clFFT 8000-point FFT kernel compilation failure on Raspberry Pi compute module 5(RPi5 CM5) using V3D/CLVK #253
Description
When running an 8000-point FFT on a Raspberry Pi CM5 (V3D GPU using CLVK OpenCL runtime), the clFFT library fails during kernel compilation.
The host setup and clFFT API calls work correctly, but during the clfftBakePlan step or first transform execution, the following error occurs:
Running clFFT on device: V3D 7.1.10.2
error: 601: Expected input to have different bit width from Result Type: UConvert
%309 = OpUConvert %uint %308
Following warnings are generated when running the code.
BUILD LOG
************************************************
clvk-uqMihz/source.cl:18:19: warning: no previous extern declaration for non-static variable 'twiddles'
18 | __constant float2 twiddles[124] = {
| ^
clvk-uqMihz/source.cl:18:12: note: declare 'static' if the variable is not intended to be used outside of this translation unit
18 | __constant float2 twiddles[124] = {
| ^
clvk-uqMihz/source.cl:157:1: warning: no previous prototype for function 'FwdRad5B1'
157 | FwdRad5B1(float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:156:32: note: declare 'static' if the function is not intended to be used outside of this translation unit
156 | __attribute__((always_inline)) void
| ^
| static
clvk-uqMihz/source.cl:183:1: warning: no previous prototype for function 'InvRad5B1'
183 | InvRad5B1(float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:182:32: note: declare 'static' if the function is not intended to be used outside of this translation unit
182 | __attribute__((always_inline)) void
| ^
| static
clvk-uqMihz/source.cl:209:24: warning: unused parameter 'b'
209 | FwdPass0(uint rw, uint b, uint me, uint inOffset, uint outOffset, __global float2 *bufIn, __local float *bufOutRe, __local float *bufOutIm, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:209:1: warning: no previous prototype for function 'FwdPass0'
209 | FwdPass0(uint rw, uint b, uint me, uint inOffset, uint outOffset, __global float2 *bufIn, __local float *bufOutRe, __local float *bufOutIm, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:208:32: note: declare 'static' if the function is not intended to be used outside of this translation unit
208 | __attribute__((always_inline)) void
| ^
| static
clvk-uqMihz/source.cl:287:24: warning: unused parameter 'b'
287 | FwdPass1(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __local float *bufOutRe, __local float *bufOutIm, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:287:41: warning: unused parameter 'inOffset'
287 | FwdPass1(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __local float *bufOutRe, __local float *bufOutIm, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:287:82: warning: unused parameter 'bufInRe'
287 | FwdPass1(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __local float *bufOutRe, __local float *bufOutIm, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:287:106: warning: unused parameter 'bufInIm'
287 | FwdPass1(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __local float *bufOutRe, __local float *bufOutIm, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:287:1: warning: no previous prototype for function 'FwdPass1'
287 | FwdPass1(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __local float *bufOutRe, __local float *bufOutIm, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:286:32: note: declare 'static' if the function is not intended to be used outside of this translation unit
286 | __attribute__((always_inline)) void
| ^
| static
clvk-uqMihz/source.cl:383:24: warning: unused parameter 'b'
383 | FwdPass2(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __global float2 *bufOut, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:383:41: warning: unused parameter 'inOffset'
383 | FwdPass2(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __global float2 *bufOut, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:383:82: warning: unused parameter 'bufInRe'
383 | FwdPass2(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __global float2 *bufOut, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:383:106: warning: unused parameter 'bufInIm'
383 | FwdPass2(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __global float2 *bufOut, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:383:1: warning: no previous prototype for function 'FwdPass2'
383 | FwdPass2(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __global float2 *bufOut, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:382:32: note: declare 'static' if the function is not intended to be used outside of this translation unit
382 | __attribute__((always_inline)) void
| ^
| static
clvk-uqMihz/source.cl:440:24: warning: unused parameter 'b'
440 | InvPass0(uint rw, uint b, uint me, uint inOffset, uint outOffset, __global float2 *bufIn, __local float *bufOutRe, __local float *bufOutIm, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:440:1: warning: no previous prototype for function 'InvPass0'
440 | InvPass0(uint rw, uint b, uint me, uint inOffset, uint outOffset, __global float2 *bufIn, __local float *bufOutRe, __local float *bufOutIm, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:439:32: note: declare 'static' if the function is not intended to be used outside of this translation unit
439 | __attribute__((always_inline)) void
| ^
| static
clvk-uqMihz/source.cl:518:24: warning: unused parameter 'b'
518 | InvPass1(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __local float *bufOutRe, __local float *bufOutIm, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:518:41: warning: unused parameter 'inOffset'
518 | InvPass1(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __local float *bufOutRe, __local float *bufOutIm, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:518:82: warning: unused parameter 'bufInRe'
518 | InvPass1(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __local float *bufOutRe, __local float *bufOutIm, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:518:106: warning: unused parameter 'bufInIm'
518 | InvPass1(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __local float *bufOutRe, __local float *bufOutIm, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:518:1: warning: no previous prototype for function 'InvPass1'
518 | InvPass1(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __local float *bufOutRe, __local float *bufOutIm, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:517:32: note: declare 'static' if the function is not intended to be used outside of this translation unit
517 | __attribute__((always_inline)) void
| ^
| static
clvk-uqMihz/source.cl:614:24: warning: unused parameter 'b'
614 | InvPass2(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __global float2 *bufOut, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:614:41: warning: unused parameter 'inOffset'
614 | InvPass2(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __global float2 *bufOut, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:614:82: warning: unused parameter 'bufInRe'
614 | InvPass2(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __global float2 *bufOut, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:614:106: warning: unused parameter 'bufInIm'
614 | InvPass2(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __global float2 *bufOut, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:614:1: warning: no previous prototype for function 'InvPass2'
614 | InvPass2(uint rw, uint b, uint me, uint inOffset, uint outOffset, __local float *bufInRe, __local float *bufInIm, __global float2 *bufOut, float2 *R0, float2 *R1, float2 *R2, float2 *R3, float2 *R4)
| ^
clvk-uqMihz/source.cl:613:32: note: declare 'static' if the function is not intended to be used outside of this translation unit
613 | __attribute__((always_inline)) void
| ^
| static
clvk-uqMihz/source.cl:673:49: warning: unknown attribute 'max_constant_size' ignored
673 | void fft_fwd(__constant cb_t *cb __attribute__((max_constant_size(32))), __global const float2 * restrict gbIn, __global float2 * restrict gbOut)
| ^~~~~~~~~~~~~~~~~~~~~
clvk-uqMihz/source.cl:693:8: warning: assigning to '__global float2 *__private' from 'const __global float2 *' discards qualifiers
693 | lwbIn = gbIn + iOffset;
| ^ ~~~~~~~~~~~~~~
clvk-uqMihz/source.cl:702:50: warning: unknown attribute 'max_constant_size' ignored
702 | void fft_back(__constant cb_t *cb __attribute__((max_constant_size(32))), __global const float2 * restrict gbIn, __global float2 * restrict gbOut)
| ^~~~~~~~~~~~~~~~~~~~~
clvk-uqMihz/source.cl:722:8: warning: assigning to '__global float2 *__private' from 'const __global float2 *' discards qualifiers
722 | lwbIn = gbIn + iOffset;
| ^ ~~~~~~~~~~~~~~
clvk-uqMihz/source.cl:146:9: warning: macro is not used
146 | #define fptype float
| ^
clvk-uqMihz/source.cl:154:9: warning: macro is not used
154 | #define C5QE 0.80901699437494742410229341718282f
| ^
************************************************
FFTGeneratedStockhamAction::compileKernels failed
Device Info
- Device: Raspberry Pi CM5 (VideoCore VII / V3D GPU)
- Driver: V3D 7.1.10.2
- OpenCL Runtime: CLVK
- clFFT from the master branch
When I set the problem size to the next nearest power of 2, (8192) clFFT generates correct output.
Additionally, when I run my code on NVIDIA RTX 3090, it generates expected outputs for both 8000 and 8192 points.
I have attached my code for your reference.
fft8k.zip
I understand this issue could be out of scope for this repository if the error is due to the limitations of CLVK and/or RPI CM5. I would appreciate if you can let me know a workaround for this issue to get FFT 8000 running with clFFT.