Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / drivers / video / fbdev / core / cfbimgblt.c
diff --git a/kernel/drivers/video/fbdev/core/cfbimgblt.c b/kernel/drivers/video/fbdev/core/cfbimgblt.c
new file mode 100644 (file)
index 0000000..a2bb276
--- /dev/null
@@ -0,0 +1,313 @@
+/*
+ *  Generic BitBLT function for frame buffer with packed pixels of any depth.
+ *
+ *      Copyright (C)  June 1999 James Simmons
+ *
+ *  This file is subject to the terms and conditions of the GNU General Public
+ *  License.  See the file COPYING in the main directory of this archive for
+ *  more details.
+ *
+ * NOTES:
+ *
+ *    This function copys a image from system memory to video memory. The
+ *  image can be a bitmap where each 0 represents the background color and
+ *  each 1 represents the foreground color. Great for font handling. It can
+ *  also be a color image. This is determined by image_depth. The color image
+ *  must be laid out exactly in the same format as the framebuffer. Yes I know
+ *  their are cards with hardware that coverts images of various depths to the
+ *  framebuffer depth. But not every card has this. All images must be rounded
+ *  up to the nearest byte. For example a bitmap 12 bits wide must be two 
+ *  bytes width. 
+ *
+ *  Tony: 
+ *  Incorporate mask tables similar to fbcon-cfb*.c in 2.4 API.  This speeds 
+ *  up the code significantly.
+ *  
+ *  Code for depths not multiples of BITS_PER_LONG is still kludgy, which is
+ *  still processed a bit at a time.   
+ *
+ *  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.
+ */
+#include <linux/module.h>
+#include <linux/string.h>
+#include <linux/fb.h>
+#include <asm/types.h>
+#include "fb_draw.h"
+
+#define DEBUG
+
+#ifdef DEBUG
+#define DPRINTK(fmt, args...) printk(KERN_DEBUG "%s: " fmt,__func__,## args)
+#else
+#define DPRINTK(fmt, args...)
+#endif
+
+static const u32 cfb_tab8_be[] = {
+    0x00000000,0x000000ff,0x0000ff00,0x0000ffff,
+    0x00ff0000,0x00ff00ff,0x00ffff00,0x00ffffff,
+    0xff000000,0xff0000ff,0xff00ff00,0xff00ffff,
+    0xffff0000,0xffff00ff,0xffffff00,0xffffffff
+};
+
+static const u32 cfb_tab8_le[] = {
+    0x00000000,0xff000000,0x00ff0000,0xffff0000,
+    0x0000ff00,0xff00ff00,0x00ffff00,0xffffff00,
+    0x000000ff,0xff0000ff,0x00ff00ff,0xffff00ff,
+    0x0000ffff,0xff00ffff,0x00ffffff,0xffffffff
+};
+
+static const u32 cfb_tab16_be[] = {
+    0x00000000, 0x0000ffff, 0xffff0000, 0xffffffff
+};
+
+static const u32 cfb_tab16_le[] = {
+    0x00000000, 0xffff0000, 0x0000ffff, 0xffffffff
+};
+
+static const u32 cfb_tab32[] = {
+       0x00000000, 0xffffffff
+};
+
+#define FB_WRITEL fb_writel
+#define FB_READL  fb_readl
+
+static inline void color_imageblit(const struct fb_image *image, 
+                                  struct fb_info *p, u8 __iomem *dst1, 
+                                  u32 start_index,
+                                  u32 pitch_index)
+{
+       /* Draw the penguin */
+       u32 __iomem *dst, *dst2;
+       u32 color = 0, val, shift;
+       int i, n, bpp = p->var.bits_per_pixel;
+       u32 null_bits = 32 - bpp;
+       u32 *palette = (u32 *) p->pseudo_palette;
+       const u8 *src = image->data;
+       u32 bswapmask = fb_compute_bswapmask(p);
+
+       dst2 = (u32 __iomem *) dst1;
+       for (i = image->height; i--; ) {
+               n = image->width;
+               dst = (u32 __iomem *) dst1;
+               shift = 0;
+               val = 0;
+               
+               if (start_index) {
+                       u32 start_mask = ~fb_shifted_pixels_mask_u32(p,
+                                               start_index, bswapmask);
+                       val = FB_READL(dst) & start_mask;
+                       shift = start_index;
+               }
+               while (n--) {
+                       if (p->fix.visual == FB_VISUAL_TRUECOLOR ||
+                           p->fix.visual == FB_VISUAL_DIRECTCOLOR )
+                               color = palette[*src];
+                       else
+                               color = *src;
+                       color <<= FB_LEFT_POS(p, bpp);
+                       val |= FB_SHIFT_HIGH(p, color, shift ^ bswapmask);
+                       if (shift >= null_bits) {
+                               FB_WRITEL(val, dst++);
+       
+                               val = (shift == null_bits) ? 0 : 
+                                       FB_SHIFT_LOW(p, color, 32 - shift);
+                       }
+                       shift += bpp;
+                       shift &= (32 - 1);
+                       src++;
+               }
+               if (shift) {
+                       u32 end_mask = fb_shifted_pixels_mask_u32(p, shift,
+                                               bswapmask);
+
+                       FB_WRITEL((FB_READL(dst) & end_mask) | val, dst);
+               }
+               dst1 += p->fix.line_length;
+               if (pitch_index) {
+                       dst2 += p->fix.line_length;
+                       dst1 = (u8 __iomem *)((long __force)dst2 & ~(sizeof(u32) - 1));
+
+                       start_index += pitch_index;
+                       start_index &= 32 - 1;
+               }
+       }
+}
+
+static inline void slow_imageblit(const struct fb_image *image, struct fb_info *p, 
+                                 u8 __iomem *dst1, u32 fgcolor,
+                                 u32 bgcolor, 
+                                 u32 start_index,
+                                 u32 pitch_index)
+{
+       u32 shift, color = 0, bpp = p->var.bits_per_pixel;
+       u32 __iomem *dst, *dst2;
+       u32 val, pitch = p->fix.line_length;
+       u32 null_bits = 32 - bpp;
+       u32 spitch = (image->width+7)/8;
+       const u8 *src = image->data, *s;
+       u32 i, j, l;
+       u32 bswapmask = fb_compute_bswapmask(p);
+
+       dst2 = (u32 __iomem *) dst1;
+       fgcolor <<= FB_LEFT_POS(p, bpp);
+       bgcolor <<= FB_LEFT_POS(p, bpp);
+
+       for (i = image->height; i--; ) {
+               shift = val = 0;
+               l = 8;
+               j = image->width;
+               dst = (u32 __iomem *) dst1;
+               s = src;
+
+               /* write leading bits */
+               if (start_index) {
+                       u32 start_mask = ~fb_shifted_pixels_mask_u32(p,
+                                               start_index, bswapmask);
+                       val = FB_READL(dst) & start_mask;
+                       shift = start_index;
+               }
+
+               while (j--) {
+                       l--;
+                       color = (*s & (1 << l)) ? fgcolor : bgcolor;
+                       val |= FB_SHIFT_HIGH(p, color, shift ^ bswapmask);
+                       
+                       /* Did the bitshift spill bits to the next long? */
+                       if (shift >= null_bits) {
+                               FB_WRITEL(val, dst++);
+                               val = (shift == null_bits) ? 0 :
+                                       FB_SHIFT_LOW(p, color, 32 - shift);
+                       }
+                       shift += bpp;
+                       shift &= (32 - 1);
+                       if (!l) { l = 8; s++; }
+               }
+
+               /* write trailing bits */
+               if (shift) {
+                       u32 end_mask = fb_shifted_pixels_mask_u32(p, shift,
+                                               bswapmask);
+
+                       FB_WRITEL((FB_READL(dst) & end_mask) | val, dst);
+               }
+               
+               dst1 += pitch;
+               src += spitch;  
+               if (pitch_index) {
+                       dst2 += pitch;
+                       dst1 = (u8 __iomem *)((long __force)dst2 & ~(sizeof(u32) - 1));
+                       start_index += pitch_index;
+                       start_index &= 32 - 1;
+               }
+               
+       }
+}
+
+/*
+ * fast_imageblit - optimized monochrome color expansion
+ *
+ * Only if:  bits_per_pixel == 8, 16, or 32
+ *           image->width is divisible by pixel/dword (ppw);
+ *           fix->line_legth is divisible by 4;
+ *           beginning and end of a scanline is dword aligned
+ */
+static inline void fast_imageblit(const struct fb_image *image, struct fb_info *p, 
+                                 u8 __iomem *dst1, u32 fgcolor, 
+                                 u32 bgcolor) 
+{
+       u32 fgx = fgcolor, bgx = bgcolor, bpp = p->var.bits_per_pixel;
+       u32 ppw = 32/bpp, spitch = (image->width + 7)/8;
+       u32 bit_mask, end_mask, eorx, shift;
+       const char *s = image->data, *src;
+       u32 __iomem *dst;
+       const u32 *tab = NULL;
+       int i, j, k;
+
+       switch (bpp) {
+       case 8:
+               tab = fb_be_math(p) ? cfb_tab8_be : cfb_tab8_le;
+               break;
+       case 16:
+               tab = fb_be_math(p) ? cfb_tab16_be : cfb_tab16_le;
+               break;
+       case 32:
+       default:
+               tab = cfb_tab32;
+               break;
+       }
+
+       for (i = ppw-1; i--; ) {
+               fgx <<= bpp;
+               bgx <<= bpp;
+               fgx |= fgcolor;
+               bgx |= bgcolor;
+       }
+       
+       bit_mask = (1 << ppw) - 1;
+       eorx = fgx ^ bgx;
+       k = image->width/ppw;
+
+       for (i = image->height; i--; ) {
+               dst = (u32 __iomem *) dst1, shift = 8; src = s;
+               
+               for (j = k; j--; ) {
+                       shift -= ppw;
+                       end_mask = tab[(*src >> shift) & bit_mask];
+                       FB_WRITEL((end_mask & eorx)^bgx, dst++);
+                       if (!shift) { shift = 8; src++; }               
+               }
+               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;
+       u32 bpl = sizeof(u32), bpp = p->var.bits_per_pixel;
+       u32 width = image->width;
+       u32 dx = image->dx, dy = image->dy;
+       u8 __iomem *dst1;
+
+       if (p->state != FBINFO_STATE_RUNNING)
+               return;
+
+       bitstart = (dy * p->fix.line_length * 8) + (dx * bpp);
+       start_index = bitstart & (32 - 1);
+       pitch_index = (p->fix.line_length & (bpl - 1)) * 8;
+
+       bitstart /= 8;
+       bitstart &= ~(bpl - 1);
+       dst1 = p->screen_base + bitstart;
+
+       if (p->fbops->fb_sync)
+               p->fbops->fb_sync(p);
+
+       if (image->depth == 1) {
+               if (p->fix.visual == FB_VISUAL_TRUECOLOR ||
+                   p->fix.visual == FB_VISUAL_DIRECTCOLOR) {
+                       fgcolor = ((u32*)(p->pseudo_palette))[image->fg_color];
+                       bgcolor = ((u32*)(p->pseudo_palette))[image->bg_color];
+               } else {
+                       fgcolor = image->fg_color;
+                       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 
+                       slow_imageblit(image, p, dst1, fgcolor, bgcolor,
+                                       start_index, pitch_index);
+       } else
+               color_imageblit(image, p, dst1, start_index, pitch_index);
+}
+
+EXPORT_SYMBOL(cfb_imageblit);
+
+MODULE_AUTHOR("James Simmons <jsimmons@users.sf.net>");
+MODULE_DESCRIPTION("Generic software accelerated imaging drawing");
+MODULE_LICENSE("GPL");
+