openwrtv3/target/linux/brcm2708/patches-4.1/0029-Speed-up-console-framebuffer-imageblit-function.patch

From ff592a233049e2d96b036deab9a5025818809580 Mon Sep 17 00:00:00 2001
From: Harm Hanemaaijer <fgenfb@yahoo.com>
Date: Thu, 20 Jun 2013 20:21:39 +0200
Subject: [PATCH 029/222] Speed up console framebuffer imageblit function

Especially on platforms with a slower CPU but a relatively high
framebuffer fill bandwidth, like current ARM devices, the existing
console monochrome imageblit function used to draw console text is
suboptimal for common pixel depths such as 16bpp and 32bpp. The existing
code is quite general and can deal with several pixel depths. By creating
special case functions for 16bpp and 32bpp, by far the most common pixel
formats used on modern systems, a significant speed-up is attained
which can be readily felt on ARM-based devices like the Raspberry Pi
and the Allwinner platform, but should help any platform using the
fb layer.

The special case functions allow constant folding, eliminating a number
of instructions including divide operations, and allow the use of an
unrolled loop, eliminating instructions with a variable shift size,
reducing source memory access instructions, and eliminating excessive
branching. These unrolled loops also allow much better code optimization
by the C compiler. The code that selects which optimized variant is used
is also simplified, eliminating integer divide instructions.

The speed-up, measured by timing 'cat file.txt' in the console, varies
between 40% and 70%, when testing on the Raspberry Pi and Allwinner
ARM-based platforms, depending on font size and the pixel depth, with
the greater benefit for 32bpp.

Signed-off-by: Harm Hanemaaijer <fgenfb@yahoo.com>
---
 drivers/video/fbdev/core/cfbimgblt.c | 152 +++++++++++++++++++++++++++++++++--
 1 file changed, 147 insertions(+), 5 deletions(-)

--- a/drivers/video/fbdev/core/cfbimgblt.c
+++ b/drivers/video/fbdev/core/cfbimgblt.c
@@ -28,6 +28,11 @@
  *
  *  Also need to add code to deal with cards endians that are different than
  *  the native cpu endians. I also need to deal with MSB position in the word.
+ *  Modified by Harm Hanemaaijer (fgenfb@yahoo.com) 2013:
+ *  - Provide optimized versions of fast_imageblit for 16 and 32bpp that are
+ *    significantly faster than the previous implementation.
+ *  - Simplify the fast/slow_imageblit selection code, avoiding integer
+ *    divides.
  */
 #include <linux/module.h>
 #include <linux/string.h>
@@ -262,6 +267,133 @@ static inline void fast_imageblit(const
 	}
 }	
 	
+/*
+ * Optimized fast_imageblit for bpp == 16. ppw = 2, bit_mask = 3 folded
+ * into the code, main loop unrolled.
+ */
+
+static inline void fast_imageblit16(const struct fb_image *image,
+				    struct fb_info *p, u8 __iomem * dst1,
+				    u32 fgcolor, u32 bgcolor)
+{
+	u32 fgx = fgcolor, bgx = bgcolor;
+	u32 spitch = (image->width + 7) / 8;
+	u32 end_mask, eorx;
+	const char *s = image->data, *src;
+	u32 __iomem *dst;
+	const u32 *tab = NULL;
+	int i, j, k;
+
+	tab = fb_be_math(p) ? cfb_tab16_be : cfb_tab16_le;
+
+	fgx <<= 16;
+	bgx <<= 16;
+	fgx |= fgcolor;
+	bgx |= bgcolor;
+
+	eorx = fgx ^ bgx;
+	k = image->width / 2;
+
+	for (i = image->height; i--;) {
+		dst = (u32 __iomem *) dst1;
+		src = s;
+
+		j = k;
+		while (j >= 4) {
+			u8 bits = *src;
+			end_mask = tab[(bits >> 6) & 3];
+			FB_WRITEL((end_mask & eorx) ^ bgx, dst++);
+			end_mask = tab[(bits >> 4) & 3];
+			FB_WRITEL((end_mask & eorx) ^ bgx, dst++);
+			end_mask = tab[(bits >> 2) & 3];
+			FB_WRITEL((end_mask & eorx) ^ bgx, dst++);
+			end_mask = tab[bits & 3];
+			FB_WRITEL((end_mask & eorx) ^ bgx, dst++);
+			src++;
+			j -= 4;
+		}
+		if (j != 0) {
+			u8 bits = *src;
+			end_mask = tab[(bits >> 6) & 3];
+			FB_WRITEL((end_mask & eorx) ^ bgx, dst++);
+			if (j >= 2) {
+				end_mask = tab[(bits >> 4) & 3];
+				FB_WRITEL((end_mask & eorx) ^ bgx, dst++);
+				if (j == 3) {
+					end_mask = tab[(bits >> 2) & 3];
+					FB_WRITEL((end_mask & eorx) ^ bgx, dst);
+				}
+			}
+		}
+		dst1 += p->fix.line_length;
+		s += spitch;
+	}
+}
+
+/*
+ * Optimized fast_imageblit for bpp == 32. ppw = 1, bit_mask = 1 folded
+ * into the code, main loop unrolled.
+ */
+
+static inline void fast_imageblit32(const struct fb_image *image,
+				    struct fb_info *p, u8 __iomem * dst1,
+				    u32 fgcolor, u32 bgcolor)
+{
+	u32 fgx = fgcolor, bgx = bgcolor;
+	u32 spitch = (image->width + 7) / 8;
+	u32 end_mask, eorx;
+	const char *s = image->data, *src;
+	u32 __iomem *dst;
+	const u32 *tab = NULL;
+	int i, j, k;
+
+	tab = cfb_tab32;
+
+	eorx = fgx ^ bgx;
+	k = image->width;
+
+	for (i = image->height; i--;) {
+		dst = (u32 __iomem *) dst1;
+		src = s;
+
+		j = k;
+		while (j >= 8) {
+			u8 bits = *src;
+			end_mask = tab[(bits >> 7) & 1];
+			FB_WRITEL((end_mask & eorx) ^ bgx, dst++);
+			end_mask = tab[(bits >> 6) & 1];
+			FB_WRITEL((end_mask & eorx) ^ bgx, dst++);
+			end_mask = tab[(bits >> 5) & 1];
+			FB_WRITEL((end_mask & eorx) ^ bgx, dst++);
+			end_mask = tab[(bits >> 4) & 1];
+			FB_WRITEL((end_mask & eorx) ^ bgx, dst++);
+			end_mask = tab[(bits >> 3) & 1];
+			FB_WRITEL((end_mask & eorx) ^ bgx, dst++);
+			end_mask = tab[(bits >> 2) & 1];
+			FB_WRITEL((end_mask & eorx) ^ bgx, dst++);
+			end_mask = tab[(bits >> 1) & 1];
+			FB_WRITEL((end_mask & eorx) ^ bgx, dst++);
+			end_mask = tab[bits & 1];
+			FB_WRITEL((end_mask & eorx) ^ bgx, dst++);
+			src++;
+			j -= 8;
+		}
+		if (j != 0) {
+			u32 bits = (u32) * src;
+			while (j > 1) {
+				end_mask = tab[(bits >> 7) & 1];
+				FB_WRITEL((end_mask & eorx) ^ bgx, dst++);
+				bits <<= 1;
+				j--;
+			}
+			end_mask = tab[(bits >> 7) & 1];
+			FB_WRITEL((end_mask & eorx) ^ bgx, dst);
+		}
+		dst1 += p->fix.line_length;
+		s += spitch;
+	}
+}
+
 void cfb_imageblit(struct fb_info *p, const struct fb_image *image)
 {
 	u32 fgcolor, bgcolor, start_index, bitstart, pitch_index = 0;
@@ -294,11 +426,21 @@ void cfb_imageblit(struct fb_info *p, co
 			bgcolor = image->bg_color;
 		}	
 		
-		if (32 % bpp == 0 && !start_index && !pitch_index && 
-		    ((width & (32/bpp-1)) == 0) &&
-		    bpp >= 8 && bpp <= 32) 			
-			fast_imageblit(image, p, dst1, fgcolor, bgcolor);
-		else 
+		if (!start_index && !pitch_index) {
+			if (bpp == 32)
+				fast_imageblit32(image, p, dst1, fgcolor,
+						 bgcolor);
+			else if (bpp == 16 && (width & 1) == 0)
+				fast_imageblit16(image, p, dst1, fgcolor,
+						 bgcolor);
+			else if (bpp == 8 && (width & 3) == 0)
+				fast_imageblit(image, p, dst1, fgcolor,
+					       bgcolor);
+			else
+				slow_imageblit(image, p, dst1, fgcolor,
+					       bgcolor,
+					       start_index, pitch_index);
+		} else
 			slow_imageblit(image, p, dst1, fgcolor, bgcolor,
 					start_index, pitch_index);
 	} else