SnoopPrecompile

Author: chriselrod

Project.toml

@@ -1,7 +1,7 @@
 name = "LoopVectorization"
 uuid = "bdcacae8-1622-11e9-2a5c-532679323890"
 authors = ["Chris Elrod <[email protected]>"]
-version = "0.12.120"
+version = "0.12.121"
 
 [deps]
 ArrayInterface = "4fba245c-0d91-5ea0-9b3e-6abc04ee57a9"
@@ -22,6 +22,7 @@ PolyesterWeave = "1d0040c9-8b98-4ee7-8388-3f51789ca0ad"
 SIMDDualNumbers = "3cdde19b-5bb0-4aaf-8931-af3e248e098b"
 SIMDTypes = "94e857df-77ce-4151-89e5-788b33177be4"
 SLEEFPirates = "476501e8-09a2-5ece-8869-fb82de89a1fa"
+SnoopPrecompile = "66db9d55-30c0-4569-8b51-7e840670fc0c"
 SpecialFunctions = "276daf66-3868-5448-9aa4-cd146d93841b"
 Static = "aedffcd0-7271-4cad-89d0-dc628f76c6d3"
 ThreadingUtilities = "8290d209-cae3-49c0-8002-c8c24d57dab5"

src/LoopVectorization.jl

@@ -231,9 +231,9 @@ loop-reordering so as to improve performance:
 LoopVectorization
 
 include("precompile.jl")
-_precompile_()
+# _precompile_()
 
-_vreduce(+, Float64[1.0])
+# _vreduce(+, Float64[1.0])
 # matmul_params(64, 32, 64)
 
 # import ChainRulesCore, ForwardDiff

src/precompile.jl

@@ -1,3 +1,10 @@
-function _precompile_()
-  ccall(:jl_generating_output, Cint, ()) == 1 || return nothing
+
+using SnoopPrecompile
+
+@precompile_setup begin
+  x = rand(10);
+  @precompile_all_calls begin
+    _vreduce(+, x)
+  end
 end
+

utils/generate_precompiles.jl

@@ -1,6 +1,8 @@
 
 using LoopVectorization, SnoopCompile
-tinf = @snoopi_deep include(joinpath(pkgdir(LoopVectorization), "test", "runtests.jl"))
+LOOPVECTORIZATION_TEST = "all"
+tinf = @snoopi_deep include(joinpath(pkgdir(LoopVectorization), "test", "testsetup.jl"))
+tinf = @snoopi_deep include(joinpath(pkgdir(LoopVectorization), "test", "grouptests.jl"))
 
 ttot, pcs = SnoopCompile.parcel(tinf);