This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

release 下浮点运算速度问题



我在使用编程时发现浮点运算在debug 模式和release模式所有的时间相差太大,不知问题出在什么地方,请高手给以解答,谢谢。

我的程序中有这个函数:


    void timetest()
{
 float  fii_r,fuu_r,fuu_x,fii_x,fu_f,ii_x;
 int k;
fii_r=5678;
fuu_x=3478;
fii_x=7896;
fuu_r=3456.00;


while(1)
{
   

fu_f =sqrt(fuu_r);  //debug模式下用时 1us,release模式下,代码考入RAM运行 用时 35us


GpioDataRegs.GPASET.bit.GPIO17 = 1;   

fu_f  =-703*(fuu_r*fii_r+fuu_x*fii_x)/(fii_r*fii_r+fii_x*fii_x);  //debug模式下用时 2.5us,release模式下,代码考入RAM运行 用时 12us
  
GpioDataRegs.GPACLEAR.bit.GPIO17 = 1;
   
}   
}

map文件如下:

.text      0    00328000    00001b8e     RUN ADDR = 00009000
                  00328000    000014de     Example_2833xGpioToggle.obj (.text)
                  003294de    0000018b     DSP2833x_DefaultIsr.obj (.text)
                  00329669    0000010b     DSP2833x_SysCtrl.obj (.text)
                  00329774    000000b2     dsp28335_xintf.obj (.text)
                  00329826    00000083     rts2800_ml.lib : fs_div.obj (.text)
                  003298a9    00000078                    : fs_add.obj (.text)
                  00329921    0000005a                    : fs_mpy.obj (.text)
                  0032997b    00000050                    : sqrt.obj (.text)
                  003299cb    00000044                    : boot.obj (.text)
                  00329a0f    00000041                    : ldexp.obj (.text)
                  00329a50    00000028     DSP2833x_PieCtrl.obj (.text)
                  00329a78    00000025     rts2800_ml.lib : fs_toi.obj (.text)
                  00329a9d    00000022                    : i_div.obj (.text)
                  00329abf    00000020     DSP2833x_PieVect.obj (.text)
                  00329adf    0000001d     rts2800_ml.lib : fs_tou.obj (.text)
                  00329afc    0000001b                    : args_main.obj (.text)
                  00329b17    0000001a                    : i_tofs.obj (.text)
                  00329b31    00000019                    : exit.obj (.text)
                  00329b4a    00000017                    : fs_cmp.obj (.text)
                  00329b61    00000015     DSP2833x_MemCopy.obj (.text)
                  00329b76    0000000b     rts2800_ml.lib : u_div.obj (.text)
                  00329b81    00000009                    : _lock.obj (.text)
                  00329b8a    00000004     DSP2833x_usDelay.obj (.text)

.econst    0    00338000    00000400    
                  00338000    00000100     DSP2833x_PieVect.obj (.econst)
                  00338100    00000100     Example_2833xGpioToggle.obj (.econst:_auchCRCHi16)
                  00338200    00000100     Example_2833xGpioToggle.obj (.econst:_auchCRCLo16)
                  00338300    00000080     Example_2833xGpioToggle.obj (.econst:_fcos)
                  00338380    00000080     Example_2833xGpioToggle.obj (.econst:_fsin)

.cinit     0    00338400    00000032     RUN ADDR = 0000ab8e
                  00338400    00000018     Example_2833xGpioToggle.obj (.cinit)
                  00338418    0000000a     rts2800_ml.lib : _lock.obj (.cinit)
                  00338422    0000000a                    : exit.obj (.cinit)
                  0033842c    00000004                    : errno.obj (.cinit)
                  00338430    00000002     --HOLE-- [fill = 0]

00009ce7   _timetest               //temitest   函数地址

下面是cmd文件

MEMORY
{
PAGE 0:    /* Program Memory */
           /* Memory (RAM/FLASH/OTP) blocks can be moved to PAGE1 for data allocation */

   ZONE0       : origin = 0x004000, length = 0x001000     /* XINTF zone 0 */
   RAML0       : origin = 0x008000, length = 0x001000     /* on-chip RAM block L0 */
   RAML112     : origin = 0x009000, length = 0x003000     /* on-chip RAM block L1 */
   ZONE6       : origin = 0x0100000, length = 0x100000    /* XINTF zone 6 */
   ZONE7A      : origin = 0x0200000, length = 0x00FC00    /* XINTF zone 7 - program space */
   FLASHH      : origin = 0x300000, length = 0x008000     /* on-chip FLASH */
   FLASHG      : origin = 0x308000, length = 0x008000     /* on-chip FLASH */
   FLASHF      : origin = 0x310000, length = 0x008000     /* on-chip FLASH */
   FLASHE      : origin = 0x318000, length = 0x008000     /* on-chip FLASH */
   FLASHD      : origin = 0x320000, length = 0x008000     /* on-chip FLASH */
   FLASHC      : origin = 0x328000, length = 0x008000     /* on-chip FLASH */
   FLASHA      : origin = 0x338000, length = 0x007F80     /* on-chip FLASH */
   CSM_RSVD    : origin = 0x33FF80, length = 0x000076     /* Part of FLASHA.  Program with all 0x0000 when CSM is in use. */
   BEGIN       : origin = 0x33FFF6, length = 0x000002     /* Part of FLASHA.  Used for "boot to Flash" bootloader mode. */
   CSM_PWL     : origin = 0x33FFF8, length = 0x000008     /* Part of FLASHA.  CSM password locations in FLASHA */
   OTP         : origin = 0x380400, length = 0x000400     /* on-chip OTP */
   ADC_CAL     : origin = 0x380080, length = 0x000009     /* ADC_cal function in Reserved memory */
   
   IQTABLES    : origin = 0x3FE000, length = 0x000b50     /* IQ Math Tables in Boot ROM */
   IQTABLES2   : origin = 0x3FEB50, length = 0x00008c     /* IQ Math Tables in Boot ROM */  
   FPUTABLES   : origin = 0x3FEBDC, length = 0x0006A0     /* FPU Tables in Boot ROM */
   ROM         : origin = 0x3FF27C, length = 0x000D44     /* Boot ROM */        
   RESET       : origin = 0x3FFFC0, length = 0x000002     /* part of boot ROM  */
   VECTORS     : origin = 0x3FFFC2, length = 0x00003E     /* part of boot ROM  */

PAGE 1 :   /* Data Memory */
           /* Memory (RAM/FLASH/OTP) blocks can be moved to PAGE0 for program allocation */
           /* Registers remain on PAGE1                                                  */
   
   BOOT_RSVD   : origin = 0x000000, length = 0x000050     /* Part of M0, BOOT rom will use this for stack */
   RAMM0       : origin = 0x000050, length = 0x0003B0     /* on-chip RAM block M0 */
   RAMM1       : origin = 0x000400, length = 0x000400     /* on-chip RAM block M1 */
   RAML4       : origin = 0x00C000, length = 0x003000     /* on-chip RAM block L1 */
   RAML5       : origin = 0x00F000, length = 0x001000     /* on-chip RAM block L1 */
   ZONE6B      : origin = 0x100000, length = 0x100000     /* XINTF zone 6 - data space */
   ZONE7B      : origin = 0x20FC00, length = 0x000400     /* XINTF zone 7 - data space */
   FLASHB      : origin = 0x330000, length = 0x008000     /* on-chip FLASH */
}

/* Allocate sections to memory blocks.
   Note:
         codestart user defined section in DSP28_CodeStartBranch.asm used to redirect code
                   execution when booting to flash
         ramfuncs  user defined section to store functions that will be copied from Flash into RAM
*/
 
SECTIONS
{
 
   /* Allocate program areas: */
   .cinit              : LOAD=FLASHA,     
                         RUN = RAML112,
                         LOAD_START(_cinit_loadstart),
                         RUN_START(_cinit_runstart),
                         SIZE(_cinit_size)
                         PAGE = 0
   .pinit              : > FLASHA,     PAGE = 0
   codestart           : > BEGIN       PAGE = 0
   wddisable            : > FLASHA      PAGE = 0
   copysections         : > FLASHA      PAGE = 0
   
     .text             : LOAD = FLASHC,
                         RUN = RAML112,
                         LOAD_START(_text_loadstart),
                         RUN_START(_text_runstart),
                         SIZE(_text_size)
                         PAGE = 0
                         
                  
   csmpasswds          : > CSM_PWL     PAGE = 0
   csm_rsvd            : > CSM_RSVD    PAGE = 0
   
   /* Allocate uninitalized data sections: */
   .stack              : > RAMM1       PAGE = 1
   .ebss               : > RAML4       PAGE = 1
   .esysmem            : > RAMM1       PAGE = 1

   /* Initalized sections go in Flash */
   /* For SDFlash to program these, they must be allocated to page 0 */
   .econst             : > FLASHA      PAGE = 0
   .switch             : > FLASHA      PAGE = 0      

   /* Allocate IQ math areas: */
   IQmath              : > FLASHC      PAGE = 0                  /* Math Code */
   IQmathTables     : > IQTABLES,  PAGE = 0, TYPE = NOLOAD
   
   /* Uncomment the section below if calling the IQNexp() or IQexp()
      functions from the IQMath.lib library in order to utilize the
      relevant IQ Math table in Boot ROM (This saves space and Boot ROM
      is 1 wait-state). If this section is not uncommented, IQmathTables2
      will be loaded into other memory (SARAM, Flash, etc.) and will take
      up space, but 0 wait-state is possible.
   */
   /*
   IQmathTables2    : > IQTABLES2, PAGE = 0, TYPE = NOLOAD
   {
   
              IQmath.lib<IQNexpTable.obj> (IQmathTablesRam)
   
   }
   */
   
   FPUmathTables    : > FPUTABLES, PAGE = 0, TYPE = NOLOAD
         
   /* Allocate DMA-accessible RAM sections: */
   DMARAML4         : > RAML4,     PAGE = 1
   DMARAML5         : > RAML5,     PAGE = 1
   /*DMARAML6         : > RAML6,     PAGE = 1
   DMARAML7         : > RAML7,     PAGE = 1*/
   
   /* Allocate 0x400 of XINTF Zone 7 to storing data */
   ZONE6DATA        : > ZONE6B,    PAGE = 1    
   ZONE7DATA        : > ZONE7B,    PAGE = 1

   /* .reset is a standard section used by the compiler.  It contains the */
   /* the address of the start of _c_int00 for C Code.   /*
   /* When using the boot ROM this section and the CPU vector */
   /* table is not needed.  Thus the default type is set here to  */
   /* DSECT  */
   .reset              : > RESET,      PAGE = 0, TYPE = DSECT
   vectors             : > VECTORS     PAGE = 0, TYPE = DSECT
   
   /* Allocate ADC_cal function (pre-programmed by factory into TI reserved memory) */
   .adc_cal     : load = ADC_CAL,   PAGE = 0, TYPE = NOLOAD

}

/*
//===========================================================================
// End of file.
//===========================================================================
*/