.

Quafadas · Oct 2, 2024 · 93df12f · 93df12f
1 parent f1ca3f8
commit 93df12f
Show file tree

Hide file tree

Showing 2 changed files with 3 additions and 3 deletions.
diff --git a/benchmark/src/Benchmark.scala b/benchmark/src/Benchmark.scala
@@ -11,8 +11,8 @@ import java.util.concurrent.TimeUnit;
 @OutputTimeUnit(TimeUnit.SECONDS)
 @State(Scope.Thread)
 @Fork(value = 1)
-@Warmup(iterations = 3)
-@Measurement(iterations = 3)
+@Warmup(iterations = 1)
+@Measurement(iterations = 1)
 abstract class BLASBenchmark:
 
   var blas: BLAS = _;

diff --git a/vecxt/jvm/src/arrays.scala b/vecxt/jvm/src/arrays.scala
@@ -301,7 +301,7 @@ object arrays:
 
     // https://developer.nvidia.com/gpugems/gpugems3/part-vi-gpu-computing/chapter-39-parallel-prefix-sum-scan-cuda#:~:text=A%20simple%20and%20common%20parallel%20algorithm#:~:text=A%20simple%20and%20common%20parallel%20algorithm
     inline def cumsum: Unit =
-      val spd: VectorSpecies[java.lang.Double] = DoubleVector.SPECIES_128
+      val spd: VectorSpecies[java.lang.Double] = DoubleVector.SPECIES_PREFERRED
       val spi: VectorSpecies[java.lang.Integer] =
         VectorSpecies.of(java.lang.Integer.TYPE, VectorShape.forBitSize(spd.vectorBitSize() / 2))