diff options
| author | Thomas Lively <7121787+tlively@users.noreply.github.com> | 2018-12-19 19:22:09 -0800 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2018-12-19 19:22:09 -0800 |
| commit | 0f41b0708384c1f5d85304d5ed94d9edd57d38c9 (patch) | |
| tree | b79bc880af6e8534125fa9fb75a89932af4cd7d1 /test | |
| parent | fcbcf3bd670a9ee793a836be49d825b944baf501 (diff) | |
| download | binaryen-0f41b0708384c1f5d85304d5ed94d9edd57d38c9.tar.gz binaryen-0f41b0708384c1f5d85304d5ed94d9edd57d38c9.tar.bz2 binaryen-0f41b0708384c1f5d85304d5ed94d9edd57d38c9.zip | |
Do not precompute v128 expressions (#1839)
Without this change, sequences like `i32.const 0, i32x4.splat` will
get precomputed to v128.const ops, which are much larger and also not
implemented in V8 yet. Until we have SIMD-aware optimization passes or
at least engine support for v128.const, do not perform such
transformations.
Diffstat (limited to 'test')
| -rw-r--r-- | test/passes/precompute.txt | 6 | ||||
| -rw-r--r-- | test/passes/precompute.wast | 5 |
2 files changed, 11 insertions, 0 deletions
diff --git a/test/passes/precompute.txt b/test/passes/precompute.txt index 1c060fbef..b199da603 100644 --- a/test/passes/precompute.txt +++ b/test/passes/precompute.txt @@ -3,6 +3,7 @@ (type $1 (func (result i32))) (type $2 (func)) (type $3 (func (result f64))) + (type $4 (func (result v128))) (memory $0 0) (global $global i32 (i32.const 1)) (global $global-mut (mut i32) (i32.const 2)) @@ -211,4 +212,9 @@ (i32.const 2) ) ) + (func $no-simd-precompute (; 11 ;) (type $4) (result v128) + (i32x4.splat + (i32.const 0) + ) + ) ) diff --git a/test/passes/precompute.wast b/test/passes/precompute.wast index 2c87d3706..54c8cf6d2 100644 --- a/test/passes/precompute.wast +++ b/test/passes/precompute.wast @@ -306,4 +306,9 @@ ) ) ) + (func $no-simd-precompute (result v128) + (i32x4.splat + (i32.const 0) + ) + ) ) |