diff options
Diffstat (limited to 'drivers/gpu/drm/udl/udl_transfer.c')
| -rw-r--r-- | drivers/gpu/drm/udl/udl_transfer.c | 49 |
1 files changed, 23 insertions, 26 deletions
diff --git a/drivers/gpu/drm/udl/udl_transfer.c b/drivers/gpu/drm/udl/udl_transfer.c index 0c87b1ac6b68..ce87661e544f 100644 --- a/drivers/gpu/drm/udl/udl_transfer.c +++ b/drivers/gpu/drm/udl/udl_transfer.c @@ -13,7 +13,6 @@ #include <linux/module.h> #include <linux/slab.h> #include <linux/fb.h> -#include <linux/prefetch.h> #include <asm/unaligned.h> #include <drm/drmP.h> @@ -51,9 +50,6 @@ static int udl_trim_hline(const u8 *bback, const u8 **bfront, int *width_bytes) int start = width; int end = width; - prefetch((void *) front); - prefetch((void *) back); - for (j = 0; j < width; j++) { if (back[j] != front[j]) { start = j; @@ -83,12 +79,12 @@ static inline u16 pixel32_to_be16(const uint32_t pixel) ((pixel >> 8) & 0xf800)); } -static inline u16 get_pixel_val16(const uint8_t *pixel, int bpp) +static inline u16 get_pixel_val16(const uint8_t *pixel, int log_bpp) { - u16 pixel_val16 = 0; - if (bpp == 2) + u16 pixel_val16; + if (log_bpp == 1) pixel_val16 = *(const uint16_t *)pixel; - else if (bpp == 4) + else pixel_val16 = pixel32_to_be16(*(const uint32_t *)pixel); return pixel_val16; } @@ -125,8 +121,9 @@ static void udl_compress_hline16( const u8 *const pixel_end, uint32_t *device_address_ptr, uint8_t **command_buffer_ptr, - const uint8_t *const cmd_buffer_end, int bpp) + const uint8_t *const cmd_buffer_end, int log_bpp) { + const int bpp = 1 << log_bpp; const u8 *pixel = *pixel_start_ptr; uint32_t dev_addr = *device_address_ptr; uint8_t *cmd = *command_buffer_ptr; @@ -139,8 +136,6 @@ static void udl_compress_hline16( const u8 *cmd_pixel_start, *cmd_pixel_end = NULL; uint16_t pixel_val16; - prefetchw((void *) cmd); /* pull in one cache line at least */ - *cmd++ = 0xaf; *cmd++ = 0x6b; *cmd++ = (uint8_t) ((dev_addr >> 16) & 0xFF); @@ -153,12 +148,11 @@ static void udl_compress_hline16( raw_pixels_count_byte = cmd++; /* we'll know this later */ raw_pixel_start = pixel; - cmd_pixel_end = pixel + (min(MAX_CMD_PIXELS + 1, - min((int)(pixel_end - pixel) / bpp, - (int)(cmd_buffer_end - cmd) / 2))) * bpp; + cmd_pixel_end = pixel + (min3(MAX_CMD_PIXELS + 1UL, + (unsigned long)(pixel_end - pixel) >> log_bpp, + (unsigned long)(cmd_buffer_end - 1 - cmd) / 2) << log_bpp); - prefetch_range((void *) pixel, (cmd_pixel_end - pixel) * bpp); - pixel_val16 = get_pixel_val16(pixel, bpp); + pixel_val16 = get_pixel_val16(pixel, log_bpp); while (pixel < cmd_pixel_end) { const u8 *const start = pixel; @@ -170,7 +164,7 @@ static void udl_compress_hline16( pixel += bpp; while (pixel < cmd_pixel_end) { - pixel_val16 = get_pixel_val16(pixel, bpp); + pixel_val16 = get_pixel_val16(pixel, log_bpp); if (pixel_val16 != repeating_pixel_val16) break; pixel += bpp; @@ -179,10 +173,10 @@ static void udl_compress_hline16( if (unlikely(pixel > start + bpp)) { /* go back and fill in raw pixel count */ *raw_pixels_count_byte = (((start - - raw_pixel_start) / bpp) + 1) & 0xFF; + raw_pixel_start) >> log_bpp) + 1) & 0xFF; /* immediately after raw data is repeat byte */ - *cmd++ = (((pixel - start) / bpp) - 1) & 0xFF; + *cmd++ = (((pixel - start) >> log_bpp) - 1) & 0xFF; /* Then start another raw pixel span */ raw_pixel_start = pixel; @@ -192,11 +186,14 @@ static void udl_compress_hline16( if (pixel > raw_pixel_start) { /* finalize last RAW span */ - *raw_pixels_count_byte = ((pixel-raw_pixel_start) / bpp) & 0xFF; + *raw_pixels_count_byte = ((pixel - raw_pixel_start) >> log_bpp) & 0xFF; + } else { + /* undo unused byte */ + cmd--; } - *cmd_pixels_count_byte = ((pixel - cmd_pixel_start) / bpp) & 0xFF; - dev_addr += ((pixel - cmd_pixel_start) / bpp) * 2; + *cmd_pixels_count_byte = ((pixel - cmd_pixel_start) >> log_bpp) & 0xFF; + dev_addr += ((pixel - cmd_pixel_start) >> log_bpp) * 2; } if (cmd_buffer_end <= MIN_RLX_CMD_BYTES + cmd) { @@ -219,19 +216,19 @@ static void udl_compress_hline16( * (that we can only write to, slowly, and can never read), and (optionally) * our shadow copy that tracks what's been sent to that hardware buffer. */ -int udl_render_hline(struct drm_device *dev, int bpp, struct urb **urb_ptr, +int udl_render_hline(struct drm_device *dev, int log_bpp, struct urb **urb_ptr, const char *front, char **urb_buf_ptr, u32 byte_offset, u32 device_byte_offset, u32 byte_width, int *ident_ptr, int *sent_ptr) { const u8 *line_start, *line_end, *next_pixel; - u32 base16 = 0 + (device_byte_offset / bpp) * 2; + u32 base16 = 0 + (device_byte_offset >> log_bpp) * 2; struct urb *urb = *urb_ptr; u8 *cmd = *urb_buf_ptr; u8 *cmd_end = (u8 *) urb->transfer_buffer + urb->transfer_buffer_length; - BUG_ON(!(bpp == 2 || bpp == 4)); + BUG_ON(!(log_bpp == 1 || log_bpp == 2)); line_start = (u8 *) (front + byte_offset); next_pixel = line_start; @@ -241,7 +238,7 @@ int udl_render_hline(struct drm_device *dev, int bpp, struct urb **urb_ptr, udl_compress_hline16(&next_pixel, line_end, &base16, - (u8 **) &cmd, (u8 *) cmd_end, bpp); + (u8 **) &cmd, (u8 *) cmd_end, log_bpp); if (cmd >= cmd_end) { int len = cmd - (u8 *) urb->transfer_buffer; |