head.s分析(2):SYSCFG配置

本文介绍如何在ADI BF561 DSP上配置和使用Cycle Counter进行性能统计。通过对SYSCFG寄存器的设置,启用64位周期计数器,并详细解释了CYCLES和CYCLES2寄存器的功能及读取方法。

 

快乐虾

http://blog.youkuaiyun.com/lights_joy/

lights@hb165.com

  

本文适用于

ADI bf561 DSP

uclinux-2008r1.5-rc3 (移植到vdsp5)

Visual DSP++ 5.0(update 5)

 

欢迎转载,但请保留作者信息

 

     /* Enable Cycle Counter and Nesting Of Interrupts */

#ifdef CONFIG_BFIN_SCRATCH_REG_CYCLES

     R0 = SYSCFG_SNEN;

#else

     R0 = SYSCFG_SNEN | SYSCFG_CCEN;

#endif

     SYSCFG = R0;

这几行代码用于设置SYSCFG的值,关于SYSCFG这个寄存器,vdsp文件这样说:

The System Configuration Register (SYSCFG) shown in Figure 4-3 controls the configuration of the processor

对于BF561而言,只有CCENSSSTEP两位有效,在这里将CCEN位设置为1,表示要启用64位的cycle counter。而对SNEN的置位将被忽略。

System Configuration  Register

在启用了CCEN之后,CYCLESCYCLES2两个寄存器将开始统计CCLK的个数。

下面是vdsp文档中对CYCLESCYCLES2这两个寄存器的一个说明:

The cycle counter counts CCLK cycles while the program is executing. All cycles, including execution, wait state, interrupts, and events, are counted while the processor is in User or Supervisor mode, but the cycle counter stops counting in Emulator mode.

The cycle counter is 64 bits wide and increments every cycle. The count value is stored in two 32-bit registers, CYCLES and CYCLES2. The least significant 32 bits (LSBs) are stored in CYCLES. The most significant 32 bits (MSBs) are stored in CYCLES2.

Note: To ensure read coherency, first read CYCLES, then CYCLES2, and then CYCLES again, to detect if an overflow has occurred in the LSBs during the read operations.

In User mode, these two registers may be read, but not written. In Supervisor and Emulator modes, they are read/write registers.

To enable the cycle counters, set the CCEN bit in the SYSCFG register. The following example shows how to use the cycle counter:

R2 = 0;

CYCLES = R2;

CYCLES2 = R2;

R2 = SYSCFG;

BITSET(R2,1);

SYSCFG = R2;

/* Insert code to be benchmarked here. */

R2 = SYSCFG;

BITCLR(R2,1);

SYSCFG = R2;

通常这两个寄存器可以用于性能统计。但是由于这两个寄存器在几种模式下均是可写的,因此也可以将它们当成通用寄存器使用。在uclinux内核中,提供了一个叫CONFIG_BFIN_SCRATCH_REG_CYCLES的选项来控制CYCLES的用途:

config BFIN_SCRATCH_REG_CYCLES

     bool "CYCLES"

     help

       Use the CYCLES register in the Blackfin exception handler

       as a stack scratch register.  This means you cannot

       safely use the CYCLES performance registers on a Blackfin

       board at anytime, but you can debug the system with a JTAG

       ICE and use the NMI.

当设置了CONFIG_BFIN_SCRATCH_REG_CYCLES后,在linux-2.6.x/arch/blackfin/mach-common/entry.S文件中将CYCLES做为普通寄存器使用:

#if defined(CONFIG_BFIN_SCRATCH_REG_RETN)

# define EX_SCRATCH_REG RETN

#elif defined(CONFIG_BFIN_SCRATCH_REG_RETE)

# define EX_SCRATCH_REG RETE

#else

# define EX_SCRATCH_REG CYCLES

#endif

自然,在此时CYCLES是不应该自动计数的,因此SYSCFG_CCEN应该保持为0

uclinux内核代码中搜索CYCLES,可以发现有以下几个文件使用了它:

linux-2.6.x/drivers/char/bfin_timer_latency.c

linux-2.6.x/drivers/media/video/blackfin/blackfin_cam.c

linux-2.6.x/drivers/zaptel/bfsi-spi-framework.c

linux-2.6.x/drivers/zaptel/bfsi.c

当要使用这几个驱动时,就一定需要使SYSCFG_CCEN=1

 

1       参考资料

head.s分析(1):保存u-boot传递过来的指针(2009-1-19)

我这样写也还是只进入一次 #include "ti_msp_dl_config.h" #include "stdio.h" #include <string.h> // 定义一个结构体来存储UART缓冲区的信息 typedef struct { uint8_t Head_addr; // 缓冲区头部地址 uint8_t Tail_addr; // 缓冲区尾部地址 char buff[256]; // 存储数据的缓冲区 uint8_t len; // 当前缓冲区中数据长度 } uart_buff_struct; // 声明一个全局的uart_buff_struct实例 uart_buff_struct uart1_buff; // 自定义函数strchrnul1,用于查找字符c在字符串s中的位置 int strchrnul1(const char *s, int c) { uint32_t temp = 0; // 循环直到找到字符c或到达字符串末尾'\0' while (*(s + temp) != '\0' && *(s + temp) != (char)c) { temp++; } // 返回字符c的位置,如果未找到则返回-1 return (*(s + temp) == (char)c) ? temp : -1; } // 延时函数,单位为毫秒 void delay_ms(unsigned int ms) { unsigned int i, j; // 外层循环控制总的延时时间 for (i = 0; i < ms; i++) { // 内层循环通过执行nop指令进行精确延时 for (j = 0; j < 8000; j++) { __asm__("nop"); // nop是“无操作”指令,消耗CPU周期以实现延时 } } } // 全局变量,用于存储当前曲线类型 char current_curve; // 全局变量,用于临时存储某些值(具体用途不明) int32_t temp; // 主函数 int main(void) { // 初始化系统配置 SYSCFG_DL_init(); // 清除UART中断挂起标志 NVIC_ClearPendingIRQ(UART_1_INST_INT_IRQN); // 使能UART中断 NVIC_EnableIRQ(UART_1_INST_INT_IRQN); // 初始化UART缓冲区的相关字段 uart1_buff.len = 0; uart1_buff.Tail_addr = 1; uart1_buff.Head_addr = 0; uart1_buff.buff[0] = 1; // 主循环 while (1) { // 在缓冲区中查找字符'J' char *result = strchr(&uart1_buff.buff[uart1_buff.Head_addr], 'J'); if (result) { // 找到字符'J'后,先清除LED引脚上的电平 DL_GPIO_clearPins(LED_PORT, LED_LED1_PIN); delay_ms(200); // 设置LED引脚上的电平 DL_GPIO_setPins(LED_PORT, LED_LED1_PIN); delay_ms(200); } // UART中断服务例程 void UART_1_INST_IRQHandler(void) { // 检查是否发生了接收中断 if (DL_UART_getEnabledInterruptStatus(UART_1_INST, DL_UART_INTERRUPT_RX)) { uart1_buff.buff[uart1_buff.Tail_addr] = DL_UART_receiveData(UART_1_INST);// 接收新数据并存入缓冲区 uart1_buff.Tail_addr++;// 更新尾部地址指针,并处理溢出情况 uart1_buff.Tail_addr %= 256; uart1_buff.len++;// 增加缓冲区中数据的长度计数 DL_UART_clearInterruptStatus(UART_1_INST, DL_UART_INTERRUPT_RX); // 清除接收中断状态 } }
07-28
09:46:37.1626 [INFO] Loaded monitor commands from: E:\Software\Renode\scripts/monitor.py 09:46:42.1768 [INFO] Including script: E:\Software\Renode\scripts\single-node\stm32f4_discovery.resc 09:46:42.2197 [INFO] System bus created. 09:46:44.2268 [INFO] sysbus: Loaded SVD: E:\Software\Renode\platforms\cpus\STM32F40x.svd. Name: STM32F40x. Description: STM32F40x. 09:46:44.9408 [INFO] sysbus: Loading segment of 78080 bytes length at 0x8000000. 09:46:44.9671 [INFO] sysbus: Loading segment of 25952 bytes length at 0x8013100. 09:46:45.0388 [INFO] cpu: Guessing VectorTableOffset value to be 0x8000000. 09:46:45.0530 [INFO] cpu: Setting initial values: PC = 0x8000CFD, SP = 0x20020000. 09:46:45.0611 [INFO] STM32F4_Discovery: Machine started. 09:46:45.1743 [WARNING] rcc: Unhandled write to offset 0xC, value 0x0. 09:46:45.1905 [WARNING] rcc: Unhandled write to offset 0x4. Unhandled bits: [29] when writing value 0x7405408. Tags: RESERVED (0x0). 09:46:45.1915 [WARNING] flash_controller: Unhandled write to offset 0x0. Unhandled bits: [9-10] when writing value 0x605. Tags: ICEN (0x1), DCEN (0x1). 09:46:45.2371 [WARNING] uart4: Unhandled read from offset 0x14. (2) 09:46:45.2382 [WARNING] uart4: Unhandled write to offset 0x14, value 0x0. 09:46:45.2969 [WARNING] sysbus: Read from an unimplemented register SYSCFG:EXTICR3 (0x40013810), returning a value from SVD: 0x0. 09:46:45.2989 [WARNING] sysbus: Write of value 0x0 to an unimplemented register SYSCFG:EXTICR3 (0x40013810) generated from SVD. 09:46:45.2989 [WARNING] sysbus: Read from an unimplemented register SYSCFG:EXTICR3 (0x40013810), returning a value from SVD: 0x0. 09:46:45.2999 [WARNING] sysbus: Write of value 0x0 to an unimplemented register SYSCFG:EXTICR3 (0x40013810) generated from SVD. 09:46:45.2999 [WARNING] sysbus: Read from an unimplemented register SYSCFG:EXTICR1 (0x40013808), returning a value from SVD: 0x0. 09:46:45.2999 [WARNING] sysbus: Write of value 0x0 to an unimplemented register SYSCFG:EXTICR1 (0x40013808) generated from SVD. 09:46:45.2999 [WARNING] sysbus: Read from an unimplemented register SYSCFG:EXTICR1 (0x40013808), returning a value from SVD: 0x0. 09:46:45.3060 [WARNING] sysbus: Write of value 0x0 to an unimplemented register SYSCFG:EXTICR1 (0x40013808) generated from SVD. 09:46:45.3221 [WARNING] sysbus: Read from an unimplemented register SYSCFG:PMC (0x40013804), returning a value from SVD: 0x0. 09:46:45.3232 [WARNING] sysbus: Write of value 0x800000 to an unimplemented register SYSCFG:PMC (0x40013804) generated from SVD. 10:15:41.2883 [WARNING] ethernet: Access to unknown phy 0 (327681) 10:15:41.2883 [INFO] ethernet: Setting RDLA to 0x20003A6C.
03-29
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包

打赏作者

嵌云阁主

你的鼓励将是我创作的最大动力

¥1 ¥2 ¥4 ¥6 ¥10 ¥20
扫码支付:¥1
获取中
扫码支付

您的余额不足,请更换扫码支付或充值

打赏作者

实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值