[llvm] [AMDGPU] Add cl option to relax lds dma waitcnt (PR #131842)

[Austin Kerbow via llvm-commits]

https://github.com/kerbowa created https://github.com/llvm/llvm-project/pull/131842

None

>From d491d2b02a40a8aedc1e54c2b2c8fe1bfb8e8cf1 Mon Sep 17 00:00:00 2001
From: Austin Kerbow <Austin.Kerbow at amd.com>
Date: Fri, 14 Mar 2025 17:44:37 -0700
Subject: [PATCH] [AMDGPU] Add cl option to relax lds dma waitcnt

---
 llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp b/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
index 239f2664f59f3..10316bf6eb872 100644
--- a/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
@@ -58,6 +58,11 @@ static cl::opt<bool> ForceEmitZeroLoadFlag(
     cl::desc("Force all waitcnt load counters to wait until 0"),
     cl::init(false), cl::Hidden);
 
+static cl::opt<bool>
+    RelaxLDSDMA("amdgpu-relax-lds-dma-waitcnt",
+                cl::desc("Relax the waitcnt for LDS DMA instructions"),
+                cl::init(false), cl::Hidden);
+
 namespace {
 // Class of object that encapsulates latest instruction counter score
 // associated with the operand.  Used for determining whether
@@ -1748,7 +1753,7 @@ bool SIInsertWaitcnts::generateWaitcntInstBefore(MachineInstr &MI,
             }
           }
         }
-        if (!FoundAliasingStore)
+        if (!FoundAliasingStore || RelaxLDSDMA)
           ScoreBrackets.determineWait(LOAD_CNT, RegNo, Wait);
         if (Memop->isStore()) {
           ScoreBrackets.determineWait(EXP_CNT, RegNo, Wait);