From: Ahmed Bougacha <ahmed.bougacha@gmail.com>
Date: Thu, 6 Nov 2014 22:04:15 +0000 (+0000)
Subject: [X86] Add VFMADDSUB cases for the 213->231 custom inserter.
X-Git-Tag: llvmorg-3.6.0-rc1~4141
X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=b5367eeea3bcbd82b09547a49321a29f53dffafb;p=platform%2Fupstream%2Fllvm.git

[X86] Add VFMADDSUB cases for the 213->231 custom inserter.

Also add tests for vfmadd/vfmsub.

llvm-svn: 221488
---

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 7aa834cea512..ff56fd92d6b1 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -20492,6 +20492,11 @@ X86TargetLowering::emitFMA3Instr(MachineInstr *MI,
         case X86::VFNMSUBPSr213r: NewFMAOpc = X86::VFNMSUBPSr231r; break;
         case X86::VFNMSUBSDr213r: NewFMAOpc = X86::VFNMSUBSDr231r; break;
         case X86::VFNMSUBSSr213r: NewFMAOpc = X86::VFNMSUBSSr231r; break;
+        case X86::VFMADDSUBPDr213r: NewFMAOpc = X86::VFMADDSUBPDr231r; break;
+        case X86::VFMADDSUBPSr213r: NewFMAOpc = X86::VFMADDSUBPSr231r; break;
+        case X86::VFMSUBADDPDr213r: NewFMAOpc = X86::VFMSUBADDPDr231r; break;
+        case X86::VFMSUBADDPSr213r: NewFMAOpc = X86::VFMSUBADDPSr231r; break;
+
         case X86::VFMADDPDr213rY: NewFMAOpc = X86::VFMADDPDr231rY; break;
         case X86::VFMADDPSr213rY: NewFMAOpc = X86::VFMADDPSr231rY; break;
         case X86::VFMSUBPDr213rY: NewFMAOpc = X86::VFMSUBPDr231rY; break;
@@ -20500,6 +20505,10 @@ X86TargetLowering::emitFMA3Instr(MachineInstr *MI,
         case X86::VFNMADDPSr213rY: NewFMAOpc = X86::VFNMADDPSr231rY; break;
         case X86::VFNMSUBPDr213rY: NewFMAOpc = X86::VFNMSUBPDr231rY; break;
         case X86::VFNMSUBPSr213rY: NewFMAOpc = X86::VFNMSUBPSr231rY; break;
+        case X86::VFMADDSUBPDr213rY: NewFMAOpc = X86::VFMADDSUBPDr231rY; break;
+        case X86::VFMADDSUBPSr213rY: NewFMAOpc = X86::VFMADDSUBPSr231rY; break;
+        case X86::VFMSUBADDPDr213rY: NewFMAOpc = X86::VFMSUBADDPDr231rY; break;
+        case X86::VFMSUBADDPSr213rY: NewFMAOpc = X86::VFMSUBADDPSr231rY; break;
         default: llvm_unreachable("Unrecognized FMA variant.");
       }
 
diff --git a/llvm/test/CodeGen/X86/fma-phi-213-to-231.ll b/llvm/test/CodeGen/X86/fma-phi-213-to-231.ll
new file mode 100644
index 000000000000..29bfb92a038b
--- /dev/null
+++ b/llvm/test/CodeGen/X86/fma-phi-213-to-231.ll
@@ -0,0 +1,246 @@
+; RUN: llc < %s -mcpu=core-avx2 | FileCheck %s
+target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.10.0"
+
+; CHECK-LABEL: fmaddsubpd_loop
+; CHECK: [[BODYLBL:LBB.+]]:
+; CHECK:   vfmaddsub231pd        %ymm{{[0-9]+}}, %ymm{{[0-9]+}}, %ymm{{[0-9]+}}
+; CHECK: [[INCLBL:LBB.+]]:
+; CHECK:   incl  [[INDREG:%[a-z0-9]+]]
+; CHECK:   cmpl  {{%.+}}, [[INDREG]]
+; CHECK:   jl    [[BODYLBL]]
+define <4 x double> @fmaddsubpd_loop(i32 %iter, <4 x double> %a, <4 x double> %b, <4 x double> %c) {
+entry:
+  br label %for.cond
+
+for.cond:
+  %c.addr.0 = phi <4 x double> [ %c, %entry ], [ %0, %for.inc ]
+  %i.0 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
+  %cmp = icmp slt i32 %i.0, %iter
+  br i1 %cmp, label %for.body, label %for.end
+
+for.body:
+  br label %for.inc
+
+for.inc:
+  %0 = call <4 x double> @llvm.x86.fma.vfmaddsub.pd.256(<4 x double> %a, <4 x double> %b, <4 x double> %c.addr.0)
+  %inc = add nsw i32 %i.0, 1
+  br label %for.cond
+
+for.end:
+  ret <4 x double> %c.addr.0
+}
+
+; CHECK-LABEL: fmsubaddpd_loop
+; CHECK: [[BODYLBL:LBB.+]]:
+; CHECK:   vfmsubadd231pd        %ymm{{[0-9]+}}, %ymm{{[0-9]+}}, %ymm{{[0-9]+}}
+; CHECK: [[INCLBL:LBB.+]]:
+; CHECK:   incl  [[INDREG:%[a-z0-9]+]]
+; CHECK:   cmpl  {{%.+}}, [[INDREG]]
+; CHECK:   jl    [[BODYLBL]]
+define <4 x double> @fmsubaddpd_loop(i32 %iter, <4 x double> %a, <4 x double> %b, <4 x double> %c) {
+entry:
+  br label %for.cond
+
+for.cond:
+  %c.addr.0 = phi <4 x double> [ %c, %entry ], [ %0, %for.inc ]
+  %i.0 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
+  %cmp = icmp slt i32 %i.0, %iter
+  br i1 %cmp, label %for.body, label %for.end
+
+for.body:
+  br label %for.inc
+
+for.inc:
+  %0 = call <4 x double> @llvm.x86.fma.vfmsubadd.pd.256(<4 x double> %a, <4 x double> %b, <4 x double> %c.addr.0)
+  %inc = add nsw i32 %i.0, 1
+  br label %for.cond
+
+for.end:
+  ret <4 x double> %c.addr.0
+}
+
+; CHECK-LABEL: fmaddpd_loop
+; CHECK: [[BODYLBL:LBB.+]]:
+; CHECK:   vfmadd231pd        %ymm{{[0-9]+}}, %ymm{{[0-9]+}}, %ymm{{[0-9]+}}
+; CHECK: [[INCLBL:LBB.+]]:
+; CHECK:   incl  [[INDREG:%[a-z0-9]+]]
+; CHECK:   cmpl  {{%.+}}, [[INDREG]]
+; CHECK:   jl    [[BODYLBL]]
+define <4 x double> @fmaddpd_loop(i32 %iter, <4 x double> %a, <4 x double> %b, <4 x double> %c) {
+entry:
+  br label %for.cond
+
+for.cond:
+  %c.addr.0 = phi <4 x double> [ %c, %entry ], [ %0, %for.inc ]
+  %i.0 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
+  %cmp = icmp slt i32 %i.0, %iter
+  br i1 %cmp, label %for.body, label %for.end
+
+for.body:
+  br label %for.inc
+
+for.inc:
+  %0 = call <4 x double> @llvm.x86.fma.vfmadd.pd.256(<4 x double> %a, <4 x double> %b, <4 x double> %c.addr.0)
+  %inc = add nsw i32 %i.0, 1
+  br label %for.cond
+
+for.end:
+  ret <4 x double> %c.addr.0
+}
+
+; CHECK-LABEL: fmsubpd_loop
+; CHECK: [[BODYLBL:LBB.+]]:
+; CHECK:   vfmsub231pd        %ymm{{[0-9]+}}, %ymm{{[0-9]+}}, %ymm{{[0-9]+}}
+; CHECK: [[INCLBL:LBB.+]]:
+; CHECK:   incl  [[INDREG:%[a-z0-9]+]]
+; CHECK:   cmpl  {{%.+}}, [[INDREG]]
+; CHECK:   jl    [[BODYLBL]]
+define <4 x double> @fmsubpd_loop(i32 %iter, <4 x double> %a, <4 x double> %b, <4 x double> %c) {
+entry:
+  br label %for.cond
+
+for.cond:
+  %c.addr.0 = phi <4 x double> [ %c, %entry ], [ %0, %for.inc ]
+  %i.0 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
+  %cmp = icmp slt i32 %i.0, %iter
+  br i1 %cmp, label %for.body, label %for.end
+
+for.body:
+  br label %for.inc
+
+for.inc:
+  %0 = call <4 x double> @llvm.x86.fma.vfmsub.pd.256(<4 x double> %a, <4 x double> %b, <4 x double> %c.addr.0)
+  %inc = add nsw i32 %i.0, 1
+  br label %for.cond
+
+for.end:
+  ret <4 x double> %c.addr.0
+}
+
+declare <4 x double> @llvm.x86.fma.vfmaddsub.pd.256(<4 x double>, <4 x double>, <4 x double>)
+declare <4 x double> @llvm.x86.fma.vfmsubadd.pd.256(<4 x double>, <4 x double>, <4 x double>)
+declare <4 x double> @llvm.x86.fma.vfmadd.pd.256(<4 x double>, <4 x double>, <4 x double>)
+declare <4 x double> @llvm.x86.fma.vfmsub.pd.256(<4 x double>, <4 x double>, <4 x double>)
+
+
+; CHECK-LABEL: fmaddsubps_loop
+; CHECK: [[BODYLBL:LBB.+]]:
+; CHECK:   vfmaddsub231ps        %ymm{{[0-9]+}}, %ymm{{[0-9]+}}, %ymm{{[0-9]+}}
+; CHECK: [[INCLBL:LBB.+]]:
+; CHECK:   incl  [[INDREG:%[a-z0-9]+]]
+; CHECK:   cmpl  {{%.+}}, [[INDREG]]
+; CHECK:   jl    [[BODYLBL]]
+define <8 x float> @fmaddsubps_loop(i32 %iter, <8 x float> %a, <8 x float> %b, <8 x float> %c) {
+entry:
+  br label %for.cond
+
+for.cond:
+  %c.addr.0 = phi <8 x float> [ %c, %entry ], [ %0, %for.inc ]
+  %i.0 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
+  %cmp = icmp slt i32 %i.0, %iter
+  br i1 %cmp, label %for.body, label %for.end
+
+for.body:
+  br label %for.inc
+
+for.inc:
+  %0 = call <8 x float> @llvm.x86.fma.vfmaddsub.ps.256(<8 x float> %a, <8 x float> %b, <8 x float> %c.addr.0)
+  %inc = add nsw i32 %i.0, 1
+  br label %for.cond
+
+for.end:
+  ret <8 x float> %c.addr.0
+}
+
+; CHECK-LABEL: fmsubaddps_loop
+; CHECK: [[BODYLBL:LBB.+]]:
+; CHECK:   vfmsubadd231ps        %ymm{{[0-9]+}}, %ymm{{[0-9]+}}, %ymm{{[0-9]+}}
+; CHECK: [[INCLBL:LBB.+]]:
+; CHECK:   incl  [[INDREG:%[a-z0-9]+]]
+; CHECK:   cmpl  {{%.+}}, [[INDREG]]
+; CHECK:   jl    [[BODYLBL]]
+define <8 x float> @fmsubaddps_loop(i32 %iter, <8 x float> %a, <8 x float> %b, <8 x float> %c) {
+entry:
+  br label %for.cond
+
+for.cond:
+  %c.addr.0 = phi <8 x float> [ %c, %entry ], [ %0, %for.inc ]
+  %i.0 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
+  %cmp = icmp slt i32 %i.0, %iter
+  br i1 %cmp, label %for.body, label %for.end
+
+for.body:
+  br label %for.inc
+
+for.inc:
+  %0 = call <8 x float> @llvm.x86.fma.vfmsubadd.ps.256(<8 x float> %a, <8 x float> %b, <8 x float> %c.addr.0)
+  %inc = add nsw i32 %i.0, 1
+  br label %for.cond
+
+for.end:
+  ret <8 x float> %c.addr.0
+}
+
+; CHECK-LABEL: fmaddps_loop
+; CHECK: [[BODYLBL:LBB.+]]:
+; CHECK:   vfmadd231ps        %ymm{{[0-9]+}}, %ymm{{[0-9]+}}, %ymm{{[0-9]+}}
+; CHECK: [[INCLBL:LBB.+]]:
+; CHECK:   incl  [[INDREG:%[a-z0-9]+]]
+; CHECK:   cmpl  {{%.+}}, [[INDREG]]
+; CHECK:   jl    [[BODYLBL]]
+define <8 x float> @fmaddps_loop(i32 %iter, <8 x float> %a, <8 x float> %b, <8 x float> %c) {
+entry:
+  br label %for.cond
+
+for.cond:
+  %c.addr.0 = phi <8 x float> [ %c, %entry ], [ %0, %for.inc ]
+  %i.0 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
+  %cmp = icmp slt i32 %i.0, %iter
+  br i1 %cmp, label %for.body, label %for.end
+
+for.body:
+  br label %for.inc
+
+for.inc:
+  %0 = call <8 x float> @llvm.x86.fma.vfmadd.ps.256(<8 x float> %a, <8 x float> %b, <8 x float> %c.addr.0)
+  %inc = add nsw i32 %i.0, 1
+  br label %for.cond
+
+for.end:
+  ret <8 x float> %c.addr.0
+}
+
+; CHECK-LABEL: fmsubps_loop
+; CHECK: [[BODYLBL:LBB.+]]:
+; CHECK:   vfmsub231ps        %ymm{{[0-9]+}}, %ymm{{[0-9]+}}, %ymm{{[0-9]+}}
+; CHECK: [[INCLBL:LBB.+]]:
+; CHECK:   incl  [[INDREG:%[a-z0-9]+]]
+; CHECK:   cmpl  {{%.+}}, [[INDREG]]
+; CHECK:   jl    [[BODYLBL]]
+define <8 x float> @fmsubps_loop(i32 %iter, <8 x float> %a, <8 x float> %b, <8 x float> %c) {
+entry:
+  br label %for.cond
+
+for.cond:
+  %c.addr.0 = phi <8 x float> [ %c, %entry ], [ %0, %for.inc ]
+  %i.0 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
+  %cmp = icmp slt i32 %i.0, %iter
+  br i1 %cmp, label %for.body, label %for.end
+
+for.body:
+  br label %for.inc
+
+for.inc:
+  %0 = call <8 x float> @llvm.x86.fma.vfmsub.ps.256(<8 x float> %a, <8 x float> %b, <8 x float> %c.addr.0)
+  %inc = add nsw i32 %i.0, 1
+  br label %for.cond
+
+for.end:
+  ret <8 x float> %c.addr.0
+}
+
+declare <8 x float> @llvm.x86.fma.vfmaddsub.ps.256(<8 x float>, <8 x float>, <8 x float>)
+declare <8 x float> @llvm.x86.fma.vfmsubadd.ps.256(<8 x float>, <8 x float>, <8 x float>)
+declare <8 x float> @llvm.x86.fma.vfmadd.ps.256(<8 x float>, <8 x float>, <8 x float>)
+declare <8 x float> @llvm.x86.fma.vfmsub.ps.256(<8 x float>, <8 x float>, <8 x float>)