[AMDGPU] Make s_setprio_inc_wg a scheduling boundary

[rampitec]

No description provided.

[rampitec]

• [AMDGPU] Make s_setprio_inc_wg a scheduling boundary #154188 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[llvmbot]

@llvm/pr-subscribers-backend-amdgpu

Author: Stanislav Mekhanoshin (rampitec)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/154188.diff

1 Files Affected:

• (modified) llvm/lib/Target/AMDGPU/SIInstrInfo.cpp (+1)

diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
index 41885e45b4101..1f3943f6e1b27 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -4241,6 +4241,7 @@ bool SIInstrInfo::isSchedulingBoundary(const MachineInstr &MI,
          MI.getOpcode() == AMDGPU::S_SETREG_IMM32_B32 ||
          MI.getOpcode() == AMDGPU::S_SETREG_B32 ||
          MI.getOpcode() == AMDGPU::S_SETPRIO ||
+         MI.getOpcode() == AMDGPU::S_SETPRIO_INC_WG ||
          changesVGPRIndexingMode(MI);
 }
 

[shiltian]

no test?

[rampitec]

no test?

That one is hard to test like most of the scheduling, but basically an user of it wants to have a portion of the code to execute with a different priority, so we do not want to mix the code into that portion. That is similar to the s_setprio instruction just above.