http://www.datapower.com/XSLTMark/
BenchMark Tests
Understanding The Benchmark

The XSLTMark test cases have been designed to challenge processors with a variety of
tasks and input conditions in order to provide a well-rounded benchmark and to
facilitate analysis of which processors perform best for various applications.
XSLTMark measures performance in four major categories:

o Pattern Matching - this category covers XSLT template pattern matching and template
instantiation. This performance category is important to stylesheets with many
template rules and with many expected apply-template invocations.

o XPath Selection - this category covers nodeset selection through the evaluation of
XPath path expressions. This performance category is crucial for stylesheets
that contain lots of xpath expressions, particularly ones with predicates.

o XPath Library Functions - this category covers the execution of XPath library
functions, particularly the frequently used string functions. This category is
most important to stylesheets that perform a lot of string processing.

o XSLT Control - this category covers the control structures defined by XSLT
elements, including variable and parameter handling. This category is most
relevant for stylesheets that perform tricky calculations involving calling
templates with parameters.

The size and composition of the source document is also an important factor in
determining the peformance of a transform. Larger source documents are always
slower to process, but for some, the performance of the engines scales much
better with increasing input size. Also, certain qualities of the input tree
can have a profound effect, such as the breadth and depth of the tree, the
number of text nodes, etc. The test bed includes input documents of a
variety of lengths and compositions.


xpath	small	(size)	match		Performs complex pattern matching.


Data exerpted from
http://www.datapower.com/XSLTMark/res_2001_04_01.html

                          prep  exec
Test Component     iter   ms    ms      KB In     KB Out    KB/s      status
Fourthought 4Suite
xpath               100   83    77175   28        20        0.32      OK

Gnome XSLT 0.5.0
xpath               100   2     235    28        20        104.72    OK

Microsoft XML Parser 3.0
xpath               100   80    170    30        13        129.54    CHK OUTPUT

Oracle XSLT 2.0
xpath               100   4     245    28        13        85.90     CHK OUTPUT

Sablotron 0.51
xpath               100   0     580    28        20        42.27     OK

Saxon 6.2.1
xpath               100   6     432    28        20        56.74     OK

TransforMiiX 0.8
xpath               100   1     799    28        18        29.70     OK

Xalan C 1.1
xpath               100   9     1365   28        20        17.99     OK

XalanJ 2.0,0
xpath               100   165   603    28        20        40.73     OK

Sun XSLTC alpha 4
xpath               100   79    100    28        17        229.98    CHK OUTPUT

XT 19991105
xpath               100   4     180    28        20        136.45    OK

Key:

iter
The number of iterations conducted of this test.

prep ms
The preparation time in milliseconds for this test's stylesheet.

exec ms
The execution time in milliseconds for this test. This time includes time spent
writing the output to a file, and may or may not include the time spent parsing
the input .xml file (depending on the processor, see note above).

KB In
Total kilobytes of source data processed (equals size of input document
* iterations)

KB Out
Total kilobytes of data output (equals size of output * iterations)

KB/s
Ihroughput, calculated as (KB in + KB out) / (2 * time).

Result
If output verification is available on your platform, this column contains the
result of a simple compliance check: "OK" indicates success, "CHK OUTPUT"
indicates a likely compliance violation, and "ERROR" indicates a failed test.
"NO REF" indicates that no reference exists for the test case.

total-results
This row contains the aggregated results for all tests whose result was "OK"
or "NO REFERENCE". Note that the aggregated KB/s number is computed as
(total KB in + total KB out) / (2 * total time), and not by averaging
the KB/s numbers of each test case.

Green Text
Annotations and corrections of the raw XSLTMark output (for example,
erroneous CHK OUTPUT reports) have been indicated in green, along
with corrected total results.


(C) Copyright 2000 by DataPower Technology, Inc.
http://www.datapower.com