The vectorization report can provide information about loops that could take advantage of Intel® Streaming SIMD Extensions (Intel® SSE3, SSE2, and SSE) vectorization, and it is available on systems based on IA-32 and Intel® 64 architectures.
See Using Parallelism for information on other vectorization options.
The -vec-report (Linux* and Mac OS* X) or /Qvec-report (Windows*) option directs the compiler to generate the vectorization reports with different levels of information. Specify a value of 3 to generate the maximum diagnostic details.
Operating System |
Command |
|---|---|
Linux and Mac OS X |
ifort -c -xSSSE3 -vec-report3 sample.f90 |
Windows |
ifort /c /QxSSSE3 /Qvec-report:3 sample.f90 |
where -c (Linux and Mac OS X) or /c (Windows) instructs the compiler to compile the example without generating an executable.
Note
Linux and Mac OS X: The space between the option and the phase is optional.
Windows: The colon between the option and phase is optional.
Some vectorization report messages and their explanations are provided below:
- Low trip count: The loop does not have sufficient iterations for vectorization to be worthwhile.
- Not inner loop: Only the inner loop of a loop nest may be vectorized.
- Existence of vector dependence: The compiler did not vectorize the loop because of a proven or potential dependence. If you are sure that any potential dependencies are not realized, you may invite the compiler to ignore them using #pragma ivdep.
- Vectorization possible but seems inefficient: The compiler concludes that vectorization may not improve the performance of this loop. You may use #pragma vector always to override the compiler’s performance assessment, and ask the compiler to vectorize anyway if it is safe to do so.
- Condition may protect exception: When the compiler tries to vectorize a loop containing an if statement, it typically evaluates the RHS expressions for all values of the loop index, but only makes the final assignment in those cases where the conditional evaluates to TRUE. In some cases, the compiler may not vectorize because the condition may be protecting against accessing an illegal memory address. You can use the #pragma ivdep to reassure the compiler that the conditional is not protecting against a memory exception in such cases.
- Data type unsupported on given target architecture: This message might occur when compiling a loop containing complex arithmetic for a target processor that supported only the Streaming SIMD Extensions 2 (SSE2) instruction set. SSE3 instructions are needed for effective vectorization of arithmetic involving complex data types.
- Statement cannot be vectorized: Certain statements, such as switch statements, cannot be vectorized.
- Subscript too complex: An array subscript may be too complicated for the compiler to decipher the memory access pattern. Try to write subscripts as an explicit function of the main loop counter.
- Unsupported loop structure or Top test could not be found: Loops that do not fulfill the requirements of countability, single entry and exit, and so on, may generate these messages.
- Operator unsuited for vectorization: Certain operators, such as the % (modulus) operator, cannot be vectorized.
The following example results illustrate the type of information generated by the vectorization report:
Example results |
|---|
sample.f90(27) : (col. 9) remark: loop was not vectorized: not inner loop. sample.f90(28) : (col. 11) remark: LOOP WAS VECTORIZED. sample.f90(31) : (col. 9) remark: loop was not vectorized: not inner loop. sample.f90(32) : (col. 11) remark: LOOP WAS VECTORIZED. sample.f90(37) : (col. 10) remark: loop was not vectorized: not inner loop. sample.f90(38) : (col. 12) remark: loop was not vectorized: not inner loop. sample.f90(40) : (col. 14) remark: loop was not vectorized: vectorization possible but seems inefficient. sample.f90(46) : (col. 10) remark: loop was not vectorized: not inner loop. sample.f90(47) : (col. 12) remark: loop was not vectorized: contains unvectorizable statement at line 48. |
If the compiler reports "remark: loop was not vectorized" because of the existence of vector dependence, then you should analyze the loop for vector dependence. If you determine there is no legitimate vector dependence, then the message indicates that the compiler was assuming the pointers or arrays in the loop were dependent, which implies the pointers or arrays were aliased. Use memory disambiguation techniques to resolve these cases.
There are three major types of vector dependence: FLOW, ANTI, and OUTPUT.
There are a number of situations where the vectorization report may indicate vector dependencies. The following situations will sometimes be reported as vector dependencies, non-unit stride, low trip count, and complex subscript expression.
Non-Unit Stride
The report might indicate that a loop could not be vectorized when the memory is accessed in a non-Unit Stride manner. This means that nonconsecutive memory locations are being accessed in the loop. In such cases, see if loop interchange can help or if it is practical. If not then you can force vectorization sometimes through vector always directive; however, you should verify improvement.