[ARM] Add ARMv8.2 FP16 vmul/vmla/vmls (by scalar)

gas/
  * config/tc-arm.c (do_neon_mac_maybe_scalar): Allow F16.
  * testsuite/gas/arm/armv8-2-fp16-simd.s: New tests.
  * testsuite/gas/arm/armv8-2-fp16-simd.d: New expected results.
  * testsuite/gas/arm/armv8-2-fp16-simd-thum.d: Likewise for Thumb.
  * testsuite/gas/arm/armv8-2-fp16-simd-warning.l: New warning results.
  * testsuite/gas/arm/simd_by_scalar_low_regbank.s: New test source.
  * testsuite/gas/arm/simd_by_scalar_low_regbank.d: New testcase.
  * testsuite/gas/arm/simd_by_scalar_low_regbank_thumb.d: Likewise for Thumb.
  * testsuite/gas/arm/simd_by_scalar_low_regbank.l: New warning results.

opcodes/
  * arm-dis.c: Support FP16 vmul, vmla, vmls (by scalar).

8 files changed, 145 insertions, 0 deletions

diff --git a/gas/testsuite/gas/arm/armv8-2-fp16-simd-thumb.d b/gas/testsuite/gas/arm/armv8-2-fp16-simd-thumb.d
index 5578b9b..9470497 100644
--- a/gas/testsuite/gas/arm/armv8-2-fp16-simd-thumb.d
+++ b/gas/testsuite/gas/arm/armv8-2-fp16-simd-thumb.d
@@ -145,3 +145,15 @@ Disassembly of section .text:
  214:	fff5 c5c4 	vcle.f16	q14, q2, #0
  218:	ffb5 e602 	vclt.f16	d14, d2, #0
  21c:	fff5 c644 	vclt.f16	q14, q2, #0
+ 220:	ef90 7941 	vmul.f16	d7, d0, d1\[0\]
+ 224:	ef98 4966 	vmul.f16	d4, d8, d6\[2\]
+ 228:	ff90 49c8 	vmul.f16	q2, q8, d0\[1\]
+ 22c:	ff90 49ef 	vmul.f16	q2, q8, d7\[3\]
+ 230:	ef94 2141 	vmla.f16	d2, d4, d1\[0\]
+ 234:	ff98 4141 	vmla.f16	q2, q4, d1\[0\]
+ 238:	ef94 2541 	vmls.f16	d2, d4, d1\[0\]
+ 23c:	ff98 4541 	vmls.f16	q2, q4, d1\[0\]
+ 240:	ef98 116f 	vmla.f16	d1, d8, d7\[3\]
+ 244:	ff90 21ef 	vmla.f16	q1, q8, d7\[3\]
+ 248:	ef98 156f 	vmls.f16	d1, d8, d7\[3\]
+ 24c:	ff90 25ef 	vmls.f16	q1, q8, d7\[3\]
diff --git a/gas/testsuite/gas/arm/armv8-2-fp16-simd-warning.l b/gas/testsuite/gas/arm/armv8-2-fp16-simd-warning.l
index ba27f7c..9027cee 100644
--- a/gas/testsuite/gas/arm/armv8-2-fp16-simd-warning.l
+++ b/gas/testsuite/gas/arm/armv8-2-fp16-simd-warning.l
@@ -135,3 +135,15 @@
 [^:]*:224: Error: selected processor does not support fp16 instruction -- `vcle.f16 q14,q2,#0'
 [^:]*:224: Error: selected processor does not support fp16 instruction -- `vclt.f16 d14,d2,#0'
 [^:]*:224: Error: selected processor does not support fp16 instruction -- `vclt.f16 q14,q2,#0'
+[^:]*:242: Error: selected processor does not support fp16 instruction -- `vmul.f16 d7,d0,d1\[0\]'
+[^:]*:243: Error: selected processor does not support fp16 instruction -- `vmul.f16 d4,d8,d6\[2\]'
+[^:]*:244: Error: selected processor does not support fp16 instruction -- `vmul.f16 q2,q8,d0\[1\]'
+[^:]*:245: Error: selected processor does not support fp16 instruction -- `vmul.f16 q2,q8,d7\[3\]'
+[^:]*:248: Error: selected processor does not support fp16 instruction -- `vmla.f16 d2,d4,d1\[0\]'
+[^:]*:248: Error: selected processor does not support fp16 instruction -- `vmla.f16 q2,q4,d1\[0\]'
+[^:]*:248: Error: selected processor does not support fp16 instruction -- `vmls.f16 d2,d4,d1\[0\]'
+[^:]*:248: Error: selected processor does not support fp16 instruction -- `vmls.f16 q2,q4,d1\[0\]'
+[^:]*:249: Error: selected processor does not support fp16 instruction -- `vmla.f16 d1,d8,d7\[3\]'
+[^:]*:249: Error: selected processor does not support fp16 instruction -- `vmla.f16 q1,q8,d7\[3\]'
+[^:]*:249: Error: selected processor does not support fp16 instruction -- `vmls.f16 d1,d8,d7\[3\]'
+[^:]*:249: Error: selected processor does not support fp16 instruction -- `vmls.f16 q1,q8,d7\[3\]'
diff --git a/gas/testsuite/gas/arm/armv8-2-fp16-simd.d b/gas/testsuite/gas/arm/armv8-2-fp16-simd.d
index 1a97f39..2dc4054 100644
--- a/gas/testsuite/gas/arm/armv8-2-fp16-simd.d
+++ b/gas/testsuite/gas/arm/armv8-2-fp16-simd.d
@@ -145,3 +145,15 @@ Disassembly of section .text:
  214:	f3f5c5c4 	vcle.f16	q14, q2, #0
  218:	f3b5e602 	vclt.f16	d14, d2, #0
  21c:	f3f5c644 	vclt.f16	q14, q2, #0
+ 220:	f2907941 	vmul.f16	d7, d0, d1\[0\]
+ 224:	f2984966 	vmul.f16	d4, d8, d6\[2\]
+ 228:	f39049c8 	vmul.f16	q2, q8, d0\[1\]
+ 22c:	f39049ef 	vmul.f16	q2, q8, d7\[3\]
+ 230:	f2942141 	vmla.f16	d2, d4, d1\[0\]
+ 234:	f3984141 	vmla.f16	q2, q4, d1\[0\]
+ 238:	f2942541 	vmls.f16	d2, d4, d1\[0\]
+ 23c:	f3984541 	vmls.f16	q2, q4, d1\[0\]
+ 240:	f298116f 	vmla.f16	d1, d8, d7\[3\]
+ 244:	f39021ef 	vmla.f16	q1, q8, d7\[3\]
+ 248:	f298156f 	vmls.f16	d1, d8, d7\[3\]
+ 24c:	f39025ef 	vmls.f16	q1, q8, d7\[3\]
diff --git a/gas/testsuite/gas/arm/armv8-2-fp16-simd.s b/gas/testsuite/gas/arm/armv8-2-fp16-simd.s
index 7758f24..172d604 100644
--- a/gas/testsuite/gas/arm/armv8-2-fp16-simd.s
+++ b/gas/testsuite/gas/arm/armv8-2-fp16-simd.s
@@ -222,3 +222,28 @@ func:
 
 	# neon_fcmp_imm0
 	f16_dq_fcmp_imm0 14, 2
+
+	.macro f16_d_by_scalar op reg0 reg1 reg2 idx
+		\op d\reg0, d\reg1, d\reg2[\idx]
+	.endm
+
+	.macro f16_q_by_scalar op reg0 reg1 reg2 idx
+		\op q\reg0, q\reg1, d\reg2[\idx]
+	.endm
+
+	.macro f16_dq_fmacmaybe_by_scalar reg0 reg1 reg2 idx
+	.irp op, vmla.f16, vmls.f16
+		\op d\reg0, d\reg1, d\reg2[\idx]
+		\op q\reg0, q\reg1, d\reg2[\idx]
+	.endr
+	.endm
+
+	# neon_mul (by scalar)
+	f16_d_by_scalar vmul.f16 7 0 1 0
+	f16_d_by_scalar vmul.f16 4 8 6 2
+	f16_q_by_scalar vmul.f16 2 8 0 1
+	f16_q_by_scalar vmul.f16 2 8 7 3
+
+	# neon_mac_maybe_scalar (by scalar)
+	f16_dq_fmacmaybe_by_scalar 2 4 1 0
+	f16_dq_fmacmaybe_by_scalar 1 8 7 3
diff --git a/gas/testsuite/gas/arm/simd_by_scalar_low_regbank.d b/gas/testsuite/gas/arm/simd_by_scalar_low_regbank.d
new file mode 100644
index 0000000..28b99f1
--- /dev/null
+++ b/gas/testsuite/gas/arm/simd_by_scalar_low_regbank.d
@@ -0,0 +1,4 @@
+#name: VMUL/VMLA/VMLS by scalar reg restriction
+#source: simd_by_scalar_low_regbank.s
+#as: -march=armv8.2-a+fp16 -mfpu=neon-fp-armv8
+#error-output: simd_by_scalar_low_regbank.l
diff --git a/gas/testsuite/gas/arm/simd_by_scalar_low_regbank.l b/gas/testsuite/gas/arm/simd_by_scalar_low_regbank.l
new file mode 100644
index 0000000..d6435cc
--- /dev/null
+++ b/gas/testsuite/gas/arm/simd_by_scalar_low_regbank.l
@@ -0,0 +1,49 @@
+[^:]*: Assembler messages:
+[^:]*:21: Error: scalar out of range for multiply instruction -- `vmul.i32 d3,d12,d7\[2\]'
+[^:]*:21: Error: scalar out of range for multiply instruction -- `vmul.i32 q3,q12,d7\[2\]'
+[^:]*:21: Error: scalar out of range for multiply instruction -- `vmul.f32 d3,d12,d7\[2\]'
+[^:]*:21: Error: scalar out of range for multiply instruction -- `vmul.f32 q3,q12,d7\[2\]'
+[^:]*:22: Error: scalar out of range for multiply instruction -- `vmul.i16 d4,d9,d8\[1\]'
+[^:]*:22: Error: scalar out of range for multiply instruction -- `vmul.i16 q4,q9,d8\[1\]'
+[^:]*:22: Error: scalar out of range for multiply instruction -- `vmul.f16 d4,d9,d8\[1\]'
+[^:]*:22: Error: scalar out of range for multiply instruction -- `vmul.f16 q4,q9,d8\[1\]'
+[^:]*:23: Error: scalar out of range for multiply instruction -- `vmul.i16 d13,d6,d15\[3\]'
+[^:]*:23: Error: scalar out of range for multiply instruction -- `vmul.i16 q13,q6,d15\[3\]'
+[^:]*:23: Error: scalar out of range for multiply instruction -- `vmul.f16 d13,d6,d15\[3\]'
+[^:]*:23: Error: scalar out of range for multiply instruction -- `vmul.f16 q13,q6,d15\[3\]'
+[^:]*:23: Error: scalar out of range for multiply instruction -- `vmul.i32 d13,d6,d15\[3\]'
+[^:]*:23: Error: scalar out of range for multiply instruction -- `vmul.i32 q13,q6,d15\[3\]'
+[^:]*:23: Error: scalar out of range for multiply instruction -- `vmul.f32 d13,d6,d15\[3\]'
+[^:]*:23: Error: scalar out of range for multiply instruction -- `vmul.f32 q13,q6,d15\[3\]'
+[^:]*:25: Error: scalar out of range for multiply instruction -- `vmla.i32 d5,d4,d6\[2\]'
+[^:]*:25: Error: scalar out of range for multiply instruction -- `vmla.i32 q5,q4,d6\[2\]'
+[^:]*:25: Error: scalar out of range for multiply instruction -- `vmla.f32 d5,d4,d6\[2\]'
+[^:]*:25: Error: scalar out of range for multiply instruction -- `vmla.f32 q5,q4,d6\[2\]'
+[^:]*:25: Error: scalar out of range for multiply instruction -- `vmls.i32 d5,d4,d6\[2\]'
+[^:]*:25: Error: scalar out of range for multiply instruction -- `vmls.i32 q5,q4,d6\[2\]'
+[^:]*:25: Error: scalar out of range for multiply instruction -- `vmls.f32 d5,d4,d6\[2\]'
+[^:]*:25: Error: scalar out of range for multiply instruction -- `vmls.f32 q5,q4,d6\[2\]'
+[^:]*:26: Error: scalar out of range for multiply instruction -- `vmla.i16 d4,d13,d10\[1\]'
+[^:]*:26: Error: scalar out of range for multiply instruction -- `vmla.i16 q4,q13,d10\[1\]'
+[^:]*:26: Error: scalar out of range for multiply instruction -- `vmla.f16 d4,d13,d10\[1\]'
+[^:]*:26: Error: scalar out of range for multiply instruction -- `vmla.f16 q4,q13,d10\[1\]'
+[^:]*:26: Error: scalar out of range for multiply instruction -- `vmls.i16 d4,d13,d10\[1\]'
+[^:]*:26: Error: scalar out of range for multiply instruction -- `vmls.i16 q4,q13,d10\[1\]'
+[^:]*:26: Error: scalar out of range for multiply instruction -- `vmls.f16 d4,d13,d10\[1\]'
+[^:]*:26: Error: scalar out of range for multiply instruction -- `vmls.f16 q4,q13,d10\[1\]'
+[^:]*:27: Error: scalar out of range for multiply instruction -- `vmla.i16 d12,d6,d13\[3\]'
+[^:]*:27: Error: scalar out of range for multiply instruction -- `vmla.i16 q12,q6,d13\[3\]'
+[^:]*:27: Error: scalar out of range for multiply instruction -- `vmla.i32 d12,d6,d13\[3\]'
+[^:]*:27: Error: scalar out of range for multiply instruction -- `vmla.i32 q12,q6,d13\[3\]'
+[^:]*:27: Error: scalar out of range for multiply instruction -- `vmla.f16 d12,d6,d13\[3\]'
+[^:]*:27: Error: scalar out of range for multiply instruction -- `vmla.f16 q12,q6,d13\[3\]'
+[^:]*:27: Error: scalar out of range for multiply instruction -- `vmla.f32 d12,d6,d13\[3\]'
+[^:]*:27: Error: scalar out of range for multiply instruction -- `vmla.f32 q12,q6,d13\[3\]'
+[^:]*:27: Error: scalar out of range for multiply instruction -- `vmls.i16 d12,d6,d13\[3\]'
+[^:]*:27: Error: scalar out of range for multiply instruction -- `vmls.i16 q12,q6,d13\[3\]'
+[^:]*:27: Error: scalar out of range for multiply instruction -- `vmls.i32 d12,d6,d13\[3\]'
+[^:]*:27: Error: scalar out of range for multiply instruction -- `vmls.i32 q12,q6,d13\[3\]'
+[^:]*:27: Error: scalar out of range for multiply instruction -- `vmls.f16 d12,d6,d13\[3\]'
+[^:]*:27: Error: scalar out of range for multiply instruction -- `vmls.f16 q12,q6,d13\[3\]'
+[^:]*:27: Error: scalar out of range for multiply instruction -- `vmls.f32 d12,d6,d13\[3\]'
+[^:]*:27: Error: scalar out of range for multiply instruction -- `vmls.f32 q12,q6,d13\[3\]'
diff --git a/gas/testsuite/gas/arm/simd_by_scalar_low_regbank.s b/gas/testsuite/gas/arm/simd_by_scalar_low_regbank.s
new file mode 100644
index 0000000..ed73b74
--- /dev/null
+++ b/gas/testsuite/gas/arm/simd_by_scalar_low_regbank.s
@@ -0,0 +1,27 @@
+	.macro vmul_iter reg0 reg1 reg2 idx
+	.irp op, vmul.i16 vmul.f16 vmul.i32 vmul.f32
+		\op d\reg0, d\reg1, d\reg2[\idx]
+		\op q\reg0, q\reg1, d\reg2[\idx]
+	.endr
+	.endm
+
+	.macro vmul_acc_iter reg0 reg1 reg2 idx
+	.irp op, vmla.i16 vmla.i32 vmla.f16 vmla.f32 vmls.i16 vmls.i32 vmls.f16 vmls.f32
+		\op d\reg0, d\reg1, d\reg2[\idx]
+		\op q\reg0, q\reg1, d\reg2[\idx]
+	.endr
+	.endm
+
+	# There are two restriction on the scalar operand:
+	#   * The scalar operand is restricted to D0-D7 if size is 16bit wide,
+	#     or D0 - D15 otherwise.
+	#   * The scalar index should within range, 0-3 if size is 16bit wide,
+	#     0-1 if size is 32bit wide.
+	vmul_iter 0 1 3 0
+	vmul_iter 3 12 7 2
+	vmul_iter 4 9 8 1
+	vmul_iter 13 6 15 3
+	vmul_acc_iter 2 7 1 0
+	vmul_acc_iter 5 4 6 2
+	vmul_acc_iter 4 13 10 1
+	vmul_acc_iter 12 6 13 3
diff --git a/gas/testsuite/gas/arm/simd_by_scalar_low_regbank_thumb.d b/gas/testsuite/gas/arm/simd_by_scalar_low_regbank_thumb.d
new file mode 100644
index 0000000..f4e1980
--- /dev/null
+++ b/gas/testsuite/gas/arm/simd_by_scalar_low_regbank_thumb.d
@@ -0,0 +1,4 @@
+#name: VMUL/VMLA/VMLS by scalar reg restriction (Thumb)
+#source: simd_by_scalar_low_regbank.s
+#as: -march=armv8.2-a+fp16 -mfpu=neon-fp-armv8 -mthumb
+#error-output: simd_by_scalar_low_regbank.l