1 files changed, 180 insertions, 0 deletions
diff --git a/tests/gc/gcbench.nim b/tests/gc/gcbench.nim
new file mode 100644
index 000000000..e29ea762d
--- /dev/null
+++ b/tests/gc/gcbench.nim
@@ -0,0 +1,180 @@
+discard """
+  outputsub: "Success!"
+"""
+
+# This is adapted from a benchmark written by John Ellis and Pete Kovac
+# of Post Communications.
+# It was modified by Hans Boehm of Silicon Graphics.
+#
+# This is no substitute for real applications. No actual application
+# is likely to behave in exactly this way. However, this benchmark was
+# designed to be more representative of real applications than other
+# Java GC benchmarks of which we are aware.
+# It attempts to model those properties of allocation requests that
+# are important to current GC techniques.
+# It is designed to be used either to obtain a single overall performance
+# number, or to give a more detailed estimate of how collector
+# performance varies with object lifetimes. It prints the time
+# required to allocate and collect balanced binary trees of various
+# sizes. Smaller trees result in shorter object lifetimes. Each cycle
+# allocates roughly the same amount of memory.
+# Two data structures are kept around during the entire process, so
+# that the measured performance is representative of applications
+# that maintain some live in-memory data. One of these is a tree
+# containing many pointers. The other is a large array containing
+# double precision floating point numbers. Both should be of comparable
+# size.
+#
+# The results are only really meaningful together with a specification
+# of how much memory was used. It is possible to trade memory for
+# better time performance. This benchmark should be run in a 32 MB
+# heap, though we don't currently know how to enforce that uniformly.
+#
+# Unlike the original Ellis and Kovac benchmark, we do not attempt
+# measure pause times. This facility should eventually be added back
+# in. There are several reasons for omitting it for now.  The original
+# implementation depended on assumptions about the thread scheduler
+# that don't hold uniformly. The results really measure both the
+# scheduler and GC. Pause time measurements tend to not fit well with
+# current benchmark suites. As far as we know, none of the current
+# commercial Java implementations seriously attempt to minimize GC pause
+# times.
+#
+# Known deficiencies:
+# - No way to check on memory use
+# - No cyclic data structures
+# - No attempt to measure variation with object size
+# - Results are sensitive to locking cost, but we don't
+#   check for proper locking
+#
+
+import
+  strutils, times
+
+type
+  PNode = ref TNode
+  TNode {.final, acyclic.} = object
+    left, right: PNode
+    i, j: int
+
+proc newNode(L, r: sink PNode): PNode =
+  new(result)
+  result.left = L
+  result.right = r
+
+const
+  kStretchTreeDepth = 18 # about 16Mb
+  kLongLivedTreeDepth = 16  # about 4Mb
+  kArraySize = 500000  # about 4Mb
+  kMinTreeDepth = 4
+  kMaxTreeDepth = 16
+
+when not declared(withScratchRegion):
+  template withScratchRegion(body: untyped) = body
+
+# Nodes used by a tree of a given size
+proc treeSize(i: int): int = return ((1 shl (i + 1)) - 1)
+
+# Number of iterations to use for a given tree depth
+proc numIters(i: int): int =
+  return 2 * treeSize(kStretchTreeDepth) div treeSize(i)
+
+# Build tree top down, assigning to older objects.
+proc populate(iDepth: int, thisNode: PNode) =
+  if iDepth <= 0:
+    return
+  else:
+    new(thisNode.left)
+    new(thisNode.right)
+    populate(iDepth-1, thisNode.left)
+    populate(iDepth-1, thisNode.right)
+
+# Build tree bottom-up
+proc makeTree(iDepth: int): PNode =
+  if iDepth <= 0:
+    new(result)
+  else:
+    return newNode(makeTree(iDepth-1), makeTree(iDepth-1))
+
+proc printDiagnostics() =
+  echo("Total memory available: " & formatSize(getTotalMem()) & " bytes")
+  echo("Free memory: " & formatSize(getFreeMem()) & " bytes")
+
+proc timeConstruction(depth: int) =
+  var
+    root, tempTree: PNode
+    iNumIters: int
+
+  iNumIters = numIters(depth)
+
+  echo("Creating " & $iNumIters & " trees of depth " & $depth)
+  var t = epochTime()
+  for i in 0..iNumIters-1:
+    new(tempTree)
+    populate(depth, tempTree)
+    tempTree = nil
+  echo("\tTop down construction took " & $(epochTime() - t) & "msecs")
+  t = epochTime()
+  for i in 0..iNumIters-1:
+    tempTree = makeTree(depth)
+    tempTree = nil
+  echo("\tBottom up construction took " & $(epochTime() - t) & "msecs")
+
+type
+  tMyArray = seq[float]
+
+proc main() =
+  var
+    root, longLivedTree, tempTree: PNode
+    myarray: tMyArray
+
+  echo("Garbage Collector Test")
+  echo(" Stretching memory with a binary tree of depth " & $kStretchTreeDepth)
+  printDiagnostics()
+  var t = epochTime()
+
+  # Stretch the memory space quickly
+  withScratchRegion:
+    tempTree = makeTree(kStretchTreeDepth)
+    tempTree = nil
+
+  # Create a long lived object
+  echo(" Creating a long-lived binary tree of depth " &
+        $kLongLivedTreeDepth)
+  new(longLivedTree)
+  populate(kLongLivedTreeDepth, longLivedTree)
+
+  # Create long-lived array, filling half of it
+  echo(" Creating a long-lived array of " & $kArraySize & " doubles")
+  withScratchRegion:
+    newSeq(myarray, kArraySize)
+    for i in 0..kArraySize div 2 - 1:
+      myarray[i] = 1.0 / toFloat(i)
+
+    printDiagnostics()
+
+    var d = kMinTreeDepth
+    while d <= kMaxTreeDepth:
+      withScratchRegion:
+        timeConstruction(d)
+      inc(d, 2)
+
+    if longLivedTree == nil or myarray[1000] != 1.0/1000.0:
+      echo("Failed")
+      # fake reference to LongLivedTree
+      # and array to keep them from being optimized away
+
+  var elapsed = epochTime() - t
+  printDiagnostics()
+  echo("Completed in " & $elapsed & "s. Success!")
+  when declared(getMaxMem):
+    echo "Max memory ", formatSize getMaxMem()
+
+when defined(GC_setMaxPause):
+  GC_setMaxPause 2_000
+
+when defined(gcDestructors):
+  let mem = getOccupiedMem()
+main()
+when defined(gcDestructors):
+  doAssert getOccupiedMem() == mem