WOLFRAM

Wolfram 语言与系统参考资料中心

DistributeDefinitions[s1,s2,…]

将符号 si 的所有定义分配给所有并行内核.

DistributeDefinitions["context`"]

分配指定上下文下的所有符号定义.

更多信息和选项
Details and Options Details and Options
范例  
基本范例  
范围  
推广和延伸  
属性和关系  
可能存在的问题  
巧妙范例  
参见
相关指南
历史
按以下格式引用：

DistributeDefinitions

DistributeDefinitions[s1,s2,…]

将符号 si 的所有定义分配给所有并行内核.

DistributeDefinitions["context`"]

分配指定上下文下的所有符号定义.

更多信息和选项

范例

打开所有单元 关闭所有单元

基本范例  (2)

确保平行子内核已启动：

Wolfram Language code: LaunchKernels[];$KernelCount

必须首先分配在子内核中使用的值：

Wolfram Language code: ctx`y = 42;
Wolfram Language code: ParallelEvaluate[ToString[ctx`y]]
Wolfram Language code: $DistributedDefinitions
Wolfram Language code: DistributeDefinitions[ctx`y]
Wolfram Language code: ParallelEvaluate[ToString[ctx`y]]
Wolfram Language code: $DistributedDefinitions

会自动分配默认上下文中的符号：

Wolfram Language code: y = 43;
Wolfram Language code: ParallelEvaluate[ToString[y]]
Wolfram Language code: $DistributedDefinitions

范围  (5)

变量值：

Wolfram Language code: ni = 5;nd := $KernelID DistributeDefinitions[ni, nd]
Wolfram Language code: ParallelEvaluate[{ni, nd}]

函数：

Wolfram Language code: nfactors[n_] := Total[FactorInteger[(10 ^ n - 1/9)][[All, 2]]] DistributeDefinitions[nfactors]
Wolfram Language code: ParallelMap[nfactors, Range[50, 70]]

上值：

Wolfram Language code: g/:f[g[x_]] := fg[x, $KernelID] DistributeDefinitions[g]
Wolfram Language code: ParallelMap[f, {g[1], h[2], g[3]}]

属性：

Wolfram Language code: np = 5;Protect[np] DistributeDefinitions[np]
Wolfram Language code: ParallelEvaluate[np = $KernelID, First[Kernels[]]]

分配当前上下文中所有符号的定义：

Wolfram Language code: f[x_] := N[Pi ^ x, $prec];$prec = 50;
Wolfram Language code: DistributeDefinitions[Evaluate[Context[]]];
Wolfram Language code: ParallelEvaluate[f[$KernelID]]

推广和延伸  (1)

已分配的定义所依赖的辅助定义也被自动分配：

Wolfram Language code: prec = 18; mtest[n_] := Timing[Inverse[RandomReal[{-1, 1}, {n, n}, WorkingPrecision -> prec]]][[1]]
Wolfram Language code: DistributeDefinitions[mtest]
Wolfram Language code: ParallelTable[mtest[i], {i, 10, 100, 10}]

属性和关系  (7)

DistributeDefinitions 会覆盖任何前面出现的值和属性：

Wolfram Language code: ParallelEvaluate[f[0] = "Zero";f[x_] := x ^ 2, DistributedContexts -> None];
Wolfram Language code: ParallelMap[f, {0, 1, 2, 3, 4}, DistributedContexts -> None]
Wolfram Language code: f[x_] := x ^ 2 DistributeDefinitions[f]
Wolfram Language code: ParallelMap[f, {0, 1, 2, 3, 4}, DistributedContexts -> None]

通过清除该函数并且再次分配它，删除一个已分配的定义：

Wolfram Language code: f[x_] := Labeled[Framed[x ^ 2], $KernelID] DistributeDefinitions[f];
Wolfram Language code: ParallelMap[f, {1, 2, 3, 4}]
Wolfram Language code: ClearAll[f];DistributeDefinitions[f];
Wolfram Language code: ParallelMap[f, {1, 2, 3, 4}]

DistributeDefinitions 使用 ParallelEvaluate 来把定义传输给所有内核：

Wolfram Language code: f[n_] := PrimeQ[2 ^ n - 1]
Wolfram Language code: DistributeDefinitions[f];

一个明确的 ParallelEvaluate 做同样的事情：

Wolfram Language code: ParallelEvaluate[g[n_] := PrimeQ[2 ^ n + 1]];
Wolfram Language code: ParallelEvaluate[Through[{Identity, f, g}[$KernelID]]]

对新内核记录分配好的定义；ParallelEvaluate 的效果则不是：

Wolfram Language code: CloseKernels[];LaunchKernels[];
Wolfram Language code: ParallelEvaluate[Through[{Identity, f, g}[$KernelID]]]

对于高层次的并行命令，自动分配交互式定义的函数：

Wolfram Language code: ftest[n_] := Labeled[Framed[PrimeQ[2 ^ n - 1]], $KernelID]
Wolfram Language code: ParallelMap[ftest, Range[1275, 1285]]

手动分配定义，并且禁止使用自动分配：

Wolfram Language code: gtest[n_] := Labeled[Framed[PrimeQ[2 ^ n - 1]], $KernelID]
Wolfram Language code: DistributeDefinitions[gtest];
Wolfram Language code: ParallelMap[gtest, Range[1275, 1285], DistributedContexts -> None]

仅仅在子内核上具有值的符号不被分配：

Wolfram Language code: ParallelEvaluate[xr = 5]
Wolfram Language code: DistributeDefinitions[xr]

在子内核上的值不改动：

Wolfram Language code: ParallelTable[Sqrt[i] / xr, {i, 5}]

一旦一个符号得到一个局部值，它将在下一次并行计算中被分配：

Wolfram Language code: xr = 6;
Wolfram Language code: ParallelTable[Sqrt[i] / xr, {i, 5}]

使用 ParallelNeeds 在所有并行内核上设置程序包：

Wolfram Language code: Needs["GraphUtilities`"] ParallelNeeds["GraphUtilities`"]
Wolfram Language code: ParallelEvaluate[HamiltonianCycles[CompleteGraph[3], All]]

使用 DistributeDefinitions 设置您自己的定义：

Wolfram Language code: hcg[n_] := HamiltonianCycles[CompleteGraph[n], All]
Wolfram Language code: DistributeDefinitions[hcg];
Wolfram Language code: ParallelMap[hcg, Range[4]]

$DistributedDefinitions 给出了所有已分配定义的符号列表：

Wolfram Language code: a = 2;b = 3;
Wolfram Language code: DistributeDefinitions[a]
Wolfram Language code: $DistributedDefinitions

列表中还包括自动分配的符号：

Wolfram Language code: ParallelEvaluate[b]
Wolfram Language code: $DistributedDefinitions

可能存在的问题  (5)

并行内核上未知的一个函数可能导致串行计算：

Wolfram Language code: ftest[n_] := Labeled[Framed[PrimeQ[2 ^ n - 1]], $KernelID]
Wolfram Language code: ParallelEvaluate[ftest[$KernelID], DistributedContexts -> None]

在所有并行内核上定义函数：

Wolfram Language code: DistributeDefinitions[ftest];

现在在并行内核上计算该函数：

Wolfram Language code: ParallelEvaluate[ftest[$KernelID], DistributedContexts -> None]

DistributeDefinitions 的使用并不会阻止定义的自动分配：

Wolfram Language code: f[i_] := {i, $KernelID} DistributeDefinitions[f];

修改定义：

Wolfram Language code: Clear[f]

自动分配修改好的定义：

Wolfram Language code: Parallelize[Map[f, Range[8]]]

阻止定义的自动分配：

Wolfram Language code: g[i_] := {i, $KernelID} DistributeDefinitions[g];
Wolfram Language code: Clear[g]
Wolfram Language code: Parallelize[Map[g, Range[8]], DistributedContexts -> None]

仅仅在子内核上定义的符号不会被自动分配：

Wolfram Language code: ParallelEvaluate[h[i_] := {i, $KernelID}];
Wolfram Language code: Parallelize[Map[h, Range[8]]]

当分配时，某些具有内部状态的对象可能无法有效工作：

Wolfram Language code: data = RandomReal[{-1, 1}, 1000]; fdata = ListInterpolation[data]
Wolfram Language code: Timing[Table[fdata[r], {r, 1.0, 1000, 0.1}];]
Wolfram Language code: DistributeDefinitions[fdata];
Wolfram Language code: ParallelEvaluate[Timing[Table[fdata[r], {r, 1.0, 1000, 0.1}];], First[Kernels[]]]

重新在所有子内核中计算以获得更好的性能：

Wolfram Language code: DistributeDefinitions[data, fdata]
Wolfram Language code: ParallelEvaluate[fdata = ListInterpolation[data];];
Wolfram Language code: ParallelEvaluate[Timing[Table[fdata[r], {r, 1.0, 1000, 0.1}];], First[Kernels[]]]

或者，在所有子内核上重新计算数据：

Wolfram Language code: fdata1 = fdata; DistributeDefinitions[fdata1]
Wolfram Language code: ParallelEvaluate[fdata1 = fdata1];
Wolfram Language code: ParallelEvaluate[Timing[Table[fdata1[r], {r, 1.0, 1000, 0.1}];], Last[Kernels[]]]

带有 ReadProtected 属性的符号定义不能被分发：

Wolfram Language code: ftest[n_] := Labeled[Framed[PrimeQ[2 ^ n - 1]], $KernelID]; SetAttributes[ftest, ReadProtected]
Wolfram Language code: DistributeDefinitions[ftest]
Wolfram Language code: ParallelMap[ftest, Range[$KernelCount]]

定义只在需要时分配；如果定义没有变化，则不再分配：

Wolfram Language code: a = 2;b = 3;
Wolfram Language code: DistributeDefinitions[a, b]
Wolfram Language code: b = 4;
Wolfram Language code: DistributeDefinitions[a, b]

依然分配了这两个符号：

Wolfram Language code: $DistributedDefinitions

巧妙范例  (1)

快速可视化高斯素数：

Wolfram Language code: test[x_] := Boole[PrimeQ[x, GaussianIntegers -> True]]
Wolfram Language code: data = With[{n = 100}, ParallelTable[test[x + y I], {x, -n, n}, {y, -n, n}]]; ArrayPlot[data]
Wolfram Research (2008)，DistributeDefinitions，Wolfram 语言函数，https://reference.wolfram.com/language/ref/DistributeDefinitions.html （更新于 2025 年）.

文本

Wolfram Research (2008)，DistributeDefinitions，Wolfram 语言函数，https://reference.wolfram.com/language/ref/DistributeDefinitions.html （更新于 2025 年）.

CMS

Wolfram 语言. 2008. "DistributeDefinitions." Wolfram 语言与系统参考资料中心. Wolfram Research. 最新版本 2025. https://reference.wolfram.com/language/ref/DistributeDefinitions.html.

APA

Wolfram 语言. (2008). DistributeDefinitions. Wolfram 语言与系统参考资料中心. 追溯自 https://reference.wolfram.com/language/ref/DistributeDefinitions.html 年

BibTeX

@misc{reference.wolfram_2026_distributedefinitions, author="Wolfram Research", title="{DistributeDefinitions}", year="2025", howpublished="\url{https://reference.wolfram.com/language/ref/DistributeDefinitions.html}", note=[Accessed: 13-September-2026]}

BibLaTeX

@online{reference.wolfram_2026_distributedefinitions, organization={Wolfram Research}, title={DistributeDefinitions}, year={2025}, url={https://reference.wolfram.com/language/ref/DistributeDefinitions.html}, note=[Accessed: 13-September-2026]}