[llvm] r292985 - AMDGPU: Remove spurious out branches after a kill

[Matt Arsenault via llvm-commits]

Author: arsenm
Date: Tue Jan 24 16:18:39 2017
New Revision: 292985

URL: http://llvm.org/viewvc/llvm-project?rev=292985&view=rev
Log:
AMDGPU: Remove spurious out branches after a kill

The sequence like this:
  v_cmpx_le_f32_e32 vcc, 0, v0
  s_branch BB0_30
  s_cbranch_execnz BB0_30
  ; BB#29:
  exp null off, off, off, off done vm
  s_endpgm
  BB0_30:
  ; %endif110

is likely wrong. The s_branch instruction will unconditionally jump
to BB0_30 and the skip block (exp done + endpgm) inserted for
performing the kill instruction will never be executed. This results
in a GPU hang with Star Ruler 2.

The s_branch instruction is added during the "Control Flow Optimizer"
pass which seems to re-organize the basic blocks, and we assume
that SI_KILL_TERMINATOR is always the last instruction inside a
basic block. Thus, after inserting a skip block we just go to the
next BB without looking at the subsequent instructions after the
kill, and the s_branch op is never removed.

Instead, we should remove the unconditional out branches and let
skip the two instructions if the exec mask is non-zero.

This patch fixes the GPU hang and doesn't introduce any regressions
with "make check".

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99019

Patch by Samuel Pitoiset <samuel.pitoiset at gmail.com>

Added:
    llvm/trunk/test/CodeGen/AMDGPU/insert-skips-kill-uncond.mir
Modified:
    llvm/trunk/lib/Target/AMDGPU/SIInsertSkips.cpp

Modified: llvm/trunk/lib/Target/AMDGPU/SIInsertSkips.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/AMDGPU/SIInsertSkips.cpp?rev=292985&r1=292984&r2=292985&view=diff
==============================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIInsertSkips.cpp (original)
+++ llvm/trunk/lib/Target/AMDGPU/SIInsertSkips.cpp Tue Jan 24 16:18:39 2017
@@ -263,6 +263,7 @@ bool SIInsertSkips::runOnMachineFunction
        BI != BE; BI = NextBB) {
     NextBB = std::next(BI);
     MachineBasicBlock &MBB = *BI;
+    bool HaveSkipBlock = false;
 
     if (!ExecBranchStack.empty() && ExecBranchStack.back() == &MBB) {
       // Reached convergence point for last divergent branch.
@@ -290,8 +291,14 @@ bool SIInsertSkips::runOnMachineFunction
       case AMDGPU::S_BRANCH:
         // Optimize out branches to the next block.
         // FIXME: Shouldn't this be handled by BranchFolding?
-        if (MBB.isLayoutSuccessor(MI.getOperand(0).getMBB()))
+        if (MBB.isLayoutSuccessor(MI.getOperand(0).getMBB())) {
           MI.eraseFromParent();
+        } else if (HaveSkipBlock) {
+          // Remove the given unconditional branch when a skip block has been
+          // inserted after the current one and let skip the two instructions
+          // performing the kill if the exec mask is non-zero.
+          MI.eraseFromParent();
+        }
         break;
 
       case AMDGPU::SI_KILL_TERMINATOR:
@@ -300,9 +307,9 @@ bool SIInsertSkips::runOnMachineFunction
 
         if (ExecBranchStack.empty()) {
           if (skipIfDead(MI, *NextBB)) {
+            HaveSkipBlock = true;
             NextBB = std::next(BI);
             BE = MF.end();
-            Next = MBB.end();
           }
         } else {
           HaveKill = true;

Added: llvm/trunk/test/CodeGen/AMDGPU/insert-skips-kill-uncond.mir
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AMDGPU/insert-skips-kill-uncond.mir?rev=292985&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/AMDGPU/insert-skips-kill-uncond.mir (added)
+++ llvm/trunk/test/CodeGen/AMDGPU/insert-skips-kill-uncond.mir Tue Jan 24 16:18:39 2017
@@ -0,0 +1,40 @@
+# RUN: llc -march=amdgcn -mcpu=polaris10 -run-pass si-insert-skips -amdgpu-skip-threshold=1 %s -o - | FileCheck %s
+# https://bugs.freedesktop.org/show_bug.cgi?id=99019
+--- |
+  define amdgpu_ps void @kill_uncond_branch() {
+    ret void
+  }
+...
+---
+
+# CHECK-LABEL: name: kill_uncond_branch
+
+# CHECK: bb.0:
+# CHECK: S_CBRANCH_VCCNZ %bb.1, implicit %vcc
+
+# CHECK: bb.1:
+# CHECK: V_CMPX_LE_F32_e32
+# CHECK-NEXT: S_CBRANCH_EXECNZ %bb.2, implicit %exec
+
+# CHECK: bb.3:
+# CHECK-NEXT: EXP_DONE
+# CHECK: S_ENDPGM
+
+# CHECK: bb.2:
+# CHECK: S_ENDPGM
+
+name: kill_uncond_branch
+
+body: |
+  bb.0:
+    successors: %bb.1
+    S_CBRANCH_VCCNZ %bb.1, implicit %vcc
+
+  bb.1:
+    successors: %bb.2
+    %vgpr0 = V_MOV_B32_e32 0, implicit %exec
+    SI_KILL_TERMINATOR %vgpr0, implicit-def %exec, implicit-def %vcc, implicit %exec
+    S_BRANCH %bb.2
+
+  bb.2:
+    S_ENDPGM