Iriver remote LCD driver: * Split out assembler parts. * Reintroduced the 2 transfer routine variants (low/high CPU clock), and made the version for high CPU clock a little slower because there were problems reported. The function can be slowed down more if there are still problems.

git-svn-id: svn://svn.rockbox.org/rockbox/trunk@17006 a1c6a512-1295-4272-9138-f99709370657

2 files changed, 309 insertions, 204 deletions

diff --git a/firmware/target/coldfire/iriver/lcd-remote-as-iriver.S b/firmware/target/coldfire/iriver/lcd-remote-as-iriver.S
new file mode 100644
index 0000000000..0325e5176c
--- /dev/null
+++ b/firmware/target/coldfire/iriver/lcd-remote-as-iriver.S
@@ -0,0 +1,302 @@
+/***************************************************************************
+ *             __________               __   ___.
+ *   Open      \______   \ ____   ____ |  | _\_ |__   _______  ___
+ *   Source     |       _//  _ \_/ ___\|  |/ /| __ \ /  _ \  \/  /
+ *   Jukebox    |    |   (  <_> )  \___|    < | \_\ (  <_> > <  <
+ *   Firmware   |____|_  /\____/ \___  >__|_ \|___  /\____/__/\_ \
+ *                     \/            \/     \/    \/            \/
+ * $Id$
+ *
+ * Copyright (C) 2008 by Jens Arnold
+ *
+ * All files in this archive are subject to the GNU General Public License.
+ * See the file COPYING in the source tree root for full license agreement.
+ *
+ * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
+ * KIND, either express or implied.
+ *
+ ****************************************************************************/
+ 
+#include "config.h"
+
+#define RS_MASK        0x00010000
+#define CLOCK_MASK     0x10000000
+#define GPIO_OUT_ADDR  0x80000004
+
+#define CS_MASK        0x00000004  /* used in moveq.l */
+#define DATA_MASK      0x00040000
+#define GPIO1_OUT_ADDR 0x800000b4
+
+#define CS_TIMEOUT     10 /* HZ/10 */
+
+    .extern cpu_frequency       /* Global variable from system.c */
+    .extern remote_byte_delay   /* Global variable from lcd-remote-iriver.c */
+    .extern remote_cs_countdown /* Global variable from lcd-remote-iriver.c */
+
+    .section    .icode,"ax",@progbits
+
+    /* Output 8 bits to the LCD. Instruction order is devised to maximize the
+     * delay between changing the data line and the CLK L->H transition, which
+     * makes the LCD controller sample DATA.
+     * Requires CLK = 0 on entry.
+     *
+     * Custom calling convention:
+     *   %a0 - GPIO_OUT_ADDR
+     *   %a1 - GPIO1_OUT_ADDR
+     *   %d4 - data byte
+     *   %d6 - DATA_MASK
+      * Clobbers:
+     *   %d0..%d4
+     */
+#ifdef HAVE_REMOTE_LCD_TICKING
+.write_byte_delayed:
+    move.l  remote_byte_delay, %d0
+1:
+    subq.l  #1, %d0
+    bne.s   1b
+#endif
+
+.write_byte:
+    move.w  %sr, %d3            /* Get current interrupt level */
+    move.w  #0x2700, %sr        /* Disable interrupts */
+
+    move.l  (%a1), %d0          /* Get current state of data port */
+    move.l  %d0, %d1          
+    and.l   %d6, %d1            /* Check current state of data line */
+    beq.s   1f                  /*   and set it as previous-state bit */
+    bset    #8, %d4
+1:
+    move.l  %d4, %d1            /* Compute the 'bit derivative', i.e. a value */
+    lsr.l   #1, %d1             /*   with 1's where the data changes from the */
+    eor.l   %d1, %d4            /*   previous state, and 0's where it doesn't */
+    swap    %d4                 /* Shift data to upper byte */
+    lsl.l   #8, %d4
+
+    move.l  (%a0),%d1           /* Get current state of clock port */
+    move.l  %d1, %d2            /* Precalculate opposite state of clock line */
+    eor.l   #CLOCK_MASK, %d2
+    
+    lsl.l   #1, %d4             /* Invert data line for bit7 ? */
+    bcc.s   1f                  /*   no: skip */
+    eor.l   %d6, %d0            /* invert data bit */
+    move.l  %d0, (%a1)          /*   output data bit7 */
+    nop
+1:
+
+.macro bit_out
+    lsl.l   #1, %d4             /* Invert data line for bit6 ? */
+    bcc.s   1f                  /*   no: skip */
+    eor.l   %d6, %d0            /* Invert data bit */
+    move.l  %d2, (%a0)          /* Bit7: set CLK = 1 */
+    move.l  %d1, (%a0)          /*       set CLK = 0 */
+    move.l  %d0, (%a1)          /* Output data bit6 */
+    bra.s   2f                  /* slower than trapf.l - required here */
+1:                              /*   else */
+    move.l  %d2, (%a0)          /* Bit7: set CLK = 1 */
+    move.l  %d1, (%a0)          /*       set CLK = 0 */
+2:
+.endm
+    bit_out
+    bit_out
+    bit_out
+    bit_out
+    bit_out
+    bit_out
+    bit_out
+
+    nop                         /* Let data line settle */
+    move.l  %d2, (%a0)          /* Bit0: Set CLK = 1 */
+    move.l  %d1, (%a0)          /*       Set CLK = 0 */
+
+    move.w  %d3, %sr            /* Restore interrupt level */
+    rts
+
+    /* Output 8 bits to the LCD as fast as possible. Use only at < 60MHz.
+     *
+     * Custom calling convention:
+     *   %a0 - GPIO_OUT_ADDR
+     *   %a1 - GPIO1_OUT_ADDR
+     *   %d4 - data word
+     *   %d6 - DATA_MASK
+     * Clobbers:
+     *   %d0..%d4
+     */
+#ifdef HAVE_REMOTE_LCD_TICKING
+.write_byte_fast_delayed:
+    move.l  remote_byte_delay, %d0
+1:
+    subq.l  #1, %d0
+    bne.s   1b
+#endif
+
+.write_byte_fast:
+    move.w  %sr, %d3            /* Get current interrupt level */
+    move.w  #0x2700,%sr         /* Disable interrupts */
+
+    move.l  (%a1), %d0          /* Get current state of data port */
+    move.l  %d0, %d1          
+    and.l   %d6, %d1            /* Check current state of data line */
+    beq.s   1f                  /*   and set it as previous-state bit */
+    bset    #8, %d4
+1:
+    move.l  %d4, %d1            /* Compute the 'bit derivative', i.e. a value */
+    lsr.l   #1, %d1             /*   with 1's where the data changes from the */
+    eor.l   %d1, %d4            /*   previous state, and 0's where it doesn't */
+    swap    %d4                 /* Shift data to upper byte */
+    lsl.l   #8, %d4
+
+    move.l  (%a0), %d1          /* Get current state of clock port */
+    move.l  %d1, %d2            /* Precalculate opposite state of clock line */
+    eor.l   #CLOCK_MASK, %d2
+
+.macro bit_out_fast
+    lsl.l   #1,%d4              /* Shift out MSB */
+    bcc.s   1f                  
+    eor.l   %d6, %d0            /* 1: flip data bit */
+    move.l  %d0, (%a1)          /*   and output new DATA state */
+1:
+    move.l  %d2, (%a0)          /* Set CLK */
+    move.l  %d1, (%a0)          /* Reset CLK */
+.endm
+    bit_out_fast
+    bit_out_fast
+    bit_out_fast
+    bit_out_fast
+    bit_out_fast
+    bit_out_fast
+    bit_out_fast
+    bit_out_fast
+
+    move.w  %d3, %sr            /* Restore interrupt level */
+    rts
+    
+    
+    .global lcd_remote_write_command
+    .type   lcd_remote_write_command, @function
+
+lcd_remote_write_command:
+    lea.l   (-4*4, %sp), %sp
+    movem.l %d2-%d4/%d6, (%sp)
+
+    move.l  (4*4+4, %sp), %d4   /* cmd */
+
+    lea.l   GPIO_OUT_ADDR, %a0
+    lea.l   GPIO1_OUT_ADDR, %a1
+    move.l  #DATA_MASK, %d6
+    
+    clr.l   remote_cs_countdown
+
+    move.l  #~RS_MASK, %d0
+    and.l   %d0, (%a0)
+    moveq.l #~CS_MASK, %d0
+    and.l   %d0, (%a1)
+    
+#ifdef HAVE_REMOTE_LCD_TICKING
+    tst.l   remote_byte_delay
+    ble.s   1f
+    bsr.w   .write_byte_delayed
+    bra.s   2f
+1:
+#endif
+    bsr.w   .write_byte
+2:
+
+    moveq.l #CS_TIMEOUT, %d0
+    move.l  %d0, remote_cs_countdown
+
+    movem.l (%sp), %d2-%d4/%d6
+    lea.l   (4*4, %sp), %sp
+    rts
+
+
+    .global lcd_remote_write_command_ex
+    .type   lcd_remote_write_command_ex, @function
+
+lcd_remote_write_command_ex:
+    lea.l   (-4*4, %sp), %sp
+    movem.l %d2-%d4/%d6, (%sp)
+    
+    lea.l   GPIO_OUT_ADDR, %a0
+    lea.l   GPIO1_OUT_ADDR, %a1
+    move.l  #DATA_MASK, %d6
+
+    clr.l   remote_cs_countdown
+
+    move.l  #~RS_MASK, %d0
+    and.l   %d0, (%a0)
+    moveq.l #~CS_MASK, %d0
+    and.l   %d0, (%a1)
+
+    move.l  (4*4+4, %sp), %d4
+
+#ifdef HAVE_REMOTE_LCD_TICKING
+    tst.l   remote_byte_delay
+    ble.s   1f
+    bsr.w   .write_byte_delayed
+    move.l  (4*4+8, %sp), %d4
+    bsr.w   .write_byte_delayed
+    bra.s   2f
+1:
+#endif
+    bsr.w   .write_byte
+    move.l  (4*4+8, %sp), %d4
+    bsr.w   .write_byte
+2:
+
+    moveq.l #CS_TIMEOUT, %d0
+    move.l  %d0, remote_cs_countdown
+
+    movem.l (%sp), %d2-%d4/%d6
+    lea.l   (4*4, %sp), %sp
+    rts
+
+
+    .global lcd_remote_write_data
+    .type   lcd_remote_write_data, @function
+
+lcd_remote_write_data:
+    lea.l   (-7*4, %sp), %sp
+    movem.l %d2-%d6/%a2-%a3, (%sp)
+
+    move.l  (7*4+4, %sp), %a2   /* p_bytes */
+    move.l  (7*4+8, %sp), %d5   /* count */
+
+    lea.l   GPIO_OUT_ADDR, %a0
+    lea.l   GPIO1_OUT_ADDR, %a1
+    move.l  #DATA_MASK, %d6
+
+    lea.l   .write_byte, %a3
+    move.l  cpu_frequency, %d0
+    cmp.l   #60000000, %d0
+    bhi.b   1f
+    lea.l   .write_byte_fast, %a3
+1:
+
+#ifdef HAVE_REMOTE_LCD_TICKING
+    tst.l   remote_byte_delay
+    ble.s   1f
+    moveq.l #(.write_byte_delayed - .write_byte), %d0
+    add.l   %d0, %a3
+1:
+#endif
+
+    clr.l   remote_cs_countdown
+
+    move.l  #RS_MASK, %d0
+    or.l    %d0, (%a0)
+    moveq.l #~CS_MASK, %d0
+    and.l   %d0, (%a1)
+
+.wd_loop:                
+    clr.l   %d4
+    move.b  (%a2)+, %d4
+    jsr     (%a3)
+    subq.l  #1, %d5
+    bne.s   .wd_loop
+
+    moveq.l #CS_TIMEOUT, %d0
+    move.l  %d0, remote_cs_countdown
+
+    movem.l (%sp), %d2-%d6/%a2-%a3
+    lea.l   (7*4, %sp), %sp
+    rts
diff --git a/firmware/target/coldfire/iriver/lcd-remote-iriver.c b/firmware/target/coldfire/iriver/lcd-remote-iriver.c
index 6ff76f3dd6..85ba0fef7a 100644
--- a/firmware/target/coldfire/iriver/lcd-remote-iriver.c
+++ b/firmware/target/coldfire/iriver/lcd-remote-iriver.c
@@ -56,13 +56,13 @@
 static int xoffset; /* needed for flip */
 
 /* timeout counter for deasserting /CS after access, <0 means not counting */
-static int cs_countdown IDATA_ATTR = 0;
+int remote_cs_countdown IDATA_ATTR = 0;
 #define CS_TIMEOUT (HZ/10)
 
 #ifdef HAVE_REMOTE_LCD_TICKING
 /* If set to true, will prevent "ticking" to headphones. */
 static bool emireduce = false;
-static int byte_delay = 0;
+int remote_byte_delay = 0; /* used in lcd-remote-as-iriver.S */
 #endif
 
 bool remote_initialized = false;
@@ -74,203 +74,6 @@ static bool cached_flip = false;
 static int cached_contrast = DEFAULT_REMOTE_CONTRAST_SETTING;
 
 
-#ifdef HAVE_REMOTE_LCD_TICKING
-static inline void _byte_delay(int delay)
-{
-    asm (
-        "move.l  %[dly], %%d0    \n"
-        "ble.s   2f              \n"
-    "1:                          \n"
-        "subq.l  #1, %%d0        \n"
-        "bne.s   1b              \n"
-    "2:                          \n"
-        : /* outputs */
-        : /* inputs */
-        [dly]"d"(delay)
-        : /* clobbers */
-        "d0"
-    );
-}
-#endif /* HAVE_REMOTE_LCD_TICKING */
-
-/* Low-level byte writer. Instruction order is devised to maximize the delay
- * between changing the data line and the CLK L->H transition, which makes
- * the LCD controller sample DATA, so it's fast yet usable even when boosted.
- * Requires CLK low on entry */
-static inline void _write_byte(unsigned data)
-{
-    asm volatile (
-        "move.w  %%sr, %%d3          \n"  /* Get current interrupt level */
-        "move.w  #0x2700, %%sr       \n"  /* Disable interrupts */
-
-        "move.l  (%[gpo1]), %%d0     \n"  /* Get current state of data port */
-        "move.l  %%d0, %%d1          \n"
-        "and.l   %[dbit], %%d1       \n"  /* Check current state of data line */
-        "beq.s   1f                  \n"  /*   and set it as previous-state bit */
-        "bset    #8, %[data]         \n"
-    "1:                              \n"
-        "move.l  %[data], %%d1       \n"  /* Compute the 'bit derivative', i.e. a value */
-        "lsr.l   #1, %%d1            \n"  /*   with 1's where the data changes from the */
-        "eor.l   %%d1, %[data]       \n"  /*   previous state, and 0's where it doesn't */
-        "swap    %[data]             \n"  /* Shift data to upper byte */
-        "lsl.l   #8, %[data]         \n"
-
-        "move.l  (%[gpo0]),%%d1      \n"  /* Get current state of clock port */
-        "move.l  %%d1, %%d2          \n"  /* Precalculate opposite state of clock line */
-        "eor.l   %[cbit], %%d2       \n"
-        
-        "lsl.l   #1, %[data]         \n"  /* Invert data line for bit7 ? */
-        "bcc.s   1f                  \n"  /*   no: skip */
-        "eor.l   %[dbit], %%d0       \n"  /* invert data bit */
-        "move.l  %%d0, (%[gpo1])     \n"  /*   output data bit7 */
-    "1:                              \n"
-
-        "lsl.l   #1, %[data]         \n"  /* Invert data line for bit6 ? */
-        "bcc.s   1f                  \n"  /*   no: skip */
-        "eor.l   %[dbit], %%d0       \n"  /* Invert data bit */
-        "move.l  %%d2, (%[gpo0])     \n"  /* Bit7: set CLK = 1 */
-        "move.l  %%d1, (%[gpo0])     \n"  /*       set CLK = 0 */
-        "move.l  %%d0, (%[gpo1])     \n"  /* Output data bit6 */
-        ".word   0x51fb              \n"  /* trapf.l - skip next 2 insns */
-    "1:                              \n"  /*   else */
-        "move.l  %%d2, (%[gpo0])     \n"  /* Bit7: set CLK = 1 */
-        "move.l  %%d1, (%[gpo0])     \n"  /*       set CLK = 0 */
-
-        "lsl.l   #1, %[data]         \n"  /* Unrolled */
-        "bcc.s   1f                  \n"
-        "eor.l   %[dbit], %%d0       \n"
-        "move.l  %%d2, (%[gpo0])     \n"
-        "move.l  %%d1, (%[gpo0])     \n"
-        "move.l  %%d0, (%[gpo1])     \n"
-        ".word   0x51fb              \n"
-    "1:                              \n"
-        "move.l  %%d2, (%[gpo0])     \n"
-        "move.l  %%d1, (%[gpo0])     \n"
-
-        "lsl.l   #1, %[data]         \n"
-        "bcc.s   1f                  \n"
-        "eor.l   %[dbit], %%d0       \n"
-        "move.l  %%d2, (%[gpo0])     \n"
-        "move.l  %%d1, (%[gpo0])     \n"
-        "move.l  %%d0, (%[gpo1])     \n"
-        ".word   0x51fb              \n"
-    "1:                              \n"
-        "move.l  %%d2, (%[gpo0])     \n"
-        "move.l  %%d1, (%[gpo0])     \n"
-
-        "lsl.l   #1, %[data]         \n"
-        "bcc.s   1f                  \n"
-        "eor.l   %[dbit], %%d0       \n"
-        "move.l  %%d2, (%[gpo0])     \n"
-        "move.l  %%d1, (%[gpo0])     \n"
-        "move.l  %%d0, (%[gpo1])     \n"
-        ".word   0x51fb              \n"
-    "1:                              \n"
-        "move.l  %%d2, (%[gpo0])     \n"
-        "move.l  %%d1, (%[gpo0])     \n"
-
-        "lsl.l   #1, %[data]         \n"
-        "bcc.s   1f                  \n"
-        "eor.l   %[dbit], %%d0       \n"
-        "move.l  %%d2, (%[gpo0])     \n"
-        "move.l  %%d1, (%[gpo0])     \n"
-        "move.l  %%d0, (%[gpo1])     \n"
-        ".word   0x51fb              \n"
-    "1:                              \n"
-        "move.l  %%d2, (%[gpo0])     \n"
-        "move.l  %%d1, (%[gpo0])     \n"
-
-        "lsl.l   #1, %[data]         \n"
-        "bcc.s   1f                  \n"
-        "eor.l   %[dbit], %%d0       \n"
-        "move.l  %%d2, (%[gpo0])     \n"
-        "move.l  %%d1, (%[gpo0])     \n"
-        "move.l  %%d0, (%[gpo1])     \n"
-        ".word   0x51fb              \n"
-    "1:                              \n"
-        "move.l  %%d2, (%[gpo0])     \n"
-        "move.l  %%d1, (%[gpo0])     \n"
-
-        "lsl.l   #1, %[data]         \n"
-        "bcc.s   1f                  \n"
-        "eor.l   %[dbit], %%d0       \n"
-        "move.l  %%d2, (%[gpo0])     \n"
-        "move.l  %%d1, (%[gpo0])     \n"
-        "move.l  %%d0, (%[gpo1])     \n"
-        ".word   0x51fb              \n"
-    "1:                              \n"
-        "move.l  %%d2, (%[gpo0])     \n"
-        "move.l  %%d1, (%[gpo0])     \n"
-
-        "nop                         \n"  /* Let data line settle */
-        "move.l  %%d2, (%[gpo0])     \n"  /* Bit0: Set CLK = 1 */
-        "move.l  %%d1, (%[gpo0])     \n"  /*       Set CLK = 0 */
-
-        "move.w  %%d3, %%sr          \n"  /* Restore interrupt level */
-        : /* outputs */
-        [data]"+d"(data)
-        : /* inputs */
-        [gpo0]"a"(&GPIO_OUT),
-        [cbit]"i"(0x10000000),
-        [gpo1]"a"(&GPIO1_OUT),
-        [dbit]"d"(0x00040000)
-        : /* clobbers */
-        "d0", "d1", "d2", "d3"
-    );
-}
-
-void lcd_remote_write_command(int cmd)
-{
-    cs_countdown = 0;
-    RS_LO;
-    CS_LO;
-
-    _write_byte(cmd);
-#ifdef HAVE_REMOTE_LCD_TICKING
-     _byte_delay(byte_delay);
-#endif
-
-    cs_countdown = CS_TIMEOUT;
-}
-
-void lcd_remote_write_command_ex(int cmd, int data)
-{
-    cs_countdown = 0;
-    RS_LO;
-    CS_LO;
-
-    _write_byte(cmd);
-#ifdef HAVE_REMOTE_LCD_TICKING
-    _byte_delay(byte_delay);
-#endif
-    _write_byte(data);
-#ifdef HAVE_REMOTE_LCD_TICKING
-    _byte_delay(byte_delay);
-#endif
-
-    cs_countdown = CS_TIMEOUT;
-}
-
-void lcd_remote_write_data(const unsigned char* p_bytes, int count) ICODE_ATTR;
-void lcd_remote_write_data(const unsigned char* p_bytes, int count)
-{
-    const unsigned char *p_end = p_bytes + count;
-
-    cs_countdown = 0;
-    RS_HI;
-    CS_LO;
-
-    while (p_bytes < p_end)
-    {
-        _write_byte(*p_bytes++);
-#ifdef HAVE_REMOTE_LCD_TICKING
-        _byte_delay(byte_delay);
-#endif
-    }
-
-    cs_countdown = CS_TIMEOUT;
-}
-
 /*** hardware configuration ***/
 
 int lcd_remote_default_contrast(void)
@@ -448,9 +251,9 @@ static void remote_tick(void)
     }
 
     /* handle chip select timeout */
-    if (cs_countdown >= 0)
-        cs_countdown--;
-    if (cs_countdown == 0)
+    if (remote_cs_countdown >= 0)
+        remote_cs_countdown--;
+    if (remote_cs_countdown == 0)
         CS_HI;
 }
 #endif
@@ -496,7 +299,7 @@ void lcd_remote_update(void)
 
 #ifdef HAVE_REMOTE_LCD_TICKING
     /* Adjust byte delay for emi reduction. */
-    byte_delay = emireduce ? cpu_frequency / 192800 - 90: 0;
+    remote_byte_delay = emireduce ? cpu_frequency / 192800 - 100: 0;
 #endif
 
     /* Copy display bitmap to hardware */
@@ -530,7 +333,7 @@ void lcd_remote_update_rect(int x, int y, int width, int height)
 
 #ifdef HAVE_REMOTE_LCD_TICKING
     /* Adjust byte delay for emi reduction */
-    byte_delay = emireduce ? cpu_frequency / 192800 - 90: 0;
+    remote_byte_delay = emireduce ? cpu_frequency / 192800 - 100: 0;
 #endif
 
     /* Copy specified rectange bitmap to hardware */