From 7ca2a7d7421c28422871dab575322bc45c848542 Mon Sep 17 00:00:00 2001
From: Hal Finkel <hfinkel@anl.gov>
Date: Sat, 23 Aug 2014 23:21:04 +0000
Subject: [PATCH] [PowerPC] Add support for dcbtst and icbt (prefetch)

Adds code generation support for dcbtst (data cache prefetch for write) and
icbt (instruction cache prefetch for read - Book E cores only).

We still end up with a 'cannot select' error for the non-supported prefetch
intrinsic forms. This will be fixed in a later commit.

Fixes PR20692.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@216339 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/PowerPC/PPCInstrFormats.td         | 15 +++++++++++
 lib/Target/PowerPC/PPCInstrInfo.td            |  9 ++++++-
 test/CodeGen/PowerPC/ppc64-prefetch.ll        | 25 ++++++++++++++++---
 .../PowerPC/ppc64-encoding-bookII.txt         |  3 +++
 test/MC/PowerPC/ppc64-encoding-bookII.s       |  4 +++
 5 files changed, 52 insertions(+), 4 deletions(-)

diff --git a/lib/Target/PowerPC/PPCInstrFormats.td b/lib/Target/PowerPC/PPCInstrFormats.td
index 2da4257d9f7..aa6849744d2 100644
--- a/lib/Target/PowerPC/PPCInstrFormats.td
+++ b/lib/Target/PowerPC/PPCInstrFormats.td
@@ -478,6 +478,21 @@ class XForm_16<bits<6> opcode, bits<10> xo, dag OOL, dag IOL, string asmstr,
   let Inst{31}    = 0;
 }
 
+class XForm_icbt<bits<6> opcode, bits<10> xo, dag OOL, dag IOL, string asmstr,
+                 InstrItinClass itin>
+         : I<opcode, OOL, IOL, asmstr, itin> {
+  bits<4> CT;
+  bits<5> RA;
+  bits<5> RB;
+
+  let Inst{6} = 0;
+  let Inst{7-10} = CT;
+  let Inst{11-15} = RA;
+  let Inst{16-20} = RB;
+  let Inst{21-30} = xo;
+  let Inst{31} = 0;
+}
+
 class XForm_sr<bits<6> opcode, bits<10> xo, dag OOL, dag IOL, string asmstr,
                 InstrItinClass itin>
          : I<opcode, OOL, IOL, asmstr, itin> {
diff --git a/lib/Target/PowerPC/PPCInstrInfo.td b/lib/Target/PowerPC/PPCInstrInfo.td
index 3b2c7cb3d81..38020efb58b 100644
--- a/lib/Target/PowerPC/PPCInstrInfo.td
+++ b/lib/Target/PowerPC/PPCInstrInfo.td
@@ -1303,8 +1303,15 @@ def DCBZL  : DCB_Form<1014, 1, (outs), (ins memrr:$dst), "dcbzl $dst",
                       IIC_LdStDCBF, [(int_ppc_dcbzl xoaddr:$dst)]>,
                       PPC970_DGroup_Single;
 
+def ICBT  : XForm_icbt<31, 22, (outs), (ins u4imm:$CT, memrr:$src),
+                       "icbt $CT, $src", IIC_LdStLoad>, Requires<[IsBookE]>;
+
 def : Pat<(prefetch xoaddr:$dst, (i32 0), imm, (i32 1)),
-          (DCBT xoaddr:$dst)>;
+          (DCBT xoaddr:$dst)>;   // data prefetch for loads
+def : Pat<(prefetch xoaddr:$dst, (i32 1), imm, (i32 1)),
+          (DCBTST xoaddr:$dst)>; // data prefetch for stores
+def : Pat<(prefetch xoaddr:$dst, (i32 0), imm, (i32 0)),
+          (ICBT 0, xoaddr:$dst)>; // inst prefetch (for read)
 
 // Atomic operations
 let usesCustomInserter = 1 in {
diff --git a/test/CodeGen/PowerPC/ppc64-prefetch.ll b/test/CodeGen/PowerPC/ppc64-prefetch.ll
index b2f37097f92..b2f6e7d6652 100644
--- a/test/CodeGen/PowerPC/ppc64-prefetch.ll
+++ b/test/CodeGen/PowerPC/ppc64-prefetch.ll
@@ -1,15 +1,34 @@
 target datalayout = "E-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v128:128:128-n32:64"
 target triple = "powerpc64-unknown-linux-gnu"
-; RUN: llc < %s | FileCheck %s
+; RUN: llc -mcpu=a2 < %s | FileCheck %s
 
 define void @test1(i8* %a, ...) nounwind {
 entry:
   call void @llvm.prefetch(i8* %a, i32 0, i32 3, i32 1)
   ret void
+
+; CHECK-LABEL: @test1
+; CHECK: dcbt
 }
 
 declare void @llvm.prefetch(i8*, i32, i32, i32)
 
-; CHECK: @test1
-; CHECK: dcbt
+define void @test2(i8* %a, ...) nounwind {
+entry:
+  call void @llvm.prefetch(i8* %a, i32 1, i32 3, i32 1)
+  ret void
+
+; CHECK-LABEL: @test2
+; CHECK: dcbtst
+}
+
+define void @test3(i8* %a, ...) nounwind {
+entry:
+  call void @llvm.prefetch(i8* %a, i32 0, i32 3, i32 0)
+  ret void
+
+; CHECK-LABEL: @test3
+; CHECK: icbt
+}
+
 
diff --git a/test/MC/Disassembler/PowerPC/ppc64-encoding-bookII.txt b/test/MC/Disassembler/PowerPC/ppc64-encoding-bookII.txt
index f0ef9024648..7a30b5cb2b6 100644
--- a/test/MC/Disassembler/PowerPC/ppc64-encoding-bookII.txt
+++ b/test/MC/Disassembler/PowerPC/ppc64-encoding-bookII.txt
@@ -3,6 +3,9 @@
 # CHECK: icbi 2, 3                       
 0x7c 0x02 0x1f 0xac
 
+# CHECK: icbt 0, 5, 31
+0x7c 0x05 0xf8 0x2c
+
 # CHECK: dcbt 2, 3                       
 0x7c 0x02 0x1a 0x2c
 
diff --git a/test/MC/PowerPC/ppc64-encoding-bookII.s b/test/MC/PowerPC/ppc64-encoding-bookII.s
index fea2a3e46f2..20eba7083f6 100644
--- a/test/MC/PowerPC/ppc64-encoding-bookII.s
+++ b/test/MC/PowerPC/ppc64-encoding-bookII.s
@@ -8,6 +8,10 @@
 # CHECK-LE: icbi 2, 3                       # encoding: [0xac,0x1f,0x02,0x7c]
             icbi 2, 3
 
+# CHECK-BE: icbt 0, 5, 31                   # encoding: [0x7c,0x05,0xf8,0x2c]
+# CHECK-LE: icbt 0, 5, 31                   # encoding: [0x2c,0xf8,0x05,0x7c]
+            icbt 0, 5, 31
+
 # FIXME:    dcbt 2, 3, 10
 # CHECK-BE: dcbt 2, 3                       # encoding: [0x7c,0x02,0x1a,0x2c]
 # CHECK-LE: dcbt 2, 3                       # encoding: [0x2c,0x1a,0x02,0x7c]