std.mem: add countScalar and countAny #24104
Open
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
CI fails were #24109 |
nektro
commented
Aug 15, 2025
|
@mrjbq7, I see we've moved over here before I was able to finish my idea and post it! 😄 const std = @import("std");
const N = std.simd.suggestVectorLength(u8) orelse @sizeOf(u8);
const V = @Vector(N, u8);
const Int = std.meta.Int(.unsigned, N);
pub fn countScalarButGood(haystack: []const u8, needle: u8) usize {
var found: usize = 0;
if (haystack.len < N) {
for (haystack) |item| {
if (item == needle) found += 1;
}
return found;
}
const broad: V = @splat(needle);
for (0..haystack.len / N) |i| {
const h: V = @bitCast(haystack[i * N ..][0..N].*);
const integer: Int = @bitCast(h == broad);
found += @popCount(integer);
}
const remaining = (haystack.len % N);
if (remaining == 0) return found;
const overlapped: std.math.Log2Int(Int) = @intCast(N - remaining);
const mask: Int = (@as(Int, 1) << overlapped) - 1;
const last: V = @bitCast(haystack[haystack.len - N ..][0..N].*);
const integer: Int = @bitCast(last == broad);
found += @popCount(integer & ~mask);
return found;
} |
|
I think you could do a branch tree and force unroll to make it even faster... |
slower! no needle in 1mb: all needle in 1mb: |
|
it's possible i need to properly randomize the data for correct benchmarking, anyway... more to come |
|
could you share your benchmark quickly? im curious to get |
well, i tried a couple of them, but basically i used a 1mb array with every index set to a constant value and then looked for it, and then looked for something that wasn't equal to that value, and then i tried this version with not many finds: pub fn main() !void {
const allocator = std.heap.smp_allocator;
const n = 1_000_000;
const bytes = try allocator.alloc(u8, n);
defer allocator.free(bytes);
for (0..n) |i| {
bytes[i] = @intCast(i % 256);
}
const count = countScalar(u8, bytes, 0);
const expected: usize = @intFromFloat(@ceil(@as(f64, @floatFromInt(n)) / 256.0));
if (count != expected) {
std.debug.print("oops {d}\n", .{count});
std.process.exit(1);
}
}this one produces less varied results: |
|
Here are my results from running my benchmark script on a I'd be very curious for input from some people with different hardware, or if there are any mistakes in my benchmark script :). Just as a sanity check, I have confirmed that using random bytes for the input does generate matches on basically all runs, usually hundreds/thousands of them, so "stuff" is happening. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
this brings 'count' into sync with the available 'indexOf*Pos' options