[ARM] Add MVE integer vector min/max instructions.

Summary:
These form a small family of their own, to go with the floating-point
VMINNM/VMAXNM instructions added in a previous commit.

They introduce the first of many special cases in the mnemonic
recognition code, because VMIN with the E suffix used by the VPT
predication system needs to avoid being interpreted as the nonexistent
instruction 'VMI' with an ordinary 'NE' condition suffix.

Reviewers: dmgreen, samparker, SjoerdMeijer, t.p.northover

Subscribers: javed.absar, kristof.beyls, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D62671

llvm-svn: 363695

diff --git a/llvm/lib/Target/ARM/ARMInstrMVE.td b/llvm/lib/Target/ARM/ARMInstrMVE.td
index 50b868186429aa501c7285dbe39719df1fa90320..5964e64522465d84c66814b34e9c5d0fafc00880 100644 (file)
--- a/llvm/lib/Target/ARM/ARMInstrMVE.td
+++ b/llvm/lib/Target/ARM/ARMInstrMVE.td
@@ -642,6 +642,32 @@ def MVE_VMAXNMf16 : MVE_VMINMAXNM<"vmaxnm", "f16", 0b1, 0b0>;
 def MVE_VMINNMf32 : MVE_VMINMAXNM<"vminnm", "f32", 0b0, 0b1>;
 def MVE_VMINNMf16 : MVE_VMINMAXNM<"vminnm", "f16", 0b1, 0b1>;
 
+class MVE_VMINMAX<string iname, string suffix, bit U, bits<2> size,
+              bit bit_4, list<dag> pattern=[]>
+  : MVE_comp<NoItinerary, iname, suffix, "", pattern> {
+
+  let Inst{28} = U;
+  let Inst{25-24} = 0b11;
+  let Inst{23} = 0b0;
+  let Inst{21-20} = size{1-0};
+  let Inst{11} = 0b0;
+  let Inst{8} = 0b0;
+  let Inst{6} = 0b1;
+  let Inst{4} = bit_4;
+}
+
+multiclass MVE_VMINMAX_all_sizes<string iname, bit bit_4> {
+  def s8  : MVE_VMINMAX<iname, "s8",  0b0, 0b00, bit_4>;
+  def s16 : MVE_VMINMAX<iname, "s16", 0b0, 0b01, bit_4>;
+  def s32 : MVE_VMINMAX<iname, "s32", 0b0, 0b10, bit_4>;
+  def u8  : MVE_VMINMAX<iname, "u8",  0b1, 0b00, bit_4>;
+  def u16 : MVE_VMINMAX<iname, "u16", 0b1, 0b01, bit_4>;
+  def u32 : MVE_VMINMAX<iname, "u32", 0b1, 0b10, bit_4>;
+}
+
+defm MVE_VMAX : MVE_VMINMAX_all_sizes<"vmax", 0b0>;
+defm MVE_VMIN : MVE_VMINMAX_all_sizes<"vmin", 0b1>;
+
 // end of mve_comp instructions
 
 class MVE_VPT<string suffix, bits<2> size, dag iops, string asm, list<dag> pattern=[]>

diff --git a/llvm/lib/Target/ARM/AsmParser/ARMAsmParser.cpp b/llvm/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
index b76924003aa41ac88eae0fa0c53ee9bb77bcaef3..91a48b040385bf3c22f232d8e82d2dd1c7456a7a 100644 (file)
--- a/llvm/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
+++ b/llvm/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
@@ -5933,7 +5933,8 @@ StringRef ARMAsmParser::splitMnemonic(StringRef Mnemonic,
   if (Mnemonic != "adcs" && Mnemonic != "bics" && Mnemonic != "movs" &&
       Mnemonic != "muls" && Mnemonic != "smlals" && Mnemonic != "smulls" &&
       Mnemonic != "umlals" && Mnemonic != "umulls" && Mnemonic != "lsls" &&
-      Mnemonic != "sbcs" && Mnemonic != "rscs") {
+      Mnemonic != "sbcs" && Mnemonic != "rscs" &&
+      !(hasMVE() && Mnemonic == "vmine")) {
     unsigned CC = ARMCondCodeFromString(Mnemonic.substr(Mnemonic.size()-2));
     if (CC != ~0U) {
       Mnemonic = Mnemonic.slice(0, Mnemonic.size() - 2);

diff --git a/llvm/test/MC/ARM/mve-minmax.s b/llvm/test/MC/ARM/mve-minmax.s
index 4de210434284ca7d2a17e69a87d28d9764c77835..1c52cbc71633ce5eed139c322fb76b4b99d61e7c 100644 (file)
--- a/llvm/test/MC/ARM/mve-minmax.s
+++ b/llvm/test/MC/ARM/mve-minmax.s
@@ -13,3 +13,61 @@ vmaxnm.f32 q0, q1, q4
 # CHECK-NOFP-NOT: vminnm.f16 q3, q0, q1  @ encoding: [0x30,0xff,0x52,0x6f]
 # ERROR-NOFP: [[@LINE+1]]:{{[0-9]+}}: error: instruction requires: mve.fp
 vminnm.f16 q3, q0, q1
+
+# CHECK: vmin.s8 q3, q0, q7 @ encoding: [0x00,0xef,0x5e,0x66]
+# CHECK-NOFP: vmin.s8 q3, q0, q7 @ encoding: [0x00,0xef,0x5e,0x66]
+vmin.s8 q3, q0, q7
+
+# CHECK: vmin.s16 q0, q1, q2 @ encoding: [0x12,0xef,0x54,0x06]
+# CHECK-NOFP: vmin.s16 q0, q1, q2 @ encoding: [0x12,0xef,0x54,0x06]
+vmin.s16 q0, q1, q2
+
+# CHECK: vmin.s32 q0, q1, q2 @ encoding: [0x22,0xef,0x54,0x06]
+# CHECK-NOFP: vmin.s32 q0, q1, q2 @ encoding: [0x22,0xef,0x54,0x06]
+vmin.s32 q0, q1, q2
+
+# CHECK: vmin.u8 q0, q1, q2 @ encoding: [0x02,0xff,0x54,0x06]
+# CHECK-NOFP: vmin.u8 q0, q1, q2 @ encoding: [0x02,0xff,0x54,0x06]
+vmin.u8 q0, q1, q2
+
+# CHECK: vmin.u16 q0, q1, q2 @ encoding: [0x12,0xff,0x54,0x06]
+# CHECK-NOFP: vmin.u16 q0, q1, q2 @ encoding: [0x12,0xff,0x54,0x06]
+vmin.u16 q0, q1, q2
+
+# CHECK: vmin.u32 q0, q1, q2 @ encoding: [0x22,0xff,0x54,0x06]
+# CHECK-NOFP: vmin.u32 q0, q1, q2 @ encoding: [0x22,0xff,0x54,0x06]
+vmin.u32 q0, q1, q2
+
+# CHECK: vmax.s8 q3, q0, q7 @ encoding: [0x00,0xef,0x4e,0x66]
+# CHECK-NOFP: vmax.s8 q3, q0, q7 @ encoding: [0x00,0xef,0x4e,0x66]
+vmax.s8 q3, q0, q7
+
+# CHECK: vmax.s16 q0, q1, q2 @ encoding: [0x12,0xef,0x44,0x06]
+# CHECK-NOFP: vmax.s16 q0, q1, q2 @ encoding: [0x12,0xef,0x44,0x06]
+vmax.s16 q0, q1, q2
+
+# CHECK: vmax.s32 q0, q1, q2 @ encoding: [0x22,0xef,0x44,0x06]
+# CHECK-NOFP: vmax.s32 q0, q1, q2 @ encoding: [0x22,0xef,0x44,0x06]
+vmax.s32 q0, q1, q2
+
+# CHECK: vmax.u8 q0, q1, q2 @ encoding: [0x02,0xff,0x44,0x06]
+# CHECK-NOFP: vmax.u8 q0, q1, q2 @ encoding: [0x02,0xff,0x44,0x06]
+vmax.u8 q0, q1, q2
+
+# CHECK: vmax.u16 q0, q1, q2 @ encoding: [0x12,0xff,0x44,0x06]
+# CHECK-NOFP: vmax.u16 q0, q1, q2 @ encoding: [0x12,0xff,0x44,0x06]
+vmax.u16 q0, q1, q2
+
+# CHECK: vmax.u32 q0, q1, q2 @ encoding: [0x22,0xff,0x44,0x06]
+# CHECK-NOFP: vmax.u32 q0, q1, q2 @ encoding: [0x22,0xff,0x44,0x06]
+vmax.u32 q0, q1, q2
+
+vpste
+vmint.s8 q0, q1, q2
+vmine.s16 q0, q1, q2
+# CHECK: vpste @ encoding: [0x71,0xfe,0x4d,0x8f]
+# CHECK-NOFP: vpste @ encoding: [0x71,0xfe,0x4d,0x8f]
+# CHECK: vmint.s8 q0, q1, q2 @ encoding: [0x02,0xef,0x54,0x06]
+# CHECK-NOFP: vmint.s8 q0, q1, q2 @ encoding: [0x02,0xef,0x54,0x06]
+# CHECK: vmine.s16 q0, q1, q2 @ encoding: [0x12,0xef,0x54,0x06]
+# CHECK-NOFP: vmine.s16 q0, q1, q2 @ encoding: [0x12,0xef,0x54,0x06]

diff --git a/llvm/test/MC/Disassembler/ARM/mve-minmax.txt b/llvm/test/MC/Disassembler/ARM/mve-minmax.txt
index 24287573b6cc1ef88c491bbd42f29ba569e066ba..21d531583cc7531dc6e7ef10896ce2d2284fac0c 100644 (file)
--- a/llvm/test/MC/Disassembler/ARM/mve-minmax.txt
+++ b/llvm/test/MC/Disassembler/ARM/mve-minmax.txt
@@ -9,3 +9,51 @@
 # CHECK: vminnm.f16 q3, q0, q1  @ encoding: [0x30,0xff,0x52,0x6f]
 # CHECK-NOMVE: [[@LINE+1]]:2: warning: invalid instruction encoding
 [0x30,0xff,0x52,0x6f]
+
+# CHECK: vmin.s8 q3, q0, q7 @ encoding: [0x00,0xef,0x5e,0x66]
+# CHECK-NOMVE: [[@LINE+1]]:2: warning: invalid instruction encoding
+[0x00,0xef,0x5e,0x66]
+
+# CHECK: vmin.s16 q0, q1, q2 @ encoding: [0x12,0xef,0x54,0x06]
+# CHECK-NOMVE: [[@LINE+1]]:2: warning: invalid instruction encoding
+[0x12,0xef,0x54,0x06]
+
+# CHECK: vmin.s32 q0, q1, q2 @ encoding: [0x22,0xef,0x54,0x06]
+# CHECK-NOMVE: [[@LINE+1]]:2: warning: invalid instruction encoding
+[0x22,0xef,0x54,0x06]
+
+# CHECK: vmin.u8 q0, q1, q2 @ encoding: [0x02,0xff,0x54,0x06]
+# CHECK-NOMVE: [[@LINE+1]]:2: warning: invalid instruction encoding
+[0x02,0xff,0x54,0x06]
+
+# CHECK: vmin.u16 q0, q1, q2 @ encoding: [0x12,0xff,0x54,0x06]
+# CHECK-NOMVE: [[@LINE+1]]:2: warning: invalid instruction encoding
+[0x12,0xff,0x54,0x06]
+
+# CHECK: vmin.u32 q0, q1, q2 @ encoding: [0x22,0xff,0x54,0x06]
+# CHECK-NOMVE: [[@LINE+1]]:2: warning: invalid instruction encoding
+[0x22,0xff,0x54,0x06]
+
+# CHECK: vmax.s8 q3, q0, q7 @ encoding: [0x00,0xef,0x4e,0x66]
+# CHECK-NOMVE: [[@LINE+1]]:2: warning: invalid instruction encoding
+[0x00,0xef,0x4e,0x66]
+
+# CHECK: vmax.s16 q0, q1, q2 @ encoding: [0x12,0xef,0x44,0x06]
+# CHECK-NOMVE: [[@LINE+1]]:2: warning: invalid instruction encoding
+[0x12,0xef,0x44,0x06]
+
+# CHECK: vmax.s32 q0, q1, q2 @ encoding: [0x22,0xef,0x44,0x06]
+# CHECK-NOMVE: [[@LINE+1]]:2: warning: invalid instruction encoding
+[0x22,0xef,0x44,0x06]
+
+# CHECK: vmax.u8 q0, q1, q2 @ encoding: [0x02,0xff,0x44,0x06]
+# CHECK-NOMVE: [[@LINE+1]]:2: warning: invalid instruction encoding
+[0x02,0xff,0x44,0x06]
+
+# CHECK: vmax.u16 q0, q1, q2 @ encoding: [0x12,0xff,0x44,0x06]
+# CHECK-NOMVE: [[@LINE+1]]:2: warning: invalid instruction encoding
+[0x12,0xff,0x44,0x06]
+
+# CHECK: vmax.u32 q0, q1, q2 @ encoding: [0x22,0xff,0x44,0x06]
+# CHECK-NOMVE: [[@LINE+1]]:2: warning: invalid instruction encoding
+[0x22,0xff,0x44,0x06]