aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorMihail Ionescu <mihail.ionescu@arm.com>2019-11-12 13:57:20 +0000
committerKyrylo Tkachov <kyrylo.tkachov@arm.com>2019-11-12 14:01:10 +0000
commitef8f595f73a6b42f745bc76a716f45079eae1075 (patch)
tree718987e71df5f0b8198b8b9dfef3ab9839d5ecdf
parent[binutils][arm] Update the decoding of MVE VMOV, VMVN (diff)
downloadbinutils-gdb-ef8f595f73a6b42f745bc76a716f45079eae1075.tar.gz
binutils-gdb-ef8f595f73a6b42f745bc76a716f45079eae1075.tar.bz2
binutils-gdb-ef8f595f73a6b42f745bc76a716f45079eae1075.zip
[gas][arm] Enable VLDM, VSTM, VPUSH, VPOP for MVE
This patch enables a few instructions for Armv8.1-M MVE. Currently VLDM, VSTM, VSTR, VLDR, VPUSH and VPOP are enabled only when the Armv8-M Floating-point Extension is enabled. According to the ARMv8.1-M ARM, section A.1.4.2[1], they can be enabled by having "Armv8-M Floating-point Extension and/or Armv8.1-M MVE". [1]https://developer.arm.com/docs/ddi0553/bh/armv81-m-architecture-reference-manual 2019-11-12 Mihail Ionescu <mihail.ionescu@arm.com> * config/tc-arm.c (do_vfp_nsyn_push): Move in order to enable it for both fpu_vfp_ext_v1xd and mve_ext and add call to the aliased vstm instruction for mve_ext. (do_vfp_nsyn_pop): Move in order to enable it for both fpu_vfp_ext_v1xd and mve_ext and add call to the aliased vldm instruction for mve_ext. (do_neon_ldm_stm): Add fpu_vfp_ext_v1 and mve_ext checks. (insns): Enable vldm, vldmia, vldmdb, vstm, vstmia, vstmdb, vpop, vpush, and fldd, fstd, flds, fsts for arm_ext_v6t2 instead of fpu_vfp_ext_v1xd. * testsuite/gas/arm/v8_1m-mve.s: New. * testsuite/gas/arm/v8_1m-mve.d: New.
-rw-r--r--gas/ChangeLog15
-rw-r--r--gas/config/tc-arm.c109
-rw-r--r--gas/testsuite/gas/arm/v8_1m-mve.d27
-rw-r--r--gas/testsuite/gas/arm/v8_1m-mve.s24
4 files changed, 131 insertions, 44 deletions
diff --git a/gas/ChangeLog b/gas/ChangeLog
index f84de0f25c8..d0ce992223c 100644
--- a/gas/ChangeLog
+++ b/gas/ChangeLog
@@ -1,5 +1,20 @@
2019-11-12 Mihail Ionescu <mihail.ionescu@arm.com>
+ * config/tc-arm.c (do_vfp_nsyn_push): Move in order to enable it for
+ both fpu_vfp_ext_v1xd and mve_ext and add call to the aliased vstm
+ instruction for mve_ext.
+ (do_vfp_nsyn_pop): Move in order to enable it for both
+ fpu_vfp_ext_v1xd and mve_ext and add call to the aliased vldm
+ instruction for mve_ext.
+ (do_neon_ldm_stm): Add fpu_vfp_ext_v1 and mve_ext checks.
+ (insns): Enable vldm, vldmia, vldmdb, vstm, vstmia, vstmdb, vpop,
+ vpush, and fldd, fstd, flds, fsts for arm_ext_v6t2 instead
+ of fpu_vfp_ext_v1xd.
+ * testsuite/gas/arm/v8_1m-mve.s: New.
+ * testsuite/gas/arm/v8_1m-mve.d: New.
+
+2019-11-12 Mihail Ionescu <mihail.ionescu@arm.com>
+
* gas/config/tc-arm.c (do_neon_mvn): Allow mve_ext cmode=0xd.
* testsuite/gas/arm/mve-vmov-vmvn-vorr-vbic.s: New test.
* testsuite/gas/arm/mve-vmov-vmvn-vorr-vbic.d: Likewise.
diff --git a/gas/config/tc-arm.c b/gas/config/tc-arm.c
index a76aaf3621a..641ce7e3a64 100644
--- a/gas/config/tc-arm.c
+++ b/gas/config/tc-arm.c
@@ -16527,36 +16527,6 @@ nsyn_insert_sp (void)
inst.operands[0].present = 1;
}
-static void
-do_vfp_nsyn_push (void)
-{
- nsyn_insert_sp ();
-
- constraint (inst.operands[1].imm < 1 || inst.operands[1].imm > 16,
- _("register list must contain at least 1 and at most 16 "
- "registers"));
-
- if (inst.operands[1].issingle)
- do_vfp_nsyn_opcode ("fstmdbs");
- else
- do_vfp_nsyn_opcode ("fstmdbd");
-}
-
-static void
-do_vfp_nsyn_pop (void)
-{
- nsyn_insert_sp ();
-
- constraint (inst.operands[1].imm < 1 || inst.operands[1].imm > 16,
- _("register list must contain at least 1 and at most 16 "
- "registers"));
-
- if (inst.operands[1].issingle)
- do_vfp_nsyn_opcode ("fldmias");
- else
- do_vfp_nsyn_opcode ("fldmiad");
-}
-
/* Fix up Neon data-processing instructions, ORing in the correct bits for
ARM mode or Thumb mode and moving the encoded bit 24 to bit 28. */
@@ -20638,6 +20608,9 @@ do_neon_tbl_tbx (void)
static void
do_neon_ldm_stm (void)
{
+ constraint (!ARM_CPU_HAS_FEATURE (cpu_variant, fpu_vfp_ext_v1xd)
+ && !ARM_CPU_HAS_FEATURE (cpu_variant, mve_ext),
+ _(BAD_FPU));
/* P, U and L bits are part of bitmask. */
int is_dbmode = (inst.instruction & (1 << 24)) != 0;
unsigned offsetbits = inst.operands[1].imm * 2;
@@ -20666,6 +20639,49 @@ do_neon_ldm_stm (void)
}
static void
+do_vfp_nsyn_pop (void)
+{
+ nsyn_insert_sp ();
+ if (ARM_CPU_HAS_FEATURE (cpu_variant, mve_ext)) {
+ return do_vfp_nsyn_opcode ("vldm");
+ }
+
+ constraint (!ARM_CPU_HAS_FEATURE (cpu_variant, fpu_vfp_ext_v1xd),
+ _(BAD_FPU));
+
+ constraint (inst.operands[1].imm < 1 || inst.operands[1].imm > 16,
+ _("register list must contain at least 1 and at most 16 "
+ "registers"));
+
+ if (inst.operands[1].issingle)
+ do_vfp_nsyn_opcode ("fldmias");
+ else
+ do_vfp_nsyn_opcode ("fldmiad");
+}
+
+static void
+do_vfp_nsyn_push (void)
+{
+ nsyn_insert_sp ();
+ if (ARM_CPU_HAS_FEATURE (cpu_variant, mve_ext)) {
+ return do_vfp_nsyn_opcode ("vstmdb");
+ }
+
+ constraint (!ARM_CPU_HAS_FEATURE (cpu_variant, fpu_vfp_ext_v1xd),
+ _(BAD_FPU));
+
+ constraint (inst.operands[1].imm < 1 || inst.operands[1].imm > 16,
+ _("register list must contain at least 1 and at most 16 "
+ "registers"));
+
+ if (inst.operands[1].issingle)
+ do_vfp_nsyn_opcode ("fstmdbs");
+ else
+ do_vfp_nsyn_opcode ("fstmdbd");
+}
+
+
+static void
do_neon_ldr_str (void)
{
int is_ldr = (inst.instruction & (1 << 20)) != 0;
@@ -20745,7 +20761,8 @@ do_vldr_vstr (void)
/* VLDR/VSTR. */
else
{
- if (!mark_feature_used (&fpu_vfp_ext_v1xd))
+ if (!mark_feature_used (&fpu_vfp_ext_v1xd)
+ && !ARM_CPU_HAS_FEATURE (cpu_variant, mve_ext))
as_bad (_("Instruction not permitted on this architecture"));
do_neon_ldr_str ();
}
@@ -24966,6 +24983,10 @@ static const struct asm_opcode insns[] =
#define THUMB_VARIANT & arm_ext_v6t2
mcCE(vmrs, ef00a10, 2, (APSR_RR, RVC), vmrs),
mcCE(vmsr, ee00a10, 2, (RVC, RR), vmsr),
+ mcCE(fldd, d100b00, 2, (RVD, ADDRGLDC), vfp_dp_ldst),
+ mcCE(fstd, d000b00, 2, (RVD, ADDRGLDC), vfp_dp_ldst),
+ mcCE(flds, d100a00, 2, (RVS, ADDRGLDC), vfp_sp_ldst),
+ mcCE(fsts, d000a00, 2, (RVS, ADDRGLDC), vfp_sp_ldst),
#undef THUMB_VARIANT
/* Moves and type conversions. */
@@ -24980,8 +25001,6 @@ static const struct asm_opcode insns[] =
cCE("fmxr", ee00a10, 2, (RVC, RR), rn_rd),
/* Memory operations. */
- cCE("flds", d100a00, 2, (RVS, ADDRGLDC), vfp_sp_ldst),
- cCE("fsts", d000a00, 2, (RVS, ADDRGLDC), vfp_sp_ldst),
cCE("fldmias", c900a00, 2, (RRnpctw, VRSLST), vfp_sp_ldstmia),
cCE("fldmfds", c900a00, 2, (RRnpctw, VRSLST), vfp_sp_ldstmia),
cCE("fldmdbs", d300a00, 2, (RRnpctw, VRSLST), vfp_sp_ldstmdb),
@@ -25023,8 +25042,6 @@ static const struct asm_opcode insns[] =
/* Double precision load/store are still present on single precision
implementations. */
- cCE("fldd", d100b00, 2, (RVD, ADDRGLDC), vfp_dp_ldst),
- cCE("fstd", d000b00, 2, (RVD, ADDRGLDC), vfp_dp_ldst),
cCE("fldmiad", c900b00, 2, (RRnpctw, VRDLST), vfp_dp_ldstmia),
cCE("fldmfdd", c900b00, 2, (RRnpctw, VRDLST), vfp_dp_ldstmia),
cCE("fldmdbd", d300b00, 2, (RRnpctw, VRDLST), vfp_dp_ldstmdb),
@@ -25078,6 +25095,19 @@ static const struct asm_opcode insns[] =
#undef ARM_VARIANT
#define ARM_VARIANT & fpu_vfp_ext_v1xd
#undef THUMB_VARIANT
+#define THUMB_VARIANT & arm_ext_v6t2
+
+ NCE(vldm, c900b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
+ NCE(vldmia, c900b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
+ NCE(vldmdb, d100b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
+ NCE(vstm, c800b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
+ NCE(vstmia, c800b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
+ NCE(vstmdb, d000b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
+
+ NCE(vpop, 0, 1, (VRSDLST), vfp_nsyn_pop),
+ NCE(vpush, 0, 1, (VRSDLST), vfp_nsyn_push),
+
+#undef THUMB_VARIANT
#define THUMB_VARIANT & fpu_vfp_ext_v1xd
/* These mnemonics are unique to VFP. */
@@ -25086,20 +25116,11 @@ static const struct asm_opcode insns[] =
nCE(vnmul, _vnmul, 3, (RVSD, RVSD, RVSD), vfp_nsyn_nmul),
nCE(vnmla, _vnmla, 3, (RVSD, RVSD, RVSD), vfp_nsyn_nmul),
nCE(vnmls, _vnmls, 3, (RVSD, RVSD, RVSD), vfp_nsyn_nmul),
- NCE(vpush, 0, 1, (VRSDLST), vfp_nsyn_push),
- NCE(vpop, 0, 1, (VRSDLST), vfp_nsyn_pop),
NCE(vcvtz, 0, 2, (RVSD, RVSD), vfp_nsyn_cvtz),
/* Mnemonics shared by Neon and VFP. */
nCEF(vmls, _vmls, 3, (RNSDQ, oRNSDQ, RNSDQ_RNSC), neon_mac_maybe_scalar),
- NCE(vldm, c900b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
- NCE(vldmia, c900b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
- NCE(vldmdb, d100b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
- NCE(vstm, c800b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
- NCE(vstmia, c800b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
- NCE(vstmdb, d000b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
-
mnCEF(vcvt, _vcvt, 3, (RNSDQMQ, RNSDQMQ, oI32z), neon_cvt),
nCEF(vcvtr, _vcvt, 2, (RNSDQ, RNSDQ), neon_cvtr),
MNCEF(vcvtb, eb20a40, 3, (RVSDMQ, RVSDMQ, oI32b), neon_cvtb),
diff --git a/gas/testsuite/gas/arm/v8_1m-mve.d b/gas/testsuite/gas/arm/v8_1m-mve.d
new file mode 100644
index 00000000000..4c528de073f
--- /dev/null
+++ b/gas/testsuite/gas/arm/v8_1m-mve.d
@@ -0,0 +1,27 @@
+# name: V8.1-m FP register instructions enabled by +mve
+# as: -march=armv8.1-m.main+mve
+# objdump: -dr --show-raw-insn -marmv8.1-m.main
+
+.*: +file format .*arm.*
+
+
+Disassembly of section .text:
+
+00000000 <\.text>:
+ *[0-9a-f]+: ec80 0b08 vstmia r0, {d0-d3}
+ *[0-9a-f]+: ecb7 3b04 vldmia r7!, {d3-d4}
+ *[0-9a-f]+: ecbd 0b06 vpop {d0-d2}
+ *[0-9a-f]+: ed2d 0b06 vpush {d0-d2}
+ *[0-9a-f]+: ecbd 2b08 vpop {d2-d5}
+ *[0-9a-f]+: ed2d 1b0c vpush {d1-d6}
+ *[0-9a-f]+: fe71 0f4d vpst
+ *[0-9a-f]+: fd00 3e01 vstrwt\.32 q1, \[q0, #-4\]
+ *[0-9a-f]+: ed82 2f80 vstr FPSCR, \[r2\]
+ *[0-9a-f]+: ed80 0b00 vstr d0, \[r0\]
+ *[0-9a-f]+: ed90 0b00 vldr d0, \[r0\]
+ *[0-9a-f]+: ed80 0a00 vstr s0, \[r0\]
+ *[0-9a-f]+: ed90 0a00 vldr s0, \[r0\]
+ *[0-9a-f]+: ed81 fb00 vstr d15, \[r1\]
+ *[0-9a-f]+: ed91 fb00 vldr d15, \[r1\]
+ *[0-9a-f]+: edc1 fa00 vstr s31, \[r1\]
+ *[0-9a-f]+: edd1 fa00 vldr s31, \[r1\]
diff --git a/gas/testsuite/gas/arm/v8_1m-mve.s b/gas/testsuite/gas/arm/v8_1m-mve.s
new file mode 100644
index 00000000000..cae1f93c158
--- /dev/null
+++ b/gas/testsuite/gas/arm/v8_1m-mve.s
@@ -0,0 +1,24 @@
+.syntax unified
+
+vstmia r0,{d0-d3}
+vldmia r7!, {d3-d4}
+
+vpop {d0-d2}
+vpush {d0-d2}
+vpop {d2-d5}
+vpush {d1-d6}
+
+vpst
+vstrwt.u32 q1, [q0, #-4]
+
+vstr FPSCR, [r2] @ Accepts offset variant without immediate
+
+vstr d0,[r0]
+vldr d0,[r0]
+vstr s0,[r0]
+vldr s0,[r0]
+
+vstr d15,[r1]
+vldr d15,[r1]
+vstr s31,[r1]
+vldr s31,[r1]