• State Accepted Answer
  • Locked Locked
  • Replies 5 replies
  • Subscribers 21 subscribers
  • Views 9343 views
  • Users 0 members are here
2025 survey

We are running a survey to help us improve the experience for all of our members. If you see the survey appear, please take the time to tell us about your experience if you can.

Options
Related
How was your experience today?
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

Convert arm_bitreversal2.S file to c code

Yura Cobain

Hi,

I am working on porting FFT functionality from CMSIS DSP library to another architecture.
Unfortunate I faced with problem that arm_bitreversal_32 function is implemented in assembler.

Does any body tried to convert it to C code?

  • 0

    Hello,


    I cannot access the assembly source code of the arm_bitreversal_32 as of now.
    I re-wrote it by C as referring the below.

    http://www.stargate-tr.com/post/102141475474/arm-cmsis-dsp-software-library-complex-fft

    (1) original code.

       .syntax unified
    .thumb
    .global sg_bitreversal_32
sg_bitreversal_32:
    ADDS     r3,r1,#1
    CMP      r3,#1
    IT       LS
    BXLS     lr
    PUSH     {r4-r9}
    ADDS     r1,r2,#2
    LSRS     r3,r3,#2
sg_bitreversal_32_0:
    LDRH     r8,[r1,#4]
    LDRH     r9,[r1,#2]
    LDRH     r2,[r1,#0]
    LDRH     r12,[r1,#-2]
    ADD      r8,r0,r8
    ADD      r9,r0,r9
    ADD      r2,r0,r2
    ADD      r12,r0,r12
    LDR      r7,[r9,#0]
    LDR      r6,[r8,#0]
    LDR      r5,[r2,#0]
    LDR      r4,[r12,#0]
    STR      r6,[r9,#0]
    STR      r7,[r8,#0]
    STR      r5,[r12,#0]
    STR      r4,[r2,#0]
    LDR      r7,[r9,#4]
    LDR      r6,[r8,#4]
    LDR      r5,[r2,#4]
    LDR      r4,[r12,#4]
    STR      r6,[r9,#4]
    STR      r7,[r8,#4]
    STR      r5,[r12,#4]
    STR      r4,[r2,#4]
    ADDS     r1,r1,#8
    SUBS     r3,r3,#1
    BNE      sg_bitreversal_32_0
    POP      {r4-r9}
    BX       lr

    (2) re-written C code.

    void
arm_bitreversal_32 (uint32_t * pSrc, const uint16_t bitRevLen,
                    const uint16_t * pBitRevTable)
{
  uint32_t r7,r6,r5,r4,r3;
  if (bitRevLen <= 0)
    return;
  r3 = (bitRevLen >> 2);
  while (r3 >= 0)
    {
      r7 = pSrc[pBitRevTable[3]];
      r6 = pSrc[pBitRevTable[2]];
      r5 = pSrc[pBitRevTable[1]];
      r4 = pSrc[pBitRevTable[0]];
      pSrc[pBitRevTable[3]] = r6;
      pSrc[pBitRevTable[2]] = r7;
      pSrc[pBitRevTable[1]] = r4;
      pSrc[pBitRevTable[0]] = r5;
      r7 = pSrc[pBitRevTable[3] + 1];
      r6 = pSrc[pBitRevTable[2] + 1];
      r5 = pSrc[pBitRevTable[1] + 1];
      r4 = pSrc[pBitRevTable[0] + 1];
      pSrc[pBitRevTable[3] + 1] = r6;
      pSrc[pBitRevTable[2] + 1] = r7;
      pSrc[pBitRevTable[1] + 1] = r4;
      pSrc[pBitRevTable[0] + 1] = r5;
      pBitRevTable += 4;
      r3--;
    }
}

    Although I don't know the functional specification of the arm_bitreversal_32, I think I could interpret it well.

    However, there might be some bugs.
    I hope someone correct my code properly.

    Best regards,

    Yasuhiko Koumoto.

  • 0 in reply to Yasuhiko Koumoto

    Hello,

    I'm sorry that the C code was not eqivalent to the original code.

    The below is the corrected version.

    Best regards,

    Yasuhilko Koumoto.

    void
arm_bitreversal_32 (uint32_t * pSrc, const uint16_t bitRevLen,
                    const uint16_t * pBitRevTable)
{
  uint32_t r7,r6,r5,r4,r3;
  if (bitRevLen <= 0)
    return;
  r3 = ((bitRevLen + 1) >> 2);
  while (r3 > 0)
    {
      r7 = pSrc[(pBitRevTable[3]>>2)];
      r6 = pSrc[(pBitRevTable[2]>>2)];
      r5 = pSrc[(pBitRevTable[1]>>2)];
      r4 = pSrc[(pBitRevTable[0]>>2)];
      pSrc[(pBitRevTable[3]>>2)] = r6;
      pSrc[(pBitRevTable[2]>>2)] = r7;
      pSrc[(pBitRevTable[1]>>2)] = r4;
      pSrc[(pBitRevTable[0]>>2)] = r5;
      r7 = pSrc[(pBitRevTable[3]>>2) + 1];
      r6 = pSrc[(pBitRevTable[2]>>2) + 1];
      r5 = pSrc[(pBitRevTable[1]>>2) + 1];
      r4 = pSrc[(pBitRevTable[0]>>2) + 1];
      pSrc[(pBitRevTable[3]>>2) + 1] = r6;
      pSrc[(pBitRevTable[2]>>2) + 1] = r7;
      pSrc[(pBitRevTable[1]>>2) + 1] = r4;
      pSrc[(pBitRevTable[0]>>2) + 1] = r5;
      pBitRevTable += 4;
      r3--;
    }
}
  • 0 in reply to Yasuhiko Koumoto

    Hello,

    I'm very sorry. I revised the code again.

    This revision would be identical to the original.

    Best regards.

    Yasuhiko Koumoto.

    void
arm_bitreversal_32 (uint32_t * pSrc, const uint16_t bitRevLen,
                    const uint16_t * pBitRevTable)
{
  uint32_t r7,r6,r5,r4,r3;
  if (bitRevLen <= 0)
    return;
  r3 = ((bitRevLen+1) >> 2);
  while (r3 > 0)
    {
      r7 = *(uint32_t*)((uint8_t*) pSrc + pBitRevTable[3]);
      r6 = *(uint32_t*)((uint8_t*) pSrc + pBitRevTable[2]);
      r5 = *(uint32_t*)((uint8_t*) pSrc + pBitRevTable[1]);
      r4 = *(uint32_t*)((uint8_t*) pSrc + pBitRevTable[0]);
      *(uint32_t*)((uint8_t*) pSrc + pBitRevTable[3]) = r6;
      *(uint32_t*)((uint8_t*) pSrc + pBitRevTable[2]) = r7;
      *(uint32_t*)((uint8_t*) pSrc + pBitRevTable[1]) = r4;
      *(uint32_t*)((uint8_t*) pSrc + pBitRevTable[0]) = r5;
      r7 = *(uint32_t*)((uint8_t*) pSrc + pBitRevTable[3] + 4);
      r6 = *(uint32_t*)((uint8_t*) pSrc + pBitRevTable[2] + 4);
      r5 = *(uint32_t*)((uint8_t*) pSrc + pBitRevTable[1] + 4);
      r4 = *(uint32_t*)((uint8_t*) pSrc + pBitRevTable[0] + 4);
      *(uint32_t*)((uint8_t*) pSrc + pBitRevTable[3] + 4) = r6;
      *(uint32_t*)((uint8_t*) pSrc + pBitRevTable[2] + 4) = r7;
      *(uint32_t*)((uint8_t*) pSrc + pBitRevTable[1] + 4) = r4;
      *(uint32_t*)((uint8_t*) pSrc + pBitRevTable[0] + 4) = r5;
      pBitRevTable += 4;
      r3--;
    }
}
  • 0 in reply to Yasuhiko Koumoto

    Hello,

    Thank you for all your assistance.

    BR,

    Iurii

  • 0

    If you'd like other strange ways of doing this plus a number of other hacks in C try

    Bit Twiddling Hacks

    I feel a bit like I'm a drug peddler pushing that!