AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add gfx908 VM model (#418)

Browse Source
This commit is contained in:
Wenkai Du
2021-08-10 08:55:11 -07:00
committed by GitHub
parent 1ed272e5f0
commit 5f15ed6e3e
4 changed files with 83 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/graph/rome_models.cc
+12
View File
@@ -294,6 +294,17 @@ static struct rcclRomeModel rome_model_43 = {
.ringBase = "0 1 2 3 4 5 6 7|0 2 5 7 4 6 1 3|0 3 1 6 4 7 5 2|0 7 6 5 4 3 2 1",
};
static struct rcclRomeModel rome_model_55 = {
.nGpus = 8, .nCpus = 4, .nNics = 0, .nLinks = 3,
.gpuIds = { 0x100000, 0x200000, 0x300000, 0x400000, 0x500000, 0x600000, 0x700000, 0x800000, },
.nicIds = { },
.gpuNuma = { 0, 0, 1, 1, 2, 2, 3, 3, },
.nicNuma = { },
.connMatrix = { 0, 1, 1, 1, 0, 0, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, 0, 0, 1, 1, 1, 0, },
.pattern = "20202020",
.ringBase = "0 1 2 3 4 5 6 7|7 6 5 4 3 2 1 0|2 3 0 1 6 7 4 5|5 4 7 6 1 0 3 2",
};
static struct rcclRomeModel romeTopoModels[] = {
rome_model_22,
rome_model_25,
@@ -318,6 +329,7 @@ static struct rcclRomeModel romeTopoModels[] = {
rome_model_52,
rome_model_53,
rome_model_43,
rome_model_55,
};
/* Parse user defined rings. Format is like :
tools/scripts/topo_val.sh
+1 -1
View File
@@ -21,7 +21,7 @@
DIR="$(cd -P "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
for i in {0..54}
for i in {0..55}
do
if [[ $i -eq 50 ]] || [[ $i -eq 51 ]]
then
tools/topo_expl/models/topo_8p_rome_vm1.xml
+69
View File
@@ -0,0 +1,69 @@
<system version="2">
<cpu numaid="0" affinity="000000,00000000,00000000,3fffffff" arch="x86_64" vendor="AuthenticAMD" familyid="143" modelid="49">
<pci busid="0001:00:00.0" class="0x038000" vendor="0x1002" device="0x738c" subsystem_vendor="0x1002" subsystem_device="0x0c34" link_speed="16 GT/s" link_width="16">
<gpu dev="0" sm="90" gcn="908" arch="38911" rank="0" gdr="0">
<xgmi target="0002:00:00.0" count="1" tclass="0x038000"/>
<xgmi target="0003:00:00.0" count="1" tclass="0x038000"/>
<xgmi target="0004:00:00.0" count="1" tclass="0x038000"/>
</gpu>
</pci>
<pci busid="0002:00:00.0" class="0x038000" vendor="0x1002" device="0x738c" subsystem_vendor="0x1002" subsystem_device="0x0c34" link_speed="16 GT/s" link_width="16">
<gpu dev="1" sm="90" gcn="908" arch="38911" rank="1" gdr="0">
<xgmi target="0001:00:00.0" count="1" tclass="0x038000"/>
<xgmi target="0003:00:00.0" count="1" tclass="0x038000"/>
<xgmi target="0004:00:00.0" count="1" tclass="0x038000"/>
</gpu>
</pci>
<nic>
<net name="eth0" dev="0" speed="100000" port="0" guid="0x0" maxconn="65536" gdr="0"/>
</nic>
</cpu>
<cpu numaid="1" affinity="000000,00000000,0fffffff,c0000000" arch="x86_64" vendor="AuthenticAMD" familyid="143" modelid="49">
<pci busid="0003:00:00.0" class="0x038000" vendor="0x1002" device="0x738c" subsystem_vendor="0x1002" subsystem_device="0x0c34" link_speed="16 GT/s" link_width="16">
<gpu dev="2" sm="90" gcn="908" arch="38911" rank="2" gdr="0">
<xgmi target="0001:00:00.0" count="1" tclass="0x038000"/>
<xgmi target="0002:00:00.0" count="1" tclass="0x038000"/>
<xgmi target="0004:00:00.0" count="1" tclass="0x038000"/>
</gpu>
</pci>
<pci busid="0004:00:00.0" class="0x038000" vendor="0x1002" device="0x738c" subsystem_vendor="0x1002" subsystem_device="0x0c34" link_speed="16 GT/s" link_width="16">
<gpu dev="3" sm="90" gcn="908" arch="38911" rank="3" gdr="0">
<xgmi target="0001:00:00.0" count="1" tclass="0x038000"/>
<xgmi target="0002:00:00.0" count="1" tclass="0x038000"/>
<xgmi target="0003:00:00.0" count="1" tclass="0x038000"/>
</gpu>
</pci>
</cpu>
<cpu numaid="2" affinity="000000,03ffffff,f0000000,00000000" arch="x86_64" vendor="AuthenticAMD" familyid="143" modelid="49">
<pci busid="0005:00:00.0" class="0x038000" vendor="0x1002" device="0x738c" subsystem_vendor="0x1002" subsystem_device="0x0c34" link_speed="16 GT/s" link_width="16">
<gpu dev="4" sm="90" gcn="908" arch="38911" rank="4" gdr="0">
<xgmi target="0006:00:00.0" count="1" tclass="0x038000"/>
<xgmi target="0007:00:00.0" count="1" tclass="0x038000"/>
<xgmi target="0008:00:00.0" count="1" tclass="0x038000"/>
</gpu>
</pci>
<pci busid="0006:00:00.0" class="0x038000" vendor="0x1002" device="0x738c" subsystem_vendor="0x1002" subsystem_device="0x0c34" link_speed="16 GT/s" link_width="16">
<gpu dev="5" sm="90" gcn="908" arch="38911" rank="5" gdr="0">
<xgmi target="0005:00:00.0" count="1" tclass="0x038000"/>
<xgmi target="0007:00:00.0" count="1" tclass="0x038000"/>
<xgmi target="0008:00:00.0" count="1" tclass="0x038000"/>
</gpu>
</pci>
</cpu>
<cpu numaid="3" affinity="ffffff,fc000000,00000000,00000000" arch="x86_64" vendor="AuthenticAMD" familyid="143" modelid="49">
<pci busid="0007:00:00.0" class="0x038000" vendor="0x1002" device="0x738c" subsystem_vendor="0x1002" subsystem_device="0x0c34" link_speed="16 GT/s" link_width="16">
<gpu dev="6" sm="90" gcn="908" arch="38911" rank="6" gdr="0">
<xgmi target="0005:00:00.0" count="1" tclass="0x038000"/>
<xgmi target="0006:00:00.0" count="1" tclass="0x038000"/>
<xgmi target="0008:00:00.0" count="1" tclass="0x038000"/>
</gpu>
</pci>
<pci busid="0008:00:00.0" class="0x038000" vendor="0x1002" device="0x738c" subsystem_vendor="0x1002" subsystem_device="0x0c34" link_speed="16 GT/s" link_width="16">
<gpu dev="7" sm="90" gcn="908" arch="38911" rank="7" gdr="0">
<xgmi target="0005:00:00.0" count="1" tclass="0x038000"/>
<xgmi target="0006:00:00.0" count="1" tclass="0x038000"/>
<xgmi target="0007:00:00.0" count="1" tclass="0x038000"/>
</gpu>
</pci>
</cpu>
</system>
tools/topo_expl/topo_expl.cpp
+1
View File
@@ -124,6 +124,7 @@ NodeModelDesc model_descs[] = {
{1, "topo_8p_90a.xml", "single node gfx90a"},
{4, "topo_8p_rome_4n_1.xml", "4 nodes node 8 gfx908 Rome 4 NICs NPS=4"},
{4, "topo_collnet_n4.xml", "4 nodes collnet 4 NICs for multiple SAT"},
{1, "topo_8p_rome_vm1.xml", "single node 8 gfx908 Rome VM"},
};
int main(int argc,char* argv[])