Added loop support for single precision.

Author: chriselrod

Manifest.toml

@@ -67,11 +67,11 @@ version = "0.1.0"
 
 [[SLEEFPirates]]
 deps = ["SIMDPirates", "VectorizationBase"]
-git-tree-sha1 = "01ff5ddb2fe743e93a6d80b072a02cceb90592bf"
+git-tree-sha1 = "1c5b6827da87a12bdb7a4c893f44c3adbce3389d"
 repo-rev = "master"
 repo-url = "https://github.com/chriselrod/SLEEFPirates.jl"
 uuid = "476501e8-09a2-5ece-8869-fb82de89a1fa"
-version = "0.1.0"
+version = "0.1.1"
 
 [[Serialization]]
 uuid = "9e88b42a-f829-5b0c-bbe9-9e923198166b"
@@ -85,6 +85,6 @@ uuid = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 
 [[VectorizationBase]]
 deps = ["CpuId", "LinearAlgebra"]
-git-tree-sha1 = "30dd7fd08829bfa0fa6c57bf84a7daeac2e9462b"
+git-tree-sha1 = "5781ee225e9eee92b7463076411c35d26c444570"
 uuid = "3d5dd08c-fd9d-11e8-17fa-ed2836048c2f"
-version = "0.1.1"
+version = "0.1.3"

Project.toml

@@ -15,8 +15,8 @@ VectorizationBase = "3d5dd08c-fd9d-11e8-17fa-ed2836048c2f"
 MacroTools = "0.5"
 Parameters = "0.12.0"
 SIMDPirates = "0.1.0"
-SLEEFPirates = "0.1.0"
-VectorizationBase = "0.1.2"
+SLEEFPirates = "0.1.1"
+VectorizationBase = "0.1.3"
 julia = "1.0.0"
 
 [extras]

benchmarks/driver.jl

@@ -1,3 +1,7 @@
+# pkgdir(pkg::String) = abspath(joinpath(dirname(Base.find_package(pkg)), ".."))
+# const LOOPVECBENCHDIR = joinpath(pkgdir("LoopVectorization"), "benchmarks")
+# includet(joinpath(LOOPVECBENCHDIR, "driver.jl"))
+
 pkgdir(pkg::String) = abspath(joinpath(dirname(Base.find_package(pkg)), ".."))
 const LOOPVECBENCHDIR = joinpath(pkgdir("LoopVectorization"), "benchmarks")
 include(joinpath(LOOPVECBENCHDIR, "looptests.jl"))

src/LoopVectorization.jl

@@ -1,7 +1,7 @@
 module LoopVectorization
 
 using VectorizationBase, SIMDPirates, SLEEFPirates, MacroTools, Parameters
-using VectorizationBase: REGISTER_SIZE, REGISTER_COUNT, extract_data, num_vector_load_expr, mask
+using VectorizationBase: REGISTER_SIZE, REGISTER_COUNT, extract_data, num_vector_load_expr, mask, masktable, pick_vector_width_val, valmul, valrem, valmuladd
 using SIMDPirates: VECTOR_SYMBOLS, evadd, evmul, vrange, reduced_add, reduced_prod
 using Base.Broadcast: Broadcasted, DefaultArrayStyle
 using LinearAlgebra: Adjoint

src/graphs.jl

@@ -165,6 +165,15 @@ function looprange(ls::LoopSet, s::Symbol, incr::Int = 1, mangledname::Symbol =
         Expr(:call, :<, mangledname, loop.hintexact ? loop.rangehint - incr : Expr(:call, :-, loop.rangesym, incr))
     end
 end
+function looprange(ls::LoopSet, s::Symbol, incr::Expr, mangledname::Symbol = s, loop = ls.loops[s])
+    increxpr = Expr(:call, :-, incr, 1)
+    increxpr = if loop.hintexact
+        Expr(:call, :-, loop.rangehint, increxpr)
+    else
+        Expr(:call, :-, loop.rangesym, increxpr)
+    end
+    Expr(:call, :<, mangledname, increxpr)
+end
 function Base.length(ls::LoopSet, is::Symbol)
     ls.loops[is].rangehint
 end