[PATCH] D19428: AMDGPU: Define priorities for register classes

Matthias Braun via llvm-commits llvm-commits at lists.llvm.org
Fri Apr 22 12:03:57 PDT 2016


MatzeB created this revision.
MatzeB added reviewers: tstellarAMD, arsenm.
MatzeB added a subscriber: llvm-commits.
MatzeB set the repository for this revision to rL LLVM.
Herald added subscribers: mcrosier, MatzeB.

Allocating larger register classes first should give better allocation results (and more importantly for myself, make the lit tests more stable with respect to scheduler changes).

This is a suggestion to work on, I did not bother to fix the lit tests.

Repository:
  rL LLVM

http://reviews.llvm.org/D19428

Files:
  lib/CodeGen/MachineScheduler.cpp
  test/CodeGen/AArch64/arm64-misched-basic-A53.ll

Index: test/CodeGen/AArch64/arm64-misched-basic-A53.ll
===================================================================
--- test/CodeGen/AArch64/arm64-misched-basic-A53.ll
+++ test/CodeGen/AArch64/arm64-misched-basic-A53.ll
@@ -1,5 +1,6 @@
 ; REQUIRES: asserts
 ; RUN: llc < %s -mtriple=arm64-linux-gnu -mcpu=cortex-a53 -pre-RA-sched=source -enable-misched -verify-misched -debug-only=misched -o - 2>&1 > /dev/null | FileCheck %s
+; RUN: llc < %s -mtriple=arm64-linux-gnu -mcpu=cortex-a53 -pre-RA-sched=source -enable-misched -verify-misched -debug-only=misched -o - -misched-limit=2 2>&1 > /dev/null | FileCheck %s
 ;
 ; The Cortex-A53 machine model will cause the MADD instruction to be scheduled
 ; much higher than the ADD instructions in order to hide latency. When not
Index: lib/CodeGen/MachineScheduler.cpp
===================================================================
--- lib/CodeGen/MachineScheduler.cpp
+++ lib/CodeGen/MachineScheduler.cpp
@@ -64,6 +64,11 @@
 static bool ViewMISchedDAGs = false;
 #endif // NDEBUG
 
+/// Avoid quadratic complexity in unusually large basic blocks by limiting the
+/// size of the ready lists.
+static cl::opt<unsigned> ReadyListLimit("misched-limit", cl::Hidden,
+  cl::desc("Limit ready list to N instructions"), cl::init(256));
+
 static cl::opt<bool> EnableRegPressure("misched-regpressure", cl::Hidden,
   cl::desc("Enable register pressure scheduling."), cl::init(true));
 
@@ -1956,7 +1961,8 @@
   // Check for interlocks first. For the purpose of other heuristics, an
   // instruction that cannot issue appears as if it's not in the ReadyQueue.
   bool IsBuffered = SchedModel->getMicroOpBufferSize() != 0;
-  if ((!IsBuffered && ReadyCycle > CurrCycle) || checkHazard(SU))
+  if ((!IsBuffered && ReadyCycle > CurrCycle) || checkHazard(SU) ||
+      Available.size() >= ReadyListLimit)
     Pending.push(SU);
   else
     Available.push(SU);
@@ -2211,6 +2217,9 @@
     if (checkHazard(SU))
       continue;
 
+    if (Available.size() >= ReadyListLimit)
+      break;
+
     Available.push(SU);
     Pending.remove(Pending.begin()+i);
     --i; --e;


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D19428.54697.patch
Type: text/x-patch
Size: 2112 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160422/cb66d7e6/attachment.bin>


More information about the llvm-commits mailing list