Feature completed cspan.

1 files changed, 39 insertions, 81 deletions

diff --git a/include/stc/cspan.h b/include/stc/cspan.h
index 78c92520..94415df7 100644
--- a/include/stc/cspan.h
+++ b/include/stc/cspan.h
@@ -74,10 +74,10 @@ int demo2() {
     STC_INLINE Self Self##_from_n(Self##_raw* raw, const intptr_t n) { \
         return (Self){.data=raw, .dim={(int32_t)n}}; \
     } \
-    STC_INLINE Self Self##_subslice_(Self##_value* v, const int32_t dim[], const int32_t stri[], \
+    STC_INLINE Self Self##_slice_(Self##_value* v, const int32_t dim[], const int32_t stri[], \
                                      const int rank, const int32_t a[][2]) { \
         Self s = {.data=v}; int outrank; \
-        s.data += _cspan_subslice(s.dim, s.stride.d, &outrank, dim, stri, rank, a); \
+        s.data += _cspan_slice(s.dim, s.stride.d, &outrank, dim, stri, rank, a); \
         c_ASSERT(outrank == RANK); \
         return s; \
     } \
@@ -90,7 +90,7 @@ int demo2() {
         return it; \
     } \
     STC_INLINE void Self##_next(Self##_iter* it) { \
-        it->ref += _cspan_next_##RANK(RANK, it->pos, it->_s->dim, it->_s->stride.d); \
+        it->ref += _cspan_next##RANK(RANK, it->pos, it->_s->dim, it->_s->stride.d); \
         if (it->pos[0] == it->_s->dim[0]) it->ref = NULL; \
     } \
     struct stc_nostruct
@@ -104,8 +104,7 @@ typedef struct { int32_t d[3]; } cspan_idx3;
 typedef struct { int32_t d[4]; } cspan_idx4;
 typedef struct { int32_t d[5]; } cspan_idx5;
 typedef struct { int32_t d[6]; } cspan_idx6;
-#define c_END -1
-#define c_ALL 0,c_END
+#define c_ALL 0,-1
 
 #define cspan_md(array, ...) \
     {.data=array, .dim={__VA_ARGS__}, .stride={.d={__VA_ARGS__}}}
@@ -114,10 +113,10 @@ typedef struct { int32_t d[6]; } cspan_idx6;
 #define cspan_make(SpanType, ...) \
     {.data=(SpanType##_value[])__VA_ARGS__, .dim={sizeof((SpanType##_value[])__VA_ARGS__)/sizeof(SpanType##_value)}}
 
-#define cspan_subslice(OutSpan, parent, ...) \
-    OutSpan##_subslice_((parent)->data, (parent)->dim, (parent)->stride.d, cspan_rank(parent) + \
-                        c_static_assert(cspan_rank(parent) == sizeof((int32_t[][2]){__VA_ARGS__})/sizeof(int32_t[2])), \
-                        (const int32_t[][2]){__VA_ARGS__})
+#define cspan_slice(OutSpan, parent, ...) \
+    OutSpan##_slice_((parent)->data, (parent)->dim, (parent)->stride.d, cspan_rank(parent) + \
+                     c_static_assert(cspan_rank(parent) == sizeof((int32_t[][2]){__VA_ARGS__})/sizeof(int32_t[2])), \
+                     (const int32_t[][2]){__VA_ARGS__})
      
 /* create a cspan from a cvec, cstack, cdeq, cqueue, or cpque (heap) */
 #define cspan_from(container) \
@@ -129,29 +128,33 @@ typedef struct { int32_t d[6]; } cspan_idx6;
 #define cspan_size(self) _cspan_size((self)->dim, cspan_rank(self))
 #define cspan_rank(self) c_ARRAYLEN((self)->dim)
 
-#define cspan_idx(self, ...) \
-    c_PASTE(_cspan_i, c_NUMARGS(__VA_ARGS__))((self)->dim, (self)->stride, __VA_ARGS__)
-
-#define cspan_index(self, ...) \
-    (_cspan_index(c_NUMARGS(__VA_ARGS__), (self)->dim, (self)->stride.d, (int32_t[]){__VA_ARGS__}) + \
+#define cspan_index(self, ...) c_PASTE(cspan_idx_, c_NUMARGS(__VA_ARGS__))(self, __VA_ARGS__)
+#define cspan_idx_1 cspan_idx_4
+#define cspan_idx_2 cspan_idx_4
+#define cspan_idx_3 cspan_idx_4
+#define cspan_idx_4(self, ...) \
+    c_PASTE(_cspan_idx, c_NUMARGS(__VA_ARGS__))((self)->dim, (self)->stride, __VA_ARGS__)
+#define cspan_idx_5(self, ...) \
+    (_cspan_idxN(c_NUMARGS(__VA_ARGS__), (self)->dim, (self)->stride.d, (int32_t[]){__VA_ARGS__}) + \
      c_static_assert(cspan_rank(self) == c_NUMARGS(__VA_ARGS__)))
+#define cspan_idx_6 cspan_idx_5
 
 #define cspan_at(self, ...) ((self)->data + cspan_index(self, __VA_ARGS__))
-#define cspan_item(self, ...) ((self)->data + cspan_idx(self, __VA_ARGS__)) // same as cspan_at(), only for rank <= 5
 #define cspan_front(self) ((self)->data)
 #define cspan_back(self) ((self)->data + cspan_size(self) - 1)
 
-// cspan_subspanN. for N > 3, use cspan_slice(&ms4, {offset, offset + count}, {0}, {0}, {0});
-
+// cspan_subspanN. (N<5) Optimized, but same as e.g. cspan_slice(Span2, &ms4, {offset, offset + count}, {c_ALL}, {c_ALL});
 #define cspan_subspan(self, offset, count) \
     {.data=cspan_at(self, offset), .dim={count}}
 #define cspan_subspan2(self, offset, count) \
     {.data=cspan_at(self, offset, 0), .dim={count, (self)->dim[1]}, .stride={(self)->stride}}
 #define cspan_subspan3(self, offset, count) \
     {.data=cspan_at(self, offset, 0, 0), .dim={count, (self)->dim[1], (self)->dim[2]}, .stride={(self)->stride}}
+#define cspan_subspan4(self, offset, count) \
+    {.data=cspan_at(self, offset, 0, 0, 0), .dim={count, (self)->dim[1], (self)->dim[2], (self)->dim[3]}, \
+                                            .stride={(self)->stride}}
 
 // cspan_submdN: return reduced rank
-
 #define cspan_submd4(...) c_MACRO_OVERLOAD(cspan_submd4, __VA_ARGS__)
 #define cspan_submd3(...) c_MACRO_OVERLOAD(cspan_submd3, __VA_ARGS__)
 #define cspan_submd2(self, x) \
@@ -169,50 +172,24 @@ typedef struct { int32_t d[6]; } cspan_idx6;
 #define cspan_submd4_4(self, x, y, z) \
     {.data=cspan_at(self, x, y, z, 0), .dim={(self)->dim[3]}}
 
-#define cspan_submd5_2(self, x) \
-    {.data=cspan_at(self, x, 0, 0, 0, 0), .dim={(self)->dim[1], (self)->dim[2], (self)->dim[3], (self)->dim[4]}, \
-                                          .stride={.d={0, (self)->stride.d[2], (self)->stride.d[3], (self)->stride.d[4]}}}
-#define cspan_submd5_3(self, x, y) \
-    {.data=cspan_at(self, x, y, 0, 0, 0), .dim={(self)->dim[2], (self)->dim[3], (self)->dim[4]}, \
-                                          .stride={.d={0, (self)->stride.d[3], (self)->stride.d[4]}}}
-#define cspan_submd5_4(self, x, y, z) \
-    {.data=cspan_at(self, x, y, z, 0, 0), .dim={(self)->dim[3], (self)->dim[4]}, .stride={.d={0, (self)->stride.d[4]}}}
-#define cspan_submd5_5(self, x, y, z, w) \
-    {.data=cspan_at(self, x, y, z, w, 0), .dim={(self)->dim[4]}}
-
-// cspan_slice:
-//  e.g.: cspan_slice(&ms3, {1,3}, {0,-1}, {1,4});
-
-#define cspan_slice(self, ...) \
-    ((void)((self)->data += _cspan_slice((self)->dim, (self)->stride.d, \
-                                         cspan_rank(self), (const int32_t[][2]){__VA_ARGS__}) + \
-                            c_static_assert(cspan_rank(self) == \
-                            sizeof((const int32_t[][2]){__VA_ARGS__})/sizeof(int32_t[2]))))
-
 // FUNCTIONS
 
-STC_INLINE intptr_t _cspan_i1(const int32_t dim[1], const cspan_idx1 stri, int32_t x)
+STC_INLINE intptr_t _cspan_idx1(const int32_t dim[1], const cspan_idx1 stri, int32_t x)
     { c_ASSERT(c_LTu(x, dim[0])); return x; }
 
-STC_INLINE intptr_t _cspan_i2(const int32_t dim[2], const cspan_idx2 stri, int32_t x, int32_t y)
+STC_INLINE intptr_t _cspan_idx2(const int32_t dim[2], const cspan_idx2 stri, int32_t x, int32_t y)
     { c_ASSERT(c_LTu(x, dim[0]) && c_LTu(y, dim[1])); return (intptr_t)stri.d[1]*x + y; }
 
-STC_INLINE intptr_t _cspan_i3(const int32_t dim[3], const cspan_idx3 stri, int32_t x, int32_t y, int32_t z) {
+STC_INLINE intptr_t _cspan_idx3(const int32_t dim[3], const cspan_idx3 stri, int32_t x, int32_t y, int32_t z) {
     c_ASSERT(c_LTu(x, dim[0]) && c_LTu(y, dim[1]) && c_LTu(z, dim[2]));
     return (intptr_t)stri.d[2]*(stri.d[1]*x + y) + z;
 }
-STC_INLINE intptr_t _cspan_i4(const int32_t dim[4], const cspan_idx4 stri, int32_t x, int32_t y,
-                                                                           int32_t z, int32_t w) {
+STC_INLINE intptr_t _cspan_idx4(const int32_t dim[4], const cspan_idx4 stri, int32_t x, int32_t y,
+                                                                             int32_t z, int32_t w) {
     c_ASSERT(c_LTu(x, dim[0]) && c_LTu(y, dim[1]) && c_LTu(z, dim[2]) && c_LTu(w, dim[3]));
     return (intptr_t)stri.d[3]*(stri.d[2]*(stri.d[1]*x + y) + z) + w;
 }
-STC_INLINE intptr_t _cspan_i5(const int32_t dim[4], const cspan_idx4 stri, int32_t x, int32_t y, int32_t z,
-                                                                           int32_t w, int32_t v) {
-    c_ASSERT(c_LTu(x, dim[0]) && c_LTu(y, dim[1]) && c_LTu(z, dim[2]) && c_LTu(w, dim[3]) && c_LTu(v, dim[4]));
-    return (intptr_t)stri.d[4]*(stri.d[3]*(stri.d[2]*(stri.d[1]*x + y) + z) + w) + v;
-}
-
-STC_INLINE intptr_t _cspan_index(int rank, const int32_t dim[], const int32_t stri[], const int32_t a[]) {
+STC_INLINE intptr_t _cspan_idxN(int rank, const int32_t dim[], const int32_t stri[], const int32_t a[]) {
     intptr_t off = a[0];
     bool ok = c_LTu(a[0], dim[0]);
     for (int i = 1; i < rank; ++i) {
@@ -230,12 +207,13 @@ STC_INLINE intptr_t _cspan_size(const int32_t dim[], int rank) {
     return sz;
 }
 
-#define _cspan_next_1(r, pos, d, s) (++pos[0], 1)
-#define _cspan_next_3 _cspan_next_2
-#define _cspan_next_4 _cspan_next_2
-#define _cspan_next_5 _cspan_next_2
+#define _cspan_next1(r, pos, d, s) (++pos[0], 1)
+#define _cspan_next3 _cspan_next2
+#define _cspan_next4 _cspan_next2
+#define _cspan_next5 _cspan_next2
+#define _cspan_next6 _cspan_next2
 
-STC_INLINE intptr_t _cspan_next_2(int rank, int32_t pos[], const int32_t dim[], const int32_t stride[]) {
+STC_INLINE intptr_t _cspan_next2(int rank, int32_t pos[], const int32_t dim[], const int32_t stride[]) {
     intptr_t off = 1, rs = 1;
     ++pos[rank - 1];
     while (--rank && pos[rank] == dim[rank]) {
@@ -247,30 +225,10 @@ STC_INLINE intptr_t _cspan_next_2(int rank, int32_t pos[], const int32_t dim[],
     return off;
 }
 
-STC_INLINE intptr_t _cspan_slice(int32_t dim[], const int32_t stri[], const int rank, const int32_t a[][2]) {
-    intptr_t off = 0;
-    bool ok = true;
-    for (int i = 0; i < rank; ++i) {
-        off *= stri[i];
-        off += a[i][0];
-        int32_t t;
-        switch (a[i][1]) { 
-            case 0: t = a[i][0] + 1; break;
-            case -1: t = dim[i]; break;
-            default: t = a[i][1];
-        }
-        dim[i] = t - a[i][0];
-        ok &= c_LTu(0, dim[i]);
-    }
-    c_ASSERT(ok);
-    return off;
-}
-
-STC_INLINE intptr_t _cspan_subslice(int32_t odim[], int32_t ostri[], int* orank, 
-                                    const int32_t dim[], const int32_t stri[], 
-                                    int rank, const int32_t a[][2]) {
-    intptr_t off = a[0][0];
-    intptr_t s = 1;
+STC_INLINE intptr_t _cspan_slice(int32_t odim[], int32_t ostri[], int* orank, 
+                                 const int32_t dim[], const int32_t stri[], 
+                                 int rank, const int32_t a[][2]) {
+    intptr_t off = 0, s = 1;
     int i = 0, j = 0, ok = true;
     for (; i < rank; ++i) {
         off *= stri[i];
@@ -283,11 +241,11 @@ STC_INLINE intptr_t _cspan_subslice(int32_t odim[], int32_t ostri[], int* orank,
         }
         odim[j] = t - a[i][0];
         ostri[j] = s*stri[i];
+        ok &= c_LTu(0, odim[j]) & !c_LTu(dim[i], t);
         s = 1; ++j;
-        ok &= c_LTu(0, odim[i]) & !c_LTu(dim[i], t);
     }
     *orank = j;
-    c_ASSERT(ok); /* input indices ok */
+    c_ASSERT(ok);
     return off;
 }
 #endif