Put records that come in order straight onto the line, a frame at once

A RECORDS frame of records in order, with nothing held and no record's
tail waiting, went through hold, on_line and line_append a record: the
same checks each time (heard, credit, the line's end, held, the memory
bound). Now they are made once for the frame and its records are copied
onto the line one after another. The receiver's CPU a 61-byte record in
M5 went from 0.15 to 0.133-0.143 us; both sides together 0.28 us, beside
ngtcp2's 0.27 and quinn's 0.31 in the same rounds (the M5 host here also
compares every record with what was sent). test-link's results are the
same to the byte.

1 files changed, 13 insertions, 1 deletions

diff --git a/link/link-protocol.c b/link/link-protocol.c
index 44aa3e9..7bc4b00 100644
--- a/link/link-protocol.c
+++ b/link/link-protocol.c
@@ -2187,9 +2187,21 @@ static bool take_from(PureLinkReceiver *s, Peer *peer, const unsigned char *pack
             unsigned count = plain[at + 8];
             at += RECORDS_HEAD - 1;
             unsigned i = 0;
+            /* In order, nothing held and no tail waiting: straight onto the
+             * line, the frame at once (what hold does a record, for all). */
+            bool straight = st && st->known && first == line_end(st) && first + count - 1 <= st->heard + CREDIT &&
+                            !st->tail && !st->held_count &&
+                            st->line_size - st->line_head + (size - at) <= HELD_MOST &&
+                            line_room(st, size - at + 4 * (size_t)count);
             for (size_t used; i < count && (used = get_varint(plain + at, size - at, &length)) &&
                               length <= size - at - used; ++i) {
-                if (st) hold(st, first + i, (uint32_t)length, 0, plain + at + used, (uint32_t)length);
+                if (straight) {
+                    put32(st->line + st->line_size, (uint32_t)length);
+                    memcpy(st->line + st->line_size + 4, plain + at + used, length);
+                    st->line_size += 4 + length;
+                    ++st->line_count;
+                } else if (st)
+                    hold(st, first + i, (uint32_t)length, 0, plain + at + used, (uint32_t)length);
                 at += used + length;
             }
             if (st) mark_ready(s, st);