Identifying JVM SIMD and SSE Usage with the VTune™ Performance Analyzer

Product Index
Search

Intel® Software Network

Language:
- English

	September 21, 2010 6:29 PM PDT

by Levent Akyil

Leveraging SIMD and SSE (Streaming SIMD Extensions) support available on target processors is one of the key optimization techniques JVMs use (or should use). The question is how to identify which Jitted methods use SSE? This has been a common question from many Java developers.

The VTune analyzer’s event based sampling can help users pinpoint exactly which methods are optimized to use SSE. One can simply use SIMD_INST_RETIRED.ANY event (Retired Streaming SIMD instructions (precise event)) to count the overall number of SIMD instructions retired.

The following events can give further break-down of this one event. These are the events that are available on Core™ architecture based processors.

Symbol Name[VTune(TM) Analyzer help]	Description
FP_MMX_TRANS.TO_FP	Transitions from MMX (TM) Instructions to Floating Point Instructions.
FP_MMX_TRANS.TO_MMX	Transitions from Floating Point to MMX (TM) Instructions.
SIMD_ASSIST	SIMD assists invoked.
SIMD_COMP_INST_RETIRED.PACKED_DOUBLE	Retired computational Streaming SIMD Extensions 2 (SSE2) packed-double instructions.
SIMD_COMP_INST_RETIRED.PACKED_SINGLE	Retired computational Streaming SIMD Extensions (SSE) packed-single instructions.
SIMD_COMP_INST_RETIRED.SCALAR_DOUBLE	Retired computational Streaming SIMD Extensions 2 (SSE2) scalar-double instructions.
SIMD_COMP_INST_RETIRED.SCALAR_SINGLE	Retired computational Streaming SIMD Extensions (SSE) scalar-single instructions.
SIMD_INSTR_RETIRED	SIMD Instructions retired.
SIMD_INST_RETIRED.ANY	Retired Streaming SIMD instructions (precise event).
SIMD_INST_RETIRED.PACKED_DOUBLE	Retired Streaming SIMD Extensions 2 (SSE2) packed-double instructions.
SIMD_INST_RETIRED.PACKED_SINGLE	Retired Streaming SIMD Extensions (SSE) packed-single instructions.
SIMD_INST_RETIRED.SCALAR_DOUBLE	Retired Streaming SIMD Extensions 2 (SSE2) scalar-double instructions.
SIMD_INST_RETIRED.SCALAR_SINGLE	Retired Streaming SIMD Extensions (SSE) scalar-single instructions.
SIMD_INST_RETIRED.VECTOR	Retired Streaming SIMD Extensions 2 (SSE2) vector integer instructions.
SIMD_SAT_INSTR_RETIRED	Saturated arithmetic instructions retired.
SIMD_SAT_UOP_EXEC	SIMD saturated arithmetic micro-ops executed.
SIMD_UOPS_EXEC	SIMD micro-ops executed (excluding stores).
SIMD_UOP_TYPE_EXEC.ARITHMETIC	SIMD packed arithmetic micro-ops executed
SIMD_UOP_TYPE_EXEC.LOGICAL	SIMD packed logical micro-ops executed
SIMD_UOP_TYPE_EXEC.MUL	SIMD packed multiply micro-ops executed
SIMD_UOP_TYPE_EXEC.PACK	SIMD pack micro-ops executed
SIMD_UOP_TYPE_EXEC.SHIFT	SIMD packed shift micro-ops executed
SIMD_UOP_TYPE_EXEC.UNPACK	SIMD unpack micro-ops executed

If I collect some of these events on SciMark2 (http://math.nist.gov/scimark2/), it can be seen that actually all our benchmarks are using SIMD instructions. It is also important to note that SIMD_INST_RETIRED.ANY should be equal to the total of all the sub events represented as SIMD_INST_RETIRED.xyz.

Do you need more help?

Search knowledge base
Visit user forums
Find additional support

This article applies to: Tools, Intel® VTune™ Performance Analyzer for Linux* Knowledge Base, Intel® VTune™ Performance Analyzer for Windows* Knowledge Base

For more complete information about compiler optimizations, see our Optimization Notice.

Comments (1)

March 15, 2011 8:17 AM PDT

Jonathan

It seems as though Java developers have always had to take a back seat with regards to SSE and SIMD in general. The article at http://queue.acm.org/detail.cfm?id=1945954 pushes for this to change. I shouldn't have to profile and obfuscate my code to ensure that SSE gets used to a respectable utilization level in interpreted languages.