Processors Command - Revision history

Nairnj: /* Custom Patching */

2023-02-02T22:02:36Z

Custom Patching

← Older revision		Revision as of 22:02, 2 February 2023
Line 29:		Line 29:
	where <tt>xnum</tt>, <tt>ynum</tt>, and <tt>znum</tt> are the number of patches in the x, y, and z direction (z only used for 3D patching). Their product must be equal to the number of processors used in the calculations. This command can be used in scripted files by using the [[XMLData Command]] for the <tt>MPMHeader</tt> section.		where <tt>xnum</tt>, <tt>ynum</tt>, and <tt>znum</tt> are the number of patches in the x, y, and z direction (z only used for 3D patching). Their product must be equal to the number of processors used in the calculations. This command can be used in scripted files by using the [[XMLData Command]] for the <tt>MPMHeader</tt> section.

	By default, the patches will be equally sized along each axis. If unequal patch sizes can improve parallel scaling, they can be created using <tt>Xpatches</tt>, <tt>Ypatches</tt>, or <tt>Zpatches</tt> elements. Each patches element must contain exactly <tt>inum-1</tt> (i = x, y, or z) monotonically increasing numbers between zero and one for the fractional break points between patches. If an axis has <tt>N</tt> elements, the <tt>k<sup>th</sup></tt> patch (for <tt>k</tt> from 1 to <tt>inum</tt>) will include elements <tt>int(n<sub>k-1</sub>N)</tt> to <tt>int(n<sub>k</sub>N)</tt> (where n<sub>0</sub> is set to 0 and n<sub>inum</sub> is set to 1). All patches ~~mush~~ have at least two elements or must have <tt>int(n<sub>k</sub>N)</tt>><tt>int(n<sub>k-1</sub>N)</tt> for all <tt>k</tt>. The patches elements are optional. Any omitted axes will use equally sizes patches corresponding to <tt>n<sub>i</sub>=i/inum</tt>.		By default, the patches will be equally sized along each axis. If unequal patch sizes can improve parallel scaling, they can be created using <tt>Xpatches</tt>, <tt>Ypatches</tt>, or <tt>Zpatches</tt> elements. Each patches element must contain exactly <tt>inum-1</tt> (i = x, y, or z) monotonically increasing numbers between zero and one for the fractional break points between patches. If an axis has <tt>N</tt> elements, the <tt>k<sup>th</sup></tt> patch (for <tt>k</tt> from 1 to <tt>inum</tt>) will include elements <tt>int(n<sub>k-1</sub>N)</tt> to <tt>int(n<sub>k</sub>N)</tt> (where n<sub>0</sub> is set to 0 and n<sub>inum</sub> is set to 1). All patches must have at least two elements or must have <tt>int(n<sub>k</sub>N)</tt>><tt>int(n<sub>k-1</sub>N)</tt> for all <tt>k</tt>. The patches elements are optional. Any omitted axes will use equally sizes patches corresponding to <tt>n<sub>i</sub>=i/inum</tt>.

	== Hyperthreading ==		== Hyperthreading ==

Nairnj: /* Custom Patching */

2023-01-26T20:47:03Z

Custom Patching

← Older revision		Revision as of 20:47, 26 January 2023
Line 29:		Line 29:
	where <tt>xnum</tt>, <tt>ynum</tt>, and <tt>znum</tt> are the number of patches in the x, y, and z direction (z only used for 3D patching). Their product must be equal to the number of processors used in the calculations. This command can be used in scripted files by using the [[XMLData Command]] for the <tt>MPMHeader</tt> section.		where <tt>xnum</tt>, <tt>ynum</tt>, and <tt>znum</tt> are the number of patches in the x, y, and z direction (z only used for 3D patching). Their product must be equal to the number of processors used in the calculations. This command can be used in scripted files by using the [[XMLData Command]] for the <tt>MPMHeader</tt> section.

	By default, the patches will be equally sized along each axis. If unequal patch sizes can improve parallel scaling, they can be created using <tt>Xpatches</tt>, <tt>Ypatches</tt>, or <tt>Zpatches</tt> elements. Each patches element must contain exactly <tt>inum-1</tt> (i = x, y, or z) monotonically increasing numbers between zero and one for the fractional break points between patches. If an axis has <tt>N</tt> elements, the <tt>k<sup>th</sup></tt> patch (for <tt>k</tt> from 1 to <tt>inum</tt>) will include elements <tt>int(n<sub>k-1</sub>N)</tt> to <tt>int(n<sub>k</sub>N)</tt> (where n<sub>0</sub> is set to 0 and n<sub>inum</sub> is set to 1). All patches mush have at least two elements or must have <tt>int(n<sub>k</sub>N)</tt>><tt>int(n<sub>k-1</sub>N)</tt>. The patches elements are optional. Any omitted axes will use equally sizes patches corresponding to <tt>n<sub>i</sub>=i/inum</tt>.		By default, the patches will be equally sized along each axis. If unequal patch sizes can improve parallel scaling, they can be created using <tt>Xpatches</tt>, <tt>Ypatches</tt>, or <tt>Zpatches</tt> elements. Each patches element must contain exactly <tt>inum-1</tt> (i = x, y, or z) monotonically increasing numbers between zero and one for the fractional break points between patches. If an axis has <tt>N</tt> elements, the <tt>k<sup>th</sup></tt> patch (for <tt>k</tt> from 1 to <tt>inum</tt>) will include elements <tt>int(n<sub>k-1</sub>N)</tt> to <tt>int(n<sub>k</sub>N)</tt> (where n<sub>0</sub> is set to 0 and n<sub>inum</sub> is set to 1). All patches mush have at least two elements or must have <tt>int(n<sub>k</sub>N)</tt>><tt>int(n<sub>k-1</sub>N)</tt> for all <tt>k</tt>. The patches elements are optional. Any omitted axes will use equally sizes patches corresponding to <tt>n<sub>i</sub>=i/inum</tt>.

	== Hyperthreading ==		== Hyperthreading ==

Nairnj: /* Custom Patching */

2023-01-26T20:46:26Z

Custom Patching

← Older revision		Revision as of 20:46, 26 January 2023
Line 29:		Line 29:
	where <tt>xnum</tt>, <tt>ynum</tt>, and <tt>znum</tt> are the number of patches in the x, y, and z direction (z only used for 3D patching). Their product must be equal to the number of processors used in the calculations. This command can be used in scripted files by using the [[XMLData Command]] for the <tt>MPMHeader</tt> section.		where <tt>xnum</tt>, <tt>ynum</tt>, and <tt>znum</tt> are the number of patches in the x, y, and z direction (z only used for 3D patching). Their product must be equal to the number of processors used in the calculations. This command can be used in scripted files by using the [[XMLData Command]] for the <tt>MPMHeader</tt> section.

	By default, the patches will be equally sized along each axis. If unequal patch sizes can improve parallel scaling, they can be created using <tt>Xpatches</tt>, <tt>Ypatches</tt>, or <tt>Zpatches</tt> elements. Each patches element must contain exactly <tt>inum-1</tt> (i = x, y, or z) monotonically increasing numbers between zero and one for the fractional break points between patches. If an axis has <tt>N</tt> elements, the <tt>k<sup>th</sup></tt> patch (for <tt>k</tt> from 1 to <tt>inum</tt>) will include elements <tt>int(n<sub>k-1</sub>N)</tt> to <tt>int(n<sub>k</sub>N)</tt> (where n<sub>0</sub> is set to 0 and n<sub>inum</sub> is set to 1). All patches mush have at least two elements or must have <tt>int(n<sub>k</sub>N)</tt>><tt>int(n<sub>k-1</sub>N)</tt>. The patches elements are optional. Any omitted axes will use equally sizes patches corresponding to <tt>n<sub>i</sub>=i/N</tt>.		By default, the patches will be equally sized along each axis. If unequal patch sizes can improve parallel scaling, they can be created using <tt>Xpatches</tt>, <tt>Ypatches</tt>, or <tt>Zpatches</tt> elements. Each patches element must contain exactly <tt>inum-1</tt> (i = x, y, or z) monotonically increasing numbers between zero and one for the fractional break points between patches. If an axis has <tt>N</tt> elements, the <tt>k<sup>th</sup></tt> patch (for <tt>k</tt> from 1 to <tt>inum</tt>) will include elements <tt>int(n<sub>k-1</sub>N)</tt> to <tt>int(n<sub>k</sub>N)</tt> (where n<sub>0</sub> is set to 0 and n<sub>inum</sub> is set to 1). All patches mush have at least two elements or must have <tt>int(n<sub>k</sub>N)</tt>><tt>int(n<sub>k-1</sub>N)</tt>. The patches elements are optional. Any omitted axes will use equally sizes patches corresponding to <tt>n<sub>i</sub>=i/inum</tt>.

	== Hyperthreading ==		== Hyperthreading ==

Nairnj: /* Custom Patching */

2023-01-26T20:44:06Z

Custom Patching

← Older revision		Revision as of 20:44, 26 January 2023
Line 29:		Line 29:
	where <tt>xnum</tt>, <tt>ynum</tt>, and <tt>znum</tt> are the number of patches in the x, y, and z direction (z only used for 3D patching). Their product must be equal to the number of processors used in the calculations. This command can be used in scripted files by using the [[XMLData Command]] for the <tt>MPMHeader</tt> section.		where <tt>xnum</tt>, <tt>ynum</tt>, and <tt>znum</tt> are the number of patches in the x, y, and z direction (z only used for 3D patching). Their product must be equal to the number of processors used in the calculations. This command can be used in scripted files by using the [[XMLData Command]] for the <tt>MPMHeader</tt> section.

	By default, the patches will be equally sized along each axis. If unequal patch sizes can improve parallel scaling, they can be created using <tt>Xpatches</tt>, <tt>Ypatches</tt>, or <tt>Zpatches</tt> elements. Each patches element must contain exactly <tt>inum-1</tt> (i = x, y, or z) monotonically increasing numbers between zero and one for the fractional break points between patches. If an axis has <tt>N</tt> elements, the <tt>k<sup>th</sup></tt> patch (for <tt>k</tt> from 1 to <tt>inum</tt>) will include elements <tt>int(n<sub>k-1</sub>/N)</tt> to <tt>int(n<sub>k</sub>/N)</tt> (where n<sub>0</sub> is set to 0 and n<sub>inum</sub> is set to 1). All patches mush have at least two elements or must have <tt>int(n<sub>k</sub>/N)</tt>><tt>int(n<sub>k-1</sub>/N)</tt>. The patches elements are optional. Any omitted axes will use equally sizes patches corresponding to <tt>n<sub>i</sub>=i/N</tt>.		By default, the patches will be equally sized along each axis. If unequal patch sizes can improve parallel scaling, they can be created using <tt>Xpatches</tt>, <tt>Ypatches</tt>, or <tt>Zpatches</tt> elements. Each patches element must contain exactly <tt>inum-1</tt> (i = x, y, or z) monotonically increasing numbers between zero and one for the fractional break points between patches. If an axis has <tt>N</tt> elements, the <tt>k<sup>th</sup></tt> patch (for <tt>k</tt> from 1 to <tt>inum</tt>) will include elements <tt>int(n<sub>k-1</sub>N)</tt> to <tt>int(n<sub>k</sub>N)</tt> (where n<sub>0</sub> is set to 0 and n<sub>inum</sub> is set to 1). All patches mush have at least two elements or must have <tt>int(n<sub>k</sub>N)</tt>><tt>int(n<sub>k-1</sub>N)</tt>. The patches elements are optional. Any omitted axes will use equally sizes patches corresponding to <tt>n<sub>i</sub>=i/N</tt>.

	== Hyperthreading ==		== Hyperthreading ==

Nairnj: /* Custom Patching */

2023-01-26T20:42:37Z

Custom Patching

← Older revision		Revision as of 20:42, 26 January 2023
Line 29:		Line 29:
	where <tt>xnum</tt>, <tt>ynum</tt>, and <tt>znum</tt> are the number of patches in the x, y, and z direction (z only used for 3D patching). Their product must be equal to the number of processors used in the calculations. This command can be used in scripted files by using the [[XMLData Command]] for the <tt>MPMHeader</tt> section.		where <tt>xnum</tt>, <tt>ynum</tt>, and <tt>znum</tt> are the number of patches in the x, y, and z direction (z only used for 3D patching). Their product must be equal to the number of processors used in the calculations. This command can be used in scripted files by using the [[XMLData Command]] for the <tt>MPMHeader</tt> section.

	By default, the patches will be equally sized along each axis. If unequal patch sizes can improve parallel scaling, they can be created using <tt>Xpatches</tt>, <tt>Ypatches</tt>, or <tt>Zpatches</tt> elements. Each patches element must contain exactly <tt>inum-1</tt> (i = x, y, or z) monotonically increasing numbers between zero and one for the fractional break points between patches. If an axis has <tt>N</tt> elements, the <tt>k<sup>th</sup></tt> patch (for <tt>k</tt> from 1 to <tt>inum</tt>) will include elements <tt>int(n_<sub>k-1</sub>/N)</tt> to <tt>int(n_<sub>k</sub>/N)</tt> (where n<sub>0</sub> is set to 0 and n<sub>inum</sub> is set to 1). All patches mush have at least two elements or must have <tt>int(n_<sub>k</sub>/N)</tt>><tt>int(n_<sub>k-1</sub>/N)</tt>. The patches elements are optional. Any omitted axes will use equally sizes patches corresponding to <tt>n_<sub>i</sub>=i/N</tt>.		By default, the patches will be equally sized along each axis. If unequal patch sizes can improve parallel scaling, they can be created using <tt>Xpatches</tt>, <tt>Ypatches</tt>, or <tt>Zpatches</tt> elements. Each patches element must contain exactly <tt>inum-1</tt> (i = x, y, or z) monotonically increasing numbers between zero and one for the fractional break points between patches. If an axis has <tt>N</tt> elements, the <tt>k<sup>th</sup></tt> patch (for <tt>k</tt> from 1 to <tt>inum</tt>) will include elements <tt>int(n<sub>k-1</sub>/N)</tt> to <tt>int(n<sub>k</sub>/N)</tt> (where n<sub>0</sub> is set to 0 and n<sub>inum</sub> is set to 1). All patches mush have at least two elements or must have <tt>int(n<sub>k</sub>/N)</tt>><tt>int(n<sub>k-1</sub>/N)</tt>. The patches elements are optional. Any omitted axes will use equally sizes patches corresponding to <tt>n<sub>i</sub>=i/N</tt>.

	== Hyperthreading ==		== Hyperthreading ==

Nairnj: /* Custom Patching */

2023-01-26T20:41:12Z

Custom Patching

← Older revision		Revision as of 20:41, 26 January 2023
Line 24:		Line 24:
	<Xpatches>n<sub>1</sub>,n<sub>2</sub>,...n<sub>xnum-1</sub></Xpatches>		<Xpatches>n<sub>1</sub>,n<sub>2</sub>,...n<sub>xnum-1</sub></Xpatches>
	<Ypatches>n<sub>1</sub>,n<sub>2</sub>,...n<sub>ynum-1</sub></Ypatches>		<Ypatches>n<sub>1</sub>,n<sub>2</sub>,...n<sub>ynum-1</sub></Ypatches>
	<Zpatches>n<sub>1</sub>,n<sub>2</sub>,...n<sub>~~ynum~~-1</sub></Ypatches>		<Zpatches>n<sub>1</sub>,n<sub>2</sub>,...n<sub>znum-1</sub></Ypatches>
	</PatchGrid>		</PatchGrid>

Nairnj: /* Custom Patching */

2023-01-26T20:40:27Z

Custom Patching

← Older revision		Revision as of 20:40, 26 January 2023
Line 22:		Line 22:

	<PatchGrid x="xnum" y="ynum" z="znum">		<PatchGrid x="xnum" y="ynum" z="znum">
	<Xpatches>n<sub>1/sub>,n<sub>2</sub>,...n<sub>xnum-1</sub></Xpatches>		<Xpatches>n<sub>1</sub>,n<sub>2</sub>,...n<sub>xnum-1</sub></Xpatches>
	<Ypatches>n<sub>1/sub>,n<sub>2</sub>,...n<sub>ynum-1</sub></Ypatches>		<Ypatches>n<sub>1</sub>,n<sub>2</sub>,...n<sub>ynum-1</sub></Ypatches>
	<Zpatches>n<sub>1/sub>,n<sub>2</sub>,...n<sub>ynum-1</sub></Ypatches>		<Zpatches>n<sub>1</sub>,n<sub>2</sub>,...n<sub>ynum-1</sub></Ypatches>
	</PatchGrid>		</PatchGrid>

Nairnj: /* Custom Patching */

2023-01-26T20:40:13Z

Custom Patching

← Older revision		Revision as of 20:40, 26 January 2023
Line 19:		Line 19:
	Parallel calculations when running [[NairnMPM]] divide the orthogonal grid into orthogonal patches. 2D grids are divided into <tt>(xnum)X(ynum)</tt> patches and 3D grids are divided into <tt>(xnum)X(ynum)X(znum)</tt> where <tt>(inum)</tt> is number of patches along the <tt>i</tt> axis. The <tt>(inum)</tt> values are automatically determined by factoring the number of processors, <tt>p</tt>, with attention to putting more patches along longer grid axes. All patches will be the same size.		Parallel calculations when running [[NairnMPM]] divide the orthogonal grid into orthogonal patches. 2D grids are divided into <tt>(xnum)X(ynum)</tt> patches and 3D grids are divided into <tt>(xnum)X(ynum)X(znum)</tt> where <tt>(inum)</tt> is number of patches along the <tt>i</tt> axis. The <tt>(inum)</tt> values are automatically determined by factoring the number of processors, <tt>p</tt>, with attention to putting more patches along longer grid axes. All patches will be the same size.

	It the automatic patching calculations or the use of equally-sized patches is not ideal for a specific simulation, you can override that calculation with custom patching with the following <tt>XML</tt> element in the [[MPM_Input_Files#Main_Header\|MPM Header]].		If the automatic patching calculations or the use of equally-sized patches is not ideal for a specific simulation, you can override that calculation with custom patching with the following <tt>XML</tt> element in the [[MPM_Input_Files#Main_Header\|MPM Header]].

	<PatchGrid x="xnum" y="ynum" z="znum">		<PatchGrid x="xnum" y="ynum" z="znum">

Nairnj: /* Custom Patching */

2023-01-26T20:39:44Z

Custom Patching

← Older revision		Revision as of 20:39, 26 January 2023
Line 17:		Line 17:
	== Custom Patching ==		== Custom Patching ==

	Parallel calculations when running [[NairnMPM]] divide the orthogonal grid into orthogonal patches. 2D grids are divided into <tt>(xnum)X(ynum)</tt> patches and 3D grids are divided into <tt>(xnum)X(ynum)X(znum)</tt> where <tt>(inum)</tt> is number of patches along the <tt>i</tt> axis. The <tt>(inum)</tt> values are automatically determined by factoring the number of processors, <tt>p</tt>, with attention to ~~put~~ more patches along longer grid axes. All patches will be the same size.		Parallel calculations when running [[NairnMPM]] divide the orthogonal grid into orthogonal patches. 2D grids are divided into <tt>(xnum)X(ynum)</tt> patches and 3D grids are divided into <tt>(xnum)X(ynum)X(znum)</tt> where <tt>(inum)</tt> is number of patches along the <tt>i</tt> axis. The <tt>(inum)</tt> values are automatically determined by factoring the number of processors, <tt>p</tt>, with attention to putting more patches along longer grid axes. All patches will be the same size.

	It the automatic patching calculations or the use of equally-sized patches is not ideal for a specific simulation, you can override that calculation with custom patching with the following <tt>XML</tt> element in the [[MPM_Input_Files#Main_Header\|MPM Header]].		It the automatic patching calculations or the use of equally-sized patches is not ideal for a specific simulation, you can override that calculation with custom patching with the following <tt>XML</tt> element in the [[MPM_Input_Files#Main_Header\|MPM Header]].

Nairnj: /* Custom Patching */

2023-01-26T20:39:13Z

Custom Patching

← Older revision		Revision as of 20:39, 26 January 2023
Line 17:		Line 17:
	== Custom Patching ==		== Custom Patching ==

	Parallel calculations when running [[NairnMPM]] divide the orthogonal grid into orthogonal patches. 2D grids are divided ~~in the~~ <tt>(xnum)X(ynum)</tt> patches and 3D grids are divided into <tt>(xnum)X(ynum)X(znum)</tt> where <tt>(inum)</tt> is number of patches along the <tt>i</tt> axis. The <tt>(inum)</tt> values are automatically determined by factoring the number of processors, <tt>p</tt>, with attention to put more patches along longer grid axes. All patches will be the same size.		Parallel calculations when running [[NairnMPM]] divide the orthogonal grid into orthogonal patches. 2D grids are divided into <tt>(xnum)X(ynum)</tt> patches and 3D grids are divided into <tt>(xnum)X(ynum)X(znum)</tt> where <tt>(inum)</tt> is number of patches along the <tt>i</tt> axis. The <tt>(inum)</tt> values are automatically determined by factoring the number of processors, <tt>p</tt>, with attention to put more patches along longer grid axes. All patches will be the same size.

	It the automatic patching calculations or the use of equally-sized patches is not ideal for a specific simulation, you can override that calculation with custom patching with the following <tt>XML</tt> element in the [[MPM_Input_Files#Main_Header\|MPM Header]].		It the automatic patching calculations or the use of equally-sized patches is not ideal for a specific simulation, you can override that calculation with custom patching with the following <tt>XML</tt> element in the [[MPM_Input_Files#Main_Header\|MPM Header]].