Fixed some compiler warnings, generalized transform interface (in prep for double etc)

6 files changed, 69 insertions, 46 deletions

diff --git a/src/codegen.c b/src/codegen.c
index 5e761e3..67e9683 100644
--- a/src/codegen.c
+++ b/src/codegen.c
@@ -144,10 +144,13 @@ void insert_nops(uint8_t **p, uint32_t count) {
 	}
 }
 
+
 void align_mem16(uint8_t **p, uint32_t offset) {
+#ifdef __ARM_NEON__
 	int r = (16 - (offset & 0xf)) - ((uint32_t)(*p) & 0xf);
 	r = (16 + r) & 0xf;
 	insert_nops(p, r);	
+#endif
 }
 
 void ffts_generate_func_code(ffts_plan_t *p, size_t N, size_t leafN, int sign) {
diff --git a/src/ffts.h b/src/ffts.h
index 7cdb5e5..edf511f 100644
--- a/src/ffts.h
+++ b/src/ffts.h
@@ -58,6 +58,8 @@ typedef size_t transform_index_t;
 //typedef void (*transform_func_t)(float *data, size_t N, float *LUT);
 typedef void (*transform_func_t)(float *data, size_t N, float *LUT);
 
+typedef struct _ffts_plan_t ffts_plan_t;
+
 struct _ffts_plan_t {
 	ptrdiff_t *offsets;
 	void __attribute__ ((aligned(32))) *ws;
@@ -69,7 +71,7 @@ struct _ffts_plan_t {
 	void *lastlut;
 	transform_index_t *transforms; 
 	//transform_func_t transform;
-	void (*transform)(struct _ffts_plan_t * , const data_t * , data_t * );
+	void (*transform)(ffts_plan_t * , const void * , void * );
 	void *transform_base;
 	size_t transform_size;
 	void *constants;
@@ -82,12 +84,11 @@ struct _ffts_plan_t {
 
 	void *transpose_buf;
 
-	void (*destroy)(struct _ffts_plan_t *);
+	void (*destroy)(ffts_plan_t *);
 
 	float *A, *B;
 };
 
-typedef struct _ffts_plan_t ffts_plan_t;
 
 void ffts_free(ffts_plan_t *);
 ffts_plan_t *ffts_init_1d(size_t N, int sign); 
diff --git a/src/ffts_nd.c b/src/ffts_nd.c
index 84dba7f..b87822e 100644
--- a/src/ffts_nd.c
+++ b/src/ffts_nd.c
@@ -146,7 +146,7 @@ void ffts_transpose(uint64_t *in, uint64_t *out, int w, int h, uint64_t *buf) {
 
 }
 
-void ffts_execute_nd(ffts_plan_t *p, const data_t *  in, data_t *  out) {
+void ffts_execute_nd(ffts_plan_t *p, const void *  in, void *  out) {
 
 	uint64_t *din = (uint64_t *)in;
 	uint64_t *buf = p->buf;
diff --git a/src/ffts_nd.h b/src/ffts_nd.h
index 4d2474d..7761f52 100644
--- a/src/ffts_nd.h
+++ b/src/ffts_nd.h
@@ -46,6 +46,12 @@
 	#include <xmmintrin.h>
 #endif
 
+void ffts_free_nd(ffts_plan_t *p);
+void ffts_transpose(uint64_t *in, uint64_t *out, int w, int h, uint64_t *buf);
+
+void ffts_execute_nd(ffts_plan_t *p, const void *  in, void *  out); 
+ffts_plan_t *ffts_init_nd(int rank, size_t *Ns, int sign); 
+ffts_plan_t *ffts_init_2d(size_t N1, size_t N2, int sign); 
 
 #endif
 
diff --git a/src/ffts_real.h b/src/ffts_real.h
index 06e15bb..e904b95 100644
--- a/src/ffts_real.h
+++ b/src/ffts_real.h
@@ -46,6 +46,7 @@
 	#include <xmmintrin.h>
 #endif
 
+ffts_plan_t *ffts_init_1d_real(size_t N, int sign);
 
 #endif
 
diff --git a/src/macros.h b/src/macros.h
index e645b11..effc87d 100644
--- a/src/macros.h
+++ b/src/macros.h
@@ -221,99 +221,111 @@ L_4_2(int inv, const data_t * restrict i0, const data_t * restrict i1, const dat
 }
 
 __INLINE void 
-firstpass_16_f(ffts_plan_t *  p, const data_t *  in, data_t *  out) {
+firstpass_16_f(ffts_plan_t *  p, const void *  in, void *  out) {
+  const data_t *din = (const data_t *)in;
+	data_t *dout = (data_t *)out;
   V r0_1,r2_3,r4_5,r6_7,r8_9,r10_11,r12_13,r14_15;
   float *LUT8 = p->ws ;
 
-  L_4_4(0, in+0,in+16,in+8,in+24,&r0_1,&r2_3,&r8_9,&r10_11);
-  L_2_4(0, in+4,in+20,in+28,in+12,&r4_5,&r6_7,&r14_15,&r12_13);
+  L_4_4(0, din+0,din+16,din+8,din+24,&r0_1,&r2_3,&r8_9,&r10_11);
+  L_2_4(0, din+4,din+20,din+28,din+12,&r4_5,&r6_7,&r14_15,&r12_13);
   K_N(0, VLD(LUT8),VLD(LUT8+4),&r0_1,&r2_3,&r4_5,&r6_7);
   K_N(0, VLD(LUT8+8),VLD(LUT8+12),&r0_1,&r4_5,&r8_9,&r12_13);
-  S_4(r0_1,r4_5,r8_9,r12_13,out+0,out+8,out+16,out+24);
+  S_4(r0_1,r4_5,r8_9,r12_13,dout+0,dout+8,dout+16,dout+24);
   K_N(0, VLD(LUT8+16),VLD(LUT8+20),&r2_3,&r6_7,&r10_11,&r14_15);
-  S_4(r2_3,r6_7,r10_11,r14_15,out+4,out+12,out+20,out+28);
+  S_4(r2_3,r6_7,r10_11,r14_15,dout+4,dout+12,dout+20,dout+28);
 }
 
 __INLINE void 
-firstpass_16_b(ffts_plan_t *  p, const data_t *  in, data_t *  out) {
+firstpass_16_b(ffts_plan_t *  p, const void *  in, void *  out) {
+  const data_t *din = (const data_t *)in;
+	data_t *dout = (data_t *)out;
   V r0_1,r2_3,r4_5,r6_7,r8_9,r10_11,r12_13,r14_15;
   float *LUT8 = p->ws ;
 
-  L_4_4(1, in+0,in+16,in+8,in+24,&r0_1,&r2_3,&r8_9,&r10_11);
-  L_2_4(1, in+4,in+20,in+28,in+12,&r4_5,&r6_7,&r14_15,&r12_13);
+  L_4_4(1, din+0,din+16,din+8,din+24,&r0_1,&r2_3,&r8_9,&r10_11);
+  L_2_4(1, din+4,din+20,din+28,din+12,&r4_5,&r6_7,&r14_15,&r12_13);
   K_N(1, VLD(LUT8),VLD(LUT8+4),&r0_1,&r2_3,&r4_5,&r6_7);
   K_N(1, VLD(LUT8+8),VLD(LUT8+12),&r0_1,&r4_5,&r8_9,&r12_13);
-  S_4(r0_1,r4_5,r8_9,r12_13,out+0,out+8,out+16,out+24);
+  S_4(r0_1,r4_5,r8_9,r12_13,dout+0,dout+8,dout+16,dout+24);
   K_N(1, VLD(LUT8+16),VLD(LUT8+20),&r2_3,&r6_7,&r10_11,&r14_15);
-  S_4(r2_3,r6_7,r10_11,r14_15,out+4,out+12,out+20,out+28);
+  S_4(r2_3,r6_7,r10_11,r14_15,dout+4,dout+12,dout+20,dout+28);
 }
 
 __INLINE void 
-firstpass_8_f(ffts_plan_t *  p, const data_t *  in, data_t *  out) {
+firstpass_8_f(ffts_plan_t *  p, const void *  in, void *  out) {
+  const data_t *din = (const data_t *)in;
+	data_t *dout = (data_t *)out;
   V r0_1,r2_3,r4_5,r6_7;
   float *LUT8 = p->ws + p->ws_is[0];
-	L_4_2(0, in+0,in+8,in+4,in+12,&r0_1,&r2_3,&r4_5,&r6_7);
+	L_4_2(0, din+0,din+8,din+4,din+12,&r0_1,&r2_3,&r4_5,&r6_7);
   K_N(0, VLD(LUT8),VLD(LUT8+4),&r0_1,&r2_3,&r4_5,&r6_7);
-  S_4(r0_1,r2_3,r4_5,r6_7,out+0,out+4,out+8,out+12);
+  S_4(r0_1,r2_3,r4_5,r6_7,dout+0,dout+4,dout+8,dout+12);
 }
 
 __INLINE void 
-firstpass_8_b(ffts_plan_t *  p, const data_t *  in, data_t *  out) {
+firstpass_8_b(ffts_plan_t *  p, const void *  in, void *  out) {
+  const data_t *din = (const data_t *)in;
+	data_t *dout = (data_t *)out;
   V r0_1,r2_3,r4_5,r6_7;
   float *LUT8 = p->ws + p->ws_is[0];
-	L_4_2(1, in+0,in+8,in+4,in+12,&r0_1,&r2_3,&r4_5,&r6_7);
+	L_4_2(1, din+0,din+8,din+4,din+12,&r0_1,&r2_3,&r4_5,&r6_7);
   K_N(1, VLD(LUT8),VLD(LUT8+4),&r0_1,&r2_3,&r4_5,&r6_7);
-  S_4(r0_1,r2_3,r4_5,r6_7,out+0,out+4,out+8,out+12);
+  S_4(r0_1,r2_3,r4_5,r6_7,dout+0,dout+4,dout+8,dout+12);
 }
 
 __INLINE void 
-firstpass_4_f(ffts_plan_t *  p, const data_t *  in, data_t *  out) {
-  cdata_t *i = (cdata_t *)in, *o = (cdata_t *)out;
+firstpass_4_f(ffts_plan_t *  p, const void *  in, void *  out) {
+  const data_t *din = (const data_t *)in;
+	data_t *dout = (data_t *)out;
 	cdata_t t0, t1, t2, t3, t4, t5, t6, t7;
-	t0[0] = in[0]; t0[1] = in[1];
-	t1[0] = in[4]; t1[1] = in[5];
-	t2[0] = in[2]; t2[1] = in[3];
-	t3[0] = in[6]; t3[1] = in[7];
+	t0[0] = din[0]; t0[1] = din[1];
+	t1[0] = din[4]; t1[1] = din[5];
+	t2[0] = din[2]; t2[1] = din[3];
+	t3[0] = din[6]; t3[1] = din[7];
 	
 	t4[0] = t0[0] + t1[0]; t4[1] = t0[1] + t1[1];
 	t5[0] = t0[0] - t1[0]; t5[1] = t0[1] - t1[1];
 	t6[0] = t2[0] + t3[0]; t6[1] = t2[1] + t3[1];
 	t7[0] = t2[0] - t3[0]; t7[1] = t2[1] - t3[1];
 
-	out[0] = t4[0] + t6[0]; out[1] = t4[1] + t6[1];
-	out[4] = t4[0] - t6[0]; out[5] = t4[1] - t6[1];
-	out[2] = t5[0] + t7[1]; out[3] = t5[1] - t7[0];
-	out[6] = t5[0] - t7[1]; out[7] = t5[1] + t7[0];
+	dout[0] = t4[0] + t6[0]; dout[1] = t4[1] + t6[1];
+	dout[4] = t4[0] - t6[0]; dout[5] = t4[1] - t6[1];
+	dout[2] = t5[0] + t7[1]; dout[3] = t5[1] - t7[0];
+	dout[6] = t5[0] - t7[1]; dout[7] = t5[1] + t7[0];
 }
 __INLINE void 
-firstpass_4_b(ffts_plan_t *  p, const data_t *  in, data_t *  out) {
-  cdata_t *i = (cdata_t *)in, *o = (cdata_t *)out;
+firstpass_4_b(ffts_plan_t *  p, const void *  in, void *  out) {
+  const data_t *din = (const data_t *)in;
+	data_t *dout = (data_t *)out;
 	cdata_t t0, t1, t2, t3, t4, t5, t6, t7;
-	t0[0] = in[0]; t0[1] = in[1];
-	t1[0] = in[4]; t1[1] = in[5];
-	t2[0] = in[2]; t2[1] = in[3];
-	t3[0] = in[6]; t3[1] = in[7];
+	t0[0] = din[0]; t0[1] = din[1];
+	t1[0] = din[4]; t1[1] = din[5];
+	t2[0] = din[2]; t2[1] = din[3];
+	t3[0] = din[6]; t3[1] = din[7];
 	
 	t4[0] = t0[0] + t1[0]; t4[1] = t0[1] + t1[1];
 	t5[0] = t0[0] - t1[0]; t5[1] = t0[1] - t1[1];
 	t6[0] = t2[0] + t3[0]; t6[1] = t2[1] + t3[1];
 	t7[0] = t2[0] - t3[0]; t7[1] = t2[1] - t3[1];
 
-	out[0] = t4[0] + t6[0]; out[1] = t4[1] + t6[1];
-	out[4] = t4[0] - t6[0]; out[5] = t4[1] - t6[1];
-	out[2] = t5[0] - t7[1]; out[3] = t5[1] + t7[0];
-	out[6] = t5[0] + t7[1]; out[7] = t5[1] - t7[0];
+	dout[0] = t4[0] + t6[0]; dout[1] = t4[1] + t6[1];
+	dout[4] = t4[0] - t6[0]; dout[5] = t4[1] - t6[1];
+	dout[2] = t5[0] - t7[1]; dout[3] = t5[1] + t7[0];
+	dout[6] = t5[0] + t7[1]; dout[7] = t5[1] - t7[0];
 }
 __INLINE void 
-firstpass_2(ffts_plan_t *  p, const data_t *  in, data_t *  out) {
-  cdata_t t0, t1, r0,r1;
-	t0[0] = in[0]; t0[1] = in[1];
-	t1[0] = in[2]; t1[1] = in[3];
+firstpass_2(ffts_plan_t *  p, const void *  in, void *  out) {
+  const data_t *din = (const data_t *)in;
+	data_t *dout = (data_t *)out;
+	cdata_t t0, t1, r0,r1;
+	t0[0] = din[0]; t0[1] = din[1];
+	t1[0] = din[2]; t1[1] = din[3];
 	r0[0] = t0[0] + t1[0];
 	r0[1] = t0[1] + t1[1];
 	r1[0] = t0[0] - t1[0];
 	r1[1] = t0[1] - t1[1];
-	out[0] = r0[0]; out[1] = r0[1];
-	out[2] = r1[0]; out[3] = r1[1];
+	dout[0] = r0[0]; dout[1] = r0[1];
+	dout[2] = r1[0]; dout[3] = r1[1];
 }
 #endif