Histograms: optimize floatBucketIterator for common case

Signed-off-by: Linas Medziunas <linas.medziunas@gmail.com>

model/histogram/float_histogram.go

@@ -767,6 +767,7 @@ func (h *FloatHistogram) floatBucketIterator(
 		},
 		targetSchema:           targetSchema,
 		absoluteStartValue:     absoluteStartValue,
+		boundReachedStartValue: absoluteStartValue == 0,
 	}
 	if positive {
 		i.spans = h.PositiveSpans
@@ -824,6 +825,32 @@ func (i *floatBucketIterator) Next() bool {
 		return false
 	}
 
+	if i.schema == i.targetSchema {
+		// Fast path for the common case.
+		span := i.spans[i.spansIdx]
+		if i.bucketsIdx == 0 {
+			// Seed origIdx for the first bucket.
+			i.currIdx = span.Offset
+		} else {
+			i.currIdx++
+		}
+
+		for i.idxInSpan >= span.Length {
+			// We have exhausted the current span and have to find a new
+			// one. We even handle pathologic spans of length 0 here.
+			i.idxInSpan = 0
+			i.spansIdx++
+			if i.spansIdx >= len(i.spans) {
+				return false
+			}
+			span = i.spans[i.spansIdx]
+			i.currIdx += span.Offset
+		}
+
+		i.currCount = i.buckets[i.bucketsIdx]
+		i.idxInSpan++
+		i.bucketsIdx++
+	} else {
 		// Copy all of these into local variables so that we can forward to the
 		// next bucket and then roll back if needed.
 		origIdx, spansIdx, idxInSpan := i.origIdx, i.spansIdx, i.idxInSpan
@@ -853,7 +880,7 @@ mergeLoop: // Merge together all buckets from the original schema that fall into
 				span = i.spans[spansIdx]
 				origIdx += span.Offset
 			}
-		currIdx := i.targetIdx(origIdx)
+			currIdx := targetIdx(origIdx, i.schema, i.targetSchema)
 			switch {
 			case firstPass:
 				i.currIdx = currIdx
@@ -873,6 +900,8 @@ mergeLoop: // Merge together all buckets from the original schema that fall into
 				break mergeLoop
 			}
 		}
+	}
+
 	// Skip buckets before absoluteStartValue.
 	// TODO(beorn7): Maybe do something more efficient than this recursive call.
 	if !i.boundReachedStartValue && getBound(i.currIdx, i.targetSchema) <= i.absoluteStartValue {
@@ -882,17 +911,6 @@ mergeLoop: // Merge together all buckets from the original schema that fall into
 	return true
 }
 
-// targetIdx returns the bucket index within i.targetSchema for the given bucket
-// index within i.schema.
-func (i *floatBucketIterator) targetIdx(idx int32) int32 {
-	if i.schema == i.targetSchema {
-		// Fast path for the common case. The below would yield the same
-		// result, just with more effort.
-		return idx
-	}
-	return ((idx - 1) >> (i.schema - i.targetSchema)) + 1
-}
-
 type reverseFloatBucketIterator struct {
 	baseBucketIterator[float64, float64]
 	idxInSpan int32 // Changed from uint32 to allow negative values for exhaustion detection.