VISHNU RAMADAS has submitted this change. (
https://gem5-review.googlesource.com/c/public/gem5/+/65511?usp=email )
Change subject: configs, gpu-compute: Add configurable L1 scalar latencies
......................................................................
configs, gpu-compute: Add configurable L1 scalar latencies
Previously the scalar cache path used the same latency parameter as the
vector cache path for memory requests. This commit adds new parameters
for the scalar cache path latencies. This commit also modifies the model
to use the new latency parameter to set the memory request latency in
the scalar cache. The new paramters are '--scalar-mem-req-latency' and
'--scalar-mem-resp-latency' and are set to default values of 50 and 0
respectively
Change-Id: I7483f780f2fc0cfbc320ed1fd0c2ee3e2dfc7af2
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/65511
Reviewed-by: Matt Sinclair <[email protected]>
Maintainer: Jason Lowe-Power <[email protected]>
Reviewed-by: Jason Lowe-Power <[email protected]>
Tested-by: kokoro <[email protected]>
Maintainer: Matt Sinclair <[email protected]>
---
M configs/example/apu_se.py
M src/gpu-compute/GPU.py
M src/gpu-compute/compute_unit.cc
M src/gpu-compute/compute_unit.hh
4 files changed, 60 insertions(+), 1 deletion(-)
Approvals:
Matt Sinclair: Looks good to me, approved; Looks good to me, approved
kokoro: Regressions pass
Jason Lowe-Power: Looks good to me, approved; Looks good to me, approved
diff --git a/configs/example/apu_se.py b/configs/example/apu_se.py
index b33daa5..39def02 100644
--- a/configs/example/apu_se.py
+++ b/configs/example/apu_se.py
@@ -276,6 +276,21 @@
help="Latency for responses from ruby to the cu.",
)
parser.add_argument(
+ "--scalar-mem-req-latency",
+ type=int,
+ default=50,
+ help="Latency for scalar requests from the cu to ruby.",
+)
+parser.add_argument(
+ "--scalar-mem-resp-latency",
+ type=int,
+ # Set to 0 as the scalar cache response path does not model
+ # response latency yet and this parameter is currently not used
+ default=0,
+ help="Latency for scalar responses from ruby to the cu.",
+)
+
+parser.add_argument(
"--TLB-prefetch", type=int, help="prefetch depth for" "TLBs"
)
parser.add_argument(
@@ -463,6 +478,8 @@
vrf_lm_bus_latency=args.vrf_lm_bus_latency,
mem_req_latency=args.mem_req_latency,
mem_resp_latency=args.mem_resp_latency,
+ scalar_mem_req_latency=args.scalar_mem_req_latency,
+ scalar_mem_resp_latency=args.scalar_mem_resp_latency,
localDataStore=LdsState(
banks=args.numLdsBanks,
bankConflictPenalty=args.ldsBankConflictPenalty,
diff --git a/src/gpu-compute/GPU.py b/src/gpu-compute/GPU.py
index 517d180..0fdc0b7 100644
--- a/src/gpu-compute/GPU.py
+++ b/src/gpu-compute/GPU.py
@@ -178,6 +178,19 @@
"TCP and cu as well as TCP data array "
"access. Specified in GPU clock cycles",
)
+ scalar_mem_req_latency = Param.Int(
+ 50,
+ "Latency for scalar requests from the cu to ruby. "
+ "Represents the pipeline to reach the TCP "
+ "and specified in GPU clock cycles",
+ )
+ scalar_mem_resp_latency = Param.Int(
+ 50,
+ "Latency for scalar responses from ruby to the "
+ "cu. Represents the pipeline between the "
+ "TCP and cu as well as TCP data array "
+ "access. Specified in GPU clock cycles",
+ )
system = Param.System(Parent.any, "system object")
cu_id = Param.Int("CU id")
vrf_to_coalescer_bus_width = Param.Int(
diff --git a/src/gpu-compute/compute_unit.cc
b/src/gpu-compute/compute_unit.cc
index 8498ea4..62cfbf9 100644
--- a/src/gpu-compute/compute_unit.cc
+++ b/src/gpu-compute/compute_unit.cc
@@ -98,6 +98,10 @@
countPages(p.countPages),
req_tick_latency(p.mem_req_latency * p.clk_domain->clockPeriod()),
resp_tick_latency(p.mem_resp_latency * p.clk_domain->clockPeriod()),
+ scalar_req_tick_latency(
+ p.scalar_mem_req_latency * p.clk_domain->clockPeriod()),
+ scalar_resp_tick_latency(
+ p.scalar_mem_resp_latency * p.clk_domain->clockPeriod()),
_requestorId(p.system->getRequestorId(this, "ComputeUnit")),
lds(*p.localDataStore), gmTokenPort(name() + ".gmTokenPort", this),
ldsPort(csprintf("%s-port", name()), this),
@@ -1786,7 +1790,7 @@
= new ComputeUnit::ScalarDataPort::MemReqEvent
(computeUnit->scalarDataPort, req_pkt);
computeUnit->schedule(scalar_mem_req_event, curTick() +
- computeUnit->req_tick_latency);
+ computeUnit->scalar_req_tick_latency);
return true;
}
diff --git a/src/gpu-compute/compute_unit.hh
b/src/gpu-compute/compute_unit.hh
index a080e3d..fcc4468 100644
--- a/src/gpu-compute/compute_unit.hh
+++ b/src/gpu-compute/compute_unit.hh
@@ -354,6 +354,8 @@
Tick req_tick_latency;
Tick resp_tick_latency;
+ Tick scalar_req_tick_latency;
+ Tick scalar_resp_tick_latency;
/**
* Number of WFs to schedule to each SIMD. This vector is populated
--
To view, visit
https://gem5-review.googlesource.com/c/public/gem5/+/65511?usp=email
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: I7483f780f2fc0cfbc320ed1fd0c2ee3e2dfc7af2
Gerrit-Change-Number: 65511
Gerrit-PatchSet: 5
Gerrit-Owner: VISHNU RAMADAS <[email protected]>
Gerrit-Reviewer: Jason Lowe-Power <[email protected]>
Gerrit-Reviewer: Jason Lowe-Power <[email protected]>
Gerrit-Reviewer: Matt Sinclair <[email protected]>
Gerrit-Reviewer: Matthew Poremba <[email protected]>
Gerrit-Reviewer: VISHNU RAMADAS <[email protected]>
Gerrit-Reviewer: kokoro <[email protected]>
Gerrit-MessageType: merged
_______________________________________________
gem5-dev mailing list -- [email protected]
To unsubscribe send an email to [email protected]
