代码拉取完成,页面将自动刷新
同步操作将从 锦轩/swcodegen 强制同步,此操作会覆盖自 Fork 仓库以来所做的任何修改,且无法恢复!!!
确定后同步将在后台操作,完成时将刷新页面,请耐心等待。
#ifndef SLAVE_ARRAY_TILE_H
#define SLAVE_ARRAY_TILE_H
#include <isl/aff_type.h>
#include <isl/map_type.h>
#include <isl/val.h>
/* The current index is such that if you add "shift",
* then the result is always a multiple of "stride",
* where "stride" may be equal to 1.
* Let D represent the initial tile->depth dimensions of the computed schedule.
* The spaces of "lb" and "shift" are of the form
*
* D -> [b]
*/
struct slave_array_bound {
isl_val *size;
isl_aff *lb;
isl_val *stride;
isl_aff *shift;
};
/* A tile of an outer array.
*
* requires_unroll is set if the schedule dimensions that are mapped
* to threads need to be unrolled for this (private) tile to be used.
*
* "depth" reflects the number of schedule dimensions that affect the tile.
* The copying into and/or out of the tile is performed at that depth.
*
* n is the dimension of the array.
* bound is an array of size "n" representing the lower bound
* and size for each index.
*
* tiling maps a tile in the global array to the corresponding
* shared/private memory tile and is of the form
*
* { [D[i] -> A[a]] -> T[(a + shift(i))/stride - lb(i)] }
*
* where D represents the initial "depth" dimensions
* of the computed schedule.
*/
struct slave_array_tile {
isl_ctx *ctx;
int requires_unroll;
int depth;
int n;
struct slave_array_bound *bound;
isl_multi_aff *tiling;
};
struct slave_array_tile *slave_array_tile_create(isl_ctx *ctx, int n_index);
struct slave_array_tile *slave_array_tile_free(struct slave_array_tile *tile);
__isl_give isl_val *slave_array_tile_size(struct slave_array_tile *tile);
#endif
此处可能存在不合适展示的内容,页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容,可点击提交进行申诉,我们将尽快为您处理。