From: Kyle Roarty <kyleroarty1716@gmail.com>
Date: Sat, 14 Nov 2020 19:42:12 +0000 (-0600)
Subject: gpu-compute: Add exp_cnt tracking for buffer store instructions
X-Git-Tag: develop-gem5-snapshot~404
X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=0bb385941b1b4bda9e3d73e76be41b2cc10bdf8a;p=gem5.git

gpu-compute: Add exp_cnt tracking for buffer store instructions

exp_cnt (expInstsIssued in the code) is used in the waitcnt instruction
to track that data has been read out of VGPRs in previous global
memory instructions, making it safe to overwrite the VGPRs used in said
global memory instructions.

Previously, exp_cnt wasn't being tracked at all, which lead to the
waitcnt finishing immediately, leading to the memory instruction's VPGRs
getting overwritten by subsequent instructions, causing errors.

This patch makes it so waitcnts waiting on exp_cnt will wait for MUBUF
buffer store instructions to read their VGPRs before completing

Change-Id: Idd2b59511bc086cf316217da27b7a228272b0b0f
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/37555
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Reviewed-by: Alexandru Duțu <alexandru.dutu@amd.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
---

diff --git a/src/gpu-compute/global_memory_pipeline.cc b/src/gpu-compute/global_memory_pipeline.cc
index 01f986c25..bcd93f886 100644
--- a/src/gpu-compute/global_memory_pipeline.cc
+++ b/src/gpu-compute/global_memory_pipeline.cc
@@ -193,6 +193,10 @@ GlobalMemPipeline::exec()
                 mp->disassemble(), mp->seqNum());
         mp->initiateAcc(mp);
 
+        if (mp->isStore() && mp->isGlobalSeg()) {
+            mp->wavefront()->decExpInstsIssued();
+        }
+
         if (((mp->isMemSync() && !mp->isEndOfKernel()) || !mp->isMemSync())) {
             /**
              * if we are not in out-of-order data delivery mode
diff --git a/src/gpu-compute/schedule_stage.cc b/src/gpu-compute/schedule_stage.cc
index 851cca81b..54e931306 100644
--- a/src/gpu-compute/schedule_stage.cc
+++ b/src/gpu-compute/schedule_stage.cc
@@ -144,6 +144,9 @@ ScheduleStage::exec()
                     wf->incLGKMInstsIssued();
                 }
             }
+            if (gpu_dyn_inst->isStore() && gpu_dyn_inst->isGlobalSeg()) {
+                wf->incExpInstsIssued();
+            }
         }
     }