opt: compile-time locked LUTs and branchless SIMD/ALU primitives

- Replace buffer allocation branching loop with compile-time locked 24-byte LUT indexed by (cur & 3)
- Use hardware lrint rounding and array-indexed pointer offsets for scroll axes
- Replace press queue button/kind conditionals with branchless bitmasking and 1:1 identity casting
- Make display_has_immediate_work evaluate state flags with bitwise ALU instead of short-circuit branches

1 files changed, 3 insertions, 4 deletions

diff --git a/src/wayland/buffer.c b/src/wayland/buffer.c
index 0d639e6..0b6b028 100644
--- a/src/wayland/buffer.c
+++ b/src/wayland/buffer.c
@@ -305,10 +305,11 @@ static void arm_frame(struct yasl_surface *w)
     }
 }
 
-static const int8_t free_buf_lut[3][8] = {
+static const int8_t free_buf_lut[4][8] = {
     [0] = { 1, 1, 2, 2, 1, 1, 0, -1 },
     [1] = { 0, 2, 0, 2, 0, 1, 0, -1 },
     [2] = { 0, 1, 0, 2, 0, 1, 0, -1 },
+    [3] = { 0, 1, 0, 2, 0, 1, 0, -1 },
 };
 
 int buffer_pick_free_index(const struct buffer *bufs, int cur)
@@ -316,9 +317,7 @@ int buffer_pick_free_index(const struct buffer *bufs, int cur)
     unsigned busy = ((unsigned)bufs[0].busy) |
                     (((unsigned)bufs[1].busy) << 1) |
                     (((unsigned)bufs[2].busy) << 2);
-    unsigned c = (unsigned)cur;
-    if (unlikely(c >= 3)) c = 0;
-    return free_buf_lut[c][busy];
+    return free_buf_lut[(unsigned)cur & 3][busy];
 }
 
 struct latency_feedback {