[ARM] Change the MC names for VMAXNM/VMINNM

Now the NEON ones have a prefix "NEON_", and the VFP ones have a
prefix "VFP_". This is so that the regex in ARMScheduleA57.td can be
made to match both of _those_ classes of VMAXNM without also matching
the MVE ones that are going to be introduced soon. NFCI.

Patch by Simon Tatham.

Differential Revision: https://reviews.llvm.org/D60700

llvm-svn: 362097

diff --git a/llvm/lib/Target/ARM/ARMInstrNEON.td b/llvm/lib/Target/ARM/ARMInstrNEON.td
index 3aa4431..effee0f 100644 (file)
--- a/llvm/lib/Target/ARM/ARMInstrNEON.td
+++ b/llvm/lib/Target/ARM/ARMInstrNEON.td
@@ -5537,22 +5537,22 @@ def  VMAXhq   : N3VQInt<0, 0, 0b01, 0b1111, 0, N3RegFrm, IIC_VBINQ,
 
 // VMAXNM
 let PostEncoderMethod = "NEONThumb2V8PostEncoder", DecoderNamespace = "v8NEON" in {
-  def VMAXNMNDf  : N3VDIntnp<0b00110, 0b00, 0b1111, 0, 1,
-                            N3RegFrm, NoItinerary, "vmaxnm", "f32",
-                            v2f32, v2f32, fmaxnum, 1>,
-                            Requires<[HasV8, HasNEON]>;
-  def VMAXNMNQf  : N3VQIntnp<0b00110, 0b00, 0b1111, 1, 1,
-                            N3RegFrm, NoItinerary, "vmaxnm", "f32",
-                            v4f32, v4f32, fmaxnum, 1>,
-                            Requires<[HasV8, HasNEON]>;
-  def VMAXNMNDh  : N3VDIntnp<0b00110, 0b01, 0b1111, 0, 1,
-                            N3RegFrm, NoItinerary, "vmaxnm", "f16",
-                            v4f16, v4f16, fmaxnum, 1>,
-                            Requires<[HasV8, HasNEON, HasFullFP16]>;
-  def VMAXNMNQh  : N3VQIntnp<0b00110, 0b01, 0b1111, 1, 1,
-                            N3RegFrm, NoItinerary, "vmaxnm", "f16",
-                            v8f16, v8f16, fmaxnum, 1>,
-                            Requires<[HasV8, HasNEON, HasFullFP16]>;
+  def NEON_VMAXNMNDf  : N3VDIntnp<0b00110, 0b00, 0b1111, 0, 1,
+                                  N3RegFrm, NoItinerary, "vmaxnm", "f32",
+                                  v2f32, v2f32, fmaxnum, 1>,
+                                  Requires<[HasV8, HasNEON]>;
+  def NEON_VMAXNMNQf  : N3VQIntnp<0b00110, 0b00, 0b1111, 1, 1,
+                                  N3RegFrm, NoItinerary, "vmaxnm", "f32",
+                                  v4f32, v4f32, fmaxnum, 1>,
+                                  Requires<[HasV8, HasNEON]>;
+  def NEON_VMAXNMNDh  : N3VDIntnp<0b00110, 0b01, 0b1111, 0, 1,
+                                  N3RegFrm, NoItinerary, "vmaxnm", "f16",
+                                  v4f16, v4f16, fmaxnum, 1>,
+                                  Requires<[HasV8, HasNEON, HasFullFP16]>;
+  def NEON_VMAXNMNQh  : N3VQIntnp<0b00110, 0b01, 0b1111, 1, 1,
+                                  N3RegFrm, NoItinerary, "vmaxnm", "f16",
+                                  v8f16, v8f16, fmaxnum, 1>,
+                                  Requires<[HasV8, HasNEON, HasFullFP16]>;
 }
 
 //   VMIN     : Vector Minimum
@@ -5579,22 +5579,22 @@ def  VMINhq   : N3VQInt<0, 0, 0b11, 0b1111, 0, N3RegFrm, IIC_VBINQ,
 
 // VMINNM
 let PostEncoderMethod = "NEONThumb2V8PostEncoder", DecoderNamespace = "v8NEON" in {
-  def VMINNMNDf  : N3VDIntnp<0b00110, 0b10, 0b1111, 0, 1,
-                            N3RegFrm, NoItinerary, "vminnm", "f32",
-                            v2f32, v2f32, fminnum, 1>,
-                            Requires<[HasV8, HasNEON]>;
-  def VMINNMNQf  : N3VQIntnp<0b00110, 0b10, 0b1111, 1, 1,
-                            N3RegFrm, NoItinerary, "vminnm", "f32",
-                            v4f32, v4f32, fminnum, 1>,
-                            Requires<[HasV8, HasNEON]>;
-  def VMINNMNDh  : N3VDIntnp<0b00110, 0b11, 0b1111, 0, 1,
-                            N3RegFrm, NoItinerary, "vminnm", "f16",
-                            v4f16, v4f16, fminnum, 1>,
-                            Requires<[HasV8, HasNEON, HasFullFP16]>;
-  def VMINNMNQh  : N3VQIntnp<0b00110, 0b11, 0b1111, 1, 1,
-                            N3RegFrm, NoItinerary, "vminnm", "f16",
-                            v8f16, v8f16, fminnum, 1>,
-                            Requires<[HasV8, HasNEON, HasFullFP16]>;
+  def NEON_VMINNMNDf  : N3VDIntnp<0b00110, 0b10, 0b1111, 0, 1,
+                                  N3RegFrm, NoItinerary, "vminnm", "f32",
+                                  v2f32, v2f32, fminnum, 1>,
+                                  Requires<[HasV8, HasNEON]>;
+  def NEON_VMINNMNQf  : N3VQIntnp<0b00110, 0b10, 0b1111, 1, 1,
+                                  N3RegFrm, NoItinerary, "vminnm", "f32",
+                                  v4f32, v4f32, fminnum, 1>,
+                                  Requires<[HasV8, HasNEON]>;
+  def NEON_VMINNMNDh  : N3VDIntnp<0b00110, 0b11, 0b1111, 0, 1,
+                                  N3RegFrm, NoItinerary, "vminnm", "f16",
+                                  v4f16, v4f16, fminnum, 1>,
+                                  Requires<[HasV8, HasNEON, HasFullFP16]>;
+  def NEON_VMINNMNQh  : N3VQIntnp<0b00110, 0b11, 0b1111, 1, 1,
+                                  N3RegFrm, NoItinerary, "vminnm", "f16",
+                                  v8f16, v8f16, fminnum, 1>,
+                                  Requires<[HasV8, HasNEON, HasFullFP16]>;
 }
 
 // Vector Pairwise Operations.

diff --git a/llvm/lib/Target/ARM/ARMInstrVFP.td b/llvm/lib/Target/ARM/ARMInstrVFP.td
index 88405a8..6f042d3 100644 (file)
--- a/llvm/lib/Target/ARM/ARMInstrVFP.td
+++ b/llvm/lib/Target/ARM/ARMInstrVFP.td
@@ -533,8 +533,8 @@ multiclass vmaxmin_inst<string op, bit opc, SDNode SD> {
   }
 }
 
-defm VMAXNM : vmaxmin_inst<"vmaxnm", 0, fmaxnum>;
-defm VMINNM : vmaxmin_inst<"vminnm", 1, fminnum>;
+defm VFP_VMAXNM : vmaxmin_inst<"vmaxnm", 0, fmaxnum>;
+defm VFP_VMINNM : vmaxmin_inst<"vminnm", 1, fminnum>;
 
 // Match reassociated forms only if not sign dependent rounding.
 def : Pat<(fmul (fneg DPR:$a), (f64 DPR:$b)),

diff --git a/llvm/lib/Target/ARM/ARMScheduleA57.td b/llvm/lib/Target/ARM/ARMScheduleA57.td
index 6db6eda..de970ae 100644 (file)
--- a/llvm/lib/Target/ARM/ARMScheduleA57.td
+++ b/llvm/lib/Target/ARM/ARMScheduleA57.td
@@ -1174,7 +1174,8 @@ def : InstRW<[A57Write_8cyc_1V], (instregex
 
 // ASIMD FP max/min
 def : InstRW<[A57Write_5cyc_1V], (instregex
-  "(VMAX|VMIN)(fd|fq|hd|hq)", "(VPMAX|VPMIN)(f|h)", "VMAXNM", "VMINNM")>;
+  "(VMAX|VMIN)(fd|fq|hd|hq)", "(VPMAX|VPMIN)(f|h)", "(NEON|VFP)_VMAXNM",
+  "(NEON|VFP)_VMINNM")>;
 
 // ASIMD FP multiply
 def A57WriteVMUL_VecFP  : SchedWriteRes<[A57UnitV]> { let Latency = 5;  }