diff options
Diffstat (limited to 'include')
| -rw-r--r-- | include/stc/cspan.h | 46 |
1 files changed, 23 insertions, 23 deletions
diff --git a/include/stc/cspan.h b/include/stc/cspan.h index 86aefdc0..e058c0e8 100644 --- a/include/stc/cspan.h +++ b/include/stc/cspan.h @@ -81,7 +81,7 @@ int demo2() { \ typedef struct { Self##_value *ref; int32_t pos[RANK]; const Self *_s; } Self##_iter; \ \ - STC_INLINE Self Self##_slice_(Self##_value* d, const int32_t shape[], const int32_t stri[], \ + STC_INLINE Self Self##_slice_(Self##_value* d, const int32_t shape[], const intptr_t stri[], \ const int rank, const int32_t a[][2]) { \ Self s; int outrank; \ s.data = d + _cspan_slice(s.shape, s.stride.d, &outrank, shape, stri, rank, a); \ @@ -106,7 +106,7 @@ int demo2() { #define using_cspan2(Self, T) using_cspan_2(Self, T); using_cspan_3(Self##2, T, 2) #define using_cspan3(Self, T) using_cspan2(Self, T); using_cspan_3(Self##3, T, 3) #define using_cspan4(Self, T) using_cspan3(Self, T); using_cspan_3(Self##4, T, 4) -#define using_cspan_tuple(N) typedef struct { int32_t d[N]; } cspan_tuple##N +#define using_cspan_tuple(N) typedef struct { intptr_t d[N]; } cspan_tuple##N using_cspan_tuple(1); using_cspan_tuple(2); using_cspan_tuple(3); using_cspan_tuple(4); using_cspan_tuple(5); using_cspan_tuple(6); @@ -139,7 +139,7 @@ typedef enum {c_ROWMAJOR, c_COLMAJOR} cspan_layout; #define cspan_front(self) ((self)->data) #define cspan_back(self) ((self)->data + cspan_size(self) - 1) #define cspan_index(self, ...) \ - (_cspan_index(c_NUMARGS(__VA_ARGS__), (self)->shape, (self)->stride.d, (int32_t[]){__VA_ARGS__}) + \ + (_cspan_index(c_NUMARGS(__VA_ARGS__), (self)->shape, (self)->stride.d, (const int32_t[]){__VA_ARGS__}) + \ c_static_assert(cspan_rank(self) == c_NUMARGS(__VA_ARGS__))) // general // cspan_subspanX: (X <= 3) optimized. Similar to cspan_slice(Span3, &ms3, {off,off+count}, {c_ALL}, {c_ALL}); @@ -172,7 +172,7 @@ typedef enum {c_ROWMAJOR, c_COLMAJOR} cspan_layout; #define cspan_md(array, ...) cspan_md_layout(c_ROWMAJOR, array, __VA_ARGS__) #define cspan_md_layout(layout, array, ...) \ {.data=array, .shape={__VA_ARGS__}, \ - .stride=*(c_PASTE(cspan_tuple, c_NUMARGS(__VA_ARGS__))*)_cspan_shape2stride(layout, ((int32_t[]){__VA_ARGS__}), c_NUMARGS(__VA_ARGS__))} + .stride=*(c_PASTE(cspan_tuple,c_NUMARGS(__VA_ARGS__))*)_cspan_shape2stride(layout, ((intptr_t[]){__VA_ARGS__}), c_NUMARGS(__VA_ARGS__))} #define cspan_transpose(self) \ _cspan_transpose((self)->shape, (self)->stride.d, cspan_rank(self)) @@ -191,25 +191,25 @@ STC_INLINE intptr_t _cspan_size(const int32_t shape[], int rank) { return sz; } -STC_INLINE void _cspan_transpose(int32_t shape[], int32_t stride[], int rank) { +STC_INLINE void _cspan_transpose(int32_t shape[], intptr_t stride[], int rank) { for (int i = 0; i < --rank; ++i) { c_swap(int32_t, shape + i, shape + rank); - c_swap(int32_t, stride + i, stride + rank); + c_swap(intptr_t, stride + i, stride + rank); } } -STC_INLINE intptr_t _cspan_index(int rank, const int32_t shape[], const int32_t stride[], const int32_t a[]) { +STC_INLINE intptr_t _cspan_index(int rank, const int32_t shape[], const intptr_t stride[], const int32_t a[]) { intptr_t off = 0; while (rank--) { c_assert(c_LTu(a[rank], shape[rank])); - off += (intptr_t)stride[rank]*a[rank]; + off += stride[rank]*a[rank]; } return off; } #define _cspan_next1(pos, shape, stride, rank, done) (*done = ++pos[0]==shape[0], stride[0]) STC_API intptr_t - _cspan_next2(int32_t pos[], const int32_t shape[], const int32_t stride[], int rank, int* done); + _cspan_next2(int32_t pos[], const int32_t shape[], const intptr_t stride[], int rank, int* done); #define _cspan_next3 _cspan_next2 #define _cspan_next4 _cspan_next2 #define _cspan_next5 _cspan_next2 @@ -217,53 +217,53 @@ STC_API intptr_t #define _cspan_next7 _cspan_next2 #define _cspan_next8 _cspan_next2 -STC_API intptr_t _cspan_slice(int32_t oshape[], int32_t ostride[], int* orank, - const int32_t shape[], const int32_t stride[], +STC_API intptr_t _cspan_slice(int32_t oshape[], intptr_t ostride[], int* orank, + const int32_t shape[], const intptr_t stride[], int rank, const int32_t a[][2]); -STC_API int32_t* _cspan_shape2stride(cspan_layout layout, int32_t shape[], int rank); +STC_API intptr_t* _cspan_shape2stride(cspan_layout layout, intptr_t shape[], int rank); #endif // STC_CSPAN_H_INCLUDED /* --------------------- IMPLEMENTATION --------------------- */ #if defined(i_implement) || defined(i_static) -STC_DEF intptr_t _cspan_next2(int32_t pos[], const int32_t shape[], const int32_t stride[], int r, int* done) { +STC_DEF intptr_t _cspan_next2(int32_t pos[], const int32_t shape[], const intptr_t stride[], int r, int* done) { intptr_t off = stride[--r]; ++pos[r]; for (; r && pos[r] == shape[r]; --r) { pos[r] = 0; ++pos[r - 1]; - off += stride[r - 1] - (intptr_t)stride[r]*shape[r]; + off += stride[r - 1] - stride[r]*shape[r]; } *done = pos[r] == shape[r]; return off; } -STC_DEF int32_t* _cspan_shape2stride(cspan_layout layout, int32_t shape[], int rank) { +STC_DEF intptr_t* _cspan_shape2stride(cspan_layout layout, intptr_t stride[], int rank) { int i, inc; if (layout == c_COLMAJOR) i = 0, inc = 1; else i = rank - 1, inc = -1; - int32_t k = 1, s1 = shape[i], s2; + intptr_t k = 1, s1 = stride[i], s2; - shape[i] = 1; + stride[i] = 1; while (--rank) { i += inc; - s2 = shape[i]; - shape[i] = (k *= s1); + s2 = stride[i]; + stride[i] = (k *= s1); s1 = s2; } - return shape; + return stride; } -STC_DEF intptr_t _cspan_slice(int32_t oshape[], int32_t ostride[], int* orank, - const int32_t shape[], const int32_t stride[], +STC_DEF intptr_t _cspan_slice(int32_t oshape[], intptr_t ostride[], int* orank, + const int32_t shape[], const intptr_t stride[], int rank, const int32_t a[][2]) { intptr_t off = 0; int i = 0, oi = 0; int32_t end; for (; i < rank; ++i) { - off += (intptr_t)stride[i]*a[i][0]; + off += stride[i]*a[i][0]; switch (a[i][1]) { case 0: c_assert(c_LTu(a[i][0], shape[i])); continue; case -1: end = shape[i]; break; |