Fix fft usage above 65536 floats: the in and out parameters need to be different, otherwize fft starts optimizing by always putting the result in the in parameter...

examples/starpufft/starpufftx1d.c

@@ -557,19 +557,19 @@ if (PARALLEL) {
 			/* first fft plan: one fft of size n2.
 			 * FFTW imposes that buffer pointers are known at
 			 * planning time. */
-			plan->plans[workerid].plan1_cpu = _FFTW(plan_dft_1d)(n2, NULL, NULL, sign, _FFTW_FLAGS);
+			plan->plans[workerid].plan1_cpu = _FFTW(plan_dft_1d)(n2, NULL, (void*) 1, sign, _FFTW_FLAGS);
 			STARPU_ASSERT(plan->plans[workerid].plan1_cpu);
 
 			/* second fft plan: n3 ffts of size n1 */
 			plan->plans[workerid].plan2_cpu = _FFTW(plan_many_dft)(plan->dim,
 					plan->n1, n3,
 					NULL, NULL, 1, plan->totsize1,
-					NULL, NULL, 1, plan->totsize1,
+					(void*) 1, NULL, 1, plan->totsize1,
 					sign, _FFTW_FLAGS);
 			STARPU_ASSERT(plan->plans[workerid].plan2_cpu);
 } else {
 			/* fft plan: one fft of size n. */
-			plan->plans[workerid].plan_cpu = _FFTW(plan_dft_1d)(n, NULL, NULL, sign, _FFTW_FLAGS);
+			plan->plans[workerid].plan_cpu = _FFTW(plan_dft_1d)(n, NULL, (void*) 1, sign, _FFTW_FLAGS);
 			STARPU_ASSERT(plan->plans[workerid].plan_cpu);
 }
 #else

examples/starpufft/starpufftx2d.c

@@ -559,19 +559,19 @@ if (PARALLEL) {
 #ifdef STARPU_HAVE_FFTW
 if (PARALLEL) {
 			/* first fft plan: one n2*m2 fft */
-			plan->plans[workerid].plan1_cpu = _FFTW(plan_dft_2d)(n2, m2, NULL, NULL, sign, _FFTW_FLAGS);
+			plan->plans[workerid].plan1_cpu = _FFTW(plan_dft_2d)(n2, m2, NULL, (void*) 1, sign, _FFTW_FLAGS);
 			STARPU_ASSERT(plan->plans[workerid].plan1_cpu);
 
 			/* second fft plan: n3*m3 n1*m1 ffts */
 			plan->plans[workerid].plan2_cpu = _FFTW(plan_many_dft)(plan->dim,
 					plan->n1, n3*m3,
 					NULL, NULL, 1, plan->totsize1,
-					NULL, NULL, 1, plan->totsize1,
+					(void*) 1, NULL, 1, plan->totsize1,
 					sign, _FFTW_FLAGS);
 			STARPU_ASSERT(plan->plans[workerid].plan2_cpu);
 } else {
 			/* fft plan: one fft of size n, m. */
-			plan->plans[workerid].plan_cpu = _FFTW(plan_dft_2d)(n, m, NULL, NULL, sign, _FFTW_FLAGS);
+			plan->plans[workerid].plan_cpu = _FFTW(plan_dft_2d)(n, m, NULL, (void*) 1, sign, _FFTW_FLAGS);
 			STARPU_ASSERT(plan->plans[workerid].plan_cpu);
 }
 #else

examples/starpufft/testx.c

@@ -158,7 +158,7 @@ int main(int argc, char *argv[]) {
 	if (argc == 2) {
 		plan = STARPUFFT(plan_dft_1d)(n, SIGN, 0);
 #ifdef STARPU_HAVE_FFTW
-		fftw_plan = _FFTW(plan_dft_1d)(n, NULL, NULL, SIGN, FFTW_ESTIMATE);
+		fftw_plan = _FFTW(plan_dft_1d)(n, NULL, (void*) 1, SIGN, FFTW_ESTIMATE);
 #endif
 #ifdef STARPU_USE_CUDA
 		if (cufftPlan1d(&cuda_plan, n, _CUFFT_C2C, 1) != CUFFT_SUCCESS)
@@ -168,7 +168,7 @@ int main(int argc, char *argv[]) {
 	} else if (argc == 3) {
 		plan = STARPUFFT(plan_dft_2d)(n, m, SIGN, 0);
 #ifdef STARPU_HAVE_FFTW
-		fftw_plan = _FFTW(plan_dft_2d)(n, m, NULL, NULL, SIGN, FFTW_ESTIMATE);
+		fftw_plan = _FFTW(plan_dft_2d)(n, m, NULL, (void*) 1, SIGN, FFTW_ESTIMATE);
 #endif
 #ifdef STARPU_USE_CUDA
 		STARPU_ASSERT(cufftPlan2d(&cuda_plan, n, m, _CUFFT_C2C) == CUFFT_SUCCESS);