ParallelTable[expr,{imax}]
并行产生 imax 个 expr 列表.
ParallelTable[expr,{i,imax}]
当 i 从 1 到 imax 时,并行产生 expr 的值的列表.
ParallelTable[expr,{i,imin,imax}]
从 i=imin 开始.
ParallelTable[expr,{i,imin,imax,di}]
用步长 di.
ParallelTable[expr,{i,{i1,i2,…}}]
用连续值 i1,i2,….
ParallelTable[expr,{i,imin,imax},{j,jmin,jmax},…]
给出一个嵌套列表,与 i 相关的列表在最外层.
ParallelTable
ParallelTable[expr,{imax}]
并行产生 imax 个 expr 列表.
ParallelTable[expr,{i,imax}]
当 i 从 1 到 imax 时,并行产生 expr 的值的列表.
ParallelTable[expr,{i,imin,imax}]
从 i=imin 开始.
ParallelTable[expr,{i,imin,imax,di}]
用步长 di.
ParallelTable[expr,{i,{i1,i2,…}}]
用连续值 i1,i2,….
ParallelTable[expr,{i,imin,imax},{j,jmin,jmax},…]
给出一个嵌套列表,与 i 相关的列表在最外层.
更多信息和选项
- ParallelTable 是一个并行的 Table, 在不同的内核和处理器之间自动分配 expr 的计算.
- ParallelTable 将给出和 Table 相同的结果,除了计算中的副作用.
- Parallelize[Table[expr,iter, …]] 等价于 ParallelTable[expr,iter,…].
- 对于不能并行化的 ParallelTable,用 Table 执行计算.
- 可以给出下列选项:
-
Method Automatic 并行化粒度 DistributedContexts $DistributedContexts 用于将符号分配到并行计算的上下文 ProgressReporting $ProgressReporting 是否报告计算进度 - Method 选项指定要使用的并行化方法. 可能的设置包括:
-
"CoarsestGrained" 将计算分成与可用内核数量相同的段 "FinestGrained" 将计算分成最小的子单元 "EvaluationsPerKernel"->e 将计算分成每个内核最多 e 段 "ItemsPerEvaluation"->m 将计算分成每个内核最多 m 个子单元 Automatic 在总开销和负载平衡之间的折中 - Method->"CoarsestGrained" 适用于涉及多个子单元的计算,所有子单元采用相同的时间计算. 将总开销降到最低,不提供任何负载均衡.
- Method->"FinestGrained" 适用于少量子单元的计算,子单元采用不同的时间计算. 它提高总开销,但最大化了负载均衡.
- 默认情况下,对于具有较大最外层级的嵌套表格,并行化会在最外层进行;否则,会在最内层进行. 使用Method->"CoarsestGrained" 时,将在最外层进行并行化. 使用 Method->"FinestGrained" 时,则会在最内层进行并行化.
- DistributedContexts 选项指定出现在 expr 中的哪些符号在计算之前将其定义自动分发到所有可用内核.
- 设置 $DistributedContexts:=$Context,默认值为 DistributedContexts:>$DistributedContexts,该值分发当前上下文中所有符号的定义,但不分发程序包中的符号定义.
- ProgressReporting 选项指定是否报告并行计算的进度.
- 默认值为 ProgressReporting:>$ProgressReporting.
范例
打开所有单元 关闭所有单元基本范例 (6)
ParallelTable 与 Table 工作方式相似,不同之处在于并行:
Table[Pause[1];f[i], {i, 4}]//AbsoluteTimingParallelTable[Pause[1];f[i], {i, 4}]//AbsoluteTimingParallelTable[i ^ 2, {i, 10}]ParallelTable[f[i], {i, 0, 20, 2}]ParallelTable[10i + j, {i, 4}, {j, 3}]MatrixForm[%]ListPlot[ParallelTable[Total[FactorInteger[2 ^ i - 1][[All, 2]]], {i, 150, 190}]]res = ParallelTable[PrimeQ[2 ^ i - 1], {i, 9601, 12000}];
| |
范围 (5)
ParallelTable[f[i], {i, 10, -5, -2}]ParallelTable[10i + j, {i, 5}, {j, i}]TableForm[%]ParallelTable[100i + 10j + k, {i, 3}, {j, 2}, {k, 4}]ParallelTable[Sqrt[x], {x, {1, 4, 9, 16}}]ParallelTable[j ^ (1 / i), {i, {1, 2, 4}}, {j, {1, 4, 9}}]选项 (14)
Method (7)
ParallelTable[Labeled[Framed[i], $KernelID], {i, 10}, Method -> "FinestGrained"]ParallelTable[Labeled[Framed[i], $KernelID], {i, 10}, Method -> "CoarsestGrained"]ParallelTable[Labeled[Framed[i], $KernelID], {i, 12}, Method -> "EvaluationsPerKernel" -> 2]ParallelTable[Labeled[Framed[i], $KernelID], {i, 18}, Method -> "ItemsPerEvaluation" -> 5]ParallelTable[Labeled[Framed[i], $KernelID], {i, 18}, Method -> Automatic]ParallelTable[{i, PrimeQ[2 ^ i - 1]}, {i, 4410, 4430}, Method -> "FinestGrained"]BinCounts[ParallelTable[Mod[Floor[i * Pi], 10], {i, 10000}, Method -> "CoarsestGrained"], {0, 10}]ParallelTable[{i, j, Style[$KernelID, Red]}, {i, 4}, {j, i}]如果想要只在第一层并行,可使用 Method"CoarsestGrained":
ParallelTable[{i, j, Style[$KernelID, Red]}, {i, 4}, {j, i}, Method -> "CoarsestGrained"]DistributedContexts (5)
remote[x_] := {$KernelID, x ^ 3}ParallelTable[remote[i], {i, 4}]local[x_] := {$KernelID, x ^ 2}ParallelTable[local[i], {i, 4}, DistributedContexts -> None]a`f[x_] := {$KernelID, x}ParallelTable[a`f[i], {i, 4}, DistributedContexts -> Automatic]b`g[x_] := {$KernelID, -x}ParallelTable[b`g[i], {i, 4}, DistributedContexts -> {"a`"}]恢复 DistributedContexts 选项的值为默认值:
SetOptions[ParallelTable, DistributedContexts :> $DistributedContexts]ProgressReporting (2)
res = ParallelTable[PrimeQ[2 ^ i - 1], {i, 9601, 12000}, ProgressReporting -> False];使用 Method"FinestGrained" 可得到准确度最高的进度报告:
res = ParallelTable[{i, Total[FactorInteger[10 ^ 60 + i][[All, 2]]]}, {i, 1000}, Method -> "FinestGrained"];
| |
应用 (5)
对于许多初始条件,求解并且绘制一个微分方程,并且使用动画显示结果:
eqn = D[z[t, s], t, t] == -(z[t, s]/(z[t, s]^2 + ((1/2) (1 + 0.1 Sin[2 Pi t]))^2)^3 / 2);sols = ParallelTable[localsol = NDSolve[{eqn, z[0, s] == param s, Derivative[1, 0][z][0, s] == param}, z, {t, 0, 10}, {s, 1, 2}];
Plot3D[Evaluate[z[tp, z0] /. localsol], {tp, 0, 10}, {z0, 1, 2}, PlotRange -> {-5, 5}, PlotLabel -> param],
{param, -1, 1, 0.1}];ListAnimate[sols]With[{L = 4, dz = 0.25},
sols = ParallelTable[
localsol = Quiet@NDSolve[{D[u[t, x, y], t, t] == D[u[t, x, y], x, x] + D[u[t, x, y], y, y] + Sin[u[t, x, y]], u[t, -L, y] == u[t, L, y], u[t, x, -L] == u[t, x, L], u[0, x, y] == a Exp[-(b x ^ 2 + y ^ 2)], Derivative[1, 0, 0][u][0, x, y] == 0}, u, {t, 0, L / 2}, {x, -L, L}, {y, -L, L}];
Plot3D[Evaluate[u[L / 2, x, y] /. First[localsol]], {x, -L, L}, {y, -L, L}, PlotRange -> {{-L, L}, {-L, L}, {-dz, dz}}, Axes -> None, PlotLabel -> {a, b}], {a, -0.5, 0.5, 0.2}, {b, 0.8, 1.2, .1}, Method -> "FinestGrained"]
];GraphicsGrid[sols]img = [image];filters = {Identity, Sharpen, Blur, GaussianFilter[#, 3]&, LaplacianFilter[#, 1]&, LaplacianGaussianFilter[#, 2]&,
GradientFilter[#, 2]&, MedianFilter[#, 5]&,
ImageConvolve[#, BoxMatrix[1] / 9]&};ParallelTable[Tooltip[filter[img], filter], {filter, filters}]ParallelTable[ImageEffect[img, effect], {effect, {"Charcoal", "Embossing", "Solarization"}}]frames[i_, f_, R_] := Module[{frame},
Do[frame = RevolutionPlot3D[Sin[t / (10f) * 2 * Pi]BesselJ[0, r], {r, 0, R}, PlotRange -> {{-R, R}, {-R, R}, {-1, 1}}, Axes -> None, PlotPoints -> 5];
Export["drum-" <> ToString[PaddedForm[t, 3, NumberPadding -> "0"]] <> ".gif", frame]
, {t, 10i + 1, 10(i + 1)}
];
frame
];samples = With[{batches = 8}, ParallelTable[frames[i, batches, 11], {i, 0, batches - 1}]];samplesDeleteFile[FileNames["drum-*.gif"]]data = Select[ParallelTable[Graphics3D[{Cuboid /@ Position[CellularAutomaton[{n, {2, 1}, {1, 1, 1}}, {{{{1}}}, 0}, {{{8}}}], 1]}, Boxed -> False], {n, 2, 80, 5}], Length[#[[1, -1]]] > 50&];GraphicsGrid[Partition[data, 3], Frame -> All, ImageSize -> 500]属性和关系 (10)
ParallelTable[Labeled[Framed[f[i, j]], $KernelID], {i, 4}, {j, i}]使用多个迭代规范相当于嵌套 Table 函数:
ParallelTable[i + j, {i, 3}, {j, i}]ParallelTable[Table[i + j, {j, i}], {i, 3}]ParallelDo 把表达式的相同序列作为 ParallelTable 计算:
ParallelDo[Print[i ^ i], {i, 3}]ParallelTable[i ^ i, {i, 3}]ParallelSum 有效地把 Plus 应用到 ParallelTable 所得的结果:
ParallelSum[x ^ i, {i, 3}]ParallelTable[x ^ i, {i, 3}]ParallelArray 在一些连续整数上迭代:
ParallelArray[#1 ^ #2&, {3, 4}]ParallelArray[Function[{x, y}, x ^ y], {3, 4}]ParallelTable[x ^ y, {x, 3}, {y, 4}]Map 把一个函数应用于列表中的连续元素上:
list = RandomInteger[9, 10]ParallelMap[Last[IntegerDigits[#, 2]]&, list]Table 可以把列表中的连续元素替换到一个表达式中:
ParallelTable[Last[IntegerDigits[x, 2]], {x, list}]在一个给定列表中迭代的 ParallelTable 等价于 ParallelCombine:
list = RandomInteger[9, 10]ParallelTable[x ^ 2, {x, list}]ParallelCombine[Function[x, x ^ 2], list]ParallelTable 可以通过使用 WaitAll 和 ParallelSubmit 实现:
WaitAll[Table[ParallelSubmit[{i}, Labeled[Framed[f[i]], $KernelID]], {i, 4}]]WaitAll[Table[ParallelSubmit[{i, j}, Labeled[Framed[f[i, j]], $KernelID]], {i, 4}, {j, i}]]ftest[n_] := Labeled[Framed[PrimeQ[2 ^ n - 1]], $KernelID]ParallelTable[ftest[i], {i, 1275, 1285}]gtest[n_] := Labeled[Framed[PrimeQ[2 ^ n - 1]], $KernelID]DistributeDefinitions[gtest];ParallelTable[gtest[i], {i, 1275, 1285}, DistributedContexts -> None]对于来自于程序包的函数,使用 ParallelNeeds 而不是 DistributeDefinitions:
Needs["FiniteFields`"]Table[FieldSize[GF[2, i]], {i, 10}]ParallelNeeds["FiniteFields`"]ParallelTable[FieldSize[GF[2, i]], {i, 10}]可能存在的问题 (3)
ftest[n_] := Labeled[Framed[PrimeQ[2 ^ n - 1]], $KernelID]ParallelTable[ftest[i], {i, 1275, 1285}, DistributedContexts -> None]DistributeDefinitions[ftest]ParallelTable[ftest[i], {i, 1275, 1285}, DistributedContexts -> None]gtest[n_] := Labeled[Framed[PrimeQ[2 ^ n - 1]], $KernelID]ParallelTable[gtest[i], {i, 1275, 1285}]ctx`mtest[n_] := Labeled[Framed[PrimeQ[2 ^ n - 1]], $KernelID]ParallelTable[ctx`mtest[i], {i, 1275, 1285}]使用 DistributeDefinitions 分配这样的定义:
DistributeDefinitions[ctx`mtest];ParallelTable[ctx`mtest[i], {i, 1275, 1285}]或者,设置 DistributedContexts 选项包括所有的上下文:
cty`mtest[n_] := Labeled[Framed[PrimeQ[2 ^ n - 1]], $KernelID]ParallelTable[cty`mtest[i], {i, 1275, 1285}, DistributedContexts -> Automatic]AbsoluteTiming[ParallelTable[N[Sin[x]], {x, 0, 1000}];]AbsoluteTiming[Table[N[Sin[x]], {x, 0, 1000}];]巧妙范例 (2)
mlength[z_] := Length[FixedPointList[# ^ 2 + z&, z, 20, SameTest -> (Abs[#] > 2&)]]points = ParallelTable[mlength[x + I y], {y, -1, 1, 0.005}, {x, -2, 1, 0.005}];ArrayPlot[points, ColorFunction -> Hue]计算并且显示物流地图的 Feigenbaum (或者分岔)图:
line[r_, dy_, np_, n0_, n_] := Module[{pts},
With[{logistics = Function[x, r x(1 - x)]},
pts = Join@@NestList[logistics, Nest[logistics, RandomReal[{0, 1}, np], n0], n - 1]];
Log[1.0 + BinCounts[pts, {0, 1, dy}]]]With[{w = 400, h = 250, r0 = 2.95, r1 = 4.0},
ArrayPlot[ParallelTable[line[r, 1 / (w - 1), w, 500, 50], {r, r0, r1, (r1 - r0) / (h - 1)}], ImageSize -> {w, h}]]文本
Wolfram Research (2008),ParallelTable,Wolfram 语言函数,https://reference.wolfram.com/language/ref/ParallelTable.html (更新于 2021 年).
CMS
Wolfram 语言. 2008. "ParallelTable." Wolfram 语言与系统参考资料中心. Wolfram Research. 最新版本 2021. https://reference.wolfram.com/language/ref/ParallelTable.html.
APA
Wolfram 语言. (2008). ParallelTable. Wolfram 语言与系统参考资料中心. 追溯自 https://reference.wolfram.com/language/ref/ParallelTable.html 年
BibTeX
@misc{reference.wolfram_2026_paralleltable, author="Wolfram Research", title="{ParallelTable}", year="2021", howpublished="\url{https://reference.wolfram.com/language/ref/ParallelTable.html}", note=[Accessed: 17-November-2026]}
BibLaTeX
@online{reference.wolfram_2026_paralleltable, organization={Wolfram Research}, title={ParallelTable}, year={2021}, url={https://reference.wolfram.com/language/ref/ParallelTable.html}, note=[Accessed: 17-November-2026]}