Add support for async f and j

Author: charleskawczynski

Project.toml

@@ -17,6 +17,12 @@ NVTX = "5da4648a-3479-48b8-97b9-01cb529c0a1f"
 SciMLBase = "0bca4576-84f4-4d90-8ffe-ffa030f20462"
 StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"
 
+[weakdeps]
+CUDA = "052768ef-5323-5732-b1bb-66c8b64840ba"
+
+[extensions]
+CudaExt = "CUDA"
+
 [compat]
 ClimaComms = "0.4, 0.5"
 Colors = "0.12"

ext/CudaExt.jl

@@ -0,0 +1,60 @@
+module CudaExt
+
+import CUDA
+import ClimaComms: SingletonCommsContext, CUDADevice
+import ClimaTimeSteppers: compute_fj!
+
+@inline function compute_fj!(f, j, U, f!, j!, ::SingletonCommsContext{CUDADevice})
+    # TODO: we should benchmark these two options to
+    #       see if one is preferrable over the other
+    if Base.Threads.nthreads() > 1
+        compute_fj_spawn!(f, j, U, f!, j!)
+    else
+        compute_fj_streams!(f, j, U, f!, j!)
+    end
+end
+
+@inline function compute_fj_streams!(f, j, U, f!, j!)
+    event = CUDA.CuEvent(CUDA.EVENT_DISABLE_TIMING)
+    CUDA.record(event, CUDA.stream()) # record event on main stream
+
+    stream1 = CUDA.CuStream() # make a stream
+    local event1
+    CUDA.stream!(stream1) do # work to be done by stream1
+        CUDA.wait(event, stream1) # make stream1 wait on event (host continues)
+        f!(f, U)
+        event1 = CUDA.CuEvent(CUDA.EVENT_DISABLE_TIMING)
+    end
+    CUDA.record(event1, stream1) # record event1 on stream1
+
+    stream2 = CUDA.CuStream() # make a stream
+    local event2
+    CUDA.stream!(stream2) do # work to be done by stream2
+        CUDA.wait(event, stream2) # make stream2 wait on event (host continues)
+        j!(j, U)
+        event2 = CUDA.CuEvent(CUDA.EVENT_DISABLE_TIMING)
+    end
+    CUDA.record(event2, stream2) # record event2 on stream2
+
+    CUDA.wait(event1, CUDA.stream()) # make main stream wait on event1
+    CUDA.wait(event2, CUDA.stream()) # make main stream wait on event2
+end
+
+@inline function compute_fj_spawn!(f, j, U, f!, j!)
+
+    CUDA.synchronize()
+    CUDA.@sync begin
+        Base.Threads.@spawn begin
+            f!(f, U)
+            CUDA.synchronize()
+            nothing
+        end
+        Base.Threads.@spawn begin
+            j!(j, U)
+            CUDA.synchronize()
+            nothing
+        end
+    end
+end
+
+end

src/solvers/compute_fj.jl

@@ -0,0 +1,17 @@
+@inline function compute_fj!(f, j, U, f!, j!, ::Union{Nothing, ClimaComms.AbstractCommsContext})
+    f!(f, U)
+    j!(j, U)
+end
+
+@inline function compute_fj!(f, j, U, f!, j!, ::ClimaComms.SingletonCommsContext{ClimaComms.CPUMultiThreaded})
+    Base.@sync begin
+        Base.Threads.@spawn begin
+            f!(f, U)
+            nothing
+        end
+        Base.Threads.@spawn begin
+            j!(j, U)
+            nothing
+        end
+    end
+end