Add optimization rules for some shift operations (#3099)

Specifically, truncates constant shift values that are greater than the number of bits available and optimizes out explicit masking of the shift value that is redundant with the implicit masking performed by shift operations.

1 files changed, 120 insertions, 0 deletions

diff --git a/test/passes/optimize-instructions_all-features.wast b/test/passes/optimize-instructions_all-features.wast
index 05e0e37c3..505f89ea2 100644
--- a/test/passes/optimize-instructions_all-features.wast
+++ b/test/passes/optimize-instructions_all-features.wast
@@ -4924,6 +4924,126 @@
       )
     )
   )
+  (func $optimize-shifts (param $x i32) (param $y i32) (param $z i64) (param $w i64)
+    ;; i32
+    (drop (i32.shl
+      (local.get $x)
+      (i32.const 32)
+    ))
+    (drop (i32.shr_s
+      (local.get $x)
+      (i32.const 32)
+    ))
+    (drop (i32.shr_u
+      (local.get $x)
+      (i32.const 64)
+    ))
+    (drop (i32.rotl
+      (local.get $x)
+      (i32.const 64)
+    ))
+    (drop (i32.rotr
+      (local.get $x)
+      (i32.const 64)
+    ))
+    ;; i64
+    (drop (i64.shl
+      (local.get $z)
+      (i64.const 64)
+    ))
+    (drop (i64.shr_s
+      (local.get $z)
+      (i64.const 64)
+    ))
+    (drop (i64.shr_u
+      (local.get $z)
+      (i64.const 128)
+    ))
+    (drop (i64.rotl
+      (local.get $z)
+      (i64.const 128)
+    ))
+    (drop (i64.rotr
+      (local.get $z)
+      (i64.const 128)
+    ))
+
+    ;; i32
+    (drop (i32.shl
+      (local.get $x)
+      (i32.and
+        (local.get $y)
+        (i32.const 31)
+      )
+    ))
+    (drop (i32.shl
+      (local.get $x)
+      (i32.and
+        (local.get $y)
+        (i32.const 63)
+      )
+    ))
+    (drop (i32.shr_s
+      (local.get $x)
+      (i32.and
+        (local.get $y)
+        (i32.const 31)
+      )
+    ))
+    (drop (i32.shr_u
+      (local.get $x)
+      (i32.and
+        (local.get $y)
+        (i32.const 31)
+      )
+    ))
+    ;; i64
+    (drop (i64.shl
+      (local.get $z)
+      (i64.and
+        (local.get $w)
+        (i64.const 63)
+      )
+    ))
+    (drop (i64.shl
+      (local.get $z)
+      (i64.and
+        (local.get $w)
+        (i64.const 127)
+      )
+    ))
+    (drop (i64.shr_s
+      (local.get $z)
+      (i64.and
+        (local.get $w)
+        (i64.const 63)
+      )
+    ))
+    (drop (i64.shr_u
+      (local.get $z)
+      (i64.and
+        (local.get $w)
+        (i64.const 63)
+      )
+    ))
+
+    ;; skip
+    (drop (i32.shl
+      (local.get $x)
+      (i32.and
+        (local.get $y)
+        (i32.const 32)
+      )
+    ))
+    ;; skip
+    (drop (i64.shr_u
+      (local.get $z)
+      (i64.and
+        (local.get $w)
+        (i64.const 31)
+      )
+    ))
+  )
   (func $optimize-bulk-memory-copy (param $dst i32) (param $src i32) (param $sz i32)
     (memory.copy  ;; skip
       (local.get $dst)