12 This document contains the release notes for the LLVM Compiler Infrastructure,
13 release 3.8. Here we describe the status of LLVM, including major improvements
14 from the previous release, improvements in various subprojects of LLVM, and
15 some of the current users of the code. All LLVM releases may be downloaded
16 from the `LLVM releases web site <http://llvm.org/releases/>`_.
18 For more information about LLVM, including information about the latest
19 release, please check out the `main LLVM web site <http://llvm.org/>`_. If you
20 have questions or comments, the `LLVM Developer's Mailing List
21 <http://lists.llvm.org/mailman/listinfo/llvm-dev>`_ is a good place to send
24 Non-comprehensive list of changes in this release
25 =================================================
26 * With this release, the minimum Windows version required for running LLVM is
27 Windows 7. Earlier versions, including Windows Vista and XP are no longer
30 * With this release, the autoconf build system is deprecated. It will be removed
31 in the 3.9 release. Please migrate to using CMake. For more information see:
32 `Building LLVM with CMake <CMake.html>`_
34 * The C API function LLVMLinkModules is deprecated. It will be removed in the
35 3.9 release. Please migrate to LLVMLinkModules2. Unlike the old function the
38 * Doesn't take an unused parameter.
39 * Destroys the source instead of only damaging it.
40 * Does not record a message. Use the diagnostic handler instead.
42 * The C API functions LLVMParseBitcode, LLVMParseBitcodeInContext,
43 LLVMGetBitcodeModuleInContext and LLVMGetBitcodeModule have been deprecated.
44 They will be removed in 3.9. Please migrate to the versions with a 2 suffix.
45 Unlike the old ones the new ones do not record a diagnostic message. Use
46 the diagnostic handler instead.
48 * The deprecated C APIs LLVMGetBitcodeModuleProviderInContext and
49 LLVMGetBitcodeModuleProvider have been removed.
51 * The deprecated C APIs LLVMCreateExecutionEngine, LLVMCreateInterpreter,
52 LLVMCreateJITCompiler, LLVMAddModuleProvider and LLVMRemoveModuleProvider
55 * With this release, the C API headers have been reorganized to improve build
56 time. Type specific declarations have been moved to Type.h, and error
57 handling routines have been moved to ErrorHandling.h. Both are included in
58 Core.h so nothing should change for projects directly including the headers,
59 but transitive dependencies may be affected.
61 * llvm-ar now suports thin archives.
63 * llvm doesn't produce .data.rel.ro.local or .data.rel sections anymore.
65 * aliases to available_externally globals are now rejected by the verifier.
67 * the IR Linker has been split into IRMover that moves bits from one module to
68 another and Linker proper that decides what to link.
70 * Support for dematerializing has been dropped.
72 * RegisterScheduler::setDefault was removed. Targets that used to call into the
73 command line parser to set the DAGScheduler, and that don't have enough
74 control with setSchedulingPreference, should look into overriding the
75 SubTargetHook "getDAGScheduler()".
77 * ``ilist_iterator<T>`` no longer has implicit conversions to and from ``T*``,
78 since ``ilist_iterator<T>`` may be pointing at the sentinel (which is usually
79 not of type ``T`` at all). To convert from an iterator ``I`` to a pointer,
80 use ``&*I``; to convert from a pointer ``P`` to an iterator, use
81 ``P->getIterator()``. Alternatively, explicit conversions via
82 ``static_cast<T>(U)`` are still available.
84 * ``ilist_node<T>::getNextNode()`` and ``ilist_node<T>::getPrevNode()`` now
85 fail at compile time when the node cannot access its parent list.
86 Previously, when the sentinel was was an ``ilist_half_node<T>``, this API
87 could return the sentinal instead of ``nullptr``. Frustrated callers should
88 be updated to use ``iplist<T>::getNextNode(T*)`` instead. Alternatively, if
89 the node ``N`` is guaranteed not to be the last in the list, it is safe to
90 call ``&*++N->getIterator()`` directly.
92 * The `Kaleidoscope tutorials <tutorial/index.html>`_ have been updated to use
95 * ORC now has a basic set of C bindings.
97 * Optional support for linking clang and the LLVM tools with a single libLLVM
98 shared library. To enable this, pass ``-DLLVM_LINK_LLVM_DYLIB=ON`` to CMake.
99 See `Building LLVM with CMake`_ for more details.
101 * The optimization to move the prologue and epilogue of functions in colder
102 code path (shrink-wrapping) is now enabled by default.
104 * A new target-independent gcc-compatible emulated Thread Local Storage mode
105 is added. When ``-femultated-tls`` flag is used, all accesses to TLS
106 variables are converted to calls to ``__emutls_get_address`` in the runtime
109 * MSVC compatible exception handling has been completely overhauled. New
110 instructions have been introduced to facilitate this:
111 `New exception handling instructions <ExceptionHandling.html#new-exception-handling-instructions>`_.
112 While we have done our best to test this feature thoroughly, it would
113 not be completely surprising if there were a few lingering issues that
114 early adopters might bump into.
117 For small 1-3 sentence descriptions, just add an entry at the end of
118 this list. If your description won't fit comfortably in one bullet
119 point (e.g. maybe you would like to give an example of the
120 functionality, or simply have a lot to talk about), see the `NOTE` below
121 for adding a new subsection.
123 * ... next change ...
126 If you would like to document a larger change, then you can add a
127 subsection about it right here. You can copy the following boilerplate
128 and un-indent it (the indentation causes it to be inside this comment).
133 Makes programs 10x faster by doing Special New Thing.
136 Changes to the ARM Backends
137 ---------------------------
139 During this release the AArch64 target has:
141 * Added support for more sanitizers (MSAN, TSAN) and made them compatible with
142 all VMA kernel configurations (currently tested on 39 and 42 bits).
143 * Gained initial LLD support in the new ELF back-end
144 * Extended the Load/Store optimiser and cleaned up some of the bad decisions
146 * Expanded LLDB support, including watchpoints, native building, Renderscript,
147 LLDB-server, debugging 32-bit applications.
148 * Added support for the ``Exynos M1`` chip.
150 During this release the ARM target has:
152 * Gained massive performance improvements on embedded benchmarks due to finally
153 running the stride vectorizer in full form, incrementing the performance gains
154 that we already had in the previous releases with limited stride vectorization.
155 * Expanded LLDB support, including watchpoints, unwind tables
156 * Extended the Load/Store optimiser and cleaned up some of the bad decisions
158 * Simplified code generation for global variable addresses in ELF, resulting in
159 a significant (4% in Chromium) reduction in code size.
160 * Gained some additional code size improvements, though there's still a long road
161 ahead, especially for older cores.
162 * Added some EABI floating point comparison functions to Compiler-RT
163 * Added support for Windows+GNU triple, +features in -mcpu/-march options.
166 Changes to the MIPS Target
167 --------------------------
169 During this release the MIPS target has:
171 * Significantly extended support for the Integrated Assembler. See below for
173 * Added support for the ``P5600`` processor.
174 * Added support for the ``interrupt`` attribute for MIPS32R2 and later. This
175 attribute will generate a function which can be used as a interrupt handler
176 on bare metal MIPS targets using the static relocation model.
177 * Added support for the ``ERETNC`` instruction found in MIPS32R5 and later.
178 * Added support for OpenCL. See http://portablecl.org/.
180 * Address spaces 1 to 255 are now reserved for software use and conversions
181 between them are no-op casts.
183 * Removed the ``mips16`` value for the -mcpu option since it is an :abbr:`ASE
184 (Application Specific Extension)` and not a processor. If you were using this,
185 please specify another CPU and use ``-mips16`` to enable MIPS16.
186 * Removed ``copy_u.w`` from 32-bit MSA and ``copy_u.d`` from 64-bit MSA since
187 they have been removed from the MSA specification due to forward compatibility
188 issues. For example, 32-bit MSA code containing ``copy_u.w`` would behave
189 differently on a 64-bit processor supporting MSA. The corresponding intrinsics
190 are still available and may expand to ``copy_s.[wd]`` where this is
191 appropriate for forward compatibility purposes.
192 * Relaxed the ``-mnan`` option to allow ``-mnan=2008`` on MIPS32R2/MIPS64R2 for
193 compatibility with GCC.
194 * Made MIPS64R6 the default CPU for 64-bit Android triples.
196 The MIPS target has also fixed various bugs including the following notable
199 * Fixed reversed operands on ``mthi``/``mtlo`` in the DSP :abbr:`ASE
200 (Application Specific Extension)`.
201 * The code generator no longer uses ``jal`` for calls to absolute immediate
203 * Disabled fast instruction selection on MIPS32R6 and MIPS64R6 since this is not
205 * Corrected addend for ``R_MIPS_HI16`` and ``R_MIPS_PCHI16`` in MCJIT
206 * The code generator no longer crashes when handling subregisters of an 64-bit
207 FPU register with undefined value.
208 * The code generator no longer attempts to use ``$zero`` for operands that do
209 not permit ``$zero``.
210 * Corrected the opcode used for ``ll``/``sc`` when using MIPS32R6/MIPS64R6 and
211 the Integrated Assembler.
212 * Added support for atomic load and atomic store.
213 * Corrected debug info when dynamically re-aligning the stack.
217 We have made a large number of improvements to the integrated assembler for
218 MIPS. In this release, the integrated assembler isn't quite production-ready
219 since there are a few known issues related to bare-metal support, checking
220 immediates on instructions, and the N32/N64 ABI's. However, the current support
221 should be sufficient for many users of the O32 ABI, particularly those targeting
222 MIPS32 on Linux or bare-metal MIPS32.
224 If you would like to try the integrated assembler, please use
227 Changes to the PowerPC Target
228 -----------------------------
230 During this release ...
233 Changes to the X86 Target
234 -----------------------------
236 During this release ...
238 * TLS is enabled for Cygwin as emutls.
240 * Smaller code for materializing 32-bit 1 and -1 constants at ``-Os``.
242 * More efficient code for wide integer compares. (E.g. 64-bit compares
245 * Tail call support for ``thiscall``, ``stdcall``, ``vectorcall``, and
246 ``fastcall`` functions.
248 Changes to the Hexagon Target
249 -----------------------------
251 In addition to general code size and performance improvements, Hexagon target
252 now has basic support for Hexagon V60 architecture and Hexagon Vector
255 Changes to the AVR Target
256 -------------------------
258 Slightly less than half of the AVR backend has been merged in at this point. It is still
259 missing a number large parts which cause it to be unusable, but is well on the
260 road to being completely merged and workable.
262 Changes to the OCaml bindings
263 -----------------------------
265 During this release ...
267 * The ocaml function link_modules has been replaced with link_modules' which
268 uses LLVMLinkModules2.
271 External Open Source Projects Using LLVM 3.8
272 ============================================
274 An exciting aspect of LLVM is that it is used as an enabling technology for
275 a lot of other language and tools projects. This section lists some of the
276 projects that have already been updated to work with LLVM 3.8.
278 LDC - the LLVM-based D compiler
279 -------------------------------
281 `D <http://dlang.org>`_ is a language with C-like syntax and static typing. It
282 pragmatically combines efficiency, control, and modeling power, with safety and
283 programmer productivity. D supports powerful concepts like Compile-Time Function
284 Execution (CTFE) and Template Meta-Programming, provides an innovative approach
285 to concurrency and offers many classical paradigms.
287 `LDC <http://wiki.dlang.org/LDC>`_ uses the frontend from the reference compiler
288 combined with LLVM as backend to produce efficient native code. LDC targets
289 x86/x86_64 systems like Linux, OS X and Windows and also PowerPC (32/64 bit)
290 and ARM. Ports to other architectures like AArch64 and MIPS64 are underway.
293 Additional Information
294 ======================
296 A wide variety of additional information is available on the `LLVM web page
297 <http://llvm.org/>`_, in particular in the `documentation
298 <http://llvm.org/docs/>`_ section. The web page also contains versions of the
299 API documentation which is up-to-date with the Subversion version of the source
300 code. You can access versions of these documents specific to this release by
301 going into the ``llvm/docs/`` directory in the LLVM tree.
303 If you have any questions or comments about LLVM, please feel free to contact
304 us via the `mailing lists <http://llvm.org/docs/#maillist>`_.