diff --git a/Cargo.toml b/Cargo.toml
index 5e3f6e5f..09c02f23 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -17,6 +17,7 @@ crc32fast = { version = "1.2", optional = true }
 flate2 = { version = "1", optional = true }
 indexmap = { version = "1.1", optional = true }
 wasmparser = { version = "0.57", optional = true }
+memchr = { version = "2.4", optional = true, default-features = false }
 
 # Internal feature, only used when building as part of libstd, not part of the
 # stable interface of this crate.
@@ -45,7 +46,7 @@ write = ["write_core", "coff", "elf", "macho"]
 
 # Enable things that require libstd.
 # Currently, this provides an `Error` implementation.
-std = []
+std = ["memchr/std"]
 # Enable decompression of compressed sections.
 # This feature is not required if you want to do your own decompression.
 compression = ["flate2", "std"]
@@ -58,7 +59,7 @@ unaligned = []
 
 #=======================================
 # File format features.
-archive = []
+archive = ["memchr"]
 coff = []
 elf = []
 macho = []
diff --git a/src/read/archive.rs b/src/read/archive.rs
index fb6dd362..2f507cb9 100644
--- a/src/read/archive.rs
+++ b/src/read/archive.rs
@@ -197,13 +197,13 @@ impl<'data> ArchiveMember<'data> {
             parse_bsd_extended_name(&header.name[3..], data, &mut file_offset, &mut file_size)
                 .read_error("Invalid archive extended name length")?
         } else if header.name[0] == b'/' {
-            let name_len =
-                (header.name.iter().position(|&x| x == b' ')).unwrap_or_else(|| header.name.len());
+            let name_len = memchr::memchr(b' ', &header.name).unwrap_or(header.name.len());
             &header.name[..name_len]
         } else {
-            let name_len = (header.name.iter().position(|&x| x == b'/'))
-                .or_else(|| header.name.iter().position(|&x| x == b' '))
-                .unwrap_or_else(|| header.name.len());
+            // FIXME: can `memchr::memchr2` be used here?
+            let name_len = memchr::memchr(b'/', &header.name)
+                .or_else(|| memchr::memchr(b' ', &header.name))
+                .unwrap_or(header.name.len());
             &header.name[..name_len]
         };
 
@@ -268,20 +268,19 @@ impl<'data> ArchiveMember<'data> {
 
 // Ignores bytes starting from the first space.
 fn parse_u64_digits(digits: &[u8], radix: u32) -> Option<u64> {
-    let len = digits
-        .iter()
-        .position(|&x| x == b' ')
-        .unwrap_or_else(|| digits.len());
-    let digits = &digits[..len];
-    if digits.is_empty() {
+    if let [b' ', ..] = digits {
         return None;
     }
     let mut result: u64 = 0;
     for &c in digits {
-        let x = (c as char).to_digit(radix)?;
-        result = result
-            .checked_mul(u64::from(radix))?
-            .checked_add(u64::from(x))?;
+        if c == b' ' {
+            return Some(result);
+        } else {
+            let x = (c as char).to_digit(radix)?;
+            result = result
+                .checked_mul(u64::from(radix))?
+                .checked_add(u64::from(x))?;
+        }
     }
     Some(result)
 }
@@ -290,7 +289,7 @@ fn parse_sysv_extended_name<'data>(digits: &[u8], names: &'data [u8]) -> Result<
     let offset = parse_u64_digits(digits, 10).ok_or(())?;
     let offset = offset.try_into().map_err(|_| ())?;
     let name_data = names.get(offset..).ok_or(())?;
-    let name = match name_data.iter().position(|&x| x == b'/' || x == 0) {
+    let name = match memchr::memchr2(b'/', b'\0', name_data) {
         Some(len) => &name_data[..len],
         None => name_data,
     };
@@ -307,7 +306,7 @@ fn parse_bsd_extended_name<'data, R: ReadRef<'data>>(
     let len = parse_u64_digits(digits, 10).ok_or(())?;
     *size = size.checked_sub(len).ok_or(())?;
     let name_data = data.read_bytes(offset, len)?;
-    let name = match name_data.iter().position(|&x| x == 0) {
+    let name = match memchr::memchr(b'\0', name_data) {
         Some(len) => &name_data[..len],
         None => name_data,
     };