Revision 1dbcf198f22f1b541f61a1a64f02fc8e9656755a authored by Steven Johnson on 21 December 2020, 17:08:20 UTC, committed by GitHub on 21 December 2020, 17:08:20 UTC
* Decouple wasm's +bulk-memory from threads When `wasm_threads` was added, `+bulk-memory` codegen was enabled in conjunction with this feature, due to some inscrutable error which apparently didn't get recorded. From inspection of the spec for bulk-memory, and experimentation with the most recent version of Emscripten (2.0.10), I can't find any reason that this actually needs to be enabled, so I've moved it into its own new feature flag. Also: drive-by fix in Target to format the tables in `get_runtime_compatible_target()` better, and to remove some wasm-related entries from the 'must match' table that didn't actually need to match. * Create .gitignore
1 parent ef45c87
FuzzFloatStores.h
#ifndef FUZZ_FLOAT_STORES_H
#define FUZZ_FLOAT_STORES_H
#include "Expr.h"
/** \file
* Defines a lowering pass that messes with floating point stores.
*/
namespace Halide {
namespace Internal {
/** On every store of a floating point value, mask off the
* least-significant-bit of the mantissa. We've found that whether or
* not this dramatically changes the output of a pipeline correlates
* very well with whether or not a pipeline will produce very
* different outputs on different architectures (e.g. with and without
* FMA). It's also a useful way to detect bad tests, such as those
* that expect exact floating point equality across platforms. */
Stmt fuzz_float_stores(const Stmt &s);
} // namespace Internal
} // namespace Halide
#endif
![swh spinner](/static/img/swh-spinner.gif)
Computing file changes ...