Hello all,<br><br>The performance comparison between Polly's Cloog and ISL code generator is posted on http://188.40.87.11:8000/db_default/v4/nts/59?compare_to=58&baseline=58<br><br>It seems their execution-time performance are comparable:<br><br>Performance Regressions - Execution Time  (ISL over Cloog)<br>MultiSource/Benchmarks/TSVC/ControlFlow-flt/ControlFlow-flt    8.49%<br>MultiSource/Benchmarks/TSVC/StatementReordering-flt/StatementReordering-flt    6.77%<br>MultiSource/Benchmarks/TSVC/CrossingThresholds-flt/CrossingThresholds-flt    2.65%<br>SingleSource/UnitTests/Vectorizer/gcc-loops    2.63%<br><br>Performance Improvements - Execution Time  (ISL over Cloog)<br>MultiSource/Benchmarks/TSVC/NodeSplitting-flt/NodeSplitting-flt    -6.77%<br>MultiSource/Benchmarks/ASC_Sequoia/AMGmk/AMGmk    -3.03%<br><br>However, ISL outperforms Cloog in compile-time performance.  With ISL code generator, 22 benchmarks have >10% compile-time performance improvement over Cloog. Top 10 improvements are shown as follows:<br><br>Performance Improvements - Compile Time (ISL over Cloog)<br>MultiSource/Benchmarks/ASC_Sequoia/AMGmk/AMGmk    -69.11%<br>MultiSource/Benchmarks/Trimaran/netbench-crc/netbench-crc    -44.39%<br>SingleSource/Benchmarks/Polybench/linear-algebra/kernels/3mm/3mm    -12.74%<br>SingleSource/Benchmarks/Polybench/linear-algebra/kernels/gemm/gemm    -11.21%<br>MultiSource/Benchmarks/TSVC/LoopRerolling-flt/LoopRerolling-flt    -11.14%<br>SingleSource/Benchmarks/Polybench/linear-algebra/kernels/syr2k/syr2k    -11.11%<br>MultiSource/Benchmarks/TSVC/Packing-flt/Packing-flt    -10.87%<br>MultiSource/Benchmarks/TSVC/Searching-dbl/Searching-dbl    -10.87%<br>SingleSource/Benchmarks/Polybench/linear-algebra/kernels/2mm/2mm    -10.74%<br>SingleSource/Benchmarks/Polybench/linear-algebra/kernels/doitgen/doitgen    -10.66%<br>...<br><br>Star Tan<br>