[gem5-dev] Change in gem5/gem5[develop]: configs,gpu-compute: Fixes to connect gmTokenPort
Matthew Poremba has submitted this change. (
https://gem5-review.googlesource.com/c/public/gem5/+/35096 )
Change subject: configs,gpu-compute: Fixes to connect gmTokenPort
..
configs,gpu-compute: Fixes to connect gmTokenPort
When the TokenPort was moved from the GCN3 staging branch to develop the
TokenPort was changed from being the port connecting the ComputeUnit to
Ruby's vector memory port to a sideband port which inhibits requests to
Ruby's vector memory port. As such, it needs to be explicitly connected
as a new port. This changes the getPort method in ComputeUnit to be
aware of the port as well as modifying the example config to connect to
TCPs.
The iteration to connect in the config file was modified since it was
not properly connecting to TCPs each time and Ruby.py does not
explicitly return a list of each MachineType.
Change-Id: Ia70a6756b2af54d95e94d19bec5d8aadd3c2d5c0
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/35096
Reviewed-by: Matt Sinclair
Reviewed-by: Jason Lowe-Power
Maintainer: Matt Sinclair
Tested-by: kokoro
---
M configs/example/apu_se.py
M src/gpu-compute/compute_unit.hh
2 files changed, 12 insertions(+), 2 deletions(-)
Approvals:
Jason Lowe-Power: Looks good to me, but someone else must approve
Matt Sinclair: Looks good to me, approved; Looks good to me, approved
kokoro: Regressions pass
diff --git a/configs/example/apu_se.py b/configs/example/apu_se.py
index 2e24079..18640c8 100644
--- a/configs/example/apu_se.py
+++ b/configs/example/apu_se.py
@@ -566,6 +566,16 @@
- options.num_scalar_cache
gpu_port_idx = gpu_port_idx - options.num_cp * 2
+# Connect token ports. For this we need to search through the list of all
+# sequencers, since the TCP coalescers will not necessarily be first. Only
+# TCP coalescers use a token port for back pressure.
+token_port_idx = 0
+for i in range(len(system.ruby._cpu_ports)):
+if isinstance(system.ruby._cpu_ports[i], VIPERCoalescer):
+system.cpu[shader_idx].CUs[token_port_idx].gmTokenPort = \
+system.ruby._cpu_ports[i].gmTokenPort
+token_port_idx += 1
+
wavefront_size = options.wf_size
for i in range(n_cu):
# The pipeline issues wavefront_size number of uncoalesced requests
@@ -573,8 +583,6 @@
for j in range(wavefront_size):
system.cpu[shader_idx].CUs[i].memory_port[j] = \
system.ruby._cpu_ports[gpu_port_idx].slave[j]
-system.cpu[shader_idx].CUs[i].gmTokenPort = \
-system.ruby._cpu_ports[gpu_port_idx].gmTokenPort
gpu_port_idx += 1
for i in range(n_cu):
diff --git a/src/gpu-compute/compute_unit.hh
b/src/gpu-compute/compute_unit.hh
index fe2091d..2df4807 100644
--- a/src/gpu-compute/compute_unit.hh
+++ b/src/gpu-compute/compute_unit.hh
@@ -1015,6 +1015,8 @@
return sqcTLBPort;
} else if (if_name == "ldsPort") {
return ldsPort;
+} else if (if_name == "gmTokenPort") {
+return gmTokenPort;
} else {
return ClockedObject::getPort(if_name, idx);
}
--
To view, visit https://gem5-review.googlesource.com/c/public/gem5/+/35096
To unsubscribe, or for help writing mail filters, visit
https://gem5-review.googlesource.com/settings
Gerrit-Project: public/gem5
Gerrit-Branch: develop
Gerrit-Change-Id: Ia70a6756b2af54d95e94d19bec5d8aadd3c2d5c0
Gerrit-Change-Number: 35096
Gerrit-PatchSet: 4
Gerrit-Owner: Matthew Poremba
Gerrit-Reviewer: Jason Lowe-Power
Gerrit-Reviewer: Kyle Roarty
Gerrit-Reviewer: Matt Sinclair
Gerrit-Reviewer: Matthew Poremba
Gerrit-Reviewer: kokoro
Gerrit-MessageType: merged
___
gem5-dev mailing list -- gem5-dev@gem5.org
To unsubscribe send an email to gem5-dev-le...@gem5.org
%(web_page_url)slistinfo%(cgiext)s/%(_internal_name)s
[gem5-dev] Change in gem5/gem5[develop]: configs,gpu-compute: Fixes to connect gmTokenPort
Matthew Poremba has uploaded this change for review. (
https://gem5-review.googlesource.com/c/public/gem5/+/35096 )
Change subject: configs,gpu-compute: Fixes to connect gmTokenPort
..
configs,gpu-compute: Fixes to connect gmTokenPort
When the TokenPort was moved from the GCN3 staging branch to develop the
TokenPort was changes from being the port connecting the ComputeUnit to
Ruby's vector memory port to a sideband port which inhibits requests to
Ruby's vector memory port. As such, it needs to be explicitly connected
as a new port. This changes the getPort method in ComputeUnit to be
aware of the port as well as modifying the example config to connect to
TCPs.
The iteration to connect in the config file was modified since it was
not properly connecting to TCPs each time and Ruby.py does not
explicitly return a list of each MachineType.
Change-Id: Ia70a6756b2af54d95e94d19bec5d8aadd3c2d5c0
---
M configs/example/apu_se.py
M src/gpu-compute/compute_unit.hh
2 files changed, 13 insertions(+), 3 deletions(-)
diff --git a/configs/example/apu_se.py b/configs/example/apu_se.py
index 077ce4c..1391a22 100644
--- a/configs/example/apu_se.py
+++ b/configs/example/apu_se.py
@@ -174,7 +174,7 @@
help="number of physical banks per LDS module")
parser.add_option("--ldsBankConflictPenalty", type="int", default=1,
help="number of cycles per LDS bank conflict")
-parser.add_options("--lds-size", type="int", default=65536,
+parser.add_option("--lds-size", type="int", default=65536,
help="Size of the LDS in bytes")
parser.add_option('--fast-forward-pseudo-op', action='store_true',
help = 'fast forward using kvm until the m5_switchcpu'
@@ -566,6 +566,16 @@
- options.num_scalar_cache
gpu_port_idx = gpu_port_idx - options.num_cp * 2
+# Connect token ports. For this we need to search through the list of all
+# sequencers, since the TCP coalescers will not necessarily be first. Only
+# TCP coalescers use a token port for back pressure.
+token_port_idx = 0
+for i in range(len(system.ruby._cpu_ports)):
+if type(system.ruby._cpu_ports[i]).__name__ == "VIPERCoalescer":
+system.cpu[shader_idx].CUs[token_port_idx].gmTokenPort = \
+system.ruby._cpu_ports[i].gmTokenPort
+token_port_idx += 1
+
wavefront_size = options.wf_size
for i in range(n_cu):
# The pipeline issues wavefront_size number of uncoalesced requests
@@ -573,8 +583,6 @@
for j in range(wavefront_size):
system.cpu[shader_idx].CUs[i].memory_port[j] = \
system.ruby._cpu_ports[gpu_port_idx].slave[j]
-system.cpu[shader_idx].CUs[i].gmTokenPort = \
-system.ruby._cpu_ports[gpu_port_idx].gmTokenPort
gpu_port_idx += 1
for i in range(n_cu):
diff --git a/src/gpu-compute/compute_unit.hh
b/src/gpu-compute/compute_unit.hh
index fe2091d..2df4807 100644
--- a/src/gpu-compute/compute_unit.hh
+++ b/src/gpu-compute/compute_unit.hh
@@ -1015,6 +1015,8 @@
return sqcTLBPort;
} else if (if_name == "ldsPort") {
return ldsPort;
+} else if (if_name == "gmTokenPort") {
+return gmTokenPort;
} else {
return ClockedObject::getPort(if_name, idx);
}
--
To view, visit https://gem5-review.googlesource.com/c/public/gem5/+/35096
To unsubscribe, or for help writing mail filters, visit
https://gem5-review.googlesource.com/settings
Gerrit-Project: public/gem5
Gerrit-Branch: develop
Gerrit-Change-Id: Ia70a6756b2af54d95e94d19bec5d8aadd3c2d5c0
Gerrit-Change-Number: 35096
Gerrit-PatchSet: 1
Gerrit-Owner: Matthew Poremba
Gerrit-MessageType: newchange
___
gem5-dev mailing list -- gem5-dev@gem5.org
To unsubscribe send an email to gem5-dev-le...@gem5.org
%(web_page_url)slistinfo%(cgiext)s/%(_internal_name)s
