[libclc] annotated tag upstream/0.2.0+git20150813 created (now ef4be64)
Michael Gilbert
mgilbert at moszumanska.debian.org
Sun Sep 13 23:35:14 UTC 2015
This is an automated email from the git hooks/post-receive script.
mgilbert pushed a change to annotated tag upstream/0.2.0+git20150813
in repository libclc.
at ef4be64 (tag)
tagging 63f5af3e9777bc37fa69accc2bf8c3b4fd6a15bf (commit)
tagged by Michael Gilbert
on Sun Sep 13 19:34:25 2015 -0400
- Log -----------------------------------------------------------------
upstream 0.2.0+git20150813
Aaron Watry (60):
libclc: vload/vstore disable assembly and fix offset calculation
Fix and re-enable R600 vload/vstore assembly
Fix build with LLVM 3.4
Implement generic upsample()
Added get_num_groups
Add missing integer min/max definitions
Add vload* for addrspace(2) and use as constant load for R600
Enable assembly vload3 int/uint constant/global for R600
Add intN vloadN() implementations for address spaces 3 and 4
Add hadd builtin
Add rhadd builtin
Add mul_hi implementation [v2]
Add atomic_inc and atomic_add builtins
Remove unneeded semi-colons
Add atomic_sub and atomic_dec builtin functions
Implement mad_hi built-in
Parenthesize arguments for mad_hi
Fix build with LLVM 3.5
Pass -fno-builtin flag to clang to silence warnings
clctypes.h: Don't rely on stddef.h for size_t and ptrdiff_t
Add all(igentype) builtin
relational: Add isequal(floatN) builtin
math: Implement mix builtin
Revert "clctypes.h: Don't rely on stddef.h for size_t and ptrdiff_t"
Fix definition of INFINITY and add NAN/HUGE_VAL[F]
relational: Implement signbit
relational: Fix signbit
relational: create re-usable macros for relational declarations
Fix isnan definition for vector results
relational/signbit: Refactor to use relational macros
relational: Implement isgreater
relational: Implement isgreaterequal
relational: Implement isnotequal
Add several missing double constant definitions
vload/vstore: Use casts instead of scalarizing everything in CLC version
Implement generic mad_sat
Add int3/uint3 to integer-gentype.inc
Revert "Implement generic mad_sat"
math: Add acos implementation
math: Add asin implementation
math: Add tan implementation
atomic: define extension functions for existing atomic implementations
atomic: Add generic implementation of atom[ic]_max
atomics: Add generic atom[ic]_and
atomic: Add atom[ic]_or
atomic: Add generic atom[ic]_xor
atomic: Add generic atomic_min implementation
atomic: Implement generic atom[ic]_xchg
atomic: Add generic atom[ic]_cmpxchg
R600: Map atomic_and address spaces
R600: Map address spaces for atomic_or
R600: Map addr spaces and use atomic_max
R600: Map address spaces for atomic_xor
R600: Map address spaces for atomic_min
R600: Map address spaces for atomic_xchg
R600: Map Address spaces for atomic_cmpxchg
atomic: undef macros that are included from atomic_decl.inc
libclc/math: Add cospi
Move mix from math to common
math: Implement erfc
Jan Vesely (21):
Add intptr types
relational: Add isless(floatN) builtin
relational: Add islessequal(floatN) builtin
configure: Add rpath to prepare-builtins util
Implement generic mad_sat
add isfinite builtin
add isnormal builtin
add islessgreater builtin
add isunordered builtin
add isordered builtin
Implement fmod
r600: Use llvm intrinsic to read work dimension information
r600: Fix get_work_dim range metadata
Implement log10
Fix compilation warnings without cl_khr_fp64
Implement sinpi builtin
Move ldexp soft implementation to a separate file
r600: Use __clc_ldexp on asics that don't implement the intruction
Fix ldexp fp64 build error
geometric: Limit fast_{distance,length} functions to single precision
math: limit half_sqrt to single precision
Jeroen Ketema (25):
Add definition for M_PI
Fix _F definitions
Add more log related float constants
Fix build broken by LLVM commit r209103
Remove unused include which breaks build after r210803
Implementations for exp(float) and exp(double) v2
Add files forgotten in the previous commit
Add remaining float constants
Add missing undefs
Add pown
Fix breakage after r211259
Protect functions taking double by #ifdef cl_khr_fp64
Move clcmacro.h to avoid cluttering user namespace v2
Add MAXFLOAT
Introduce CLC_VERSION macros v2
Add half limits
Add exp10
OpenCL 1.1 does not define CL_VERSION_1_2 so use hardcoded number instead
Remove redundant semi-colons
Remove more redundant semi-colons
Don't include <stddef.h>
Remove wrong semi-colons
Properly initialize Module pointer
Require LLVM >=3.7 and bump version to 0.2.0
Remove files accidentally not removed in r244310
Michel Danzer (2):
Fix build against LLVM SVN >= r216393
Fix build against LLVM SVN >= r216488
NAKAMURA Takumi (3):
Update the copyright coredits -- Happy new year 2013!
Update the copyright credits -- Happy new year 2014!
Prune CRLF.
Niels Ole Salscheider (1):
Include llvm-config.h instead of config.h
Peter Collingbourne (28):
Create libclc subdirectories.
Initial commit.
Update repository paths.
Test web page update.
Switch to BSD/MIT dual license.
Switch to the NVPTX backend.
Explicit conversions.
Add some tests which had heretofore evaded 'git add'.
Add fabs builtin.
Fix typo in double precision case.
Implement exp, exp2, log, log2, native_exp, native_exp2, native_log,
Implement mad builtin.
Add fma, hypot builtins.
Define FLOAT in floatn.inc.
Add missing dot.h include.
Enable cl_khr_fp64 when building the library, and fix several bugs
Add pow builtin.
configure.py: Add an install rule.
Fix declarations of __clc_add_sat_*. Patch by Lei Mou!
Implement sub_sat builtin. Patch by Lei Mou!
PTX: move implementations of work-item and synchronisation functions
Do not use linkonce_odr linkage in .ll files. This prevented them
Add floor builtin. Patch by Cassie Epps!
Add rsqrt builtin. Based on patch by Cassie Epps!
Add barrier.cl to SOURCES, spotted by Jin Wang.
Add native_powr builtin. Patch by Tom Stellard!
Implement any() builtin. Patch by Tom Stellard!
Fix build against recent versions of Clang. Based on patch by Alastair Donaldson!
Tom Stellard (132):
r600: Initial support
r600: Fix get_global_id implementation
r600: Add get_global_size() implementation
Move R600 headers into generic directory
PTX: move implementations of work-item and synchronisation functions
R600: Replace cl implementations with LLVM IR implementation
Make libclc more Linux FHS conform.
Allow targets to override generic implementations
r600: Add overrides file
Fix typo in include/clc/geometric/length.inc
Use brackets around include files in length.cl and normalize.cl
Remove the static keyword from the _CLC_INLINE macro
Implement fmax() and fmin() builtins
Implement ceil() builtin
Fix build with LLVM 3.3
configure: fix out-of-source build
configure: Enable building separate libraries for target variants
libclc: Add max() builtin function
libclc: Fix abs_diff builtin integer function
libclc: Add clamp() builtin for integer/floating point
libclc: Move max builtin to shared/
libclc: implement rotate builtin
Simplify rotate implementation a bit..
Add a TODO note.
Add a another TODO note.
libclc: Rename [add|sub]_sat.ll to [add|sub]_sat_if.ll
libclc: implement initial version of min()
libclc: Implement the min(vec, scalar) version of the min builtin.
libclc: Add clamp(vec, scalar, scalar) and max(vec, scalar)
libclc: Implement clz() builtin
r600: Fix implementations of get_group_id.ll and get_local_size.ll
libclc: Initial vload implementation
libclc: Initial vstore implementation
libclc: Add assembly versions of vload for global int4/8/16
libclc: Add assembly versions of vstore for global [u]int4/8/16
Add bitselect() builtin
Implement barrier() builtin
Add __CLC_ prefix to all macro definitions in headers
Implement mad24() and mul24() builtins
Add integer-gentype.inc: Missing file from r185839
configure: Fix build when clang is installed to a non-standard prefix
Implement generic rint()
Add some missing convert_* functions
Place pkg-config file in $prefix/share/pkgconfig.
Add missing as_{float,double} functions
Implement isnan() builtin
Implement nextafter() builtin
Implement sign() builtin
Port pocl's gen_convert.py script to libclc
Make C++ compiler configurable
Clean-up dependency files
R600: Set the noduplicate attribute on barrier() intrinsics
s/_CLC_DECL/_CLC_DEF/
Implement builtins for cl_khr_global_int32_base_atomics extension
Implement round builtin
R600: Add aliases for Sea Islands GPUs
Fix a C&P error in r195021 (65a950abab3cb8435ccb2646ac4773986c995c81)
Implement trunc builtin.
Add floating-point macro definitions v2
Fix build broken by LLVM commit r199279
Enforce python2 for systems that use python3 as their default.
Fixed ninja build issues relating to use of $(DESTDIR)
Fixed rules names so they are unique when aliases are present.
Updated README.TXT with information about using DESTDIR and building with Ninja.
Revert "Enforce python2 for systems that use python3 as their default."
Add generic nvptx targets
Fix build since r202052
Add cross for double3 and double4
Add sincos
Add definition for M_PI_F v3
Replace tabs by spaces
Introduce M_LOG2E_F and M_LOG2E
Remove clc/gentype.inc
Fix build broken by LLVM commit r207593
prepare-builtins: Use std:: prefix for error_code
prepare-builtins: Fix broken build due to recent LLVM API change
Implement atan builtin
Implement atan2 builtin
Implement cos builtin for float types
Implement sin builtin for float types
R600: Add aliases for hainan and mullins
Implement prefetch builtin
Add missing file from r216127
Fix implementation of copysign
Implement isinf builtin
Implement wait_group_events builtin v2
Implement async_work_group_strided_copy builtin v2
Implement async_work_group_copy builtin v3
Implement log1p builtin
Require LLVM 3.6 and bump version to 0.1.0
r600: get_work_dim: Update metadata syntax for LLVM 3.6
Use amdgcn triple for SI+ GPUs
Implement degrees builtin v2
Implement radians builtin v2
Implement smoothstep builtin v2
Implement step builtin
Fix bitselect for float/double types v2
Add __clc_ prefix to functions in sincos_helpers.cl
Fix implementation of length builtin v2
Implement distance builtin v2
Implement half_sqrt builtin v2
Implement fast_length builtin
Implement fast_distance builtin
Implement fmin using __builtin_fmin
Implement fmax using __builtin_fmax
Implement acospi builtin
Implement asinh builtin
Implement asinpi builtin
Implement atanpi builtin
Implement acosh builtin
Implement atanh builtin
configure: Add --enable-runtime-subnormal option
Implement fract builtin
Allow compilation depending to the LLVM version
Fix implementation of normalize builtin
Implement ldexp for R600/SI
math: Add ldexp implementation
Implement half_rsqrt builtin v3
Implement fast_normalize builtin v4
Implement atan2 for doubles
Implement atan2pi builtin
Implement cos for double types
Implement sin for double types
Implement exp2 using OpenCL C rather than using an intrinsic
Use a more accurate implementation for exp
prepare-builtins: Fix build with LLVM 3.7
prepare-builtins: Fix build with LLVM 3.6
Fix implementation of sqrt v2
R600: Implement accurate double precision sqrt v2
Use llvm intrinsics for native_log and native_log2
Implement accurate log2 function
Fix double implementation of log
-----------------------------------------------------------------------
No new revisions were added by this update.
--
Alioth's /usr/local/bin/git-commit-notice on /srv/git.debian.org/git/pkg-opencl/libclc.git
More information about the Pkg-opencl-commits
mailing list