Applying Performance Enhancement Strategies

Technical applications with loopy code

High

Technical applications are those programs that have some subset of functions that consume a majority of total CPU cycles in loop nests.

Target loop nests using -03 (Linux* and Mac OS*) or /O3 (Windows*) to enable more aggressive loop transformations and prefetching.

Use High-Level Optimization (HLO) reporting to determine which HLO optimizations the compiler elected to apply.

See High-Level Optimization Report.

(same as above)
Itanium® Only

High

For -O2 and -O3 (Linux) or /O2 and /O3 (Windows), use the SWP report to determine if Software Pipelining occurred on key loops, and if not, why not.

You might be able to change the code to allow software pipelining under the following conditions:

• If recurrences are listed in the report that you suspect do not exist, eliminate aliasing problems (for example, using the restrict keyword), or use ivdep pragma on the loop.

• If the loop is too large or runs out of registers, you might be able to distribute the loop into smaller segments; distribute the loop manually or by using the distribute pragma.

• If the compiler determines the Global Acyclic Scheduler can produce better results but you think the loop should still be pipelined, use the SWP pragma on the loop.

(same as above)
IA-32 and Intel® EM64T Only

High

See Vectorization Overview and the remaining topics in the Auto-Vectorization section for applicable options.

See Vectorization Report for specific details about when you can change code.

(same as above)

Medium

Use PGO profile to guide other optimizations.

See Profile-guided Optimizations Overview.

Applications with many denormalized floating-point value operations

Significant

Attempt to use flush-to-zero where appropriate.

Decide if a high degree of precision is necessary; if it's not then using flush-to-zero might help. Denormal values require hardware or operating system interventions to handle the computation. Denormalized floating point values are those which are too small to be represented in the normal manner, that is, the mantissa cannot be left-justified. Using flush-to-zero causes denormal numbers to be treated as zero by the hardware.

Remove floating-point denormals using some common strategies:

• Change the data type to a larger data type.

• Depending on the target architecture, use flush-to-zero or vectorization options.

IA-32 and Intel® EM64T:

• Flush-to-zero mode is enabled by default for SSE2 instructions. The Intel® EM64T compiler generates SSE2 instructions by default. Enable SSE2 instructions in the IA-32 compiler by using -xW, -xN, -xB or -xP (Linux) or /QxW, /QxN, /QxB or /QxP (Windows).  

• See Vectorization Support for more information.

Itanium®:

• The most common, easiest flush-to-zero strategy is to use the -ftz (Linux) or /Qftz (Windows) option on the source file containing main().

• Selecting -O3 (Linux) or /O3 (Windows) automatically enables -ftz (Linux) or /Qftz (Windows).

After using flush-to-zero, ensure that your program still gives correct results when treating denormalized values as zero.

Sparse matrix applications

Medium

See the suggested strategy for memory pointer disambiguation (below).

Use prefetch pragma or prefetch intrinsics. Experiment with different prefetching schemes on indirect arrays.

See HLO Overview or Data Prefetching starting places for using prefetching.

Server application with branch-centric code and a fairly flat profile

Medium

Flat profile applications are those applications where no single module seems to consume CPU cycles inordinately.

Use PGO to communicate typical hot paths and functions to the compiler, so the Intel® compiler can arrange code in the optimal manner.

Use PGO on as much of the application as is feasible.

See Profile-guided Optimizations Overview.

Applications with many small functions that are called from multiple locations

Low

Use -ip (Linux and Mac OS) or /Qip (Windows) to enable inter-procedural inlining within a single source module.

Streamlines code execution for simple functions by duplicating the code within the code block  that originally called the function. This will increase application size.

As a general rule, do not inline large, complicated functions.

See Interprocedural Optimizations Overview.

(same as above)

Low

Use -ipo (Linux and Mac OS) or /Qipo (Windows) to enable inter-procedural inlining both within and between multiple source modules. You might experience an additional increase over using -ip (Linux and Mac OS) or /Qip (Windows).

Using this option will increase link time due to the extended program flow analysis that occurs.

Interprocedural Optimization (IPO) can perform whole program analysis to help memory pointer disambiguation.

Cache Blocking

High

Use -O3 (Linux and Mac OS) or /O3 (Windows) to enable automatic cache blocking; use the HLO report to determine if the compiler enabled cache blocking automatically. If not consider manual cache blocking.

See Cache Blocking.

Compiler pragmas for better alias analysis

Medium

Ignore vector dependencies. Use ivdep and other pragmas to increase application speed.

See Vectorization Support.

Memory pointer disambiguation compiler keywords and options

Medium

Use restrict keyword and the -restrict (Linux) or /Qrestrict (Windows) option to disambiguate memory pointers.

If you use the restrict keyword in your source code, you must use -restrict (Linux) or /Qrestrict (Windows) option during compilation.

Instead of using the restrict keyword and option, you can use the following compiler options:

• -fno-fnalias (Linux)

• -ansi-alias (Linux) or /Qansi-alias (Windows)

• /Oa (Windows)

• /Ow (Windows)

• -alias-args (Linux) or /Qalias-args (Windows)

Math functions

Low

Use float intrinsics for single precision data type, for example, sqrtf() not sqrt().

Call Math Kernel Library (MKL) instead of user code.

Symbol preemption

Low

Linux has a less performance-friendly symbol preemption model than Windows. Linux uses full preemption, and Windows uses no preemption. Use -fminshared -fvisibility=protected.

See Symbol Visibility Attribute Options.

Memory allocation

Low

Using third-party memory management libraries can help improve performance for applications that require extensive memory allocation.

Disk

Medium

Consider using more advanced hard drive storage strategies. For example, consider using SCSI instead of IDE.

Consider using the appropriate RAID level.

Consider increasing the number hard drives in your system.

Memory

Low

You can experience performance gains by distributing memory in a system. For example, if you have four open memory slots and only two slots are populated, populating the other two slots with memory will increase performance.

Processor

For many applications, performance scales is directly affected by processor speed, the number of processors, processor core type, and cache size.