Fix broken precompilation warning, and bump SIMDPirates requirement.

Author: chriselrod

Project.toml

@@ -15,7 +15,7 @@ VectorizationBase = "3d5dd08c-fd9d-11e8-17fa-ed2836048c2f"
 [compat]
 DocStringExtensions = "0.8"
 OffsetArrays = "1"
-SIMDPirates = "0.8.4"
+SIMDPirates = "0.8.6"
 SLEEFPirates = "0.5"
 UnPack = "0,1"
 VectorizationBase = "0.12.6"

src/determinestrategy.jl

@@ -186,7 +186,7 @@ function unroll_no_reductions(ls, order, unrolled, vectorized, Wshift, size_T)
     # @show compute_rt, load_rt
     # roundpow2(min(4, round(Int, (compute_rt + load_rt + 1) / compute_rt)))
     rt = max(compute_rt, load_rt)
-    rt == 0.0 && return 4
+    iszero(rt) && return 4
     max(1, roundpow2( min( 4, round(Int, 16 / rt) ) ))
 end
 function determine_unroll_factor(
@@ -286,9 +286,10 @@ function solve_unroll(X, R, u₁L, u₂L, u₁step, u₂step)
     discriminant < 0 && return -1,-1,Inf
     u₁float = max(float(u₁step), (sqrt(discriminant) + b) / (-2a)) # must be at least 1
     u₂float = (RR - u₁float*R₂)/(u₁float*R₁)
-    if !(isfinite(u₂float) && isfinite(u₁float))
-        return 4, 4, unroll_cost(X, 4, 4, u₁L, u₂L)
-        # return itertilesize(X, u₁L, u₂L)
+    if !(isfinite(u₂float) & isfinite(u₁float)) # brute force
+        u₁low = u₂low = 1
+        u₁high = u₂high = REGISTER_COUNT == 32 ? 10 : 6#8
+        return solve_unroll_iter(X, R, u₁L, u₂L, u₁low:u₁step:u₁high, u₂low:u₂step:u₂high)
     end
     u₁low = floor(Int, u₁float)
     u₂low = max(u₂step, floor(Int, u₂float)) # must be at least 1
@@ -564,6 +565,13 @@ function load_elimination_cost_factor!(
         false
     end
 end
+function loadintostore(ls::LoopSet, op::Operation)
+    isload(op) || return false
+    for opp ∈ operations(ls)
+        isstore(opp) && opp.ref == op.ref && return true
+    end
+    false
+end
 function add_constant_offset_load_elmination_cost!(
     X, R, choose_to_inline, ls::LoopSet, op::Operation, iters, unrollsyms::UnrollSymbols, u₁reduces::Bool, u₂reduces::Bool, Wshift::Int, size_T::Int, opisininnerloop::Bool
 )
@@ -575,6 +583,9 @@ function add_constant_offset_load_elmination_cost!(
         rt, lat, rp = cost(ls, op, vectorized, Wshift, size_T)
         rt *= iters
         rp = opisininnerloop ? rp : zero(rp)
+        # if loadintostore(ls, op) # For now, let's just avoid unrolling in this way...
+        #     rt = Inf
+        # end
         # u_uid is getting eliminated
         # we treat this as the unrolled loop getting eliminated is split into 2 parts:
         # 1 a non-cost-reduced part, with factor udependent_reduction
@@ -700,7 +711,8 @@ function evaluate_cost_tile(
             prefetch_good_idea = true
         end
         # @show isunrolled₁, isunrolled₂, op rt, lat, rp
-        rp = opisininnerloop ? rp : zero(rp) # we only care about register pressure within the inner most loop
+        rp = (opisininnerloop && !(loadintostore(ls, op))) ? rp : zero(rp) # we only care about register pressure within the inner most loop
+        # rp = opisininnerloop ? rp : zero(rp) # we only care about register pressure within the inner most loop
         rt *= iters[id]
         if u₁reduces & u₂reduces
             cost_vec[4] += rt

src/mapreduce.jl

@@ -80,7 +80,8 @@ Vectorized version of `reduce`. Reduces the array `A` using the operator `op`.
 @inline vreduce(op, arg) = vmapreduce(identity, op, arg)
 
 for (op, init) in zip((:+, :max, :min), (:zero, :identity, :identity))
-    @eval function vreduce(::typeof($op), arg; dims)
+    @eval function vreduce(::typeof($op), arg; dims = nothing)
+        isnothing(dims) && return _vreduce($op, arg)
         @assert length(dims) == 1
         out = $init(arg[ntuple(d -> d == dims ? (1:1) : (1:size(arg, d)), ndims(arg))...])
         Rpre = CartesianIndices(axes(arg)[1:dims-1])
@@ -95,11 +96,11 @@ for (op, init) in zip((:+, :max, :min), (:zero, :identity, :identity))
         return out
     end
 
-    @eval function vreduce(::typeof($op), arg)
+    @eval function _vreduce(::typeof($op), arg)
         s = $init(arg[1])
         @avx for i in 1:length(arg)
             s = $op(s, arg[i])
         end
         return s
     end
-end
+end