[libc-commits] [mlir] [libcxx] [lldb] [libc] [clang] [clang-tools-extra] [llvm] [lld] [compiler-rt] [MLIR][LLVM] Add Continuous Loop Peeling transform to SCF (PR #71555)

[Matthias Springer via libc-commits]

================
@@ -81,6 +81,47 @@ void naivelyFuseParallelOps(Region &region);
 LogicalResult peelForLoopAndSimplifyBounds(RewriterBase &rewriter, ForOp forOp,
                                            scf::ForOp &partialIteration);
 
+/// Rewrite a for loop with bounds/step that potentially do not divide the
+/// iteration space evenly into a chain of for loops where the step is a
+/// power of 2 and decreases exponentially across subsequent loops.
+///
+/// E.g., assuming a lower bound of 0, the following loop
+/// ```
+/// scf.for %iv = %c0 to %ub step %c8 {
+///   (loop body)
+/// }
+/// ```
+/// is rewritten into the following pseudo IR:
+/// ```
+/// %newUb = %ub - (%ub mod %c8)
+/// scf.for %iv = %c0 to %newUb step %c8 {
+///   (loop body)
+/// }
+/// %newUb2 = %ub - (%ub mod %c4)
+/// scf.for %iv2 = %newUb to %newUb2 {
----------------
matthias-springer wrote:

Step size is missing.

https://github.com/llvm/llvm-project/pull/71555