Add a pass to lower unaligned loads and stores (#2078)

This replaces the wasm2js code that lowered them to pessimistic (1-byte aligned) loads and stores. The new pass will do the optimal thing, keeping 2-byte alignment where possible.

This is also nicer as a standalone pass, which has the simple property that after it runs all loads and stores are aligned, instead of some code scattered inside wasm2js.

1 files changed, 63 insertions, 0 deletions

diff --git a/test/passes/alignment-lowering.wast b/test/passes/alignment-lowering.wast
new file mode 100644
index 000000000..135fd281e
--- /dev/null
+++ b/test/passes/alignment-lowering.wast
@@ -0,0 +1,63 @@
+(module
+ (memory $0 1 1)
+ (func $func_4
+  (drop (i32.load (i32.const 4)))
+  (drop (i32.load align=1 (i32.const 4)))
+  (drop (i32.load align=2 (i32.const 4)))
+  (drop (i32.load align=4 (i32.const 4)))
+  (drop (i32.load offset=100 (i32.const 4)))
+  (drop (i32.load offset=100 align=1 (i32.const 4)))
+  (drop (i32.load offset=100 align=2 (i32.const 4)))
+  (drop (i32.load offset=100 align=4 (i32.const 4)))
+  (drop (i32.load offset=100 align=1 (unreachable)))
+  (i32.store (i32.const 4) (i32.const 8))
+  (i32.store align=1 (i32.const 4) (i32.const 8))
+  (i32.store align=2 (i32.const 4) (i32.const 8))
+  (i32.store align=4 (i32.const 4) (i32.const 8))
+  (i32.store offset=100 (i32.const 4) (i32.const 8))
+  (i32.store offset=100 align=1 (i32.const 4) (i32.const 8))
+  (i32.store offset=100 align=2 (i32.const 4) (i32.const 8))
+  (i32.store offset=100 align=4 (i32.const 4) (i32.const 8))
+  (i32.store offset=100 align=1 (unreachable) (i32.const 8))
+  (i32.store offset=100 align=1 (i32.const 4) (unreachable))
+ )
+ (func $func_2
+  (drop (i32.load16_u (i32.const 4)))
+  (drop (i32.load16_u align=1 (i32.const 4)))
+  (drop (i32.load16_u align=2 (i32.const 4)))
+  (drop (i32.load16_u offset=100 (i32.const 4)))
+  (drop (i32.load16_u offset=100 align=1 (i32.const 4)))
+  (drop (i32.load16_u offset=100 align=2 (i32.const 4)))
+  (drop (i32.load16_u offset=100 align=1 (unreachable)))
+  (i32.store16 (i32.const 4) (i32.const 8))
+  (i32.store16 align=1 (i32.const 4) (i32.const 8))
+  (i32.store16 align=2 (i32.const 4) (i32.const 8))
+  (i32.store16 offset=100 (i32.const 4) (i32.const 8))
+  (i32.store16 offset=100 align=1 (i32.const 4) (i32.const 8))
+  (i32.store16 offset=100 align=2 (i32.const 4) (i32.const 8))
+  (i32.store16 offset=100 align=1 (unreachable) (i32.const 8))
+  (i32.store16 offset=100 align=1 (i32.const 4) (unreachable))
+ )
+ (func $func_1
+  (drop (i32.load8_u (i32.const 4)))
+  (drop (i32.load8_u align=1 (i32.const 4)))
+  (drop (i32.load8_u offset=100 (i32.const 4)))
+  (drop (i32.load8_u offset=100 align=1 (i32.const 4)))
+  (drop (i32.load8_u offset=100 align=1 (unreachable)))
+  (i32.store8 (i32.const 4) (i32.const 8))
+  (i32.store8 align=1 (i32.const 4) (i32.const 8))
+  (i32.store8 offset=100 (i32.const 4) (i32.const 8))
+  (i32.store8 offset=100 align=1 (i32.const 4) (i32.const 8))
+  (i32.store8 offset=100 align=1 (unreachable) (i32.const 8))
+  (i32.store8 offset=100 align=1 (i32.const 4) (unreachable))
+ )
+ (func $func_signed
+  (drop (i32.load16_s (i32.const 4)))
+  (drop (i32.load16_s align=1 (i32.const 4)))
+  (drop (i32.load16_s align=2 (i32.const 4)))
+  (drop (i32.load16_s offset=100 (i32.const 4)))
+  (drop (i32.load16_s offset=100 align=1 (i32.const 4)))
+  (drop (i32.load16_s offset=100 align=2 (i32.const 4)))
+  (drop (i32.load16_s offset=100 align=1 (unreachable)))
+ )
+)