Optimize memcasecmp. Benchmarks shows slight improvement.

We are also avoiding potential cache-misses, by avoiding load.

name                         old speed            new speed            delta
BM_Searchcase                3.09GB/s ±13%        3.20GB/s ±16%   +3.69%  (p=0.039 n=20+17)
BM_SearchcaseMedium          1.08GB/s ± 7%        1.04GB/s ±14%     ~     (p=0.814 n=16+20)
BM_SearchcasePathological     618kB/s ±13%         652kB/s ± 6%   +5.55%  (p=0.043 n=20+16)
BM_Memcasematch              2.43GB/s ± 3%        2.45GB/s ± 3%     ~     (p=0.488 n=17+16)
BM_MemcasematchMedium         230MB/s ± 8%         261MB/s ±14%  +13.77%  (p=0.000 n=16+20)
BM_MemcasematchPathological   624kB/s ±14%         619kB/s ±14%     ~     (p=0.836 n=20+20)

PiperOrigin-RevId: 577919033
Change-Id: I31324e04b6a577c582ad630d171d3b41d826f1e4

1 files changed, 12 insertions, 4 deletions

diff --git a/absl/strings/internal/memutil.cc b/absl/strings/internal/memutil.cc
index e2e7347c..0bbd8aa1 100644
--- a/absl/strings/internal/memutil.cc
+++ b/absl/strings/internal/memutil.cc
@@ -27,10 +27,18 @@ int memcasecmp(const char* s1, const char* s2, size_t len) {
   const unsigned char* us2 = reinterpret_cast<const unsigned char*>(s2);
 
   for (size_t i = 0; i < len; i++) {
-    const int diff =
-        int{static_cast<unsigned char>(absl::ascii_tolower(us1[i]))} -
-        int{static_cast<unsigned char>(absl::ascii_tolower(us2[i]))};
-    if (diff != 0) return diff;
+    unsigned char c1 = us1[i];
+    unsigned char c2 = us2[i];
+    // If bytes are the same, they will be the same when converted to lower.
+    // So we only need to convert if bytes are not equal.
+    // NOTE(b/308193381): We do not use `absl::ascii_tolower` here in order
+    // to avoid its lookup table and improve performance.
+    if (c1 != c2) {
+      c1 = c1 >= 'A' && c1 <= 'Z' ? c1 - 'A' + 'a' : c1;
+      c2 = c2 >= 'A' && c2 <= 'Z' ? c2 - 'A' + 'a' : c2;
+      const int diff = int{c1} - int{c2};
+      if (diff != 0) return diff;
+    }
   }
   return 0;
 }