Fix incorrect timing results for CUDA.@Elapsed (#2118)

JuliaGPU · Oct 23, 2023 · 87fd247 · 87fd247
1 parent 8dbcb2a
commit 87fd247
Show file tree

Hide file tree

Showing 2 changed files with 9 additions and 1 deletion.
diff --git a/lib/cudadrv/events.jl b/lib/cudadrv/events.jl
@@ -118,7 +118,7 @@ macro elapsed(ex...)
     quote
         t0, t1 = CuEvent(), CuEvent()
         record(t0)
-        $(esc(ex))
+        $(esc(code))
         record(t1)
         synchronize(t1; blocking=$blocking)
         elapsed(t0, t1)

diff --git a/test/core/cudadrv.jl b/test/core/cudadrv.jl
@@ -200,6 +200,14 @@ end
 @test (CUDA.@elapsed identity(nothing)) > 0
 @test (CUDA.@elapsed blocking=true identity(nothing)) > 0
 
+let
+    empty_fun() = nothing
+
+    # Check that the benchmarked code occurs as-is in the macro expansion.
+    me = @macroexpand1 CUDA.@elapsed empty_fun()
+    @test any(arg -> arg == :(empty_fun()), me.args)
+end
+
 CuEvent(CUDA.EVENT_BLOCKING_SYNC)
 CuEvent(CUDA.EVENT_BLOCKING_SYNC | CUDA.EVENT_DISABLE_TIMING)