[AArch64][SVE2] Asm: support SVE2 Floating Point Convert Group

Summary:
Patch adds support for the following intructions:

SVE2 floating-point convert precision:
    * FCVTXNT, FCVTNT, FCVTLT

The specification can be found here:
https://developer.arm.com/docs/ddi0602/latest

Reviewed By: chill

Differential Revision: https://reviews.llvm.org/D62382

llvm-svn: 361801

llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td

@@ -1289,6 +1289,11 @@ let Predicates = [HasSVE2] in {
   // SVE2 histogram generation (vector)
   defm HISTCNT_ZPzZZ : sve2_hist_gen_vector<"histcnt">;
 
+  // SVE2 floating-point convert precision
+  defm FCVTXNT_ZPmZ : sve2_fp_convert_down_odd_rounding<"fcvtxnt">;
+  defm FCVTNT_ZPmZ  : sve2_fp_convert_down_narrow<"fcvtnt">;
+  defm FCVTLT_ZPmZ  : sve2_fp_convert_up_long<"fcvtlt">;
+
   // Predicated shifts
   defm SQSHL_ZPmI  : sve_int_bin_pred_shift_imm_left< 0b0110, "sqshl">;
   defm UQSHL_ZPmI  : sve_int_bin_pred_shift_imm_left< 0b0111, "uqshl">;

llvm/lib/Target/AArch64/SVEInstrFormats.td

@@ -1435,6 +1435,43 @@ multiclass sve_fp_fcadd<string asm> {
   def _D : sve_fp_fcadd<0b11, asm, ZPR64>;
 }
 
+//===----------------------------------------------------------------------===//
+// SVE2 Floating Point Convert Group
+//===----------------------------------------------------------------------===//
+
+class sve2_fp_convert_precision<bits<4> opc, string asm,
+                                ZPRRegOp zprty1, ZPRRegOp zprty2>
+: I<(outs zprty1:$Zd), (ins PPR3bAny:$Pg, zprty2:$Zn),
+  asm, "\t$Zd, $Pg/m, $Zn",
+  "",
+  []>, Sched<[]> {
+  bits<5> Zd;
+  bits<5> Zn;
+  bits<3> Pg;
+  let Inst{31-24} = 0b01100100;
+  let Inst{23-22} = opc{3-2};
+  let Inst{21-18} = 0b0010;
+  let Inst{17-16} = opc{1-0};
+  let Inst{15-13} = 0b101;
+  let Inst{12-10} = Pg;
+  let Inst{9-5}   = Zn;
+  let Inst{4-0}   = Zd;
+}
+
+multiclass sve2_fp_convert_down_narrow<string asm> {
+  def _StoH : sve2_fp_convert_precision<0b1000, asm, ZPR16, ZPR32>;
+  def _DtoS : sve2_fp_convert_precision<0b1110, asm, ZPR32, ZPR64>;
+}
+
+multiclass sve2_fp_convert_up_long<string asm> {
+  def _HtoS : sve2_fp_convert_precision<0b1001, asm, ZPR32, ZPR16>;
+  def _StoD : sve2_fp_convert_precision<0b1111, asm, ZPR64, ZPR32>;
+}
+
+multiclass sve2_fp_convert_down_odd_rounding<string asm> {
+  def _DtoS : sve2_fp_convert_precision<0b0010, asm, ZPR32, ZPR64>;
+}
+
 //===----------------------------------------------------------------------===//
 // SVE Stack Allocation Group
 //===----------------------------------------------------------------------===//

llvm/test/MC/AArch64/SVE2/fcvtlt-diagnostics.s

@@ -0,0 +1,69 @@
+// RUN: not llvm-mc -triple=aarch64 -show-encoding -mattr=+sve2 2>&1 < %s| FileCheck %s
+
+
+// --------------------------------------------------------------------------//
+// Invalid element width
+
+fcvtlt z0.b, p0/m, z0.b
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtlt z0.b, p0/m, z0.b
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+fcvtlt z0.h, p0/m, z0.h
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtlt z0.h, p0/m, z0.h
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+fcvtlt z0.s, p0/m, z0.s
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtlt z0.s, p0/m, z0.s
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+fcvtlt z0.d, p0/m, z0.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtlt z0.d, p0/m, z0.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+fcvtlt z0.h, p0/m, z0.b
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtlt z0.h, p0/m, z0.b
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+fcvtlt z0.q, p0/m, z0.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtlt z0.q, p0/m, z0.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+
+// --------------------------------------------------------------------------//
+// Invalid predicate operation
+
+fcvtlt z0.s, p0/z, z0.h
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid operand for instruction
+// CHECK-NEXT: fcvtlt z0.s, p0/z, z0.h
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+
+// --------------------------------------------------------------------------//
+// Predicate not in restricted predicate range
+
+fcvtlt z0.s, p8/m, z0.h
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: restricted predicate has range [0, 7].
+// CHECK-NEXT: fcvtlt z0.s, p8/m, z0.h
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+
+// --------------------------------------------------------------------------//
+// Negative tests for instructions that are incompatible with movprfx
+
+movprfx z0.s, p0/m, z7.s
+fcvtlt z0.s, p7/m, z1.h
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: instruction is unpredictable when following a movprfx, suggest replacing movprfx with mov
+// CHECK-NEXT: fcvtlt z0.s, p7/m, z1.h
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+movprfx z0, z7
+fcvtlt z0.s, p7/m, z1.h
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: instruction is unpredictable when following a movprfx, suggest replacing movprfx with mov
+// CHECK-NEXT: fcvtlt z0.s, p7/m, z1.h
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:

llvm/test/MC/AArch64/SVE2/fcvtlt.s

@@ -0,0 +1,21 @@
+// RUN: llvm-mc -triple=aarch64 -show-encoding -mattr=+sve2 < %s \
+// RUN:        | FileCheck %s --check-prefixes=CHECK-ENCODING,CHECK-INST
+// RUN: not llvm-mc -triple=aarch64 -show-encoding < %s 2>&1 \
+// RUN:        | FileCheck %s --check-prefix=CHECK-ERROR
+// RUN: llvm-mc -triple=aarch64 -filetype=obj -mattr=+sve2 < %s \
+// RUN:        | llvm-objdump -d -mattr=+sve2 - | FileCheck %s --check-prefix=CHECK-INST
+// RUN: llvm-mc -triple=aarch64 -filetype=obj -mattr=+sve2 < %s \
+// RUN:        | llvm-objdump -d - | FileCheck %s --check-prefix=CHECK-UNKNOWN
+
+
+fcvtlt z0.s, p0/m, z1.h
+// CHECK-INST: fcvtlt z0.s, p0/m, z1.h
+// CHECK-ENCODING: [0x20,0xa0,0x89,0x64]
+// CHECK-ERROR: instruction requires: sve2
+// CHECK-UNKNOWN: 20 a0 89 64 <unknown>
+
+fcvtlt z30.d, p7/m, z31.s
+// CHECK-INST: fcvtlt z30.d, p7/m, z31.s
+// CHECK-ENCODING: [0xfe,0xbf,0xcb,0x64]
+// CHECK-ERROR: instruction requires: sve2
+// CHECK-UNKNOWN: fe bf cb 64 <unknown>

llvm/test/MC/AArch64/SVE2/fcvtnt-diagnostics.s

@@ -0,0 +1,69 @@
+// RUN: not llvm-mc -triple=aarch64 -show-encoding -mattr=+sve2 2>&1 < %s| FileCheck %s
+
+
+// --------------------------------------------------------------------------//
+// Invalid element width
+
+fcvtnt z0.b, p0/m, z0.b
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtnt z0.b, p0/m, z0.b
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+fcvtnt z0.h, p0/m, z0.h
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtnt z0.h, p0/m, z0.h
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+fcvtnt z0.s, p0/m, z0.s
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtnt z0.s, p0/m, z0.s
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+fcvtnt z0.d, p0/m, z0.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtnt z0.d, p0/m, z0.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+fcvtnt z0.b, p0/m, z0.h
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtnt z0.b, p0/m, z0.h
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+fcvtnt z0.d, p0/m, z0.q
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtnt z0.d, p0/m, z0.q
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+
+// --------------------------------------------------------------------------//
+// Invalid predicate operation
+
+fcvtnt z0.h, p0/z, z0.s
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid operand for instruction
+// CHECK-NEXT: fcvtnt z0.h, p0/z, z0.s
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+
+// --------------------------------------------------------------------------//
+// Predicate not in restricted predicate range
+
+fcvtnt z0.h, p8/m, z0.s
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: restricted predicate has range [0, 7].
+// CHECK-NEXT: fcvtnt z0.h, p8/m, z0.s
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+
+// --------------------------------------------------------------------------//
+// Negative tests for instructions that are incompatible with movprfx
+
+movprfx z0.s, p0/m, z7.s
+fcvtnt z0.s, p7/m, z1.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: instruction is unpredictable when following a movprfx, suggest replacing movprfx with mov
+// CHECK-NEXT: fcvtnt z0.s, p7/m, z1.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+movprfx z0, z7
+fcvtnt z0.s, p7/m, z1.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: instruction is unpredictable when following a movprfx, suggest replacing movprfx with mov
+// CHECK-NEXT: fcvtnt z0.s, p7/m, z1.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:

llvm/test/MC/AArch64/SVE2/fcvtnt.s

@@ -0,0 +1,21 @@
+// RUN: llvm-mc -triple=aarch64 -show-encoding -mattr=+sve2 < %s \
+// RUN:        | FileCheck %s --check-prefixes=CHECK-ENCODING,CHECK-INST
+// RUN: not llvm-mc -triple=aarch64 -show-encoding < %s 2>&1 \
+// RUN:        | FileCheck %s --check-prefix=CHECK-ERROR
+// RUN: llvm-mc -triple=aarch64 -filetype=obj -mattr=+sve2 < %s \
+// RUN:        | llvm-objdump -d -mattr=+sve2 - | FileCheck %s --check-prefix=CHECK-INST
+// RUN: llvm-mc -triple=aarch64 -filetype=obj -mattr=+sve2 < %s \
+// RUN:        | llvm-objdump -d - | FileCheck %s --check-prefix=CHECK-UNKNOWN
+
+
+fcvtnt z0.h, p0/m, z1.s
+// CHECK-INST: fcvtnt z0.h, p0/m, z1.s
+// CHECK-ENCODING: [0x20,0xa0,0x88,0x64]
+// CHECK-ERROR: instruction requires: sve2
+// CHECK-UNKNOWN: 20 a0 88 64 <unknown>
+
+fcvtnt z30.s, p7/m, z31.d
+// CHECK-INST: fcvtnt z30.s, p7/m, z31.d
+// CHECK-ENCODING: [0xfe,0xbf,0xca,0x64]
+// CHECK-ERROR: instruction requires: sve2
+// CHECK-UNKNOWN: fe bf ca 64 <unknown>

llvm/test/MC/AArch64/SVE2/fcvtxnt-diagnostics.s

@@ -0,0 +1,74 @@
+// RUN: not llvm-mc -triple=aarch64 -show-encoding -mattr=+sve2 2>&1 < %s| FileCheck %s
+
+
+// --------------------------------------------------------------------------//
+// Invalid element width
+
+fcvtxnt z0.b, p0/m, z0.b
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtxnt z0.b, p0/m, z0.b
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+fcvtxnt z0.h, p0/m, z0.h
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtxnt z0.h, p0/m, z0.h
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+fcvtxnt z0.s, p0/m, z0.s
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtxnt z0.s, p0/m, z0.s
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+fcvtxnt z0.d, p0/m, z0.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtxnt z0.d, p0/m, z0.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+fcvtxnt z0.h, p0/m, z0.s
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtxnt z0.h, p0/m, z0.s
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+fcvtxnt z0.b, p0/m, z0.h
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtxnt z0.b, p0/m, z0.h
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+fcvtxnt z0.d, p0/m, z0.q
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid element width
+// CHECK-NEXT: fcvtxnt z0.d, p0/m, z0.q
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+
+// --------------------------------------------------------------------------//
+// Invalid predicate operation
+
+fcvtxnt z0.s, p0/z, z0.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid operand for instruction
+// CHECK-NEXT: fcvtxnt z0.s, p0/z, z0.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+
+// --------------------------------------------------------------------------//
+// Predicate not in restricted predicate range
+
+fcvtxnt z0.s, p8/m, z0.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: restricted predicate has range [0, 7].
+// CHECK-NEXT: fcvtxnt z0.s, p8/m, z0.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+
+// --------------------------------------------------------------------------//
+// Negative tests for instructions that are incompatible with movprfx
+
+movprfx z0.s, p0/m, z7.s
+fcvtxnt z0.s, p7/m, z1.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: instruction is unpredictable when following a movprfx, suggest replacing movprfx with mov
+// CHECK-NEXT: fcvtxnt z0.s, p7/m, z1.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:
+
+movprfx z0, z7
+fcvtxnt z0.s, p7/m, z1.d
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: instruction is unpredictable when following a movprfx, suggest replacing movprfx with mov
+// CHECK-NEXT: fcvtxnt z0.s, p7/m, z1.d
+// CHECK-NOT: [[@LINE-1]]:{{[0-9]+}}:

llvm/test/MC/AArch64/SVE2/fcvtxnt.s

@@ -0,0 +1,21 @@
+// RUN: llvm-mc -triple=aarch64 -show-encoding -mattr=+sve2 < %s \
+// RUN:        | FileCheck %s --check-prefixes=CHECK-ENCODING,CHECK-INST
+// RUN: not llvm-mc -triple=aarch64 -show-encoding < %s 2>&1 \
+// RUN:        | FileCheck %s --check-prefix=CHECK-ERROR
+// RUN: llvm-mc -triple=aarch64 -filetype=obj -mattr=+sve2 < %s \
+// RUN:        | llvm-objdump -d -mattr=+sve2 - | FileCheck %s --check-prefix=CHECK-INST
+// RUN: llvm-mc -triple=aarch64 -filetype=obj -mattr=+sve2 < %s \
+// RUN:        | llvm-objdump -d - | FileCheck %s --check-prefix=CHECK-UNKNOWN
+
+
+fcvtxnt z0.s, p0/m, z1.d
+// CHECK-INST: fcvtxnt z0.s, p0/m, z1.d
+// CHECK-ENCODING: [0x20,0xa0,0x0a,0x64]
+// CHECK-ERROR: instruction requires: sve2
+// CHECK-UNKNOWN: 20 a0 0a 64 <unknown>
+
+fcvtxnt z30.s, p7/m, z31.d
+// CHECK-INST: fcvtxnt z30.s, p7/m, z31.d
+// CHECK-ENCODING: [0xfe,0xbf,0x0a,0x64]
+// CHECK-ERROR: instruction requires: sve2
+// CHECK-UNKNOWN: fe bf 0a 64 <unknown>