renegade-project/linux-next
2
0
Fork 0
mirror of https://github.com/edk2-porting/linux-next.git synced 2024-12-27 06:34:11 +08:00
Code
a9a62b6a88
linux-next/drivers/video/pxa3xx-gcu.h

39 lines
890 B
C
Raw Normal View History

video: add driver for PXA3xx 2D graphics accelerator This adds a driver for the the 2D graphics accelerator found on PXA3xx processors. Only resource mapping, interrupt handling and a simple ioctl handler is done by the kernel part, the rest of the logic is implemented in DirectFB userspace. Graphic applications greatly benefit for line drawing, blend, and rectangle and triangle filling operations. Benchmarks done on a PXA303 using the df_dok benchmarking tool follow, where the value in square brackets show the CPU usage during that test. Without accelerator (benchmarking 256x252 on 480x262 RGB16 (16bit)): Anti-aliased Text 3.016 secs ( 65.649 KChars/sec) [ 99.6%] Fill Rectangle 3.021 secs ( 175.107 MPixel/sec) [ 98.0%] Fill Rectangle (blend) 3.582 secs ( 3.602 MPixel/sec) [ 99.7%] Fill Rectangles [10] 3.177 secs ( 182.753 MPixel/sec) [ 98.1%] Fill Rectangles [10] (blend) 18.020 secs ( 3.580 MPixel/sec) [ 98.7%] Fill Spans 3.019 secs ( 145.306 MPixel/sec) [ 98.0%] Fill Spans (blend) 3.616 secs ( 3.568 MPixel/sec) [ 99.4%] Blit 3.074 secs ( 39.874 MPixel/sec) [ 98.0%] Blit 180 3.020 secs ( 32.042 MPixel/sec) [ 98.0%] Blit with format conversion 3.005 secs ( 19.321 MPixel/sec) [ 99.6%] Blit from 32bit (blend) 4.792 secs ( 2.692 MPixel/sec) [ 98.7%] With accelerator: Anti-aliased Text 3.056 secs (* 36.518 KChars/sec) [ 21.3%] Fill Rectangle 3.015 secs (* 115.543 MPixel/sec) [ 8.9%] Fill Rectangle (blend) 3.180 secs (* 20.286 MPixel/sec) [ 1.8%] Fill Rectangles [10] 3.251 secs (* 119.062 MPixel/sec) [ 1.2%] Fill Rectangles [10] (blend) 6.293 secs (* 20.502 MPixel/sec) [ 0.3%] Fill Spans 3.051 secs (* 97.264 MPixel/sec) [ 35.7%] Fill Spans (blend) 3.377 secs (* 15.282 MPixel/sec) [ 17.8%] Blit 3.046 secs (* 27.533 MPixel/sec) [ 2.6%] Blit 180 3.098 secs (* 27.070 MPixel/sec) [ 2.2%] Blit with format conversion 3.131 secs (* 39.148 MPixel/sec) [ 2.8%] Blit from 32bit (blend) 3.346 secs (* 11.568 MPixel/sec) [ 0.8%] Signed-off-by: Daniel Mack <daniel@caiaq.de> Tested-by: Sven Neumann <s.neumann@raumfeld.com> Cc: Eric Miao <eric.y.miao@gmail.com> Cc: Denis Oliver Kropp <dok@directfb.org> Cc: Sven Neumann <s.neumann@raumfeld.com> Cc: Haojian Zhuang <haojian.zhuang@gmail.com> Signed-off-by: Eric Miao <eric.y.miao@gmail.com>
2010-11-05 02:44:00 +08:00
#ifndef __PXA3XX_GCU_H__
#define __PXA3XX_GCU_H__
#include <linux/types.h>
/* Number of 32bit words in display list (ring buffer). */
#define PXA3XX_GCU_BUFFER_WORDS ((256 * 1024 - 256) / 4)
/* To be increased when breaking the ABI */
#define PXA3XX_GCU_SHARED_MAGIC 0x30000001
#define PXA3XX_GCU_BATCH_WORDS 8192
struct pxa3xx_gcu_shared {
u32 buffer[PXA3XX_GCU_BUFFER_WORDS];
bool hw_running;
unsigned long buffer_phys;
unsigned int num_words;
unsigned int num_writes;
unsigned int num_done;
unsigned int num_interrupts;
unsigned int num_wait_idle;
unsigned int num_wait_free;
unsigned int num_idle;
u32 magic;
};
/* Initialization and synchronization.
* Hardware is started upon write(). */
#define PXA3XX_GCU_IOCTL_RESET _IO('G', 0)
#define PXA3XX_GCU_IOCTL_WAIT_IDLE _IO('G', 2)
#endif /* __PXA3XX_GCU_H__ */
Copy Permalink