Skip to content
版 本

asc_ldca

产 品 支 持 情 况

  • Ascend 950PR/Ascend 950DT:支 持
  • Atlas A3 训 练 系 列 产 品/Atlas A3 推 理 系 列 产 品:不 支 持
  • Atlas A2 训 练 系 列 产 品/Atlas A2 推 理 系 列 产 品:不 支 持
  • Atlas 200I/500 A2 推 理 产 品:不 支 持
  • Atlas 推 理 系 列 产 品AI Core:不 支 持
  • Atlas 推 理 系 列 产 品Vector Core:不 支 持
  • Atlas 训 练 系 列 产 品:不 支 持

功 能 说 明

首 先 从Data Cache加 载 缓 存 数 据,若 未 命 中,则 尝 试 从L2 Cache加 载。L2 Cache与Global Memory之 间 的 数 据 一 致 性 由 硬 件 保 证。如 果Data Cache和L2 Cache中 均 未 找 到 所 需 数 据,则 从Global Memory中 读 取 数 据,然 后 将 其 缓 存 到L2 Cache和Data Cache中。

函 数 原 型

Text
inline long int asc_ldca(long int* address)
Text
inline unsigned long int asc_ldca(unsigned long int* address)
Text
inline long long int asc_ldca(long long int* address)
Text
inline unsigned long long int asc_ldca(unsigned long long int* address)
Text
inline long2 asc_ldca(long2* address)
Text
inline ulong2 asc_ldca(ulong2* address)
Text
inline long4 asc_ldca(long4* address)
Text
inline ulong4 asc_ldca(ulong4* address)
Text
inline longlong2 asc_ldca(longlong2* address)
Text
inline ulonglong2 asc_ldca(ulonglong2* address)
Text
inline longlong4 asc_ldca(longlong4* address)
Text
inline ulonglong4 asc_ldca(ulonglong4* address)
Text
inline signed char asc_ldca(signed char* address)
Text
inline unsigned char asc_ldca(unsigned char* address)
Text
inline char2 asc_ldca(char2* address)
Text
inline uchar2 asc_ldca(uchar2* address)
Text
inline char4 asc_ldca(char4* address)
Text
inline uchar4 asc_ldca(uchar4* address)
Text
inline short asc_ldca(short* address)
Text
inline unsigned short asc_ldca(unsigned short* address)
Text
inline short2 asc_ldca(short2* address)
Text
inline ushort2 asc_ldca(ushort2* address)
Text
inline short4 asc_ldca(short4* address)
Text
inline ushort4 asc_ldca(ushort4* address)
Text
inline int asc_ldca(int* address)
Text
inline unsigned int asc_ldca(unsigned int* address)
Text
inline int2 asc_ldca(int2* address)
Text
inline uint2 asc_ldca(uint2* address)
Text
inline int4 asc_ldca(int4* address)
Text
inline uint4 asc_ldca(uint4* address)
Text
inline float asc_ldca(float* address)
Text
inline float2 asc_ldca(float2* address)
Text
inline float4 asc_ldca(float4* address)
Text
inline bfloat16_t asc_ldca(bfloat16_t* address)
Text
inline bfloat16x2_t asc_ldca(bfloat16x2_t* address)
Text
inline half asc_ldca(half* address)
Text
inline half2 asc_ldca(half2* address)

参 数 说 明

表 1 参 数 说 明

参 数 名输 入/输 出描 述
address输 入Global Memory的 地 址。

返 回 值 说 明

Data Cache或L2 Cache中 的 缓 存 数 据,或 输 入 的Global Memory地 址 上 的 数 据。

约 束 说 明

需 要 包 含 的 头 文 件

使 用 除half、half2、bfloat16_t、bfloat16x2_t类 型 之 外 的 接 口 需 要 包 含"simt_api/device_functions.h"头 文 件,使 用half和half2类 型 接 口 需 要 包 含"simt_api/asc_fp16.h"头 文 件,使 用bfloat16_t和bfloat16x2_t类 型 接 口 需 要 包 含"simt_api/asc_bf16.h"头 文 件。

Text
#include "simt_api/device_functions.h"
Text
#include "simt_api/asc_fp16.h"
Text
#include "simt_api/asc_bf16.h"

调 用 示 例

  • SIMT编 程 场 景:

    Text
    __global__ __launch_bounds__(1024) void kernel_asc_ldca(float* dst, float* src)
    {
        int idx = threadIdx.x + blockIdx.x * blockDim.x;
        dst[idx] = asc_ldca(src + idx);
    }
    
  • SIMD与SIMT混 合 编 程 场 景:

    SIMD与SIMT混 合 编 程 场 景,需 要 显 式 使 用 地 址 空 间 限 定 符 表 示 地 址 空 间:__gm__表 示Global Memory内 存 空 间。

    Text
    __simt_vf__ __launch_bounds__(1024) inline void kernel_asc_ldca(__gm__ float* dst, __gm__ float* src)
    {
        int idx = threadIdx.x + blockIdx.x * blockDim.x;
        dst[idx] = asc_ldca(src + idx);
    }
    

免 责 声 明:本 站 内 容 由 asc-devkit 仓 master 分 支 自 动 编 译 生 成,属 于 持 续 开 发 版 本,可 能 存 在 缺 陷,仅 供 预 览 与 参 考。如 需 稳 定 及 商 用 资 料,请 查 阅 官 方 昇 腾 社 区