Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore a conference presentation introducing a compiler that performs tiling optimizations for programs expressed as mathematical recurrence equations. Learn how REPTILE recursively decomposes recurrence programs into unique tiles and simplifies each into different sets of recurrences. Discover the compiler's approach to generating C code that composes compiler-generated loops with calls to external hand-optimized libraries, given declarative user specifications of recurrence equations, optimizations, and optional mappings. Examine performance evaluations demonstrating how the generated C code matches and often exceeds the performance of standard hand-optimized libraries for direct linear solvers expressible as recurrence equations. Review comparative results against Intel MKL implementations and bioinformatics algorithms including Needleman-Wunsch and Smith-Waterman, with speedup achievements ranging from 0.79x to 2.72x across various linear algebra operations including LU decomposition, Cholesky decomposition, lower triangular matrix inversion, and triangular solve operations.
