Changes between Version 38 and Version 39 of Orio
- Timestamp:
- 06/12/08 00:59:16 (15 years ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
Orio
v38 v39 62 62 }}} 63 63 64 == The Structure of Orio==64 == Structure of Orio Framework == 65 65 66 66 The picture shown below depicts at a high level the structure and the optimization process of the Orio framework. … … 150 150 /*@ begin PerfTuning ( 151 151 def build { 152 arg command = 'gcc';153 arg options = '-O0';152 arg command = 'gcc'; 153 arg options = '-O0'; 154 154 } 155 155 def performance_params { 156 param UF[] = [1,2,3,4,5,6,7,8];156 param UF[] = range(1,33); 157 157 } 158 158 def input_params { 159 param N[] = [1000,10000000];159 param N[] = [1000,10000000]; 160 160 } 161 161 def input_vars { 162 decl static double y[N] = 0;163 decl double a1 = random;164 decl double a2 = random;165 decl double a3 = random;166 decl double a4 = random;167 decl static double x1[N] = random;168 decl static double x2[N] = random;169 decl static double x3[N] = random;170 decl static double x4[N] = random;162 decl static double y[N] = 0; 163 decl double a1 = random; 164 decl double a2 = random; 165 decl double a3 = random; 166 decl double a4 = random; 167 decl static double x1[N] = random; 168 decl static double x2[N] = random; 169 decl static double x3[N] = random; 170 decl static double x4[N] = random; 171 171 } 172 172 ) @*/ 173 173 int i; 174 174 /*@ begin Loop ( 175 transform Unroll(ufactor=UF) 175 transform Unroll(ufactor=UF) 176 176 for (i=0; i<=N-1; i++) 177 177 y[i] = y[i] + a1*x1[i] + a2*x2[i] + a3*x3[i] + a4*x4[i]; … … 182 182 /*@ end @*/ 183 183 }}} 184 185 The tuned application in the given example is the same AXPY-4 used in the earlier subsection. In this example, the goal of the tuning process is to determine the most optimal value of the unroll factor parameter (represented as variable UF) for two distinct problem sizes: N=1K and N=10M. The unroll factor values under consideration extends over integers from 1 to 32, inclusively. 184 186 185 187 Because of the huge search space, a proper search heuristic becomes a critical component of an empirical tuning system. Hence, in addition to an exhaustive search and a random search, two effective and practical search heuristic strategies have been developed and integrated into the Orio’s search engine. These include the ''Nelder-Mead Simplex'' method and ''Simulated Annealing'' method.