Perf: improve sum like reductions

[wsmoses]

No description provided.

[github-actions]

Your PR requires formatting changes to meet the project's style guidelines.
Please consider running Runic (git runic main) to apply these changes.

Click here to view the suggested changes.

diff --git a/src/compiler/interpreter.jl b/src/compiler/interpreter.jl
index 14795635..d702af93 100644
--- a/src/compiler/interpreter.jl
+++ b/src/compiler/interpreter.jl
@@ -992,22 +992,22 @@ end
 @inline function override_bc_mapreduce(f, op, ::Base.IndexLinear, A::Base.AbstractArrayOrBroadcasted)
     inds = Base.LinearIndices(A)
     n = length(inds)
-    
+
     HET = Base.IteratorEltype(A)
     if HET isa Base.HasEltype
-	ET = eltype(A)
-	    if (op === Base.:+ || op === Base.add_sum) && Base.isconcretetype(ET) && (ET <: Base.IEEEFloat || ET <: Complex{<:Base.IEEEFloat})
-		@inbounds i = first(inds)
-		@inbounds l = last(inds)
-		s = zero(ET)
-		while i <= l
-		    @inbounds Ai = A[i]
-		    i+=1
-		    s = op(s, f(Ai))
-		    @simdloop
-		end
-		return s
-	    end
+        ET = eltype(A)
+        if (op === Base.:+ || op === Base.add_sum) && Base.isconcretetype(ET) && (ET <: Base.IEEEFloat || ET <: Complex{<:Base.IEEEFloat})
+            @inbounds i = first(inds)
+            @inbounds l = last(inds)
+            s = zero(ET)
+            while i <= l
+                @inbounds Ai = A[i]
+                i += 1
+                s = op(s, f(Ai))
+                @simdloop
+            end
+            return s
+        end
     end
 
     if n == 0

[github-actions]

Benchmark Results

	main	2614f6c...	main / 2614f6c...
basics/make_zero/namedtuple	0.0559 ± 0.0021 μs	0.0586 ± 0.0041 μs	0.952 ± 0.076
basics/make_zero/struct	0.265 ± 0.0066 μs	0.265 ± 0.0063 μs	0.999 ± 0.034
basics/overhead	5.26 ± 0.01 ns	4.34 ± 0.01 ns	1.21 ± 0.0036
basics/remake_zero!/namedtuple	0.257 ± 0.0094 μs	0.259 ± 0.0084 μs	0.991 ± 0.048
basics/remake_zero!/struct	0.253 ± 0.012 μs	0.256 ± 0.0081 μs	0.991 ± 0.056
fold_broadcast/multidim_sum_bcast/1D	10.3 ± 0.36 μs	10.3 ± 0.31 μs	1 ± 0.046
fold_broadcast/multidim_sum_bcast/2D	10.3 ± 0.18 μs	10.3 ± 0.23 μs	0.999 ± 0.028
time_to_load	1.1 ± 0.0097 s	1.11 ± 0.014 s	0.994 ± 0.016

Benchmark Plots

A plot of the benchmark results has been uploaded as an artifact at https://github.com/EnzymeAD/Enzyme.jl/actions/runs/20544784076/artifacts/4976395009.

[codecov]

Codecov Report

❌ Patch coverage is 0% with 15 lines in your changes missing coverage. Please review.
✅ Project coverage is 66.64%. Comparing base (89305f0) to head (2614f6c).

Files with missing lines	Patch %	Lines
src/compiler/interpreter.jl	0.00%	15 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2865      +/-   ##
==========================================
- Coverage   67.40%   66.64%   -0.76%     
==========================================
  Files          58       48      -10     
  Lines       21092    20493     -599     
==========================================
- Hits        14217    13658     -559     
+ Misses       6875     6835      -40     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.