aarch64: Add support for FEAT_SVE2p1.

Hi,

This patch add support for FEAT_SVE2p1 (SVE2.1 Extension) feature
along with +sve2p1 optional flag to enabe this feature.

Also support for following SVE2p1 instructions is added
addqv, andqv, smaxqv, sminqv, umaxqv, uminqv and uminqv.

Regression testing for aarch64-none-elf target and found no regressions.

Ok for binutils-master?

Regards,
Srinath.

diff --git a/gas/NEWS b/gas/NEWS
index 542ded4c272a5ad30d888a03f3f6a0b6de954f37..6f865a7f3ec7c78e57cd231f4981a83f02ff2f17 100644 (file)
--- a/gas/NEWS
+++ b/gas/NEWS
@@ -3,6 +3,9 @@
   hand-written asm using the new command line option --scfi=experimental on
   x86-64.  Only System V AMD64 ABI is supported.
 
+* Add support for the Arm Scalable Vector Extension version 2.1 (SVE2.1)
+  instructions.
+
 * Add support for the AArch64 Scalable Matrix Extension version 2.1 (SME2.1)
   instructions.
 

diff --git a/gas/config/tc-aarch64.c b/gas/config/tc-aarch64.c
index 34159c2168b78fe12d4a549678ae77be88e50313..04dd08a6fa71b84b3e71e0ab422fb6deb9fedb38 100644 (file)
--- a/gas/config/tc-aarch64.c
+++ b/gas/config/tc-aarch64.c
@@ -10354,6 +10354,7 @@ static const struct aarch64_option_cpu_value_table aarch64_features[] = {
                        AARCH64_FEATURE (LSE128)},
   {"b16b16",           AARCH64_FEATURE (B16B16), AARCH64_FEATURE (SVE2)},
   {"sme2p1",           AARCH64_FEATURE (SME2p1), AARCH64_FEATURE (SME2)},
+  {"sve2p1",           AARCH64_FEATURE (SVE2p1), AARCH64_FEATURE (SVE2)},
   {NULL,               AARCH64_NO_FEATURES, AARCH64_NO_FEATURES},
 };
 

diff --git a/gas/doc/c-aarch64.texi b/gas/doc/c-aarch64.texi
index 1f3a4fbcafbab652d70f202b33527ae3dde7e1bb..7a8da72c24a452b1867656acfd1ee1af3e4b56d5 100644 (file)
--- a/gas/doc/c-aarch64.texi
+++ b/gas/doc/c-aarch64.texi
@@ -278,6 +278,8 @@ automatically cause those extensions to be disabled.
  @tab Enable the 128-bit Page Descriptor Extension.  This implies @code{lse128}.
 @item @code{sme2p1} @tab N/A @tab No
  @tab Enable the SME2.1 Extension.
+@item @code{sve2p1} @tab N/A @tab No
+ @tab Enable the SVE2.1 Extension.
 @end multitable
 
 @node AArch64 Syntax

diff --git a/gas/testsuite/gas/aarch64/sve2p1-1-bad.d b/gas/testsuite/gas/aarch64/sve2p1-1-bad.d
new file mode 100644 (file)
index 0000000..a2ca49e
--- /dev/null
+++ b/gas/testsuite/gas/aarch64/sve2p1-1-bad.d
@@ -0,0 +1,4 @@
+#name: Illegal test of SVE2.1 min max instructions.
+#as: -march=armv9.4-a
+#source: sve2p1-1.s
+#error_output: sve2p1-1-bad.l

diff --git a/gas/testsuite/gas/aarch64/sve2p1-1-bad.l b/gas/testsuite/gas/aarch64/sve2p1-1-bad.l
new file mode 100644 (file)
index 0000000..6b07eee
--- /dev/null
+++ b/gas/testsuite/gas/aarch64/sve2p1-1-bad.l
@@ -0,0 +1,37 @@
+.*: Assembler messages:
+.*: Error: selected processor does not support `addqv v0.16b,p0,z16.b'
+.*: Error: selected processor does not support `addqv v1.8h,p1,z8.h'
+.*: Error: selected processor does not support `addqv v2.4s,p2,z4.s'
+.*: Error: selected processor does not support `addqv v4.2d,p3,z2.d'
+.*: Error: selected processor does not support `addqv v8.2d,p4,z1.d'
+.*: Error: selected processor does not support `addqv v16.4s,p7,z0.s'
+.*: Error: selected processor does not support `andqv v0.16b,p0,z16.b'
+.*: Error: selected processor does not support `andqv v1.8h,p1,z8.h'
+.*: Error: selected processor does not support `andqv v2.4s,p2,z4.s'
+.*: Error: selected processor does not support `andqv v4.2d,p3,z2.d'
+.*: Error: selected processor does not support `andqv v8.2d,p4,z1.d'
+.*: Error: selected processor does not support `andqv v16.4s,p7,z0.s'
+.*: Error: selected processor does not support `smaxqv v0.16b,p0,z16.b'
+.*: Error: selected processor does not support `smaxqv v1.8h,p1,z8.h'
+.*: Error: selected processor does not support `smaxqv v2.4s,p2,z4.s'
+.*: Error: selected processor does not support `smaxqv v4.2d,p3,z2.d'
+.*: Error: selected processor does not support `smaxqv v8.2d,p4,z1.d'
+.*: Error: selected processor does not support `smaxqv v16.4s,p7,z0.s'
+.*: Error: selected processor does not support `umaxqv v0.16b,p0,z16.b'
+.*: Error: selected processor does not support `umaxqv v1.8h,p1,z8.h'
+.*: Error: selected processor does not support `umaxqv v2.4s,p2,z4.s'
+.*: Error: selected processor does not support `umaxqv v4.2d,p3,z2.d'
+.*: Error: selected processor does not support `umaxqv v8.2d,p4,z1.d'
+.*: Error: selected processor does not support `umaxqv v16.4s,p7,z0.s'
+.*: Error: selected processor does not support `sminqv v0.16b,p0,z16.b'
+.*: Error: selected processor does not support `sminqv v1.8h,p1,z8.h'
+.*: Error: selected processor does not support `sminqv v2.4s,p2,z4.s'
+.*: Error: selected processor does not support `sminqv v4.2d,p3,z2.d'
+.*: Error: selected processor does not support `sminqv v8.2d,p4,z1.d'
+.*: Error: selected processor does not support `sminqv v16.4s,p7,z0.s'
+.*: Error: selected processor does not support `uminqv v0.16b,p0,z16.b'
+.*: Error: selected processor does not support `uminqv v1.8h,p1,z8.h'
+.*: Error: selected processor does not support `uminqv v2.4s,p2,z4.s'
+.*: Error: selected processor does not support `uminqv v4.2d,p3,z2.d'
+.*: Error: selected processor does not support `uminqv v8.2d,p4,z1.d'
+.*: Error: selected processor does not support `uminqv v16.4s,p7,z0.s'

diff --git a/gas/testsuite/gas/aarch64/sve2p1-1.d b/gas/testsuite/gas/aarch64/sve2p1-1.d
new file mode 100644 (file)
index 0000000..d3d14f3
--- /dev/null
+++ b/gas/testsuite/gas/aarch64/sve2p1-1.d
@@ -0,0 +1,46 @@
+#name: Test of SVE2.1 min max instructions.
+#as: -march=armv9.4-a+sve2p1
+#objdump: -dr
+
+[^:]+:     file format .*
+
+
+[^:]+:
+
+[^:]+:
+.*:    04052200        addqv   v0.16b, p0, z16.b
+.*:    04452501        addqv   v1.8h, p1, z8.h
+.*:    04852882        addqv   v2.4s, p2, z4.s
+.*:    04c52c44        addqv   v4.2d, p3, z2.d
+.*:    04c53028        addqv   v8.2d, p4, z1.d
+.*:    04853c10        addqv   v16.4s, p7, z0.s
+.*:    041e2200        andqv   v0.16b, p0, z16.b
+.*:    045e2501        andqv   v1.8h, p1, z8.h
+.*:    049e2882        andqv   v2.4s, p2, z4.s
+.*:    04de2c44        andqv   v4.2d, p3, z2.d
+.*:    04de3028        andqv   v8.2d, p4, z1.d
+.*:    049e3c10        andqv   v16.4s, p7, z0.s
+.*:    040c2200        smaxqv  v0.16b, p0, z16.b
+.*:    044c2501        smaxqv  v1.8h, p1, z8.h
+.*:    048c2882        smaxqv  v2.4s, p2, z4.s
+.*:    04cc2c44        smaxqv  v4.2d, p3, z2.d
+.*:    04cc3028        smaxqv  v8.2d, p4, z1.d
+.*:    048c3c10        smaxqv  v16.4s, p7, z0.s
+.*:    040d2200        umaxqv  v0.16b, p0, z16.b
+.*:    044d2501        umaxqv  v1.8h, p1, z8.h
+.*:    048d2882        umaxqv  v2.4s, p2, z4.s
+.*:    04cd2c44        umaxqv  v4.2d, p3, z2.d
+.*:    04cd3028        umaxqv  v8.2d, p4, z1.d
+.*:    048d3c10        umaxqv  v16.4s, p7, z0.s
+.*:    040e2200        sminqv  v0.16b, p0, z16.b
+.*:    044e2501        sminqv  v1.8h, p1, z8.h
+.*:    048e2882        sminqv  v2.4s, p2, z4.s
+.*:    04ce2c44        sminqv  v4.2d, p3, z2.d
+.*:    04ce3028        sminqv  v8.2d, p4, z1.d
+.*:    048e3c10        sminqv  v16.4s, p7, z0.s
+.*:    040f2200        uminqv  v0.16b, p0, z16.b
+.*:    044f2501        uminqv  v1.8h, p1, z8.h
+.*:    048f2882        uminqv  v2.4s, p2, z4.s
+.*:    04cf2c44        uminqv  v4.2d, p3, z2.d
+.*:    04cf3028        uminqv  v8.2d, p4, z1.d
+.*:    048f3c10        uminqv  v16.4s, p7, z0.s

diff --git a/gas/testsuite/gas/aarch64/sve2p1-1.s b/gas/testsuite/gas/aarch64/sve2p1-1.s
new file mode 100644 (file)
index 0000000..c56ebf8
--- /dev/null
+++ b/gas/testsuite/gas/aarch64/sve2p1-1.s
@@ -0,0 +1,41 @@
+addqv v0.16b, p0, z16.b
+addqv v1.8h, p1, z8.h
+addqv v2.4s, p2, z4.s
+addqv v4.2d, p3, z2.d
+addqv v8.2d, p4, z1.d
+addqv v16.4s, p7, z0.s
+
+andqv v0.16b, p0, z16.b
+andqv v1.8h, p1, z8.h
+andqv v2.4s, p2, z4.s
+andqv v4.2d, p3, z2.d
+andqv v8.2d, p4, z1.d
+andqv v16.4s, p7, z0.s
+
+smaxqv v0.16b, p0, z16.b
+smaxqv v1.8h, p1, z8.h
+smaxqv v2.4s, p2, z4.s
+smaxqv v4.2d, p3, z2.d
+smaxqv v8.2d, p4, z1.d
+smaxqv v16.4s, p7, z0.s
+
+umaxqv v0.16b, p0, z16.b
+umaxqv v1.8h, p1, z8.h
+umaxqv v2.4s, p2, z4.s
+umaxqv v4.2d, p3, z2.d
+umaxqv v8.2d, p4, z1.d
+umaxqv v16.4s, p7, z0.s
+
+sminqv v0.16b, p0, z16.b
+sminqv v1.8h, p1, z8.h
+sminqv v2.4s, p2, z4.s
+sminqv v4.2d, p3, z2.d
+sminqv v8.2d, p4, z1.d
+sminqv v16.4s, p7, z0.s
+
+uminqv v0.16b, p0, z16.b
+uminqv v1.8h, p1, z8.h
+uminqv v2.4s, p2, z4.s
+uminqv v4.2d, p3, z2.d
+uminqv v8.2d, p4, z1.d
+uminqv v16.4s, p7, z0.s

diff --git a/include/opcode/aarch64.h b/include/opcode/aarch64.h
index 648e25f3e4242bb738eee5f62079838784223b8a..1af49c406e06e79ba81a1f01887d43da37d8a625 100644 (file)
--- a/include/opcode/aarch64.h
+++ b/include/opcode/aarch64.h
@@ -226,6 +226,8 @@ enum aarch64_feature_bit {
   AARCH64_FEATURE_B16B16,
   /* SME2.1 instructions.  */
   AARCH64_FEATURE_SME2p1,
+  /* SVE2.1 instructions.  */
+  AARCH64_FEATURE_SVE2p1,
   AARCH64_NUM_FEATURES
 };
 
@@ -1000,6 +1002,7 @@ enum aarch64_insn_class
   cssc,
   gcs,
   the,
+  sve2_urqvs
 };
 
 /* Opcode enumerators.  */
@@ -1272,7 +1275,9 @@ extern const aarch64_opcode aarch64_opcode_table[];
    allow.  This impacts the constraintts on assembly but yelds no
    impact on disassembly.  */
 #define F_OPD_NARROW (1ULL << 33)
-/* Next bit is 34.  */
+/* For the instruction with size[22:23] field.  */
+#define F_OPD_SIZE (1ULL << 34)
+/* Next bit is 35.  */
 
 /* Instruction constraints.  */
 /* This instruction has a predication constraint on the instruction at PC+4.  */
@@ -1339,7 +1344,8 @@ static inline bool
 opcode_has_special_coder (const aarch64_opcode *opcode)
 {
   return (opcode->flags & (F_SF | F_LSE_SZ | F_SIZEQ | F_FPTYPE | F_SSIZE | F_T
-         | F_GPRSIZE_IN_Q | F_LDS_SIZE | F_MISC | F_N | F_COND)) != 0;
+         | F_GPRSIZE_IN_Q | F_LDS_SIZE | F_MISC | F_N | F_COND
+         | F_OPD_SIZE)) != 0;
 }
 \f
 struct aarch64_name_value_pair

diff --git a/opcodes/aarch64-asm.c b/opcodes/aarch64-asm.c
index 3fac127a5899077e2ac19c5e98df737b8ffbe147..1dfd59df42dbbe5640ece7c83f43f027a8329d06 100644 (file)
--- a/opcodes/aarch64-asm.c
+++ b/opcodes/aarch64-asm.c
@@ -1981,6 +1981,20 @@ do_special_encoding (struct aarch64_inst *inst)
       gen_sub_field (FLD_imm5, 0, num + 1, &field);
       insert_field_2 (&field, &inst->value, 1 << num, inst->opcode->mask);
     }
+
+  if ((inst->opcode->flags & F_OPD_SIZE) && inst->opcode->iclass == sve2_urqvs)
+    {
+      enum aarch64_opnd_qualifier qualifier[1];
+      aarch64_insn value1 = 0;
+      idx = 0;
+      qualifier[0] = inst->operands[idx].qualifier;
+      qualifier[1] = inst->operands[idx+2].qualifier;
+      value = aarch64_get_qualifier_standard_value (qualifier[0]);
+      value1 = aarch64_get_qualifier_standard_value (qualifier[1]);
+      assert ((value >> 1) == value1);
+      insert_field (FLD_size, &inst->value, value1, inst->opcode->mask);
+    }
+
   if (inst->opcode->flags & F_GPRSIZE_IN_Q)
     {
       /* Use Rt to encode in the case of e.g.

diff --git a/opcodes/aarch64-dis.c b/opcodes/aarch64-dis.c
index a14b2ca02d2c0fcf6c94f5bb0c587a7168594a5b..d395438966f16d1fc0fa7117a434cff50901f96e 100644 (file)
--- a/opcodes/aarch64-dis.c
+++ b/opcodes/aarch64-dis.c
@@ -2609,6 +2609,16 @@ do_special_decoding (aarch64_inst *inst)
        get_vreg_qualifier_from_value ((num << 1) | Q);
     }
 
+  if ((inst->opcode->flags & F_OPD_SIZE) && inst->opcode->iclass == sve2_urqvs)
+    {
+      unsigned size;
+      size = (unsigned) extract_field (FLD_size, inst->value,
+                                      inst->opcode->mask);
+      inst->operands[0].qualifier
+       = get_vreg_qualifier_from_value (1 + (size << 1));
+      inst->operands[2].qualifier = get_sreg_qualifier_from_value (size);
+    }
+
   if (inst->opcode->flags & F_GPRSIZE_IN_Q)
     {
       /* Use Rt to encode in the case of e.g.

diff --git a/opcodes/aarch64-tbl.h b/opcodes/aarch64-tbl.h
index 9c7648b0a6df5444cc89f52aef3d455e624eedbb..f433257634e72b6afb64d58a1f0f052164291033 100644 (file)
--- a/opcodes/aarch64-tbl.h
+++ b/opcodes/aarch64-tbl.h
@@ -1487,6 +1487,10 @@
    - P: the operand has a /[ZM] suffix and the choice of suffix is not
      the same for all variants.
 
+   - v: the operand has a V_[16B|8H|4S|2D] qualifier and the choice of
+     qualifier suffix is not the same for all variants.  This is used for
+     the same kinds of operands as [BHSD] above.
+
    The _<sizes>, if present, give the subset of [BHSD] that are accepted
    by the V entries in <operands>.  */
 #define OP_SVE_B                                        \
@@ -1911,6 +1915,13 @@
   QLF3(S_S,S_H,NIL),                                    \
   QLF3(S_D,S_S,NIL),                                    \
 }
+#define OP_SVE_vUS_BHSD_BHSD                           \
+{                                                      \
+  QLF3(V_16B,NIL,S_B),                                 \
+  QLF3(V_8H,NIL,S_H),                                  \
+  QLF3(V_4S,NIL,S_S),                                  \
+  QLF3(V_2D,NIL,S_D),                                  \
+}
 #define OP_SVE_VMV_SD                                   \
 {                                                       \
   QLF3(S_S,P_M,S_S),                                    \
@@ -2620,6 +2631,8 @@ static const aarch64_feature_set aarch64_feature_b16b16 =
   AARCH64_FEATURE (B16B16);
 static const aarch64_feature_set aarch64_feature_sme2p1 =
   AARCH64_FEATURE (SME2p1);
+static const aarch64_feature_set aarch64_feature_sve2p1 =
+  AARCH64_FEATURE (SVE2p1);
 
 #define CORE           &aarch64_feature_v8
 #define FP             &aarch64_feature_fp
@@ -2684,6 +2697,7 @@ static const aarch64_feature_set aarch64_feature_sme2p1 =
 #define D128_THE  &aarch64_feature_d128_the
 #define B16B16  &aarch64_feature_b16b16
 #define SME2p1  &aarch64_feature_sme2p1
+#define SVE2p1  &aarch64_feature_sve2p1
 
 #define CORE_INSN(NAME,OPCODE,MASK,CLASS,OP,OPS,QUALS,FLAGS) \
   { NAME, OPCODE, MASK, CLASS, OP, CORE, OPS, QUALS, FLAGS, 0, 0, NULL }
@@ -2762,6 +2776,12 @@ static const aarch64_feature_set aarch64_feature_sme2p1 =
 #define B16B16_INSNC(NAME,OPCODE,MASK,CLASS,OP,OPS,QUALS,FLAGS,CONSTRAINTS,TIED) \
   { NAME, OPCODE, MASK, CLASS, OP, B16B16, OPS, QUALS, \
     FLAGS | F_STRICT, CONSTRAINTS, TIED, NULL }
+#define SVE2p1_INSN(NAME,OPCODE,MASK,CLASS,OP,OPS,QUALS,FLAGS,TIED) \
+  { NAME, OPCODE, MASK, CLASS, OP, SVE2p1, OPS, QUALS, \
+    FLAGS | F_STRICT, 0, TIED, NULL }
+#define SVE2p1_INSNC(NAME,OPCODE,MASK,CLASS,OP,OPS,QUALS,FLAGS,CONSTRAINTS,TIED) \
+  { NAME, OPCODE, MASK, CLASS, OP, SVE2p1, OPS, QUALS, \
+    FLAGS | F_STRICT, CONSTRAINTS, TIED, NULL }
 #define SVE2AES_INSN(NAME,OPCODE,MASK,CLASS,OP,OPS,QUALS,FLAGS,TIED) \
   { NAME, OPCODE, MASK, CLASS, OP, SVE2_AES, OPS, QUALS, \
     FLAGS | F_STRICT, 0, TIED, NULL }
@@ -6309,6 +6329,15 @@ const struct aarch64_opcode aarch64_opcode_table[] =
   SME2p1_INSN ("movaz", 0xc0460200, 0xffff1f01, sme2_movaz, 0, OP2 (SME_Zdnx2, SME_ZA_array_vrsh_1), OP_SVE_HH, 0, 0),
   SME2p1_INSN ("movaz", 0xc0860200, 0xffff1f01, sme2_movaz, 0, OP2 (SME_Zdnx2, SME_ZA_array_vrss_1), OP_SVE_SS, 0, 0),
   SME2p1_INSN ("movaz", 0xc0c60200, 0xffff1f01, sme2_movaz, 0, OP2 (SME_Zdnx2, SME_ZA_array_vrsd_1), OP_SVE_DD, 0, 0),
+
+/* SVE2p1 Instructions.  */
+  SVE2p1_INSNC("addqv",0x04052000, 0xff3fe000, sve2_urqvs, 0, OP3 (Vd, SVE_Pg3, SVE_Zn), OP_SVE_vUS_BHSD_BHSD, F_OPD_SIZE, C_SCAN_MOVPRFX, 0),
+  SVE2p1_INSNC("andqv",0x041e2000, 0xff3fe000, sve2_urqvs, 0, OP3 (Vd, SVE_Pg3, SVE_Zn), OP_SVE_vUS_BHSD_BHSD, F_OPD_SIZE, C_SCAN_MOVPRFX, 0),
+  SVE2p1_INSNC("smaxqv",0x040c2000, 0xff3fe000, sve2_urqvs, 0, OP3 (Vd, SVE_Pg3, SVE_Zn), OP_SVE_vUS_BHSD_BHSD, F_OPD_SIZE, C_SCAN_MOVPRFX, 0),
+  SVE2p1_INSNC("sminqv",0x040e2000, 0xff3fe000, sve2_urqvs, 0, OP3 (Vd, SVE_Pg3, SVE_Zn), OP_SVE_vUS_BHSD_BHSD, F_OPD_SIZE, C_SCAN_MOVPRFX, 0),
+  SVE2p1_INSNC("umaxqv",0x040d2000, 0xff3fe000, sve2_urqvs, 0, OP3 (Vd, SVE_Pg3, SVE_Zn), OP_SVE_vUS_BHSD_BHSD, F_OPD_SIZE, C_SCAN_MOVPRFX, 0),
+  SVE2p1_INSNC("uminqv",0x040f2000, 0xff3fe000, sve2_urqvs, 0, OP3 (Vd, SVE_Pg3, SVE_Zn), OP_SVE_vUS_BHSD_BHSD, F_OPD_SIZE, C_SCAN_MOVPRFX, 0),
+
   {0, 0, 0, 0, 0, 0, {}, {}, 0, 0, 0, NULL},
 };