From b60993014298a4943b7d9660375814393a75519e Mon Sep 17 00:00:00 2001
From: Nuno Cruces <ncruces@users.noreply.github.com>
Date: Wed, 7 May 2025 12:30:19 +0100
Subject: [PATCH] Refactor #274.

---
 sqlite3/libc/libc.wasm | Bin 3754 -> 3761 bytes
 sqlite3/libc/libc.wat  | 533 +++++++++++++++++++++--------------------
 sqlite3/libc/string.h  | 174 +++++++-------
 3 files changed, 353 insertions(+), 354 deletions(-)

diff --git a/sqlite3/libc/libc.wasm b/sqlite3/libc/libc.wasm
index 9352b8eee6c1ccc40a36fee8779059a77875e6c0..e34936036794e6898620413504bd48177ce76ae2 100755
GIT binary patch
delta 653
zcmZ`$u}<7T5Z&3?vkwgNNfAQgLNe<Vkpd`)k_K-<l;{cyM4ci?AiK$3PS=3sAE4WY
zmWn79$x%l62KWO$fi{Jeyj?DcCR$lu&Ca~{=DlyL-&e2R5PhAl(L}$vCaoJf(dCYE
zy|E~v@D2KPp<VBT|HW??tP<QqAs_hP{MxubH1uM&)_?qWm__68pdVoNOv$A`HL^^u
zS6MKG-xJj7)ST1vGoww@IoSdKhEfIZn$oE}i?aMt@^Zi~9|G?#9TgqvpvRH0Gm}u}
z@n{=``&804GXeV^Z4mLVty{;Ah;7;5CM(Z)QDvemi-{VYvgDh&C1Ssve%!3{4CoMw
zcY*M_pb8FP21Bm-B8*kIbT^=sMA%=XO9?x|fd84=M$ce{+uSAq*IdF|UA0OnN!nLM
z+J3^05PS%@W4wS7c2zcx)`fYXi$hk=SaoEDAP`Yp4>P#K?RQZdWi7)<zu&#|ZFjtg
zPz#E2B8Rvi`SZDhhNRmb9#^8mqO)KgbACW_R0q{z#RN+Ro<gr98*=J>Kl)5CAMYdB
j3=m|I4kwO=yTMuw>}*XzKjA=NkC6z~0fwfZ=eNyYn>}~_

delta 685
zcmY+BJ&P1U5QeL&d$#x7U3Ws|T&x+813wT9O>SDa!$TI4!&LDjkXaCRT{gL3H8bzT
zRLm~%WHS>ZbAN(A!B}YV?e<2!WTvL3s^0gh`ZPJ7oLpsozq!UMnS5lCH6KZ{OTJwC
za#oAd!HASq?u9Sr=CCG2BTk8F_-S^+6?^N+N!RvYqsfh?L8VCz;>&~77l}wIJ?kPq
zh=(K5B0k%*ji>P-Rn00<_GMeK!bA%}Gu?c)@A%RiD|THR?ZVCtw99Ezb{byxF0aLp
z=e6<Q-J&kEK#fnjc9qhyT--+Dhf@~^7DEv(YNMtr>(6pUo?QusM6y)%P?iG+Q=Cl6
zhz3e9NouFbcEzTNLlzr`?mQ-Q8@Ao!{T4X*NC@0wk9LYir2P7o3I>_Re}ftAsayvQ
zdAhruopm}%PQq7LUBE80mP#DMQC3W|!!LL74xbe=XM{`^>Z~VLfMc#v+0S-(@YaT>
z{bPeIhT+z7XNH|n&G$2zoRalRZC-;gzo9H`_^|b`?Z@XY5NA+a4wPWUx&DDdivEE1
kfXWF}{(|B>_}!UH$HPVz$zwj6o3A|2_wu?okIo(1-!g4^oB#j-

diff --git a/sqlite3/libc/libc.wat b/sqlite3/libc/libc.wat
index 0f86ea1..2c62f2a 100644
--- a/sqlite3/libc/libc.wat
+++ b/sqlite3/libc/libc.wat
@@ -550,16 +550,17 @@
         (memory.size)
         (i32.const 16)
        )
-       (i32.const 16)
+       (select
+        (local.get $0)
+        (local.get $1)
+        (i32.gt_u
+         (local.get $0)
+         (local.get $1)
+        )
+       )
       )
      )
-     (local.get $0)
-    )
-   )
-   (br_if $block
-    (i32.gt_u
-     (local.get $1)
-     (local.get $2)
+     (i32.const 16)
     )
    )
    (loop $label
@@ -595,21 +596,21 @@
       (i32.const 16)
      )
     )
-    (br_if $block
-     (i32.gt_u
-      (local.tee $0
-       (i32.add
-        (local.get $0)
-        (i32.const 16)
-       )
-      )
-      (local.get $2)
+    (local.set $0
+     (i32.add
+      (local.get $0)
+      (i32.const 16)
      )
     )
     (br_if $label
-     (i32.le_u
-      (local.get $1)
-      (local.get $2)
+     (i32.gt_u
+      (local.tee $2
+       (i32.sub
+        (local.get $2)
+        (i32.const 16)
+       )
+      )
+      (i32.const 15)
      )
     )
    )
@@ -643,7 +644,10 @@
     (loop $label1
      (if
       (i32.eqz
-       (local.get $2)
+       (i32.and
+        (local.get $2)
+        (i32.const 255)
+       )
       )
       (then
        (return
@@ -691,87 +695,91 @@
   (local $3 i32)
   (local $4 i32)
   (local $5 v128)
-  (block $block1
-   (block $block
-    (br_if $block
-     (i32.lt_u
-      (local.tee $3
-       (i32.sub
-        (i32.shl
-         (memory.size)
-         (i32.const 16)
+  (block $block
+   (if
+    (i32.ge_u
+     (local.tee $2
+      (select
+       (local.tee $3
+        (i32.sub
+         (i32.shl
+          (memory.size)
+          (i32.const 16)
+         )
+         (select
+          (local.get $0)
+          (local.get $1)
+          (i32.gt_u
+           (local.get $0)
+           (local.get $1)
+          )
+         )
         )
-        (i32.const 16)
+       )
+       (local.get $2)
+       (i32.gt_u
+        (local.get $2)
+        (local.get $3)
        )
       )
-      (local.get $0)
      )
+     (i32.const 16)
     )
-    (loop $label
-     (br_if $block
-      (i32.gt_u
-       (local.get $1)
-       (local.get $3)
-      )
-     )
-     (br_if $block
-      (i32.lt_u
-       (local.get $2)
-       (i32.const 16)
-      )
-     )
-     (br_if $block1
-      (v128.any_true
-       (v128.xor
-        (v128.load align=1
-         (local.get $1)
-        )
-        (local.tee $5
+    (then
+     (loop $label
+      (br_if $block
+       (v128.any_true
+        (v128.xor
          (v128.load align=1
-          (local.get $0)
+          (local.get $1)
+         )
+         (local.tee $5
+          (v128.load align=1
+           (local.get $0)
+          )
          )
         )
        )
       )
-     )
-     (if
-      (i32.eqz
-       (i8x16.all_true
-        (local.get $5)
-       )
-      )
-      (then
-       (return
-        (i32.const 0)
-       )
-      )
-     )
-     (local.set $2
-      (i32.sub
-       (local.get $2)
-       (i32.const 16)
-      )
-     )
-     (local.set $1
-      (i32.add
-       (local.get $1)
-       (i32.const 16)
-      )
-     )
-     (br_if $label
-      (i32.le_u
-       (local.tee $0
-        (i32.add
-         (local.get $0)
-         (i32.const 16)
+      (if
+       (i32.eqz
+        (i8x16.all_true
+         (local.get $5)
         )
        )
-       (local.get $3)
+       (then
+        (return
+         (i32.const 0)
+        )
+       )
+      )
+      (local.set $1
+       (i32.add
+        (local.get $1)
+        (i32.const 16)
+       )
+      )
+      (local.set $0
+       (i32.add
+        (local.get $0)
+        (i32.const 16)
+       )
+      )
+      (br_if $label
+       (i32.gt_u
+        (local.tee $2
+         (i32.sub
+          (local.get $2)
+          (i32.const 16)
+         )
+        )
+        (i32.const 15)
+       )
       )
      )
     )
    )
-   (br_if $block1
+   (br_if $block
     (local.get $2)
    )
    (return
@@ -1077,7 +1085,7 @@
       )
       (if
        (i32.eqz
-        (local.tee $2
+        (local.tee $3
          (i32.load8_u
           (local.get $1)
          )
@@ -1093,7 +1101,7 @@
      )
      (i32.const 16)
     )
-    (i32.const 16)
+    (local.get $0)
    )
   )
   (block $block
@@ -1116,9 +1124,9 @@
       (i32.store8
        (local.tee $5
         (i32.or
-         (local.tee $3
+         (local.tee $2
           (i32.and
-           (local.get $2)
+           (local.get $3)
            (i32.const 15)
           )
          )
@@ -1131,10 +1139,10 @@
         )
         (i32.shl
          (i32.const 1)
-         (local.tee $2
+         (local.tee $3
           (i32.and
            (i32.shr_u
-            (local.get $2)
+            (local.get $3)
             (i32.const 4)
            )
            (i32.const 15)
@@ -1148,26 +1156,26 @@
        (local.get $7)
       )
       (i32.store8
-       (local.tee $3
+       (local.tee $2
         (i32.or
-         (local.get $3)
+         (local.get $2)
          (i32.const 1008)
         )
        )
        (i32.or
         (i32.load8_u
-         (local.get $3)
+         (local.get $2)
         )
         (i32.shl
          (i32.const 1)
          (i32.sub
-          (local.get $2)
+          (local.get $3)
           (i32.const 8)
          )
         )
        )
       )
-      (local.set $2
+      (local.set $3
        (i32.load8_u
         (local.get $1)
        )
@@ -1189,16 +1197,16 @@
        )
       )
       (br_if $label
-       (local.get $2)
+       (local.get $3)
       )
      )
-     (local.set $2
+     (local.set $3
       (local.get $0)
      )
      (if
-      (i32.le_u
-       (local.get $0)
+      (i32.ge_u
        (local.get $4)
+       (i32.const 16)
       )
       (then
        (local.set $1
@@ -1217,10 +1225,7 @@
                 (i8x16.shr_u
                  (local.tee $8
                   (v128.load align=1
-                   (i32.add
-                    (local.get $0)
-                    (local.get $1)
-                   )
+                   (local.get $3)
                   )
                  )
                  (i32.const 4)
@@ -1254,7 +1259,7 @@
          )
          (then
           (return
-           (i32.add
+           (i32.sub
             (i32.ctz
              (i32.xor
               (i8x16.bitmask
@@ -1268,20 +1273,24 @@
           )
          )
         )
+        (local.set $3
+         (i32.add
+          (local.get $3)
+          (i32.const 16)
+         )
+        )
         (br_if $label1
-         (i32.le_u
-          (local.tee $2
-           (i32.add
-            (local.get $0)
-            (local.tee $1
-             (i32.add
-              (local.get $1)
-              (i32.const 16)
-             )
+         (i32.gt_u
+          (i32.add
+           (local.get $4)
+           (local.tee $1
+            (i32.sub
+             (local.get $1)
+             (i32.const 16)
             )
            )
           )
-          (local.get $4)
+          (i32.const 15)
          )
         )
        )
@@ -1293,7 +1302,7 @@
         (local.get $0)
         (i32.const -1)
        )
-       (local.get $2)
+       (local.get $3)
       )
      )
      (loop $label2
@@ -1305,7 +1314,7 @@
         (i32.lt_s
          (local.tee $0
           (i32.load8_s
-           (local.get $2)
+           (local.get $3)
           )
          )
          (i32.const 0)
@@ -1318,9 +1327,9 @@
         (i32.const 1)
        )
       )
-      (local.set $2
+      (local.set $3
        (i32.add
-        (local.get $2)
+        (local.get $3)
         (i32.const 1)
        )
       )
@@ -1351,21 +1360,21 @@
      (br $block)
     )
    )
-   (local.set $3
+   (local.set $2
     (local.get $0)
    )
    (if
-    (i32.le_u
-     (local.get $0)
+    (i32.ge_u
      (local.get $4)
+     (i32.const 16)
     )
     (then
      (local.set $6
       (i8x16.splat
-       (local.get $2)
+       (local.get $3)
       )
      )
-     (local.set $1
+     (local.set $2
       (i32.const 0)
      )
      (loop $label3
@@ -1377,7 +1386,7 @@
            (v128.load align=1
             (i32.add
              (local.get $0)
-             (local.get $1)
+             (local.get $2)
             )
            )
            (local.get $6)
@@ -1396,28 +1405,35 @@
             (i32.const -1)
            )
           )
-          (local.get $1)
+          (local.get $2)
          )
         )
        )
       )
+      (local.set $2
+       (i32.add
+        (local.get $2)
+        (i32.const 16)
+       )
+      )
       (br_if $label3
-       (i32.le_u
-        (local.tee $3
-         (i32.add
-          (local.get $0)
-          (local.tee $1
-           (i32.add
-            (local.get $1)
-            (i32.const 16)
-           )
-          )
+       (i32.gt_u
+        (local.tee $4
+         (i32.sub
+          (local.get $4)
+          (i32.const 16)
          )
         )
-        (local.get $4)
+        (i32.const 15)
        )
       )
      )
+     (local.set $2
+      (i32.add
+       (local.get $0)
+       (local.get $2)
+      )
+     )
     )
    )
    (local.set $1
@@ -1426,7 +1442,7 @@
       (local.get $0)
       (i32.const -1)
      )
-     (local.get $3)
+     (local.get $2)
     )
    )
    (loop $label4
@@ -1441,18 +1457,18 @@
       (block (result i32)
        (local.set $scratch_12
         (i32.load8_u
-         (local.get $3)
+         (local.get $2)
         )
        )
-       (local.set $3
+       (local.set $2
         (i32.add
-         (local.get $3)
+         (local.get $2)
          (i32.const 1)
         )
        )
        (local.get $scratch_12)
       )
-      (local.get $2)
+      (local.get $3)
      )
     )
    )
@@ -1584,109 +1600,106 @@
   )
   (if
    (i32.ge_u
-    (local.tee $3
+    (local.tee $2
      (i32.sub
-      (block (result i32)
-       (local.set $scratch
-        (i32.shl
+      (i32.shl
+       (block (result i32)
+        (local.set $scratch
          (memory.size)
-         (i32.const 16)
         )
-       )
-       (loop $label1
-        (v128.store
-         (i32.const 1008)
-         (local.get $6)
-        )
-        (i32.store8
-         (i32.or
-          (local.tee $3
-           (i32.and
-            (local.tee $2
-             (i32.load8_u
-              (local.get $1)
+        (loop $label1
+         (v128.store
+          (i32.const 1008)
+          (local.get $6)
+         )
+         (i32.store8
+          (i32.or
+           (local.tee $3
+            (i32.and
+             (local.tee $2
+              (i32.load8_u
+               (local.get $1)
+              )
              )
+             (i32.const 15)
+            )
+           )
+           (i32.const 1008)
+          )
+          (i32.or
+           (i32.load8_u
+            (i32.or
+             (local.get $3)
+             (i32.const 1008)
+            )
+           )
+           (i32.shl
+            (i32.const 1)
+            (i32.sub
+             (local.tee $5
+              (i32.shr_u
+               (local.get $2)
+               (i32.const 4)
+              )
+             )
+             (i32.const 8)
             )
-            (i32.const 15)
            )
           )
-          (i32.const 1008)
          )
-         (i32.or
-          (i32.load8_u
+         (v128.store
+          (i32.const 992)
+          (local.get $7)
+         )
+         (i32.store8
+          (local.tee $3
            (i32.or
             (local.get $3)
-            (i32.const 1008)
+            (i32.const 992)
            )
           )
-          (i32.shl
-           (i32.const 1)
-           (i32.sub
-            (local.tee $5
-             (i32.shr_u
-              (local.get $2)
-              (i32.const 4)
-             )
-            )
-            (i32.const 8)
+          (i32.or
+           (i32.load8_u
+            (local.get $3)
+           )
+           (i32.shl
+            (i32.const 1)
+            (local.get $5)
            )
           )
          )
-        )
-        (v128.store
-         (i32.const 992)
-         (local.get $7)
-        )
-        (i32.store8
-         (local.tee $3
-          (i32.or
-           (local.get $3)
+         (local.set $1
+          (i32.add
+           (local.get $1)
+           (i32.const 1)
+          )
+         )
+         (local.set $6
+          (v128.load
+           (i32.const 1008)
+          )
+         )
+         (local.set $7
+          (v128.load
            (i32.const 992)
           )
          )
-         (i32.or
-          (i32.load8_u
-           (local.get $3)
-          )
-          (i32.shl
-           (i32.const 1)
-           (local.get $5)
-          )
+         (br_if $label1
+          (local.get $2)
          )
         )
-        (local.set $1
-         (i32.add
-          (local.get $1)
-          (i32.const 1)
-         )
-        )
-        (local.set $6
-         (v128.load
-          (i32.const 1008)
-         )
-        )
-        (local.set $7
-         (v128.load
-          (i32.const 992)
-         )
-        )
-        (br_if $label1
-         (local.get $2)
-        )
+        (local.get $scratch)
        )
-       (local.get $scratch)
+       (i32.const 16)
+      )
+      (local.tee $1
+       (local.get $0)
       )
-      (i32.const 16)
      )
     )
-    (local.tee $1
-     (local.get $0)
-    )
+    (i32.const 16)
    )
    (then
-    (local.set $2
-     (i32.const 0)
-    )
     (loop $label2
      (if
       (v128.any_true
@@ -1699,10 +1712,7 @@
             (i8x16.shr_u
              (local.tee $8
               (v128.load align=1
-               (i32.add
-                (local.get $0)
-                (local.get $2)
-               )
+               (local.get $1)
               )
              )
              (i32.const 4)
@@ -1735,43 +1745,44 @@
       )
       (then
        (return
-        (i32.add
-         (i32.ctz
-          (i8x16.bitmask
-           (local.get $8)
+        (i32.sub
+         (i32.add
+          (local.get $1)
+          (i32.ctz
+           (i8x16.bitmask
+            (local.get $8)
+           )
           )
          )
-         (local.get $2)
+         (local.get $0)
         )
        )
       )
      )
+     (local.set $1
+      (i32.add
+       (local.get $1)
+       (i32.const 16)
+      )
+     )
      (br_if $label2
-      (i32.le_u
-       (local.tee $1
-        (i32.add
-         (local.get $0)
-         (local.tee $2
-          (i32.add
-           (local.get $2)
-           (i32.const 16)
-          )
-         )
+      (i32.gt_u
+       (local.tee $2
+        (i32.sub
+         (local.get $2)
+         (i32.const 16)
         )
        )
-       (local.get $3)
+       (i32.const 15)
       )
      )
     )
    )
   )
-  (local.set $0
-   (i32.add
-    (i32.xor
-     (local.get $0)
-     (i32.const -1)
-    )
+  (local.set $2
+   (i32.sub
     (local.get $1)
+    (i32.const 1)
    )
   )
   (loop $label3
@@ -1781,27 +1792,20 @@
      (local.get $6)
      (local.get $7)
      (i32.lt_s
-      (local.tee $2
+      (local.tee $1
        (i32.load8_s
-        (local.get $1)
+        (local.tee $2
+         (i32.add
+          (local.get $2)
+          (i32.const 1)
+         )
+        )
        )
       )
       (i32.const 0)
      )
     )
    )
-   (local.set $0
-    (i32.add
-     (local.get $0)
-     (i32.const 1)
-    )
-   )
-   (local.set $1
-    (i32.add
-     (local.get $1)
-     (i32.const 1)
-    )
-   )
    (br_if $label3
     (i32.eqz
      (i32.and
@@ -1809,7 +1813,7 @@
        (i32.load8_u
         (i32.or
          (i32.and
-          (local.get $2)
+          (local.get $1)
           (i32.const 15)
          )
          (i32.const 976)
@@ -1817,7 +1821,7 @@
        )
        (i32.and
         (i32.shr_u
-         (local.get $2)
+         (local.get $1)
          (i32.const 4)
         )
         (i32.const 7)
@@ -1828,7 +1832,10 @@
     )
    )
   )
-  (local.get $0)
+  (i32.sub
+   (local.get $2)
+   (local.get $0)
+  )
  )
  (func $memccpy (param $0 i32) (param $1 i32) (param $2 i32) (param $3 i32) (result i32)
   (memory.copy
diff --git a/sqlite3/libc/string.h b/sqlite3/libc/string.h
index 3de90f7..71b97c3 100644
--- a/sqlite3/libc/string.h
+++ b/sqlite3/libc/string.h
@@ -176,17 +176,15 @@ size_t strlen(const char *s) {
 }
 
 static int __strcmp(const char *s1, const char *s2) {
-  // Set limit to the largest possible valid v128_t pointer.
-  // Unsigned modular arithmetic gives the correct result
-  // unless memory size is zero, in which case all pointers are invalid.
-  const v128_t *const limit =
-      (v128_t *)(__builtin_wasm_memory_size(0) * PAGESIZE) - 1;
+  // How many bytes can be read before pointers go out of bounds.
+  size_t N = __builtin_wasm_memory_size(0) * PAGESIZE -  //
+             (size_t)(s1 > s2 ? s1 : s2);
 
   // Unaligned loads handle the case where the strings
   // have mismatching alignments.
   const v128_t *w1 = (v128_t *)s1;
   const v128_t *w2 = (v128_t *)s2;
-  while (w1 <= limit && w2 <= limit) {
+  for (; N >= sizeof(v128_t); N -= sizeof(v128_t)) {
     // Find any single bit difference.
     if (wasm_v128_any_true(wasm_v128_load(w1) ^ wasm_v128_load(w2))) {
       // The strings may still be equal,
@@ -237,17 +235,16 @@ int strcmp(const char *s1, const char *s2) {
 
 __attribute__((weak))
 int strncmp(const char *s1, const char *s2, size_t n) {
-  // Set limit to the largest possible valid v128_t pointer.
-  // Unsigned modular arithmetic gives the correct result
-  // unless memory size is zero, in which case all pointers are invalid.
-  const v128_t *const limit =
-      (v128_t *)(__builtin_wasm_memory_size(0) * PAGESIZE) - 1;
+  // How many bytes can be read before pointers go out of bounds.
+  size_t N = __builtin_wasm_memory_size(0) * PAGESIZE -  //
+             (size_t)(s1 > s2 ? s1 : s2);
+  if (n > N) n = N;
 
   // Unaligned loads handle the case where the strings
   // have mismatching alignments.
   const v128_t *w1 = (v128_t *)s1;
   const v128_t *w2 = (v128_t *)s2;
-  for (; w1 <= limit && w2 <= limit && n >= sizeof(v128_t); n -= sizeof(v128_t)) {
+  for (; n >= sizeof(v128_t); n -= sizeof(v128_t)) {
     // Find any single bit difference.
     if (wasm_v128_any_true(wasm_v128_load(w1) ^ wasm_v128_load(w2))) {
       // The strings may still be equal,
@@ -332,21 +329,62 @@ char *strrchr(const char *s, int c) {
   return (char *)memrchr(s, c, strlen(s) + 1);
 }
 
+// http://0x80.pl/notesen/2018-10-18-simd-byte-lookup.html
+
+#define _WASM_SIMD128_BITMAP256_T                                    \
+  struct {                                                           \
+    uint8_t l __attribute__((__vector_size__(16), __aligned__(16))); \
+    uint8_t h __attribute__((__vector_size__(16), __aligned__(16))); \
+  }
+
+#define _WASM_SIMD128_SETBIT(bitmap, i)            \
+  ({                                               \
+    uint8_t _c = (uint8_t)(i);                     \
+    uint8_t _hi_nibble = _c >> 4;                  \
+    uint8_t _lo_nibble = _c & 0xf;                 \
+    bitmap.l[_lo_nibble] |= 1 << (_hi_nibble - 0); \
+    bitmap.h[_lo_nibble] |= 1 << (_hi_nibble - 8); \
+  })
+
+#define _WASM_SIMD128_CHKBIT(bitmap, i)                                   \
+  ({                                                                      \
+    uint8_t _c = (uint8_t)(i);                                            \
+    uint8_t _hi_nibble = _c >> 4;                                         \
+    uint8_t _lo_nibble = _c & 0xf;                                        \
+    uint8_t _bitmask = 1 << (_hi_nibble & 0x7);                           \
+    uint8_t _bitset = (_hi_nibble < 8 ? bitmap.l : bitmap.h)[_lo_nibble]; \
+    _bitmask & _bitset;                                                   \
+  })
+
+#define _WASM_SIMD128_CHKBITS(bitmap, v)                                    \
+  ({                                                                        \
+    v128_t _w = v;                                                          \
+    v128_t _hi_nibbles = wasm_u8x16_shr(_w, 4);                             \
+    v128_t _lo_nibbles = _w & wasm_u8x16_const_splat(0xf);                  \
+                                                                            \
+    v128_t _bitmask_lookup = wasm_u8x16_const(1, 2, 4, 8, 16, 32, 64, 128,  \
+                                              1, 2, 4, 8, 16, 32, 64, 128); \
+                                                                            \
+    v128_t _bitmask = wasm_i8x16_swizzle(_bitmask_lookup, _hi_nibbles);     \
+    v128_t _bitsets = wasm_v128_bitselect(                                  \
+        wasm_i8x16_swizzle(bitmap.l, _lo_nibbles),                          \
+        wasm_i8x16_swizzle(bitmap.h, _lo_nibbles),                          \
+        wasm_i8x16_lt(_hi_nibbles, wasm_u8x16_const_splat(8)));             \
+                                                                            \
+    wasm_i8x16_eq(_bitsets & _bitmask, _bitmask);                           \
+  })
+
 __attribute__((weak))
 size_t strspn(const char *s, const char *c) {
-  // Set limit to the largest possible valid v128_t pointer.
-  // Unsigned modular arithmetic gives the correct result
-  // unless memory size is zero, in which case all pointers are invalid.
-  const v128_t *const limit =
-      (v128_t *)(__builtin_wasm_memory_size(0) * PAGESIZE) - 1;
-
+  // How many bytes can be read before the pointer goes out of bounds.
+  size_t N = __builtin_wasm_memory_size(0) * PAGESIZE - (size_t)s;
   const v128_t *w = (v128_t *)s;
   const char *const a = s;
 
   if (!c[0]) return 0;
   if (!c[1]) {
     const v128_t wc = wasm_i8x16_splat(*c);
-    while (w <= limit) {
+    for (; N >= sizeof(v128_t); N -= sizeof(v128_t)) {
       const v128_t cmp = wasm_i8x16_eq(wasm_v128_load(w), wc);
       // Bitmask is slow on AArch64, all_true is much faster.
       if (!wasm_i8x16_all_true(cmp)) {
@@ -361,111 +399,65 @@ size_t strspn(const char *s, const char *c) {
     return s - a;
   }
 
-  // http://0x80.pl/notesen/2018-10-18-simd-byte-lookup.html
-  typedef unsigned char u8x16
-      __attribute__((__vector_size__(16), __aligned__(16)));
+  _WASM_SIMD128_BITMAP256_T bitmap = {};
 
-  u8x16 bitmap07 = {};
-  u8x16 bitmap8f = {};
   for (; *c; c++) {
-    unsigned lo_nibble = *(unsigned char *)c % 16;
-    unsigned hi_nibble = *(unsigned char *)c / 16;
-    bitmap07[lo_nibble] |= 1 << (hi_nibble - 0);
-    bitmap8f[lo_nibble] |= 1 << (hi_nibble - 8);
+    _WASM_SIMD128_SETBIT(bitmap, *c);
     // Terminator IS NOT on the bitmap.
   }
 
-  for (; w <= limit; w++) {
-    const v128_t lo_nibbles = wasm_v128_load(w) & wasm_u8x16_const_splat(0xf);
-    const v128_t hi_nibbles = wasm_u8x16_shr(wasm_v128_load(w), 4);
-
-    const v128_t bitmask_lookup =
-        wasm_u8x16_const(1, 2, 4, 8, 16, 32, 64, 128,  //
-                         1, 2, 4, 8, 16, 32, 64, 128);
-
-    const v128_t bitmask = wasm_i8x16_swizzle(bitmask_lookup, hi_nibbles);
-    const v128_t bitsets = wasm_v128_bitselect(
-        wasm_i8x16_swizzle(bitmap07, lo_nibbles),
-        wasm_i8x16_swizzle(bitmap8f, lo_nibbles),
-        wasm_i8x16_lt(hi_nibbles, wasm_u8x16_const_splat(8)));
-
-    const v128_t cmp = wasm_i8x16_eq(bitsets & bitmask, bitmask);
+  for (; N >= sizeof(v128_t); N -= sizeof(v128_t)) {
+    const v128_t cmp = _WASM_SIMD128_CHKBITS(bitmap, wasm_v128_load(w));
+    // Bitmask is slow on AArch64, all_true is much faster.
     if (!wasm_i8x16_all_true(cmp)) {
       size_t ctz = __builtin_ctz(~wasm_i8x16_bitmask(cmp));
       return (char *)w + ctz - s;
     }
+    w++;
   }
 
   // Baseline algorithm.
-  for (s = (char *)w;; s++) {
-    const unsigned lo_nibble = *(unsigned char *)s & 0xf;
-    const unsigned hi_nibble = *(unsigned char *)s >> 4;
-    const unsigned bitmask = 1 << (hi_nibble & 0x7);
-    const unsigned bitset =
-        hi_nibble < 8 ? bitmap07[lo_nibble] : bitmap8f[lo_nibble];
-    if ((bitset & bitmask) == 0) return s - a;
-  }
+  for (s = (char *)w; _WASM_SIMD128_CHKBIT(bitmap, *s); s++);
+  return s - a;
 }
 
 __attribute__((weak))
 size_t strcspn(const char *s, const char *c) {
   if (!c[0] || !c[1]) return __strchrnul(s, *c) - s;
 
-  // Set limit to the largest possible valid v128_t pointer.
-  // Unsigned modular arithmetic gives the correct result
-  // unless memory size is zero, in which case all pointers are invalid.
-  const v128_t *const limit =
-      (v128_t *)(__builtin_wasm_memory_size(0) * PAGESIZE) - 1;
-
+  // How many bytes can be read before the pointer goes out of bounds.
+  size_t N = __builtin_wasm_memory_size(0) * PAGESIZE - (size_t)s;
   const v128_t *w = (v128_t *)s;
   const char *const a = s;
 
-  // http://0x80.pl/notesen/2018-10-18-simd-byte-lookup.html
-  typedef unsigned char u8x16
-      __attribute__((__vector_size__(16), __aligned__(16)));
+  _WASM_SIMD128_BITMAP256_T bitmap = {};
 
-  u8x16 bitmap07 = {};
-  u8x16 bitmap8f = {};
-  for (;; c++) {
-    unsigned lo_nibble = *(unsigned char *)c % 16;
-    unsigned hi_nibble = *(unsigned char *)c / 16;
-    bitmap07[lo_nibble] |= 1 << (hi_nibble - 0);
-    bitmap8f[lo_nibble] |= 1 << (hi_nibble - 8);
-    if (!*c) break;  // Terminator IS on the bitmap.
+  for (;;) {
+    _WASM_SIMD128_SETBIT(bitmap, *c);
+    // Terminator IS on the bitmap.
+    if (!*c++) break;
   }
 
-  for (; w <= limit; w++) {
-    const v128_t lo_nibbles = wasm_v128_load(w) & wasm_u8x16_const_splat(0xf);
-    const v128_t hi_nibbles = wasm_u8x16_shr(wasm_v128_load(w), 4);
-
-    const v128_t bitmask_lookup =
-        wasm_u8x16_const(1, 2, 4, 8, 16, 32, 64, 128,  //
-                         1, 2, 4, 8, 16, 32, 64, 128);
-
-    const v128_t bitmask = wasm_i8x16_swizzle(bitmask_lookup, hi_nibbles);
-    const v128_t bitsets = wasm_v128_bitselect(
-        wasm_i8x16_swizzle(bitmap07, lo_nibbles),
-        wasm_i8x16_swizzle(bitmap8f, lo_nibbles),
-        wasm_i8x16_lt(hi_nibbles, wasm_u8x16_const_splat(8)));
-
-    const v128_t cmp = wasm_i8x16_eq(bitsets & bitmask, bitmask);
+  for (; N >= sizeof(v128_t); N -= sizeof(v128_t)) {
+    const v128_t cmp = _WASM_SIMD128_CHKBITS(bitmap, wasm_v128_load(w));
+    // Bitmask is slow on AArch64, any_true is much faster.
     if (wasm_v128_any_true(cmp)) {
       size_t ctz = __builtin_ctz(wasm_i8x16_bitmask(cmp));
       return (char *)w + ctz - s;
     }
+    w++;
   }
 
   // Baseline algorithm.
-  for (s = (char *)w;; s++) {
-    const unsigned lo_nibble = *(unsigned char *)s & 0xf;
-    const unsigned hi_nibble = *(unsigned char *)s >> 4;
-    const unsigned bitmask = 1 << (hi_nibble & 0x7);
-    const unsigned bitset =
-        hi_nibble < 8 ? bitmap07[lo_nibble] : bitmap8f[lo_nibble];
-    if (bitset & bitmask) return s - a;
-  }
+  for (s = (char *)w; !_WASM_SIMD128_CHKBIT(bitmap, *s); s++);
+  return s - a;
 }
 
+#undef _WASM_SIMD128_SETBIT
+#undef _WASM_SIMD128_CHKBIT
+#undef _WASM_SIMD128_CHKBITS
+#undef _WASM_SIMD128_BITMAP256_T
+
 // Given the above SIMD implementations,
 // these are best implemented as
 // small wrappers over those functions.