name IO_STANDARD \"" "1.2 V") "\""
CallStack (from HasCallStack):
error, called at src-ghc/Clash/GHC/GHC2Core.hs:775:19 in clash-ghc-1.4.0-DkSmlpJFpg8Ad9ypRSNVx8:Clash.GHC.GHC2Core
I don't see why this can't be supported; it seems that the evaluator merely assumes that the string type is already in normal form, which isn't the case here. My actual goal is to use closed type families to write a more sophisticated version of Pin, but the basic idea here is all the same.
Consider the following attempt to abstract out an unwieldly use of
Annotateinto an easy-to-usePintype:Using the tip of the 1.4 branch, I get:
I don't see why this can't be supported; it seems that the evaluator merely assumes that the string type is already in normal form, which isn't the case here. My actual goal is to use closed type families to write a more sophisticated version of
Pin, but the basic idea here is all the same.