xref: /llvm-project/llvm/docs/RISCVUsage.rst (revision 0cb7636a462a8d4209e2b6344304eb43f02853eb)
1=============================
2User Guide for RISC-V Target
3=============================
4
5.. contents::
6   :local:
7
8Introduction
9============
10
11The RISC-V target provides code generation for processors implementing
12supported variations of the RISC-V specification.  It lives in the
13``llvm/lib/Target/RISCV`` directory.
14
15Specification Documents
16=======================
17
18There have been a number of revisions to the RISC-V specifications. LLVM aims
19to implement the most recent ratified version of the standard RISC-V base ISAs
20and ISA extensions with pragmatic variances. The most recent specification can
21be found at: https://github.com/riscv/riscv-isa-manual/releases/.
22
23`The official RISC-V International specification page
24<https://riscv.org/technical/specifications/>`__. is also worth checking, but
25tends to significantly lag the specifications linked above. Make sure to check
26the `wiki for not yet integrated extensions
27<https://wiki.riscv.org/display/HOME/Recently+Ratified+Extensions>`__ and note
28that in addition, we sometimes carry support for extensions that have not yet
29been ratified (these will be marked as experimental - see below) and support
30various vendor-specific extensions (see below).
31
32The current known variances from the specification are:
33
34* Unconditionally allowing instructions from zifencei, zicsr, zicntr, and
35  zihpm without gating them on the extensions being enabled.  Previous
36  revisions of the specification included these instructions in the base
37  ISA, and we preserve this behavior to avoid breaking existing code.  If
38  a future revision of the specification reuses these opcodes for other
39  extensions, we may need to reevaluate this choice, and thus recommend
40  users migrate build systems so as not to rely on this.
41* Allowing CSRs to be named without gating on specific extensions.  This
42  applies to all CSR names, not just those in zicsr, zicntr, and zihpm.
43* The ordering of ``z*``, ``s*``, and ``x*`` prefixed extension names is not
44  enforced in user-specified ISA naming strings (e.g. ``-march``).
45
46We are actively deciding not to support multiple specification revisions
47at this time. We acknowledge a likely future need, but actively defer the
48decisions making around handling this until we have a concrete example of
49real hardware having shipped and an incompatible change to the
50specification made afterwards.
51
52Base ISAs
53=========
54
55The specification defines five base instruction sets: RV32I, RV32E, RV64I,
56RV64E, and RV128I. Currently, LLVM fully supports RV32I, and RV64I.  RV32E and
57RV64E are supported by the assembly-based tools only.  RV128I is not supported.
58
59To specify the target triple:
60
61  .. table:: RISC-V Architectures
62
63     ============ ==============================================================
64     Architecture Description
65     ============ ==============================================================
66     ``riscv32``   RISC-V with XLEN=32 (i.e. RV32I or RV32E)
67     ``riscv64``   RISC-V with XLEN=64 (i.e. RV64I or RV64E)
68     ============ ==============================================================
69
70To select an E variant ISA (e.g. RV32E instead of RV32I), use the base
71architecture string (e.g. ``riscv32``) with the extension ``e``.
72
73Profiles
74========
75
76Supported profile names can be passed using ``-march`` instead of a standard
77ISA naming string. Currently supported profiles:
78
79* ``rvi20u32``
80* ``rvi20u64``
81* ``rva20u64``
82* ``rva20s64``
83* ``rva22u64``
84* ``rva22s64``
85* ``rva23u64``
86* ``rva23s64``
87* ``rvb23u64``
88* ``rvb23s64``
89
90Note that you can also append additional extension names to be enabled, e.g.
91``rva20u64_zicond`` will enable the ``zicond`` extension in addition to those
92in the ``rva20u64`` profile.
93
94Profiles that are not yet ratified cannot be used unless
95``-menable-experimental-extensions`` (or equivalent for other tools) is
96specified. This applies to the following profiles:
97
98* ``rvm23u32``
99
100.. _riscv-extensions:
101
102Extensions
103==========
104
105The following table provides a status summary for extensions which have been
106ratified and thus have finalized specifications.  When relevant, detailed notes
107on support follow.
108
109  .. table:: Ratified Extensions by Status
110
111     ================  =================================================================
112     Extension         Status
113     ================  =================================================================
114     ``A``             Supported
115     ``B``             Supported
116     ``C``             Supported
117     ``D``             Supported
118     ``F``             Supported
119     ``E``             Supported (`See note <#riscv-rve-note>`__)
120     ``H``             Assembly Support
121     ``M``             Supported
122     ``Sha``           Supported
123     ``Shcounterenw``  Assembly Support (`See note <#riscv-profiles-extensions-note>`__)
124     ``Shgatpa``       Assembly Support (`See note <#riscv-profiles-extensions-note>`__)
125     ``Shtvala``       Assembly Support (`See note <#riscv-profiles-extensions-note>`__)
126     ``Shvsatpa``      Assembly Support (`See note <#riscv-profiles-extensions-note>`__)
127     ``Shvstvala``     Assembly Support (`See note <#riscv-profiles-extensions-note>`__)
128     ``Shvstvecd``     Assembly Support (`See note <#riscv-profiles-extensions-note>`__)
129     ``Smaia``         Supported
130     ``Smcdeleg``      Supported
131     ``Smcsrind``      Supported
132     ``Smdbltrp``      Supported
133     ``Smepmp``        Supported
134     ``Smmpm``         Supported
135     ``Smnpm``         Supported
136     ``Smrnmi``        Assembly Support
137     ``Smstateen``     Assembly Support
138     ``Ssaia``         Supported
139     ``Ssccfg``        Supported
140     ``Ssccptr``       Assembly Support (`See note <#riscv-profiles-extensions-note>`__)
141     ``Sscofpmf``      Assembly Support
142     ``Sscounterenw``  Assembly Support (`See note <#riscv-profiles-extensions-note>`__)
143     ``Sscsrind``      Supported
144     ``Ssdbltrp``      Supported
145     ``Ssnpm``         Supported
146     ``Sspm``          Supported
147     ``Ssqosid``       Assembly Support
148     ``Ssstateen``     Assembly Support (`See note <#riscv-profiles-extensions-note>`__)
149     ``Ssstrict``      Assembly Support (`See note <#riscv-profiles-extensions-note>`__)
150     ``Sstc``          Assembly Support
151     ``Sstvala``       Assembly Support (`See note <#riscv-profiles-extensions-note>`__)
152     ``Sstvecd``       Assembly Support (`See note <#riscv-profiles-extensions-note>`__)
153     ``Ssu64xl``       Assembly Support (`See note <#riscv-profiles-extensions-note>`__)
154     ``Supm``          Supported
155     ``Svade``         Assembly Support (`See note <#riscv-profiles-extensions-note>`__)
156     ``Svadu``         Assembly Support
157     ``Svbare``        Assembly Support (`See note <#riscv-profiles-extensions-note>`__)
158     ``Svinval``       Assembly Support
159     ``Svnapot``       Assembly Support
160     ``Svpbmt``        Supported
161     ``Svvptc``        Supported
162     ``V``             Supported
163     ``Za128rs``       Supported (`See note <#riscv-profiles-extensions-note>`__)
164     ``Za64rs``        Supported (`See note <#riscv-profiles-extensions-note>`__)
165     ``Zaamo``         Assembly Support
166     ``Zabha``         Supported
167     ``Zacas``         Supported (`See note <#riscv-zacas-note>`__)
168     ``Zalrsc``        Assembly Support
169     ``Zama16b``       Supported (`See note <#riscv-profiles-extensions-note>`__)
170     ``Zawrs``         Assembly Support
171     ``Zba``           Supported
172     ``Zbb``           Supported
173     ``Zbc``           Supported
174     ``Zbkb``          Supported (`See note <#riscv-scalar-crypto-note1>`__)
175     ``Zbkc``          Supported
176     ``Zbkx``          Supported (`See note <#riscv-scalar-crypto-note1>`__)
177     ``Zbs``           Supported
178     ``Zca``           Supported
179     ``Zcb``           Supported
180     ``Zcd``           Supported
181     ``Zcf``           Supported
182     ``Zcmop``         Supported
183     ``Zcmp``          Supported
184     ``Zcmt``          Assembly Support
185     ``Zdinx``         Supported
186     ``Zfa``           Supported
187     ``Zfbfmin``       Supported
188     ``Zfh``           Supported
189     ``Zfhmin``        Supported
190     ``Zfinx``         Supported
191     ``Zhinx``         Supported
192     ``Zhinxmin``      Supported
193     ``Zic64b``        Supported (`See note <#riscv-profiles-extensions-note>`__)
194     ``Zicbom``        Assembly Support
195     ``Zicbop``        Supported
196     ``Zicboz``        Assembly Support
197     ``Ziccamoa``      Supported (`See note <#riscv-profiles-extensions-note>`__)
198     ``Ziccif``        Supported (`See note <#riscv-profiles-extensions-note>`__)
199     ``Zicclsm``       Supported (`See note <#riscv-profiles-extensions-note>`__)
200     ``Ziccrse``       Supported (`See note <#riscv-profiles-extensions-note>`__)
201     ``Zicntr``        (`See Note <#riscv-i2p1-note>`__)
202     ``Zicond``        Supported
203     ``Zicsr``         (`See Note <#riscv-i2p1-note>`__)
204     ``Zifencei``      (`See Note <#riscv-i2p1-note>`__)
205     ``Zihintntl``     Supported
206     ``Zihintpause``   Assembly Support
207     ``Zihpm``         (`See Note <#riscv-i2p1-note>`__)
208     ``Zimop``         Supported
209     ``Zkn``           Supported
210     ``Zknd``          Supported (`See note <#riscv-scalar-crypto-note2>`__)
211     ``Zkne``          Supported (`See note <#riscv-scalar-crypto-note2>`__)
212     ``Zknh``          Supported (`See note <#riscv-scalar-crypto-note2>`__)
213     ``Zksed``         Supported (`See note <#riscv-scalar-crypto-note2>`__)
214     ``Zksh``          Supported (`See note <#riscv-scalar-crypto-note2>`__)
215     ``Zk``            Supported
216     ``Zkr``           Supported
217     ``Zks``           Supported
218     ``Zkt``           Supported
219     ``Zmmul``         Supported
220     ``Ztso``          Supported
221     ``Zvbb``          Supported
222     ``Zvbc``          Supported (`See note <#riscv-vector-crypto-note>`__)
223     ``Zve32x``        (`Partially <#riscv-vlen-32-note>`__) Supported
224     ``Zve32f``        (`Partially <#riscv-vlen-32-note>`__) Supported
225     ``Zve64x``        Supported
226     ``Zve64f``        Supported
227     ``Zve64d``        Supported
228     ``Zvfbfmin``      Supported
229     ``Zvfbfwma``      Supported
230     ``Zvfh``          Supported
231     ``Zvfhmin``       Supported
232     ``Zvkb``          Supported
233     ``Zvkg``          Supported (`See note <#riscv-vector-crypto-note>`__)
234     ``Zvkn``          Supported (`See note <#riscv-vector-crypto-note>`__)
235     ``Zvknc``         Supported (`See note <#riscv-vector-crypto-note>`__)
236     ``Zvkned``        Supported (`See note <#riscv-vector-crypto-note>`__)
237     ``Zvkng``         Supported (`See note <#riscv-vector-crypto-note>`__)
238     ``Zvknha``        Supported (`See note <#riscv-vector-crypto-note>`__)
239     ``Zvknhb``        Supported (`See note <#riscv-vector-crypto-note>`__)
240     ``Zvks``          Supported (`See note <#riscv-vector-crypto-note>`__)
241     ``Zvksc``         Supported (`See note <#riscv-vector-crypto-note>`__)
242     ``Zvksed``        Supported (`See note <#riscv-vector-crypto-note>`__)
243     ``Zvksg``         Supported (`See note <#riscv-vector-crypto-note>`__)
244     ``Zvksh``         Supported (`See note <#riscv-vector-crypto-note>`__)
245     ``Zvkt``          Supported
246     ``Zvl32b``        (`Partially <#riscv-vlen-32-note>`__) Supported
247     ``Zvl64b``        Supported
248     ``Zvl128b``       Supported
249     ``Zvl256b``       Supported
250     ``Zvl512b``       Supported
251     ``Zvl1024b``      Supported
252     ``Zvl2048b``      Supported
253     ``Zvl4096b``      Supported
254     ``Zvl8192b``      Supported
255     ``Zvl16384b``     Supported
256     ``Zvl32768b``     Supported
257     ``Zvl65536b``     Supported
258     ================  =================================================================
259
260Assembly Support
261  LLVM supports the associated instructions in assembly.  All assembly related tools (e.g. assembler, disassembler, llvm-objdump, etc..) are supported.  Compiler and linker will accept extension names, and linked binaries will contain appropriate ELF flags and attributes to reflect use of named extension.
262
263Supported
264  Fully supported by the compiler.  This includes everything in Assembly Support, along with - if relevant - C language intrinsics for the instructions and pattern matching by the compiler to recognize idiomatic patterns which can be lowered to the associated instructions.
265
266.. _riscv-rve-note:
267
268``E``
269  Support of RV32E/RV64E and ilp32e/lp64e ABIs are experimental. To be compatible with the implementation of ilp32e in GCC, we don't use aligned registers to pass variadic arguments. Furthermore, we set the stack alignment to 4 bytes for types with length of 2*XLEN.
270
271.. _riscv-scalar-crypto-note1:
272
273``Zbkb``, ``Zbkx``
274  Pattern matching support for these instructions is incomplete.
275
276.. _riscv-scalar-crypto-note2:
277
278``Zknd``, ``Zkne``, ``Zknh``, ``Zksed``, ``Zksh``
279  No pattern matching exists.  As a result, these instructions can only be used from assembler or via intrinsic calls.
280
281.. _riscv-vector-crypto-note:
282
283``Zvbc``, ``Zvkg``, ``Zvkn``, ``Zvknc``, ``Zvkned``, ``Zvkng``, ``Zvknha``, ``Zvknhb``, ``Zvks``, ``Zvks``, ``Zvks``, ``Zvksc``, ``Zvksed``, ``Zvksg``, ``Zvksh``.
284  No pattern matching exists. As a result, these instructions can only be used from assembler or via intrinsic calls.
285
286.. _riscv-vlen-32-note:
287
288``Zve32x``, ``Zve32f``, ``Zvl32b``
289  LLVM currently assumes a minimum VLEN (vector register width) of 64 bits during compilation, and as a result ``Zve32x`` and ``Zve32f`` are supported only for VLEN>=64.  Assembly support doesn't have this restriction.
290
291.. _riscv-i2p1-note:
292
293``Zicntr``, ``Zicsr``, ``Zifencei``, ``Zihpm``
294  Between versions 2.0 and 2.1 of the base I specification, a backwards incompatible change was made to remove selected instructions and CSRs from the base ISA.  These instructions were grouped into a set of new extensions, but were no longer required by the base ISA.  This change is partially described in "Preface to Document Version 20190608-Base-Ratified" from the specification document (the ``zicntr`` and ``zihpm`` bits are not mentioned).  LLVM currently implements version 2.1 of the base specification. To maintain compatibility, instructions from these extensions are accepted without being in the ``-march`` string.  LLVM also allows the explicit specification of the extensions in an ``-march`` string.
295
296.. _riscv-profiles-extensions-note:
297
298``Za128rs``, ``Za64rs``, ``Zama16b``, ``Zic64b``, ``Ziccamoa``, ``Ziccif``, ``Zicclsm``, ``Ziccrse``, ``Shcounterenvw``, ``Shgatpa``, ``Shtvala``, ``Shvsatpa``, ``Shvstvala``, ``Shvstvecd``, ``Ssccptr``, ``Sscounterenw``, ``Ssstateen``, ``Ssstrict``, ``Sstvala``, ``Sstvecd``, ``Ssu64xl``, ``Svade``, ``Svbare``
299  These extensions are defined as part of the `RISC-V Profiles specification <https://github.com/riscv/riscv-profiles/releases/tag/v1.0>`__.  They do not introduce any new features themselves, but instead describe existing hardware features.
300
301.. _riscv-zacas-note:
302
303``Zacas``
304  The compiler will not generate amocas.d on RV32 or amocas.q on RV64 due to ABI compatibilty. These can only be used in the assembler.
305
306Atomics ABIs
307============
308
309At the time of writing there are three atomics mappings (ABIs) `defined for RISC-V <https://github.com/riscv-non-isa/riscv-elf-psabi-doc/blob/master/riscv-elf.adoc#tag_riscv_atomic_abi-14-uleb128version>`__.  As of LLVM 19, LLVM defaults to "A6S", which is compatible with both the original "A6" and the future "A7" ABI. See `the psABI atomics document <https://github.com/riscv-non-isa/riscv-elf-psabi-doc/blob/master/riscv-atomic.adoc>`__ for more information on these mappings.
310
311Note that although the "A6S" mapping is used, the ELF attribute recording the mapping isn't currently emitted by default due to a bug causing a crash in older versions of binutils when processing files containing this attribute.
312
313Experimental Extensions
314=======================
315
316LLVM supports (to various degrees) a number of experimental extensions.  All experimental extensions have ``experimental-`` as a prefix.  There is explicitly no compatibility promised between versions of the toolchain, and regular users are strongly advised *not* to make use of experimental extensions before they reach ratification.
317
318The primary goal of experimental support is to assist in the process of ratification by providing an existence proof of an implementation, and simplifying efforts to validate the value of a proposed extension against large code bases.  Experimental extensions are expected to either transition to ratified status, or be eventually removed.  The decision on whether to accept an experimental extension is currently done on an entirely case by case basis; if you want to propose one, attending the bi-weekly RISC-V sync-up call is strongly advised.
319
320``experimental-zalasr``
321  LLVM implements the `0.0.5 draft specification <https://github.com/mehnadnerd/riscv-zalasr>`__.
322
323``experimental-zicfilp``, ``experimental-zicfiss``
324  LLVM implements the `1.0 release specification <https://github.com/riscv/riscv-cfi/releases/tag/v1.0>`__.
325
326``experimental-zvbc32e``, ``experimental-zvkgs``
327  LLVM implements the `0.7 release specification <https://github.com/user-attachments/files/16450464/riscv-crypto-spec-vector-extra_v0.0.7.pdf>`__.
328
329``experimental-sdext``, ``experimental-sdtrig``
330  LLVM implements the `1.0-rc4 specification <https://github.com/riscv/riscv-debug-spec/releases/download/1.0.0-rc4/riscv-debug-specification.pdf>`__.
331
332``experimental-smctr``, ``experimental-ssctr``
333  LLVM implements the `1.0-rc3 specification <https://github.com/riscv/riscv-control-transfer-records/releases/tag/v1.0_rc3>`__.
334
335``experimental-svukte``
336  LLVM implements the `0.3 draft specification <https://github.com/riscv/riscv-isa-manual/pull/1564>`__.
337
338To use an experimental extension from `clang`, you must add `-menable-experimental-extensions` to the command line, and specify the exact version of the experimental extension you are using.  To use an experimental extension with LLVM's internal developer tools (e.g. `llc`, `llvm-objdump`, `llvm-mc`), you must prefix the extension name with `experimental-`.  Note that you don't need to specify the version with internal tools, and shouldn't include the `experimental-` prefix with `clang`.
339
340Vendor Extensions
341=================
342
343Vendor extensions are extensions which are not standardized by RISC-V International, and are instead defined by a hardware vendor.  The term vendor extension roughly parallels the definition of a `non-standard` extension from Section 1.3 of the Volume I: RISC-V Unprivileged ISA specification.  In particular, we expect to eventually accept both `custom` extensions and `non-conforming` extensions.
344
345Inclusion of a vendor extension will be considered on a case by case basis.  All proposals should be brought to the bi-weekly RISCV sync calls for discussion.  For a general idea of the factors likely to be considered, please see the `Clang documentation <https://clang.llvm.org/get_involved.html>`__.
346
347It is our intention to follow the naming conventions described in `riscv-non-isa/riscv-toolchain-conventions <https://github.com/riscv-non-isa/riscv-toolchain-conventions#conventions-for-vendor-extensions>`__.  Exceptions to this naming will need to be strongly motivated.
348
349The current vendor extensions supported are:
350
351``XTHeadBa``
352  LLVM implements `the THeadBa (address-generation) vendor-defined instructions specified in <https://github.com/T-head-Semi/thead-extension-spec/releases/download/2.2.2/xthead-2023-01-30-2.2.2.pdf>`__ by T-HEAD of Alibaba.  Instructions are prefixed with `th.` as described in the specification.
353
354``XTHeadBb``
355  LLVM implements `the THeadBb (basic bit-manipulation) vendor-defined instructions specified in <https://github.com/T-head-Semi/thead-extension-spec/releases/download/2.2.2/xthead-2023-01-30-2.2.2.pdf>`__ by T-HEAD of Alibaba.  Instructions are prefixed with `th.` as described in the specification.
356
357``XTHeadBs``
358  LLVM implements `the THeadBs (single-bit operations) vendor-defined instructions specified in <https://github.com/T-head-Semi/thead-extension-spec/releases/download/2.2.2/xthead-2023-01-30-2.2.2.pdf>`__ by T-HEAD of Alibaba.  Instructions are prefixed with `th.` as described in the specification.
359
360``XTHeadCondMov``
361  LLVM implements `the THeadCondMov (conditional move) vendor-defined instructions specified in <https://github.com/T-head-Semi/thead-extension-spec/releases/download/2.2.2/xthead-2023-01-30-2.2.2.pdf>`__ by T-HEAD of Alibaba.  Instructions are prefixed with `th.` as described in the specification.
362
363``XTHeadCmo``
364  LLVM implements `the THeadCmo (cache management operations) vendor-defined instructions specified in <https://github.com/T-head-Semi/thead-extension-spec/releases/download/2.2.2/xthead-2023-01-30-2.2.2.pdf>`__  by T-HEAD of Alibaba.  Instructions are prefixed with `th.` as described in the specification.
365
366``XTHeadFMemIdx``
367  LLVM implements `the THeadFMemIdx (indexed memory operations for floating point) vendor-defined instructions specified in <https://github.com/T-head-Semi/thead-extension-spec/releases/download/2.2.2/xthead-2023-01-30-2.2.2.pdf>`__ by T-HEAD of Alibaba.  Instructions are prefixed with `th.` as described in the specification.
368
369``XTheadMac``
370  LLVM implements `the XTheadMac (multiply-accumulate instructions) vendor-defined instructions specified in <https://github.com/T-head-Semi/thead-extension-spec/releases/download/2.2.2/xthead-2023-01-30-2.2.2.pdf>`__ by T-HEAD of Alibaba.  Instructions are prefixed with `th.` as described in the specification.
371
372``XTHeadMemIdx``
373  LLVM implements `the THeadMemIdx (indexed memory operations) vendor-defined instructions specified in <https://github.com/T-head-Semi/thead-extension-spec/releases/download/2.2.2/xthead-2023-01-30-2.2.2.pdf>`__ by T-HEAD of Alibaba.  Instructions are prefixed with `th.` as described in the specification.
374
375``XTHeadMemPair``
376  LLVM implements `the THeadMemPair (two-GPR memory operations) vendor-defined instructions specified in <https://github.com/T-head-Semi/thead-extension-spec/releases/download/2.2.2/xthead-2023-01-30-2.2.2.pdf>`__ by T-HEAD of Alibaba.  Instructions are prefixed with `th.` as described in the specification.
377
378``XTHeadSync``
379  LLVM implements `the THeadSync (multi-core synchronization instructions) vendor-defined instructions specified in <https://github.com/T-head-Semi/thead-extension-spec/releases/download/2.2.2/xthead-2023-01-30-2.2.2.pdf>`__ by T-HEAD of Alibaba.  Instructions are prefixed with `th.` as described in the specification.
380
381``XTHeadVdot``
382  LLVM implements `version 1.0.0 of the THeadV-family custom instructions specification <https://github.com/T-head-Semi/thead-extension-spec/releases/download/2.2.0/xthead-2022-12-04-2.2.0.pdf>`__ by T-HEAD of Alibaba.  All instructions are prefixed with `th.` as described in the specification, and the riscv-toolchain-convention document linked above.
383
384``XVentanaCondOps``
385  LLVM implements `version 1.0.0 of the VTx-family custom instructions specification <https://github.com/ventanamicro/ventana-custom-extensions/releases/download/v1.0.0/ventana-custom-extensions-v1.0.0.pdf>`__ by Ventana Micro Systems.  All instructions are prefixed with `vt.` as described in the specification, and the riscv-toolchain-convention document linked above.  These instructions are only available for riscv64 at this time.
386
387``XSfvcp``
388  LLVM implements `version 1.1.0 of the SiFive Vector Coprocessor Interface (VCIX) Software Specification <https://sifive.cdn.prismic.io/sifive/Zn3m1R5LeNNTwnLS_vcix-spec-software-v1p1.pdf>`__ by SiFive.  All instructions are prefixed with `sf.vc.` as described in the specification, and the riscv-toolchain-convention document linked above.
389
390``XSfvqmaccdod``, ``XSfvqmaccqoq``
391  LLVM implements `version 1.1.0 of the SiFive Int8 Matrix Multiplication Extensions Specification <https://sifive.cdn.prismic.io/sifive/1a2ad85b-d818-49f7-ba83-f51f1731edbe_int8-matmul-spec.pdf>`__ by SiFive.  All instructions are prefixed with `sf.` as described in the specification linked above.
392
393``Xsfvfnrclipxfqf``
394  LLVM implements `version 1.0.0 of the FP32-to-int8 Ranged Clip Instructions Extension Specification <https://sifive.cdn.prismic.io/sifive/0aacff47-f530-43dc-8446-5caa2260ece0_xsfvfnrclipxfqf-spec.pdf>`__ by SiFive.  All instructions are prefixed with `sf.` as described in the specification linked above.
395
396``Xsfvfwmaccqqq``
397  LLVM implements `version 1.0.0 of the Matrix Multiply Accumulate Instruction Extension Specification <https://sifive.cdn.prismic.io/sifive/c391d53e-ffcf-4091-82f6-c37bf3e883ed_xsfvfwmaccqqq-spec.pdf>`__ by SiFive.  All instructions are prefixed with `sf.` as described in the specification linked above.
398
399``XCVbitmanip``
400  LLVM implements `version 1.0.0 of the CORE-V Bit Manipulation custom instructions specification <https://github.com/openhwgroup/cv32e40p/blob/62bec66b36182215e18c9cf10f723567e23878e9/docs/source/instruction_set_extensions.rst>`__ by OpenHW Group.  All instructions are prefixed with `cv.` as described in the specification.
401
402``XCVelw``
403  LLVM implements `version 1.0.0 of the CORE-V Event load custom instructions specification <https://github.com/openhwgroup/cv32e40p/blob/master/docs/source/instruction_set_extensions.rst>`__ by OpenHW Group.  All instructions are prefixed with `cv.` as described in the specification. These instructions are only available for riscv32 at this time.
404
405``XCVmac``
406  LLVM implements `version 1.0.0 of the CORE-V Multiply-Accumulate (MAC) custom instructions specification <https://github.com/openhwgroup/cv32e40p/blob/4f024fe4b15a68b76615b0630c07a6745c620da7/docs/source/instruction_set_extensions.rst>`__ by OpenHW Group.  All instructions are prefixed with `cv.mac` as described in the specification. These instructions are only available for riscv32 at this time.
407
408``XCVmem``
409  LLVM implements `version 1.0.0 of the CORE-V Post-Increment load and stores custom instructions specification <https://github.com/openhwgroup/cv32e40p/blob/master/docs/source/instruction_set_extensions.rst>`__ by OpenHW Group.  All instructions are prefixed with `cv.` as described in the specification. These instructions are only available for riscv32 at this time.
410
411``XCValu``
412  LLVM implements `version 1.0.0 of the Core-V ALU custom instructions specification <https://github.com/openhwgroup/cv32e40p/blob/4f024fe4b15a68b76615b0630c07a6745c620da7/docs/source/instruction_set_extensions.rst>`__ by Core-V.  All instructions are prefixed with `cv.` as described in the specification. These instructions are only available for riscv32 at this time.
413
414``XCVsimd``
415  LLVM implements `version 1.0.0 of the CORE-V SIMD custom instructions specification <https://github.com/openhwgroup/cv32e40p/blob/cv32e40p_v1.3.2/docs/source/instruction_set_extensions.rst>`__ by OpenHW Group.  All instructions are prefixed with `cv.` as described in the specification.
416
417``XCVbi``
418  LLVM implements `version 1.0.0 of the CORE-V immediate branching custom instructions specification <https://github.com/openhwgroup/cv32e40p/blob/cv32e40p_v1.3.2/docs/source/instruction_set_extensions.rst>`__ by OpenHW Group.  All instructions are prefixed with `cv.` as described in the specification. These instructions are only available for riscv32 at this time.
419
420``XSiFivecdiscarddlone``
421  LLVM implements `the SiFive sf.cdiscard.d.l1 instruction specified in <https://sifive.cdn.prismic.io/sifive/767804da-53b2-4893-97d5-b7c030ae0a94_s76mc_core_complex_manual_21G3.pdf>`_ by SiFive.
422
423``XSiFivecflushdlone``
424  LLVM implements `the SiFive sf.cflush.d.l1 instruction specified in <https://sifive.cdn.prismic.io/sifive/767804da-53b2-4893-97d5-b7c030ae0a94_s76mc_core_complex_manual_21G3.pdf>`_ by SiFive.
425
426``XSfcease``
427  LLVM implements `the SiFive sf.cease instruction specified in <https://sifive.cdn.prismic.io/sifive/767804da-53b2-4893-97d5-b7c030ae0a94_s76mc_core_complex_manual_21G3.pdf>`_ by SiFive.
428
429``Xwchc``
430  LLVM implements `the custom compressed opcodes present in some QingKe cores` by WCH / Nanjing Qinheng Microelectronics. The vendor refers to these opcodes by the name "XW".
431
432``experimental-Xqcia``
433  LLVM implements `version 0.2 of the Qualcomm uC Arithmetic extension specification <https://github.com/quic/riscv-unified-db/releases/latest>`__ by Qualcomm.  All instructions are prefixed with `qc.` as described in the specification. These instructions are only available for riscv32.
434
435``experimental-Xqciac``
436  LLVM implements `version 0.3 of the Qualcomm uC Load-Store Address Calculation extension specification <https://github.com/quic/riscv-unified-db/releases/latest>`__ by Qualcomm.  All instructions are prefixed with `qc.` as described in the specification. These instructions are only available for riscv32.
437
438``experimental-Xqcicli``
439  LLVM implements `version 0.2 of the Qualcomm uC Conditional Load Immediate extension specification <https://github.com/quic/riscv-unified-db/releases/latest>`__ by Qualcomm.  All instructions are prefixed with `qc.` as described in the specification. These instructions are only available for riscv32.
440
441``experimental-Xqcicm``
442  LLVM implements `version 0.2 of the Qualcomm uC Conditional Move extension specification <https://github.com/quic/riscv-unified-db/releases/latest>`__ by Qualcomm.  All instructions are prefixed with `qc.` as described in the specification. These instructions are only available for riscv32.
443
444``experimental-Xqcics``
445  LLVM implements `version 0.2 of the Qualcomm uC Conditional Select extension specification <https://github.com/quic/riscv-unified-db/releases/latest>`__ by Qualcomm.  All instructions are prefixed with `qc.` as described in the specification. These instructions are only available for riscv32.
446
447``experimental-Xqcicsr``
448  LLVM implements `version 0.2 of the Qualcomm uC CSR extension specification <https://github.com/quic/riscv-unified-db/releases/latest>`__ by Qualcomm.  All instructions are prefixed with `qc.` as described in the specification. These instructions are only available for riscv32.
449
450``experimental-Xqciint``
451  LLVM implements `version 0.2 of the Qualcomm uC Interrupts extension specification <https://github.com/quic/riscv-unified-db/releases/latest>`__ by Qualcomm.  All instructions are prefixed with `qc.` as described in the specification. These instructions are only available for riscv32.
452
453``experimental-Xqcilo``
454  LLVM implements `version 0.2 of the Qualcomm uC Large Offset Load Store extension specification <https://github.com/quic/riscv-unified-db/releases/latest>`__ by Qualcomm.  All instructions are prefixed with `qc.` as described in the specification. These instructions are only available for riscv32.
455
456``experimental-Xqcilsm``
457  LLVM implements `version 0.2 of the Qualcomm uC Load Store Multiple extension specification <https://github.com/quic/riscv-unified-db/releases/latest>`__ by Qualcomm.  All instructions are prefixed with `qc.` as described in the specification. These instructions are only available for riscv32.
458
459``experimental-Xqcisls``
460  LLVM implements `version 0.2 of the Qualcomm uC Scaled Load Store extension specification <https://github.com/quic/riscv-unified-db/releases/latest>`__ by Qualcomm.  All instructions are prefixed with `qc.` as described in the specification. These instructions are only available for riscv32.
461
462``Xmipscmove``
463  LLVM implements conditional move for the `p8700 processor <https://mips.com/products/hardware/p8700/>` by MIPS.
464
465``Xmipslsp``
466  LLVM implements load/store pair instructions for the `p8700 processor <https://mips.com/products/hardware/p8700/>` by MIPS.
467
468Experimental C Intrinsics
469=========================
470
471In some cases an extension is non-experimental but the C intrinsics for that
472extension are still experimental.  To use C intrinsics for such an extension
473from `clang`, you must add `-menable-experimental-extensions` to the command
474line.  This currently applies to the following extensions:
475
476No extensions have experimental intrinsics.
477
478Long (>32-bit) Instruction Support
479==================================
480
481RISC-V is a variable-length ISA, but the standard currently only defines 16- and 32-bit instructions. The specification describes longer instruction encodings, but these are not ratified.
482
483The LLVM disassembler, `llvm-objdump`, does use the longer instruction encodings described in the specification to guess the instruction length (up to 176 bits) and will group the disassembly view of encoding bytes correspondingly.
484
485The LLVM integrated assembler for RISC-V supports two different kinds of ``.insn`` directive, for assembling instructions that LLVM does not yet support:
486
487* ``.insn type, args*`` which takes a known instruction type, and a list of fields. You are strongly recommended to use this variant of the directive if your instruction fits an existing instruction type.
488* ``.insn [ length , ] encoding`` which takes an (optional) explicit length (in bytes) and a raw encoding for the instruction. When given an explicit length, this variant can encode instructions up to 64 bits long. The encoding part of the directive must be given all bits for the instruction, none are filled in for the user. When used without the optional length, this variant of the directive will use the LSBs of the raw encoding to work out if an instruction is 16 or 32 bits long. LLVM does not infer that an instruction might be longer than 32 bits - in this case, the user must give the length explicitly.
489
490It is strongly recommended to use the ``.insn`` directive for assembling unsupported instructions instead of ``.word`` or ``.hword``, because it will produce the correct mapping symbols to mark the word as an instruction, not data.
491
492Global Pointer (GP) Relaxation and the Small Data Limit
493=======================================================
494
495Some of the RISC-V psABI variants reserve ``gp`` (``x3``) for use as a "Global Pointer", to make generating data addresses more efficient.
496
497To use this functionality, you need to be doing all of the following:
498
499* Use the ``medlow`` (aka ``small``) code model;
500* Not use the ``gp`` register for any other uses (some platforms use it for the shadow stack and others as a temporary -- as denoted by the ``Tag_RISCV_x3_reg_usage`` build attribute);
501* Compile your objects with Clang's ``-mrelax`` option, to enable relaxation annotations on relocatable objects (this is the default, but ``-mno-relax`` disables these relaxation annotations);
502* Compile for a position-dependent static executable (not a shared library, and ``-fno-PIC`` / ``-fno-pic`` / ``-fno-pie``); and
503* Use LLD's ``--relax-gp`` option.
504
505LLD will relax (rewrite) any code sequences that materialize an address within 2048 bytes of ``__global_pointer$`` (which will be defined if it is used and does not already exist) to instead generate the address using ``gp`` and the correct (signed) 12-bit immediate. This usually saves at least one instruction compared to materialising a full 32-bit address value.
506
507There can only be one ``gp`` value in a process (as ``gp`` is not changed when calling into a function in a shared library), so the symbol is is only defined and this relaxation is only done for executables, and not for shared libraries. The linker expects executable startup code to put the value of ``__global_pointer$`` (from the executable) into ``gp`` before any user code is run.
508
509Arguably, the most efficient use for this addressing mode is for smaller global variables, as larger global variables likely need many more loads or stores when they are being accessed anyway, so the cost of materializing the upper bits can be shared.
510
511Therefore the compiler can place smaller global variables into sections with names starting with ``.sdata`` or ``.sbss`` (matching sections with names starting with ``.data`` and ``.bss`` respectively). LLD knows to define the ``global_pointer$`` symbol close to these sections, and to lay these sections out adjacent to the ``.data`` section.
512
513Clang's ``-msmall-data-limit=`` option controls what the threshold size is (in bytes) for a global variable to be considered small. ``-msmall-data-limit=0`` disables the use of sections starting ``.sdata`` and ``.sbss``. The ``-msmall-data-limit=`` option will not move global variables that have an explicit data section, and will keep globals in separate sections if you are using ``-fdata-sections``.
514
515The small data limit threshold is also used to separate small constants into sections with names starting with ``.srodata``. LLD does not place these with the ``.sdata`` and ``.sbss`` sections as ``.srodata`` sections are read only and the other two are writable. Instead the ``.srodata`` sections are placed adjacent to ``.rodata``.
516
517Data suggests that these options can produce significant improvements across a range of benchmarks.
518