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

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?

Parents

0 Yasuhiko Koumoto over 9 years ago

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.

Reply

0 Yasuhiko Koumoto over 9 years ago

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.

Children

0 Yasuhiko Koumoto over 9 years ago 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 Yasuhiko Koumoto over 9 years ago 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 Yura Cobain over 9 years ago in reply to Yasuhiko Koumoto

Hello,
Thank you for all your assistance.
BR,
Iurii
Cancel
Up 0 Down

Cancel