Skip all-zero passes in radix sort of results

Precompute OR of all sort keys; a byte position that is zero across every
key contributes a no-op pass. Default two-criteria setup leaves the low 32
bits zero, so 4 of 8 passes are skipped without a histogram scan. 2x on the
sort, 3-8% on high-match queries. No regression when all bytes are used.

CHANGELOG.md

@@ -3,8 +3,9 @@ CHANGELOG
 
 0.74.2
 ------
-- Up to 2x faster matching for single-character queries
+- Single-character queries are up to 2.4x faster
     - This is the most latency-sensitive case; the first keystroke scans the entire list before the result cache can help
+- Faster sorting of search results; the default two-criteria ranking skips redundant radix passes
 - Fixed nondeterministic match highlight positions
 - fzf now detects terminal resize on Windows in `--height` mode (#4790) (@Cyrus580529)
 - Fixed signal handler cleanup when fzf is used as a library; SIGINT/SIGTERM/SIGHUP and resize handlers no longer persist after `Run()` returns

src/result.go

@@ -369,9 +369,21 @@ func radixSortResults(a []Result, tac bool, scratch []Result) []Result {
 	src, dst := a, buf
 	scattered := 0
 
+	// OR of all keys: a byte position that is zero here is zero in every key,
+	// so its pass is a no-op and can be skipped without a histogram scan.
+	// With the common two-criteria setup the low 32 bits are always zero.
+	var keyOR uint64
+	for i := range src {
+		keyOR |= sortKey(&src[i])
+	}
+
 	for pass := range 8 {
 		shift := uint(pass) * 8
 
+		if byte(keyOR>>shift) == 0 {
+			continue
+		}
+
 		var count [256]int
 		for i := range src {
 			count[byte(sortKey(&src[i])>>shift)]++