Source-Changes-HG archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
[src/trunk]: src/sys/arch/luna68k/dev Pull LUNA's framebuffer improvements by...
details: https://anonhg.NetBSD.org/src/rev/f75b17b43727
branches: trunk
changeset: 332703:f75b17b43727
user: tsutsui <tsutsui%NetBSD.org@localhost>
date: Sat Oct 04 16:58:17 2014 +0000
description:
Pull LUNA's framebuffer improvements by Kenji Aoyama from OpenBSD/luna88k.
http://marc.info/?l=openbsd-cvs&m=141199909120631&w=2
>> Use raster(logic) operation, or ROP, function on LUNA frame buffer.
>> It makes 4bpp wscons putchar ~20% faster.
This Makes 4bpp wscons putchar ~30% on LUNA-II.
Also use the similar ROP in 1bpp putchar and cursor functions
and the 1bpp putchar is also ~5% faster.
While here, reduce diffs from OpenBSD a bit.
Tested on all 1bpp/4bpp/8bpp framebuffers.
diffstat:
sys/arch/luna68k/dev/lunafb.c | 9 +-
sys/arch/luna68k/dev/omrasops.c | 282 +++++++++++++++++++++---------------
sys/arch/luna68k/dev/omrasopsvar.h | 87 ++++++++---
3 files changed, 230 insertions(+), 148 deletions(-)
diffs (truncated from 711 to 300 lines):
diff -r 93fedcec6370 -r f75b17b43727 sys/arch/luna68k/dev/lunafb.c
--- a/sys/arch/luna68k/dev/lunafb.c Sat Oct 04 15:54:53 2014 +0000
+++ b/sys/arch/luna68k/dev/lunafb.c Sat Oct 04 16:58:17 2014 +0000
@@ -1,4 +1,4 @@
-/* $NetBSD: lunafb.c,v 1.35 2014/07/25 16:40:12 tsutsui Exp $ */
+/* $NetBSD: lunafb.c,v 1.36 2014/10/04 16:58:17 tsutsui Exp $ */
/*-
* Copyright (c) 2000 The NetBSD Foundation, Inc.
@@ -31,7 +31,7 @@
#include <sys/cdefs.h> /* RCS ID & Copyright macro defns */
-__KERNEL_RCSID(0, "$NetBSD: lunafb.c,v 1.35 2014/07/25 16:40:12 tsutsui Exp $");
+__KERNEL_RCSID(0, "$NetBSD: lunafb.c,v 1.36 2014/10/04 16:58:17 tsutsui Exp $");
#include <sys/param.h>
#include <sys/systm.h>
@@ -75,11 +75,8 @@
};
#define OMFB_RFCNT 0xB1000000 /* video h-origin/v-origin */
-#define OMFB_PLANEMASK 0xB1040000 /* planemask register */
-#define OMFB_FB_WADDR 0xB1080008 /* common plane */
-#define OMFB_FB_RADDR 0xB10C0008 /* plane #0 */
-#define OMFB_ROPFUNC 0xB12C0000 /* ROP function code */
#define OMFB_RAMDAC 0xC1100000 /* Bt454/Bt458 RAMDAC */
+
#define OMFB_SIZE (0xB1300000 - 0xB1080000 + PAGE_SIZE)
struct hwcmap {
diff -r 93fedcec6370 -r f75b17b43727 sys/arch/luna68k/dev/omrasops.c
--- a/sys/arch/luna68k/dev/omrasops.c Sat Oct 04 15:54:53 2014 +0000
+++ b/sys/arch/luna68k/dev/omrasops.c Sat Oct 04 16:58:17 2014 +0000
@@ -1,4 +1,4 @@
-/* $NetBSD: omrasops.c,v 1.18 2014/09/28 05:00:56 tsutsui Exp $ */
+/* $NetBSD: omrasops.c,v 1.19 2014/10/04 16:58:17 tsutsui Exp $ */
/*-
* Copyright (c) 2000 The NetBSD Foundation, Inc.
@@ -31,7 +31,7 @@
#include <sys/cdefs.h> /* RCS ID & Copyright macro defns */
-__KERNEL_RCSID(0, "$NetBSD: omrasops.c,v 1.18 2014/09/28 05:00:56 tsutsui Exp $");
+__KERNEL_RCSID(0, "$NetBSD: omrasops.c,v 1.19 2014/10/04 16:58:17 tsutsui Exp $");
/*
* Designed speficically for 'm68k bitorder';
@@ -78,13 +78,6 @@
#define ALIGNMASK (0x1f)
#define BYTESDONE (4)
-#define W(p) (*(uint32_t *)(p))
-#define R(p) (*(uint32_t *)((uint8_t *)(p) + 0x40000))
-#define P0(p) (*(uint32_t *)((uint8_t *)(p) + 0x40000))
-#define P1(p) (*(uint32_t *)((uint8_t *)(p) + 0x80000))
-#define P2(p) (*(uint32_t *)((uint8_t *)(p) + 0xc0000))
-#define P3(p) (*(uint32_t *)((uint8_t *)(p) + 0x100000))
-
/*
* macros to handle unaligned bit copy ops.
* See src/sys/dev/rasops/rasops_mask.h for MI version.
@@ -96,13 +89,13 @@
#define FASTGETBITS(psrc, x, w, dst) \
asm("bfextu %3{%1:%2},%0" \
: "=d" (dst) \
- : "di" (x), "di" (w), "o" ((uint32_t *)(psrc)))
+ : "di" (x), "di" (w), "o" (*(uint32_t *)(psrc)))
/* luna68k version PUTBITS() that puts w bits from bit x at pdst memory */
/* XXX this macro assumes (x + w) <= 32 to handle unaligned residual bits */
#define FASTPUTBITS(src, x, w, pdst) \
asm("bfins %3,%0{%1:%2}" \
- : "+o" ((uint32_t *)(pdst)) \
+ : "+o" (*(uint32_t *)(pdst)) \
: "di" (x), "di" (w), "d" (src) \
: "memory" );
@@ -137,16 +130,23 @@
rmask = ALL1BITS << (-width & ALIGNMASK);
if (width <= BLITWIDTH) {
lmask &= rmask;
+ /* set lmask as ROP mask value, with THROUGH mode */
+ ((volatile uint32_t *)OMFB_ROPFUNC)[ROP_THROUGH] = lmask;
+
while (height > 0) {
glyph = 0;
for (i = ri->ri_font->stride; i != 0; i--)
glyph = (glyph << 8) | *fb++;
glyph <<= (4 - ri->ri_font->stride) * NBBY;
glyph = (glyph >> align) ^ inverse;
- P0(p) = (P0(p) & ~lmask) | (glyph & lmask);
+
+ *W(p) = glyph;
+
p += scanspan;
height--;
}
+ /* reset mask value */
+ ((volatile uint32_t *)OMFB_ROPFUNC)[ROP_THROUGH] = ALL1BITS;
} else {
uint8_t *q = p;
uint32_t lhalf, rhalf;
@@ -157,14 +157,26 @@
glyph = (glyph << 8) | *fb++;
glyph <<= (4 - ri->ri_font->stride) * NBBY;
lhalf = (glyph >> align) ^ inverse;
- P0(p) = (P0(p) & ~lmask) | (lhalf & lmask);
+ /* set lmask as ROP mask value, with THROUGH mode */
+ ((volatile uint32_t *)OMFB_ROPFUNC)[ROP_THROUGH] =
+ lmask;
+
+ *W(p) = lhalf;
+
p += BYTESDONE;
+
rhalf = (glyph << (BLITWIDTH - align)) ^ inverse;
- P0(p) = (rhalf & rmask) | (P0(p) & ~rmask);
+ /* set rmask as ROP mask value, with THROUGH mode */
+ ((volatile uint32_t *)OMFB_ROPFUNC)[ROP_THROUGH] =
+ rmask;
+
+ *W(p) = rhalf;
p = (q += scanspan);
height--;
}
+ /* reset mask value */
+ ((volatile uint32_t *)OMFB_ROPFUNC)[ROP_THROUGH] = ALL1BITS;
}
}
@@ -201,8 +213,15 @@
width = ri->ri_font->fontwidth + align;
lmask = ALL1BITS >> align;
rmask = ALL1BITS << (-width & ALIGNMASK);
+
+ /* select all planes for later ROP function target */
+ *(volatile uint32_t *)OMFB_PLANEMASK = 0xff;
+
if (width <= BLITWIDTH) {
lmask &= rmask;
+ /* set lmask as ROP mask value, with THROUGH mode */
+ ((volatile uint32_t *)OMFB_ROPFUNC)[ROP_THROUGH] = lmask;
+
while (height > 0) {
glyph = 0;
for (i = ri->ri_font->stride; i != 0; i--)
@@ -210,21 +229,25 @@
glyph <<= (4 - ri->ri_font->stride) * NBBY;
glyph = (glyph >> align);
glyphbg = glyph ^ ALL1BITS;
+
fgpat = glyph & fgmask0;
bgpat = glyphbg & bgmask0;
- P0(p) = (P0(p) & ~lmask) | ((fgpat | bgpat) & lmask);
+ *P0(p) = (fgpat | bgpat);
fgpat = glyph & fgmask1;
bgpat = glyphbg & bgmask1;
- P1(p) = (P1(p) & ~lmask) | ((fgpat | bgpat) & lmask);
+ *P1(p) = (fgpat | bgpat);
fgpat = glyph & fgmask2;
bgpat = glyphbg & bgmask2;
- P2(p) = (P2(p) & ~lmask) | ((fgpat | bgpat) & lmask);
+ *P2(p) = (fgpat | bgpat);
fgpat = glyph & fgmask3;
bgpat = glyphbg & bgmask3;
- P3(p) = (P3(p) & ~lmask) | ((fgpat | bgpat) & lmask);
+ *P3(p) = (fgpat | bgpat);
+
p += scanspan;
height--;
}
+ /* reset mask value */
+ ((volatile uint32_t *)OMFB_ROPFUNC)[ROP_THROUGH] = ALL1BITS;
} else {
uint8_t *q = p;
uint32_t lhalf, rhalf;
@@ -237,38 +260,52 @@
glyph <<= (4 - ri->ri_font->stride) * NBBY;
lhalf = (glyph >> align);
lhalfbg = lhalf ^ ALL1BITS;
+ /* set lmask as ROP mask value, with THROUGH mode */
+ ((volatile uint32_t *)OMFB_ROPFUNC)[ROP_THROUGH] =
+ lmask;
+
fgpat = lhalf & fgmask0;
bgpat = lhalfbg & bgmask0;
- P0(p) = (P0(p) & ~lmask) | ((fgpat | bgpat) & lmask);
+ *P0(p) = (fgpat | bgpat);
fgpat = lhalf & fgmask1;
bgpat = lhalfbg & bgmask1;
- P1(p) = (P1(p) & ~lmask) | ((fgpat | bgpat) & lmask);
+ *P1(p) = (fgpat | bgpat);
fgpat = lhalf & fgmask2;
bgpat = lhalfbg & bgmask2;
- P2(p) = (P2(p) & ~lmask) | ((fgpat | bgpat) & lmask);
+ *P2(p) = (fgpat | bgpat);
fgpat = lhalf & fgmask3;
bgpat = lhalfbg & bgmask3;
- P3(p) = (P3(p) & ~lmask) | ((fgpat | bgpat) & lmask);
+ *P3(p) = (fgpat | bgpat);
+
p += BYTESDONE;
+
rhalf = (glyph << (BLITWIDTH - align));
rhalfbg = rhalf ^ ALL1BITS;
+ /* set rmask as ROP mask value, with THROUGH mode */
+ ((volatile uint32_t *)OMFB_ROPFUNC)[ROP_THROUGH] =
+ rmask;
+
fgpat = rhalf & fgmask0;
bgpat = rhalfbg & bgmask0;
- P0(p) = ((fgpat | bgpat) & rmask) | (P0(p) & ~rmask);
+ *P0(p) = (fgpat | bgpat);
fgpat = rhalf & fgmask1;
bgpat = rhalfbg & bgmask1;
- P1(p) = ((fgpat | bgpat) & rmask) | (P1(p) & ~rmask);
+ *P1(p) = (fgpat | bgpat);
fgpat = rhalf & fgmask2;
bgpat = rhalfbg & bgmask2;
- P2(p) = ((fgpat | bgpat) & rmask) | (P2(p) & ~rmask);
+ *P2(p) = (fgpat | bgpat);
fgpat = rhalf & fgmask3;
bgpat = rhalfbg & bgmask3;
- P3(p) = ((fgpat | bgpat) & rmask) | (P3(p) & ~rmask);
+ *P3(p) = (fgpat | bgpat);
p = (q += scanspan);
height--;
}
+ /* reset mask value */
+ ((volatile uint32_t *)OMFB_ROPFUNC)[ROP_THROUGH] = ALL1BITS;
}
+ /* select plane #0 only; XXX need this ? */
+ *(volatile uint32_t *)OMFB_PLANEMASK = 0x01;
}
static void
@@ -295,22 +332,22 @@
lmask &= rmask;
fill &= lmask;
while (height > 0) {
- P0(p) = (P0(p) & ~lmask) | fill;
+ *P0(p) = (*P0(p) & ~lmask) | fill;
p += scanspan;
height--;
}
} else {
uint8_t *q = p;
while (height > 0) {
- P0(p) = (P0(p) & ~lmask) | (fill & lmask);
+ *P0(p) = (*P0(p) & ~lmask) | (fill & lmask);
width -= 2 * BLITWIDTH;
while (width > 0) {
p += BYTESDONE;
- P0(p) = fill;
+ *P0(p) = fill;
width -= BLITWIDTH;
}
p += BYTESDONE;
- P0(p) = (fill & rmask) | (P0(p) & ~rmask);
+ *P0(p) = (fill & rmask) | (*P0(p) & ~rmask);
p = (q += scanspan);
width = w + align;
@@ -350,34 +387,34 @@
fill2 &= lmask;
fill3 &= lmask;
while (height > 0) {
- P0(p) = (P0(p) & ~lmask) | fill0;
- P1(p) = (P1(p) & ~lmask) | fill1;
- P2(p) = (P2(p) & ~lmask) | fill2;
- P3(p) = (P3(p) & ~lmask) | fill3;
+ *P0(p) = (*P0(p) & ~lmask) | fill0;
+ *P1(p) = (*P1(p) & ~lmask) | fill1;
+ *P2(p) = (*P2(p) & ~lmask) | fill2;
+ *P3(p) = (*P3(p) & ~lmask) | fill3;
p += scanspan;
height--;
}
} else {
uint8_t *q = p;
while (height > 0) {
- P0(p) = (P0(p) & ~lmask) | (fill0 & lmask);
- P1(p) = (P1(p) & ~lmask) | (fill1 & lmask);
- P2(p) = (P2(p) & ~lmask) | (fill2 & lmask);
- P3(p) = (P3(p) & ~lmask) | (fill3 & lmask);
+ *P0(p) = (*P0(p) & ~lmask) | (fill0 & lmask);
+ *P1(p) = (*P1(p) & ~lmask) | (fill1 & lmask);
+ *P2(p) = (*P2(p) & ~lmask) | (fill2 & lmask);
+ *P3(p) = (*P3(p) & ~lmask) | (fill3 & lmask);
width -= 2 * BLITWIDTH;
while (width > 0) {
p += BYTESDONE;
- P0(p) = fill0;
- P1(p) = fill1;
- P2(p) = fill2;
- P3(p) = fill3;
Home |
Main Index |
Thread Index |
Old Index