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, 46 insertions, 20 deletions

diff --git a/test/wasm2js/unaligned.2asm.js.opt b/test/wasm2js/unaligned.2asm.js.opt
index 448e35abd..8a069f083 100644
--- a/test/wasm2js/unaligned.2asm.js.opt
+++ b/test/wasm2js/unaligned.2asm.js.opt
@@ -51,48 +51,74 @@ function asmFunc(global, env, buffer) {
  var setTempRet0 = env.setTempRet0;
  var i64toi32_i32$HIGH_BITS = 0;
  function $0() {
-  var wasm2js_i32$0 = 0;
-  return (wasm2js_i32$0 = 0, HEAPU8[wasm2js_i32$0 | 0] | HEAPU8[wasm2js_i32$0 + 1 | 0] << 8 | HEAPU8[wasm2js_i32$0 + 2 | 0] << 16 | HEAPU8[wasm2js_i32$0 + 3 | 0] << 24) | 0;
+  var $0_1 = 0;
+  return HEAPU8[0 | 0] | HEAPU8[$0_1 + 1 | 0] << 8 | (HEAPU8[$0_1 + 2 | 0] << 16 | HEAPU8[$0_1 + 3 | 0] << 24);
+ }
+ 
+ function $1() {
+  var $0_1 = 0, $1_1 = 0;
+  $0_1 = HEAPU8[0 | 0] | HEAPU8[$0_1 + 1 | 0] << 8 | (HEAPU8[$0_1 + 2 | 0] << 16 | HEAPU8[$0_1 + 3 | 0] << 24);
+  i64toi32_i32$HIGH_BITS = HEAPU8[$1_1 + 4 | 0] | HEAPU8[$1_1 + 5 | 0] << 8 | (HEAPU8[$1_1 + 6 | 0] << 16 | HEAPU8[$1_1 + 7 | 0] << 24);
+  return $0_1 | 0;
  }
  
  function $2() {
-  var wasm2js_i32$0 = 0;
-  return Math_fround((wasm2js_scratch_store_i32(0, (wasm2js_i32$0 = 0, HEAPU8[wasm2js_i32$0 | 0] | HEAPU8[wasm2js_i32$0 + 1 | 0] << 8 | HEAPU8[wasm2js_i32$0 + 2 | 0] << 16 | HEAPU8[wasm2js_i32$0 + 3 | 0] << 24)), wasm2js_scratch_load_f32()));
+  var $0_1 = 0;
+  return Math_fround((wasm2js_scratch_store_i32(0, HEAPU8[0 | 0] | HEAPU8[$0_1 + 1 | 0] << 8 | (HEAPU8[$0_1 + 2 | 0] << 16 | HEAPU8[$0_1 + 3 | 0] << 24)), wasm2js_scratch_load_f32()));
  }
  
  function $3() {
-  var $0_1 = 0, wasm2js_i32$0 = 0;
-  $0_1 = (wasm2js_i32$0 = $0_1, HEAPU8[wasm2js_i32$0 + 4 | 0] | HEAPU8[wasm2js_i32$0 + 5 | 0] << 8 | HEAPU8[wasm2js_i32$0 + 6 | 0] << 16 | HEAPU8[wasm2js_i32$0 + 7 | 0] << 24);
-  wasm2js_scratch_store_i32(0 | 0, (wasm2js_i32$0 = 0, HEAPU8[wasm2js_i32$0 | 0] | HEAPU8[wasm2js_i32$0 + 1 | 0] << 8 | HEAPU8[wasm2js_i32$0 + 2 | 0] << 16 | HEAPU8[wasm2js_i32$0 + 3 | 0] << 24) | 0);
+  var $0_1 = 0, $1_1 = 0;
+  $1_1 = HEAPU8[0 | 0] | HEAPU8[$1_1 + 1 | 0] << 8 | (HEAPU8[$1_1 + 2 | 0] << 16 | HEAPU8[$1_1 + 3 | 0] << 24);
+  $0_1 = HEAPU8[$0_1 + 4 | 0] | HEAPU8[$0_1 + 5 | 0] << 8 | (HEAPU8[$0_1 + 6 | 0] << 16 | HEAPU8[$0_1 + 7 | 0] << 24);
+  wasm2js_scratch_store_i32(0 | 0, $1_1 | 0);
   wasm2js_scratch_store_i32(1 | 0, $0_1 | 0);
   return +wasm2js_scratch_load_f64();
  }
  
  function $4() {
-  var wasm2js_i32$0 = 0, wasm2js_i32$1 = 0;
-  (wasm2js_i32$0 = 0, wasm2js_i32$1 = 0), ((HEAP8[wasm2js_i32$0 | 0] = wasm2js_i32$1, HEAP8[wasm2js_i32$0 + 1 | 0] = wasm2js_i32$1 >>> 8), HEAP8[wasm2js_i32$0 + 2 | 0] = wasm2js_i32$1 >>> 16), HEAP8[wasm2js_i32$0 + 3 | 0] = wasm2js_i32$1 >>> 24;
+  var $0_1 = 0, $1_1 = 0;
+  HEAP8[0 | 0] = 0;
+  HEAP8[$0_1 + 1 | 0] = $1_1 >>> 8;
+  HEAP8[$0_1 + 2 | 0] = $1_1 >>> 16;
+  HEAP8[$0_1 + 3 | 0] = $1_1 >>> 24;
  }
  
  function $5() {
-  var $0_1 = 0, wasm2js_i32$0 = 0, wasm2js_i32$1 = 0;
-  (wasm2js_i32$0 = 0, wasm2js_i32$1 = 0), ((HEAP8[wasm2js_i32$0 | 0] = wasm2js_i32$1, HEAP8[wasm2js_i32$0 + 1 | 0] = wasm2js_i32$1 >>> 8), HEAP8[wasm2js_i32$0 + 2 | 0] = wasm2js_i32$1 >>> 16), HEAP8[wasm2js_i32$0 + 3 | 0] = wasm2js_i32$1 >>> 24;
-  (wasm2js_i32$0 = $0_1, wasm2js_i32$1 = 0), ((HEAP8[wasm2js_i32$0 + 4 | 0] = wasm2js_i32$1, HEAP8[wasm2js_i32$0 + 5 | 0] = wasm2js_i32$1 >>> 8), HEAP8[wasm2js_i32$0 + 6 | 0] = wasm2js_i32$1 >>> 16), HEAP8[wasm2js_i32$0 + 7 | 0] = wasm2js_i32$1 >>> 24;
+  var $0_1 = 0, $1_1 = 0, $2_1 = 0;
+  HEAP8[0 | 0] = 0;
+  HEAP8[$0_1 + 1 | 0] = $2_1 >>> 8;
+  HEAP8[$0_1 + 2 | 0] = $2_1 >>> 16;
+  HEAP8[$0_1 + 3 | 0] = $2_1 >>> 24;
+  HEAP8[$1_1 + 4 | 0] = 0;
+  HEAP8[$1_1 + 5 | 0] = $0_1 >>> 8;
+  HEAP8[$1_1 + 6 | 0] = $0_1 >>> 16;
+  HEAP8[$1_1 + 7 | 0] = $0_1 >>> 24;
  }
  
  function $7() {
-  var $0_1 = 0, $1 = 0, wasm2js_i32$0 = 0, wasm2js_i32$1 = 0;
+  var $0_1 = 0, $1_1 = 0, $2_1 = 0, $3_1 = 0;
   wasm2js_scratch_store_f64(0.0);
   $0_1 = wasm2js_scratch_load_i32(1 | 0) | 0;
-  (wasm2js_i32$0 = 0, wasm2js_i32$1 = wasm2js_scratch_load_i32(0 | 0) | 0), ((HEAP8[wasm2js_i32$0 | 0] = wasm2js_i32$1, HEAP8[wasm2js_i32$0 + 1 | 0] = wasm2js_i32$1 >>> 8), HEAP8[wasm2js_i32$0 + 2 | 0] = wasm2js_i32$1 >>> 16), HEAP8[wasm2js_i32$0 + 3 | 0] = wasm2js_i32$1 >>> 24;
-  (wasm2js_i32$0 = $1, wasm2js_i32$1 = $0_1), ((HEAP8[wasm2js_i32$0 + 4 | 0] = wasm2js_i32$1, HEAP8[wasm2js_i32$0 + 5 | 0] = wasm2js_i32$1 >>> 8), HEAP8[wasm2js_i32$0 + 6 | 0] = wasm2js_i32$1 >>> 16), HEAP8[wasm2js_i32$0 + 7 | 0] = wasm2js_i32$1 >>> 24;
+  $1_1 = wasm2js_scratch_load_i32(0 | 0) | 0;
+  HEAP8[0 | 0] = $1_1;
+  HEAP8[$3_1 + 1 | 0] = $1_1 >>> 8;
+  HEAP8[$3_1 + 2 | 0] = $1_1 >>> 16;
+  HEAP8[$3_1 + 3 | 0] = $1_1 >>> 24;
+  HEAP8[$2_1 + 4 | 0] = $0_1;
+  HEAP8[$2_1 + 5 | 0] = $0_1 >>> 8;
+  HEAP8[$2_1 + 6 | 0] = $0_1 >>> 16;
+  HEAP8[$2_1 + 7 | 0] = $0_1 >>> 24;
  }
  
  function legalstub$1() {
-  var $0_1 = 0, wasm2js_i32$0 = 0;
-  $0_1 = (wasm2js_i32$0 = 0, HEAPU8[wasm2js_i32$0 | 0] | HEAPU8[wasm2js_i32$0 + 1 | 0] << 8 | HEAPU8[wasm2js_i32$0 + 2 | 0] << 16 | HEAPU8[wasm2js_i32$0 + 3 | 0] << 24);
-  i64toi32_i32$HIGH_BITS = (wasm2js_i32$0 = 4, HEAPU8[wasm2js_i32$0 | 0] | HEAPU8[wasm2js_i32$0 + 1 | 0] << 8 | HEAPU8[wasm2js_i32$0 + 2 | 0] << 16 | HEAPU8[wasm2js_i32$0 + 3 | 0] << 24);
-  setTempRet0(i64toi32_i32$HIGH_BITS | 0);
-  return $0_1;
+  var $0_1 = 0, $1_1 = 0, $2_1 = 0, $3_1 = 0;
+  $1_1 = $1();
+  $2_1 = i64toi32_i32$HIGH_BITS;
+  $3_1 = 32;
+  $0_1 = $3_1 & 31;
+  setTempRet0((32 >>> 0 <= $3_1 >>> 0 ? $2_1 >>> $0_1 : ((1 << $0_1) - 1 & $2_1) << 32 - $0_1 | $1_1 >>> $0_1) | 0);
+  return $1_1;
  }
  
  var FUNCTION_TABLE = [];