[PATCHv2] gdb/riscv: add support for vector registers in target descriptions
Andrew Burgess
andrew.burgess@embecosm.com
Mon Jun 21 11:04:20 GMT 2021
Thanks for feedback on v1.
I agree that there is a very small risk that the vector register set
could change for the final version of the spec, but like you both seem
to say, this seems unlikely.
And even if it did change I don't think changing the requirements of
the feature for a later release of GDB would be that bad.
So, I'm proposing that once this has had a doc review, I'll get this
merged, hopefully in time for the next release (of GDB).
In v2:
- Fixed the f-register/v-register typo that Jim pointed out,
- Added documentation of the new feature to the manual,
- Added a NEWS entry,
- Updated the commit message to mention that the vector extension is
still in draft.
Thanks,
Andrew
---
commit 2a909530fd8b98ee31fb59ba9019a641029cff35
Author: Andrew Burgess <andrew.burgess@embecosm.com>
Date: Tue May 4 11:41:09 2021 +0100
gdb/riscv: add support for vector registers in target descriptions
This commit adds support to RISC-V GDB for vector registers in the
incoming target description.
The vector registers should be described in a feature called
"org.gnu.gdb.riscv.vector", and should contain the register v0 to
v31. There's no restriction on the size or type of these registers,
so the target description can set these up as it requires.
However, if the target feature is present then all of the registers
must be present, and they must all be the same size, these
requirements are, I believe, inline with the RISC-V vector extension.
The DWARF register numbers for the vector registers have been added,
and the code to map between GDB's internal numbering and the DWARF
numbering has been updated.
I have not yet added a feature/riscv/*.xml file for the vector
extension, the consequence of this is that we can't, right now, detect
vector registers on a native target, this patch is all about
supporting vectors on a remote target.
It is worth noting that I don't actually have access to a RISC-V
target with vectors, so the only testing that this patch has had has
been done using 'set tdesc filename ....' to load a target description
to which I have manually added the vector feature. This has shown
that the vector register feature can be successfully parsed, and that
the registers show up in the expected register groups.
Additionally, the RISC-V vector extension is currently at v0.10, which
is also the v1.0 draft release. However, this extension is not yet
finalised. It is possible (but unlikely I think) that the register
set could change between now and the final release of the vector
extension. If this were to happen then we would potentially end up
changing the requirements for the new org.gnu.gdb.riscv.vector
feature. I really don't think it is likely that the register set will
change this late in the process, and even if it did, changing the
feature requirements will not be a problem as far as I am
concerned (when the alternative is GDB just continues without this
feature for now).
gdb/ChangeLog:
* NEWS: Mention new target feature name.
* arch/riscv.c (riscv_create_target_description): GDB doesn't
currently create target descriptions containing vector registers.
* arch/riscv.h (struct riscv_gdbarch_features) <vlen>: New member
variable.
<operator==>: Also compare vlen.
<hash>: Also include vlen.
* riscv-tdep.c (riscv_feature_name_vector): New static global.
(struct riscv_vector_feature): New struct.
(riscv_vector_feature): New static global.
(riscv_register_reggroup_p): Ensure vector registers are part of
the 'all' group, and part of the 'vector' group.
(riscv_dwarf_reg_to_regnum): Handle vector registers.
(riscv_gdbarch_init): Check vector register feature.
* riscv-tdep.h: Add vector registers to GDB's internal register
numbers, and to the DWARF register numbers.
gdb/doc/ChangeLog:
* gdb.texinfo (RISC-V Features): Mention vector register feature.
diff --git a/gdb/NEWS b/gdb/NEWS
index 56743fc9aea..6bf5e44da99 100644
--- a/gdb/NEWS
+++ b/gdb/NEWS
@@ -78,6 +78,11 @@
and "-eiex" that allow options (that would normally appear in a
gdbearlyinit file) to be passed on the command line.
+* For RISC-V targets, the target feature "org.gnu.gdb.riscv.vector" is
+ now understood by GDB, and can be used to describe the vector
+ registers of a target. The precise requirements of this register
+ feature are documented in the GDB manual.
+
* New commands
set debug event-loop
diff --git a/gdb/arch/riscv.c b/gdb/arch/riscv.c
index 8fbcad1a8ea..85d60a3e257 100644
--- a/gdb/arch/riscv.c
+++ b/gdb/arch/riscv.c
@@ -84,6 +84,12 @@ riscv_create_target_description (const struct riscv_gdbarch_features features)
else if (features.flen == 8)
regnum = create_feature_riscv_64bit_fpu (tdesc.get (), regnum);
+ /* Currently GDB only supports vector features coming from remote
+ targets. We don't support creating vector features on native targets
+ (yet). */
+ if (features.vlen != 0)
+ error (_("unable to create vector feature"));
+
return tdesc;
}
diff --git a/gdb/arch/riscv.h b/gdb/arch/riscv.h
index faa038a713e..65a998bb2b9 100644
--- a/gdb/arch/riscv.h
+++ b/gdb/arch/riscv.h
@@ -46,13 +46,20 @@ struct riscv_gdbarch_features
that there are no f-registers. No other value is valid. */
int flen = 0;
+ /* The size of the v-registers in bytes. The value 0 indicates a target
+ with no vector registers. The minimum value for a standard compliant
+ target should be 16, but GDB doesn't currently mind, and will accept
+ any vector size. */
+ int vlen = 0;
+
/* When true this target is RV32E. */
bool embedded = false;
/* Equality operator. */
bool operator== (const struct riscv_gdbarch_features &rhs) const
{
- return (xlen == rhs.xlen && flen == rhs.flen && embedded == rhs.embedded);
+ return (xlen == rhs.xlen && flen == rhs.flen
+ && embedded == rhs.embedded && vlen == rhs.vlen);
}
/* Inequality operator. */
@@ -66,7 +73,8 @@ struct riscv_gdbarch_features
{
std::size_t val = ((embedded ? 1 : 0) << 10
| (xlen & 0x1f) << 5
- | (flen & 0x1f) << 0);
+ | (flen & 0x1f) << 0
+ | (vlen & 0xfff) << 11);
return val;
}
};
diff --git a/gdb/doc/gdb.texinfo b/gdb/doc/gdb.texinfo
index 7cdc83b9537..dfc00b120b4 100644
--- a/gdb/doc/gdb.texinfo
+++ b/gdb/doc/gdb.texinfo
@@ -46349,6 +46349,14 @@
feature if the target has the floating point control registers, but no
other floating point hardware.
+The @samp{org.gnu.gdb.riscv.vector} feature is optional. If present,
+it should contain registers @samp{v0} through @samp{v31}, all of which
+must be the same size. These requirements are based on the v0.10
+draft vector extension, as the vector extension is not yet final. In
+the event that the register set of the vector extension changes for
+the final specification, the requirements given here could change for
+future releases of @value{GDBN}.
+
@node RX Features
@subsection RX Features
@cindex target descriptions, RX Features
diff --git a/gdb/riscv-tdep.c b/gdb/riscv-tdep.c
index 7fae9140155..d942b059f55 100644
--- a/gdb/riscv-tdep.c
+++ b/gdb/riscv-tdep.c
@@ -99,6 +99,7 @@ const char *riscv_feature_name_csr = "org.gnu.gdb.riscv.csr";
static const char *riscv_feature_name_cpu = "org.gnu.gdb.riscv.cpu";
static const char *riscv_feature_name_fpu = "org.gnu.gdb.riscv.fpu";
static const char *riscv_feature_name_virtual = "org.gnu.gdb.riscv.virtual";
+static const char *riscv_feature_name_vector = "org.gnu.gdb.riscv.vector";
/* Cached information about a frame. */
@@ -575,6 +576,115 @@ struct riscv_csr_feature : public riscv_register_feature
static const struct riscv_csr_feature riscv_csr_feature;
+/* Class representing the v-registers feature set. */
+
+struct riscv_vector_feature : public riscv_register_feature
+{
+ riscv_vector_feature ()
+ : riscv_register_feature (riscv_feature_name_vector)
+ {
+ m_registers = {
+ { RISCV_V0_REGNUM + 0, { "v0" } },
+ { RISCV_V0_REGNUM + 1, { "v1" } },
+ { RISCV_V0_REGNUM + 2, { "v2" } },
+ { RISCV_V0_REGNUM + 3, { "v3" } },
+ { RISCV_V0_REGNUM + 4, { "v4" } },
+ { RISCV_V0_REGNUM + 5, { "v5" } },
+ { RISCV_V0_REGNUM + 6, { "v6" } },
+ { RISCV_V0_REGNUM + 7, { "v7" } },
+ { RISCV_V0_REGNUM + 8, { "v8" } },
+ { RISCV_V0_REGNUM + 9, { "v9" } },
+ { RISCV_V0_REGNUM + 10, { "v10" } },
+ { RISCV_V0_REGNUM + 11, { "v11" } },
+ { RISCV_V0_REGNUM + 12, { "v12" } },
+ { RISCV_V0_REGNUM + 13, { "v13" } },
+ { RISCV_V0_REGNUM + 14, { "v14" } },
+ { RISCV_V0_REGNUM + 15, { "v15" } },
+ { RISCV_V0_REGNUM + 16, { "v16" } },
+ { RISCV_V0_REGNUM + 17, { "v17" } },
+ { RISCV_V0_REGNUM + 18, { "v18" } },
+ { RISCV_V0_REGNUM + 19, { "v19" } },
+ { RISCV_V0_REGNUM + 20, { "v20" } },
+ { RISCV_V0_REGNUM + 21, { "v21" } },
+ { RISCV_V0_REGNUM + 22, { "v22" } },
+ { RISCV_V0_REGNUM + 23, { "v23" } },
+ { RISCV_V0_REGNUM + 24, { "v24" } },
+ { RISCV_V0_REGNUM + 25, { "v25" } },
+ { RISCV_V0_REGNUM + 26, { "v26" } },
+ { RISCV_V0_REGNUM + 27, { "v27" } },
+ { RISCV_V0_REGNUM + 28, { "v28" } },
+ { RISCV_V0_REGNUM + 29, { "v29" } },
+ { RISCV_V0_REGNUM + 30, { "v30" } },
+ { RISCV_V0_REGNUM + 31, { "v31" } },
+ };
+ }
+
+ /* Return the preferred name for the register with gdb register number
+ REGNUM, which must be in the inclusive range RISCV_V0_REGNUM to
+ RISCV_V0_REGNUM + 31. */
+ const char *register_name (int regnum) const
+ {
+ gdb_assert (regnum >= RISCV_V0_REGNUM
+ && regnum <= RISCV_V0_REGNUM + 31);
+ regnum -= RISCV_V0_REGNUM;
+ return m_registers[regnum].names[0];
+ }
+
+ /* Check this feature within TDESC, record the registers from this
+ feature into TDESC_DATA and update ALIASES and FEATURES. */
+ bool check (const struct target_desc *tdesc,
+ struct tdesc_arch_data *tdesc_data,
+ std::vector<riscv_pending_register_alias> *aliases,
+ struct riscv_gdbarch_features *features) const
+ {
+ const struct tdesc_feature *feature_vector = tdesc_feature (tdesc);
+
+ /* It's fine if this feature is missing. Update the architecture
+ feature set and return. */
+ if (feature_vector == nullptr)
+ {
+ features->vlen = 0;
+ return true;
+ }
+
+ /* Check all of the vector registers are present. */
+ for (const auto ® : m_registers)
+ {
+ if (!reg.check (tdesc_data, feature_vector, true, aliases))
+ return false;
+ }
+
+ /* Look through all of the vector registers and check they all have the
+ same bitsize. Use this bitsize to update the feature set for this
+ gdbarch. */
+ int vector_bitsize = -1;
+ for (const auto ® : m_registers)
+ {
+ int reg_bitsize = -1;
+ for (const char *name : reg.names)
+ {
+ if (tdesc_unnumbered_register (feature_vector, name))
+ {
+ reg_bitsize = tdesc_register_bitsize (feature_vector, name);
+ break;
+ }
+ }
+ gdb_assert (reg_bitsize != -1);
+ if (vector_bitsize == -1)
+ vector_bitsize = reg_bitsize;
+ else if (vector_bitsize != reg_bitsize)
+ return false;
+ }
+
+ features->vlen = (vector_bitsize / 8);
+ return true;
+ }
+};
+
+/* An instance of the v-register feature set. */
+
+static const struct riscv_vector_feature riscv_vector_feature;
+
/* Controls whether we place compressed breakpoints or not. When in auto
mode GDB tries to determine if the target supports compressed
breakpoints, and uses them if it does. */
@@ -1192,7 +1302,7 @@ riscv_register_reggroup_p (struct gdbarch *gdbarch, int regnum,
if (reggroup == all_reggroup)
{
- if (regnum < RISCV_FIRST_CSR_REGNUM || regnum == RISCV_PRIV_REGNUM)
+ if (regnum < RISCV_FIRST_CSR_REGNUM || regnum >= RISCV_PRIV_REGNUM)
return 1;
if (riscv_is_regnum_a_named_csr (regnum))
return 1;
@@ -1226,7 +1336,7 @@ riscv_register_reggroup_p (struct gdbarch *gdbarch, int regnum,
return 0;
}
else if (reggroup == vector_reggroup)
- return 0;
+ return (regnum >= RISCV_V0_REGNUM && regnum <= RISCV_V31_REGNUM);
else
return 0;
}
@@ -3320,6 +3430,9 @@ riscv_dwarf_reg_to_regnum (struct gdbarch *gdbarch, int reg)
else if (reg >= RISCV_DWARF_FIRST_CSR && reg <= RISCV_DWARF_LAST_CSR)
return RISCV_FIRST_CSR_REGNUM + (reg - RISCV_DWARF_FIRST_CSR);
+ else if (reg >= RISCV_DWARF_REGNUM_V0 && reg <= RISCV_DWARF_REGNUM_V31)
+ return RISCV_V0_REGNUM + (reg - RISCV_DWARF_REGNUM_V0);
+
return -1;
}
@@ -3488,7 +3601,9 @@ riscv_gdbarch_init (struct gdbarch_info info,
&& riscv_virtual_feature.check (tdesc, tdesc_data.get (),
&pending_aliases, &features)
&& riscv_csr_feature.check (tdesc, tdesc_data.get (),
- &pending_aliases, &features));
+ &pending_aliases, &features)
+ && riscv_vector_feature.check (tdesc, tdesc_data.get (),
+ &pending_aliases, &features));
if (!valid_p)
{
if (riscv_debug_gdbarch)
diff --git a/gdb/riscv-tdep.h b/gdb/riscv-tdep.h
index 154097df5d0..62bf4797d02 100644
--- a/gdb/riscv-tdep.h
+++ b/gdb/riscv-tdep.h
@@ -53,7 +53,11 @@ enum
RISCV_PRIV_REGNUM = 4161,
- RISCV_LAST_REGNUM = RISCV_PRIV_REGNUM
+ RISCV_V0_REGNUM,
+
+ RISCV_V31_REGNUM = RISCV_V0_REGNUM + 31,
+
+ RISCV_LAST_REGNUM = RISCV_V31_REGNUM
};
/* RiscV DWARF register numbers. */
@@ -63,6 +67,8 @@ enum
RISCV_DWARF_REGNUM_X31 = 31,
RISCV_DWARF_REGNUM_F0 = 32,
RISCV_DWARF_REGNUM_F31 = 63,
+ RISCV_DWARF_REGNUM_V0 = 96,
+ RISCV_DWARF_REGNUM_V31 = 127,
RISCV_DWARF_FIRST_CSR = 4096,
RISCV_DWARF_LAST_CSR = 8191,
};
More information about the Gdb-patches
mailing list