[PATCH] D75958: [AMDGPU] Disable nested endcf collapse

[Stanislav Mekhanoshin via Phabricator via llvm-commits]

This revision was automatically updated to reflect the committed changes.
Closed by commit rG9801e5469b4e: [AMDGPU] Disable nested endcf collapse (authored by rampitec).

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D75958/new/

https://reviews.llvm.org/D75958

Files:
  llvm/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp
  llvm/test/CodeGen/AMDGPU/collapse-endcf.ll
  llvm/test/CodeGen/AMDGPU/collapse-endcf.mir


Index: llvm/test/CodeGen/AMDGPU/collapse-endcf.mir
===================================================================
--- llvm/test/CodeGen/AMDGPU/collapse-endcf.mir
+++ llvm/test/CodeGen/AMDGPU/collapse-endcf.mir
@@ -1,5 +1,5 @@
 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -march=amdgcn -verify-machineinstrs -run-pass=si-optimize-exec-masking-pre-ra %s -o - | FileCheck -check-prefix=GCN %s
+# RUN: llc -march=amdgcn -verify-machineinstrs -run-pass=si-optimize-exec-masking-pre-ra -amdgpu-remove-redundant-endcf %s -o - | FileCheck -check-prefix=GCN %s
 
 # Make sure dbg_value doesn't change codeegn when collapsing end_cf
 ---
Index: llvm/test/CodeGen/AMDGPU/collapse-endcf.ll
===================================================================
--- llvm/test/CodeGen/AMDGPU/collapse-endcf.ll
+++ llvm/test/CodeGen/AMDGPU/collapse-endcf.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,ALL %s
+; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -amdgpu-remove-redundant-endcf < %s | FileCheck -enable-var-scope -check-prefixes=GCN,ALL %s
 ; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -amdgpu-opt-exec-mask-pre-ra=0 < %s | FileCheck -enable-var-scope -check-prefixes=DISABLED,ALL %s
 
 ; ALL-LABEL: {{^}}simple_nested_if:
Index: llvm/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp
+++ llvm/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp
@@ -31,6 +31,10 @@
 
 #define DEBUG_TYPE "si-optimize-exec-masking-pre-ra"
 
+static cl::opt<bool>
+RemoveRedundantEndcf("amdgpu-remove-redundant-endcf",
+    cl::init(false), cl::ReallyHidden);
+
 namespace {
 
 class SIOptimizeExecMaskingPreRA : public MachineFunctionPass {
@@ -379,7 +383,13 @@
       continue;
     }
 
+    if (!RemoveRedundantEndcf)
+      continue;
+
     // Try to collapse adjacent endifs.
+    // The assumption is that conditional regions are perfectly nested and
+    // a mask restored at the exit from the inner block will be completely
+    // covered by a mask restored in the outer.
     auto E = MBB.end();
     auto Lead = skipDebugInstructionsForward(MBB.begin(), E);
     if (MBB.succ_size() != 1 || Lead == E || !isEndCF(*Lead, TRI, ST))


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D75958.249703.patch
Type: text/x-patch
Size: 2404 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20200311/a240dcb1/attachment.bin>