[src/trunk]: src/sys/dev/rasops rasops8_putchar_aa():

[macallan <macallan%NetBSD.org@localhost>]

details:   https://anonhg.NetBSD.org/src/rev/fdf370b95b30
branches:  trunk
changeset: 773037:fdf370b95b30
user:      macallan <macallan%NetBSD.org@localhost>
date:      Wed Jan 25 16:38:27 2012 +0000

description:
rasops8_putchar_aa():
render glyphs scanline by scanline into a 64bit-aligned buffer, then memcpy()
it into video memory instead of writing directly.
This gives >10% speedup even on valkyriefb, likely more on PCIe framebuffers.

diffstat:

 sys/dev/rasops/rasops8.c |  11 ++++++-----
 1 files changed, 6 insertions(+), 5 deletions(-)

diffs (41 lines):

diff -r ed9406ab7dbc -r fdf370b95b30 sys/dev/rasops/rasops8.c
--- a/sys/dev/rasops/rasops8.c  Wed Jan 25 16:25:28 2012 +0000
+++ b/sys/dev/rasops/rasops8.c  Wed Jan 25 16:38:27 2012 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: rasops8.c,v 1.30 2012/01/04 20:16:20 macallan Exp $    */
+/*     $NetBSD: rasops8.c,v 1.31 2012/01/25 16:38:27 macallan Exp $    */
 
 /*-
  * Copyright (c) 1999 The NetBSD Foundation, Inc.
@@ -30,7 +30,7 @@
  */
 
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: rasops8.c,v 1.30 2012/01/04 20:16:20 macallan Exp $");
+__KERNEL_RCSID(0, "$NetBSD: rasops8.c,v 1.31 2012/01/25 16:38:27 macallan Exp $");
 
 #include "opt_rasops.h"
 
@@ -205,6 +205,7 @@
        struct wsdisplay_font *font = PICK_FONT(ri, uc);
        int x, y, r, g, b, aval;
        int r1, g1, b1, r0, g0, b0, fgo, bgo;
+       uint8_t scanline[32] __attribute__ ((aligned(8)));
 
        hp = hrp = NULL;
 
@@ -278,11 +279,11 @@
                                                ((g & 0xe000) >> 11) |
                                                ((b & 0xc000) >> 14);
                                }
-                               *dp = pixel;
-                               dp++;
+                               scanline[x] = pixel;
                        }
+                       memcpy(rp, scanline, width);
                        if (ri->ri_hwbits) {
-                               memcpy(rp, hrp, width);
+                               memcpy(hrp, scanline, width);
                                hrp += ri->ri_stride;
                        }
                        rp += ri->ri_stride;