diff options
| author | Christian Sigg <csigg@google.com> | 2019-10-04 04:37:14 -0700 |
|---|---|---|
| committer | A. Unique TensorFlower <gardener@tensorflow.org> | 2019-10-04 04:37:53 -0700 |
| commit | 85dcaf19c721c4c745cd4fa3972aa9093acd69d3 (patch) | |
| tree | c311b0c28846249bdfaa6f124ceda4b79d5757be /mlir/lib/Quantizer/Support | |
| parent | 5830f71a45df33e24c864bea4c5de070be47b488 (diff) | |
| download | llvm-85dcaf19c721c4c745cd4fa3972aa9093acd69d3.tar.gz | |
Fix typos, NFC.
PiperOrigin-RevId: 272851237
Diffstat (limited to 'mlir/lib/Quantizer/Support')
| -rw-r--r-- | mlir/lib/Quantizer/Support/UniformSolvers.cpp | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/mlir/lib/Quantizer/Support/UniformSolvers.cpp b/mlir/lib/Quantizer/Support/UniformSolvers.cpp index b4c14ca80ef1..db5eaa0a05c6 100644 --- a/mlir/lib/Quantizer/Support/UniformSolvers.cpp +++ b/mlir/lib/Quantizer/Support/UniformSolvers.cpp @@ -82,7 +82,7 @@ bool UniformParamsFromMinMaxSolver::compute() { auto fMid = fns::overshoot(origMinAdj, origMaxAdj, numLevelsDouble, deltaMid); if (fMid == 0 || (fMid > 0 && std::fabs(deltaMid - prevDeltaMid) < 1e-15)) { - // Solution found (or step size is infinitessimal and an overshoot). + // Solution found (or step size is infinitesimal and an overshoot). // Empirically, this seems to terminate around 30-50 steps or so. // This will find a zero point for exactly representable ranges and // will terminate on a small step size for inexact, biasing towards |