Optimize screen scrolling, SIMD wrap padding, and reflow loops

- Batch full-screen scroll rotations, eliminating row-by-row scroll loops
- Use block memmove for sub-region scrolling, eliminating heap/stack row swaps
- Vectorize CELL_WRAP_PAD flag setting with AVX2/SSE2/NEON in screen.h
- Replace scalar tab-stop loops with 64-bit word stamping
- Optimize active row backwards scan in resize and erase mode 2

1 files changed, 2 insertions, 4 deletions

diff --git a/src/selection.c b/src/selection.c
index 0eda34f..b05a43c 100644
--- a/src/selection.c
+++ b/src/selection.c
@@ -187,8 +187,7 @@ static const struct cell *line_seg_cells(const struct lines *L, uint64_t id, int
         if (got < L->cols) {
             fill_cells(buf + got, L->cols - got, TERM_COLOR_FG, TERM_COLOR_BG);
             if (wrapped)
-                for (int x = got; x < L->cols; x++)
-                    buf[x].flags |= CELL_WRAP_PAD;
+                set_cells_wrap_pad(buf + got, L->cols - got);
         }
         if (out_wrapped)
             *out_wrapped = wrapped;
@@ -678,8 +677,7 @@ static char *range_text_inner(const struct term *t,
                     if (got < t->cols) {
                         fill_cells(row_buf + got, t->cols - got, TERM_COLOR_FG, TERM_COLOR_BG);
                         if (row_wrapped)
-                            for (int x = got; x < t->cols; x++)
-                                row_buf[x].flags |= CELL_WRAP_PAD;
+                            set_cells_wrap_pad(row_buf + got, t->cols - got);
                     }
                     cells = row_buf;
                 }