Changes between Version 40 and Version 41 of OpenCL_Memo


Ignore:
Timestamp:
Dec 12, 2011 10:50:58 PM (13 years ago)
Author:
nakasato
Comment:

--

Legend:

Unmodified
Added
Removed
Modified

  • OpenCL_Memo

    v40 v41  
    11= OpenCL Notes = 
    22== Disable auto vectorization == 
    3 In the section 6.7.2 (187 page) in the OpenCL Specification Version: 1.1 (Revision: 33), "__attribute__((vec_type_hint(<typen>))" is described. This hint controls the autovectorizer in the compiler for OpenCL C. I tested this feature by dumping the assembly code for a kernel targeted for AVX instructions with Intel SDK (version 1.5). 
     3In the section 6.7.2 (187 page) in the OpenCL Specification Version: 1.1 (Revision: 33), "__attribute__((vec_type_hint(<typen>))" is described. This hint controls the autovectorizer in the compiler for OpenCL C. I tested this feature by dumping the assembly code for a kernel (using "float8" ) targeted for AVX instructions with Intel SDK (version 1.5). 
    44 
    5 ||               || lines || 
    6 || no hint       || 1567 || 
    7 || with the hint || 333   || 
     5|| attribute       || lines ||  comment    || 
     6|| no hint         || 1567  || vectorized (inner-loop is further unrolled) || 
     7|| with the hint || 333   || simple translation of the input kernel  || 
    88 
    9 This hint amazingly reduces the size of the generated kernel code. 
    10 Still working... 
     9This hint amazingly reduces the size of the generated assembly code. 
     10Without this attribute, the generated code includes two functions: (1) the code without unrolling and (2) the code over unrolled.  
     11Due to this, the generated assembly file is large but we don't know which code is really used. 
    1112 
    1213== How to use "ioc" command equipped with Intel SDK. ==