Revision 3657cf5f363fd64aeaf06432e62e3960800927b0 authored by Andrew Adams on 26 January 2024, 17:26:12 UTC, committed by GitHub on 26 January 2024, 17:26:12 UTC
* Fix bounds_of_nested_lanes bounds_of_nested_lanes assumed that one layer of nested vectorization could be removed at a time. When faced with the expression: min(ramp(x8(a), x8(b), 5), x40(27)) It panicked, because on the left hand side it reduced the bounds to x8(a) ... x8(a) + x8(b) * 4, and on the right hand side it reduced the bounds to 27. It then attempted to take a min of mismatched types. In general we can't assume that binary operators on nested vectors have the same nesting structure on both sides, so I just rewrote it to reduce directly to a scalar. Fixes #8038
1 parent 4590a09
LICM.h
#ifndef HALIDE_LICM_H
#define HALIDE_LICM_H
/** \file
* Methods for lifting loop invariants out of inner loops.
*/
#include "Expr.h"
namespace Halide {
namespace Internal {
/** Hoist loop-invariants out of inner loops. This is especially
* important in cases where LLVM would not do it for us
* automatically. For example, it hoists loop invariants out of cuda
* kernels. */
Stmt hoist_loop_invariant_values(Stmt);
/** Just hoist loop-invariant if statements as far up as
* possible. Does not lift other values. It's useful to run this
* earlier in lowering to simplify the IR. */
Stmt hoist_loop_invariant_if_statements(Stmt);
} // namespace Internal
} // namespace Halide
#endif
Computing file changes ...
