Release Information

What problems have been fixed in PGI 2011?

Each release of the PGI compilers brings a number of technical corrections to problems discovered by PGI and by users. Problems found can be emailed directly to PGI Technical Support or submitted through the web-based Support Request form. PGI refers to all such reports as Technical Problem Reports, or TPRs.

The following table is a list of problems that have been fixed in each release. Descriptions as provided may be incomplete due to proprietary nature of the sample code. In such cases simple replacement examples may not be available.

Note: ICE stands for Internal Compiler Error.

TPR Version Affected
Component
Symptoms /
Actions
18024 11.10 64-bit Fortran CUDA Fortran example fails to link when -fPIC is present
Added -lcublas when -Mcuda and -fPIC are present
18113 11.10 CUDA Fortran CUDA Fortran allocate statement fails with -mcmodel=medium
Corrected the size computation of the allocate when -mcmodel=medium is present
18151 11.10 64-bit C/C++ Header files not properly installed on systems with gcc version > 4.1
Corrected the install process when the version of gcc is > 4.1
18215 11.10 32/64-bit Fortran pgfortran reports the error, "Type bound procedure must be a module procedure or an external procedure with an explicit interface"
Corrected the processing of a procedure pointer whose interface is a module subroutine without arguments
18216 11.10 64-bit CUDA Fortran Undefined references occur with -Mcuda and -mcmodel=medium
Corrected libcudaforblas.so and libcudaforblas4.so to work with -mcmodel=medium
18220 11.10 PGI Accelerator Fortran pgfortran reports the error, "Internal compiler error. ACC CG: could not find ALV index for symbol"
Corrected the handling of a module allocatable which also appears in a mirror directive
18222 11.10 32/64-bit Fortran pgfortran reports the error, "Too many data constants in initialization statement"
Corrected the use of the intrinsic acos in a data initialization of a scalar variable
18183 11.9 Fortran #warning lines generate severe errors
Changed the error level of messages for #warning lines from severe to warning
18169 11.9 C/C++ double complex C example produces error, "Internal compiler error. 'load_retval, unexp. Dt"
Corrected returning the value of a double complex function when the function contains VLAs
18166 11.9 Fortran pgfortran example containing chararater length expressions fails to compile
Corrected the processing of character declarations within interface blocks whose lengths are arithmetic expressions
18150 11.9 Fortran pgfortran example containing an 'optional, allocatable, intent(out)' array fails
Corrected the compiler to deallocate the optional array only if present
18147 11.9 32-bit CUDA Fortran Timing information is missing for loops with -Mcuda
Corrected the compiler to collect timings with the -Mcuda interface
18145 11.9 Fortran pgfortran example fails when an allocatable array is assigned a pointer
Corrected the allocatable array assignment semantics when the right hand side is pointer
18129 11.9 Linux Installer PGI compilers do not install successfully on 32-bit Ubuntu 11.04
Corrected the installer
18127 11.9 PGI Accelerator -mcmodel=medium fails with missing references
Corrected the libraries accessed when -mcmodel=medium is present
18115 11.9 CUDA Fortran Cannot allocate 2GB+ on GPU
Corrected the cuda allocate routine to properly handle large sizes
18114 11.9 CUDA Fortran Acc_Get_Memory and Acc_Get_Free_Memory giving incorrect results
Corrected the declarations of the acc_get_memory() and acc_get_free_memory() routines
18113 11.9 CUDA Fortran Allocate statement fails with -mcmodel=medium
Corrected the CUDA Fortran support of allocate when -mcmodel=medium is present
18109 11.9 64-bit C/C++ pgcc example reports the error, "Internal compiler error. gen_aili: unrec. ili opcode"
Corrected the processing of volatile objects which are type complex
18105 11.9 PGI Accelerator Example reports the error, "Internal compiler error. gen_regarg_ili: ILI not handled'"
Corrected the handling of certain expressions used in return statements
18098 11.9 64-bit CUDA Fortran Module device variable not passed to cudamemcpy correctly
Corrected the conversion of module device argument host addresses to device addresses in all calls within CUDA host code
18080 11.9 Fortran pgfortran example produces the error, "Internal compiler error unrec dtype in cp_iregarg"
Corrected the processing of arguments when the called function contains VALUE arguments
18040 11.9 Fortran pgfortran example causes the compiler to abort
Corrected the handling of array constructors containing complex constants
17983 11.9 PGI Accelerator Fortran "call to cuMemFree returned error 1" when using !$mirror of an allocatable array
Corrected the handling of mirrored allocatables on the device
17930 11.9 Fortran pgfortran example produces the error, "Internal compiler error. size_of: attempt to size assumed size character"
Corrected the handling of array constructors containing characters of non-constant length
17817 11.9 Documentation Documentation problem with scalapack example
Corrected the example showing the use of -Mscalapack
17663 11.9 Documentation Add flexlm recommendation for "-local"
Updated the installation instructions to include the use of "-local"
16548 11.9 PGI Accelerator Missing device scalar assignment
Corrected the compiler to produce a scalar kernel for the assignment
18079 11.8 64-bit C/C++ pgcpp generates incorrect location information for debugging
Corrected the debug line number and file information when functions are inlined from multiple files
18058 11.8 64-bit C/C++ pgcpp fails with "undefined symbol .LB18086' in operation"
Corrected the optimzation of C++ code containing catch/throw
18063 11.8 Fortran OpenMPI link fails with "Symbol__fvs_exp_vex' causes overflow in R_X86_64_PC32 relocation" errors
Corrected the build of the shared library containing the AVX versions of various math routines
18053 11.8 Accelerator Fortran pgfortran reports the error, " undefined reference to '.BSS14' "
Corrected the handling of accelerator regions within contained subprograms
18029 11.8 PVF Using the IMSL library property prevents PVF from creating mod files
Updated the option -Mimsl to correctly specify the path of the IMSL mod file
18028 11.8 CUDA Fortran CUDA Fortran reports the error, "invalid type conversion"
Corrected the processing of a DEVICE scalar used as a reduction variable
18027 11.8 CUDA Fortran CUDA Fortran reports the error, "ILM file" problems, intrinsic loc not allowed, unknown symbol"
Corrected the handling of a variable named loc appearing in a data clause
18026 11.8 Windows Fortran 3F fstat64 fails on win32 and win64
Corrected the fstat64 3F routine when retrieving large file information
18025 11.8 CUDA Fortran CUDA Fortran code fails when compiled with -i8
Corrected comparing 64-bit integers within accelerator regions
18024 11.8 64-bit CUDA Fortran CUDA Fortran with -fPIC reports "undefined references" to cublasAlloc and cublasFree
Corrected the shared library libcudafor.so
18023 11.8 CUDA Fortran CUDA Fortran reports the error, "invalid type conversion"
Corrected the processing of a DEVICE scalar used as a reduction variable
18021 11.8 CUDA Fortran USE conflict when using both cudafor and cublas
Corrected the use of cuda_stream_kind in the cublas and cudafor modules
18013 11.8 64-bit C/C++ pgcc test fails at -O1/-O2
Corrected a local optimization involving copy propagation of several unsigned long values
18010 11.8 Fortran pgfortran fails to report an error when intent(out) is used with the value attribute
Added a check in the compiler to detect the use of VALUE and INTENT(OUT) as an error
18001 11.8 64-bit Windows Fortran tcsh on Windows reports "Out of memory"
Upgraded cygwin to a newer version which resolves the memory issue on Windows
18000 11.8 Fortran pgfortran derived type array of procedure pointer components returning same result
Corrected the reference of a procedure pointer component whose parent is subscripted
17984 11.8 CUDA Fortran CUDA Fortran missing the GPU_TIME intrinsic
Added the support of GPU_TIME to cudadevice module
17979 11.8 Fortran pgfortran fails to compile LAPACK files csyconv.f and dsyconv.f at -fast
Corrected how the compiler handles certain loops
17973 11.8 Fortran pgfortran fails with "Illegal selector - KIND parameter has unknown value for data type"
Corrected the recognition of complex constants that consist of constants with kind parameters
17972 11.8 64-bit CUDA Fortran CUDA Fortran -mcmodel=medium fails with "undefined references" to cublasAlloc and cublasFree
Corrected the support for -mcmodel=medium in CUDA Fortran
17966 11.8 Profiler pgprof -text reports the error, "Required tool xrefresh was not found"
Removed the dependencies on X when text mode is used
17924 11.8 Fortran pgfortran reports the error, "Lowering Error: bad ast optype in expression"
Corrected the support of non-contiguous character arrays as internal files
17845 11.8 64-bit CUDA Fortran CUDA Fortran -mcmodel=medium fails with "undefined reference to pgf90_dev_auto_alloc_i8' "
Corrected the support for -mcmodel=medium in CUDA Fortran
17828 11.8 Fortran pgfortran reports the error, "Lowering Error: array upper bound is not a symbol for datatype"
Corrected the handling of an array-valued procedure pointer whose extent is adjustable
17578 11.8 64-bit Fortran pgfortran reports the error, "Internal compiler error. unrec dtype in cp_iregarg"
Corrected passing derived type arguments by value
17460 11.8 CUDA Fortran CUDA Fortran missing support for volatile shared arrays
Allow the VOLATILE attribute with the DEVICE and SHARED attributes
16947 11.8 64-bit CUDA Fortran CUDA Fortran does not support -mcmodel=medium
Corrected the support for -mcmodel=medium in CUDA Fortran
16212 11.8 Fortran pgfortran test containing iso_c_binding and VALUE attribute fails
Corrected passing derived type arguments by value
16211 11.8 Fortran pgfortran test containing iso_c_binding with derived type valued function fails
Corrected passing derived type arguments by value
16210 11.8 64-bit Fortran Compiler crashes, ICE, from VALUE attribute
Corrected passing derived type arguments by value
17970 11.7 CUDA Fortran real(8) __fma_rn does not work
Corrected the processing of the double precision version of __fma_m
17945 11.7 64-bit C/C++ pgcpp crashes compiling openmpi 1.4.3
Corrected the handling of constructors appearing in nameless functions
17935 11.7 PGI Accelerstor Fortran pgfortran crashes on code containing module device allocatable arrays
Corrected the compiler's handling of device allocatables referenced within multiple accelerator regions
17928 11.7 Fortran pgfortran crashes on code with derived types containing pointer components
Corrected the handling of self-referential derived types containing initialized pointer components
17922 11.7 Fortran pgfortran example seg faults with -fast
Corrected calling functions which return allocatable arrays
17919 11.7 CUDA Fortran Example fails with "Internal compiler error. unsupported procedure"
Corrected the use of the complex double precision intrinsic, EXP, within global or device routines
17916 11.7 64-bit C/C++ pgcpp example produces errors of the form "Error: undefined symbol '.LB2595' in operation" with -fast
Corrected the vectorizer's processing of loops containing code to handle exceptions
17914 11.7 PVF Installer 32-bit PVF install fails when changing the destination of the installation
Corrected the installation of the 32-bit compilers and tools to a location other than the default
17912 11.7 Fortran Program fails with a false array bounds error
Corrected -Mbounds for certain adjustable array argument when sections of the array are passed to routines as assumed-shape arrays
17908 11.7 Fortran pgfortran examples produces the error, "Lowering Error: bad ast optype in expression"
Corrected the use of the ASSOCIATED intrinsic within a FORALL when its argument is a pointer component
17885 11.7 Fortran pgfortran emits the error, "Constant expression of wrong data type", for correct code.
Corrected the conversion of a named array constant within an initializer
17846 11.7 Fortran pgfortran incorrectly detects the error, "PRIVATE type must not be used for PUBLIC variable no_data"
Corrected the accessibility of use-associated name array constants
17730 11.7 Fortran pgfortran incorrectly reports the error, "Derived type has not been declared - ttt"
Corrected the accessibility of extended derived types
17725 11.7 Fortran pgfortran emits the error, "Illegal statement in the specification part of a MODULE", for final procedures
Corrected the support of final procedures declared within modules
17724 11.7 Debugger pgdbg feature request - up/down arrows to repeat previous commands
Added command-line editing support to pgdbg
17020 11.7 Fortran pgfortran reports the error, "Illegal use of symbol local_decomposition"
Corrected the accessibility of use-associated module variables
15044 11.7 Fortran pgfortran reports the error, "illegal use of symbol myrank - not public entity"
Corrected the accessibility of use-associated module variables
17884 11.6 64-bit C/C++ pgcpp compilation error with a large number of try/catch statements in templates
Corrected the processing of exception handling constructs
17873 11.6 C/C++ pgcpp2 compilation error with OpenMP reductions
Corrected the setup for certain types of reductions clauses appearing in a parallel for
17872 11.6 Fortran pgfortran -Mcuda=emu reports severe errors for character strings
Corrected the compiler to allow character strings when in CUDA emulation mode
17867 11.6 Fortran Error report: "Internal compiler error. Errors in Lowering"
Corrected the compiler's interface processing of pointers to subroutines
17853 11.6 Fortran pgf902 compilation error with deferred character components
Corrected the creation of internal data structures of deferred character members of derived types
17836 11.6 Profiler The presence of JAVA_TOOL_OPTIONS causes 'java -version' to print misinformation
Corrected the processing of the value of the environment variable, JAVA_TOOLS_OPTION
17832 11.6 Fortran Error report: extension where logical operations can be applied to integers
Corrected the handling of hexadecimal constants appearing as operands to logical operators
17831 11.6 Fortran pgfortran compilation error for an array PARAMETER
Corrected the handling of a subscripted array PARAMETER appearing in an initializer
17823 11.6 C/C++ OpenMP C test case exhibiting race-condition behavior
Corrected the scoping of private variables within OpenMP nested task constructs
17816 11.6 Fortran pgfortran -mp error in recursive subroutines with dummy VALUE arguments
Corrected referencing arguments which are passed by value within a recursive subroutine when compiling -mp
17763 11.6 Fortran Error report: "Internal compiler error. get_keyword_args, kwdnfd"
Corrected the processing of a structure constructor for an extended TYPE
17742 11.6 Fortran pgfortran -mp BOTS example produces incorrect results
Corrected the handling of shared variables within nested tasks
17286 11.6 PGI Accelerator Fortran Example gets wrong answers when using data region and sum reduction
Corrected assigning of the reduction value to subscripted arrays
17722 11.6 32-bit Fortran pgfortran reports the preprocessor error "Actual parameters too long for .."
Increased the size of the buffer used to collect the actual arguments, and possibly their expansions, when invoking a function macro
17712 11.6 C/C++ pgcc does not inline function of type complex
Corrected the C compiler to allow inlining of user-defined functions returning complex values
17638 11.6 64-bit Fortran Slow performance of dot product on newer architectures
For newer x86_64 architectures, improved the code generation for certain reductions, such as dot product.
16799 11.5 CUDA Fortran pow raised to integer not supported
Added support for X ** I, where I is integer
17577 11.5 Fortran Error report: "Internal compiler error. initialization expression: invalid array subscripts"
Corrected the processing of an initialization expression consisting of an array constructor appearing in a binary operation
17713 11.5 Fortran Error report: "Illegal selector - KIND = specified with a non-intrinsic type"
Corrected handling of the LOGICAL intrinsic when the kind argument is not default logical type
17714 11.5 Fortran Error report: "Internal compiler error. Errors in Lowering 1"
Corrected the processing of a procedure component whose name is the same as a construct
17726 11.5 Fortran Error report: "Illegal use of abstract type"
Corrected type overloading where a type is the same name as a generic type bound procedure
17731 11.5 Fortran Error report: "Incompatible PASS argument"
Corrected the TKR checking of the PASS argument
17732 11.5 Fortran Error report: "Illegal POINTER assignment - type mismatch, subscripts specified for non-array variable"
Corrected invoking a function which returns a procedure pointer
17738 11.5 PGI Accelerator Fortran Error report: "Internal compiler error. invalid redundant expression"
Corrected the compiler's processing of vector tests inside loops which have vector synchronization
17774 11.5 PGI Accelerator Fortran Mirrored directive allocating memory before allocate is called
Corrected the allocation of 'mirrored' variables when they have the allocatable attribute
17775 11.5 Fortran MAXLOC/MINLOC returns the wrong value when used in an OpenMP parallel region
Corrected the evaluation of MAXLOC and MINLOC appearing within a parallel region when the DIM argument is absent
17786 11.5 Linux Installer Install of libacml_mp.so still has dependencies for libpgbind.so
Corrected the install of ACML in include libpgbind.so when necessary
17789 11.5 C/C++ Error report: "Internal compiler error. sym_is_refd: bad sptr", "-mp -Mvect=sse"
Corrected the handling of long double OpenMP reduction variables
17805 11.5 CUDA Fortran x=2*func(b) produces wrong answers
Corrected the handling of the multiplication of 2 and a function call
17806 11.5 CUDA Fortran Double precision values not properly data initialized
Corrected the initialization of double precision variables which have the constant attribute
17808 11.5 Fortran Test fails when an array section is passed as an argument
Corrected the passing of a section of a module allocatable array
17810 11.5 CUDA Fortran Error report: "Internal compiler error. check_member: member arrived with wrong derived type"
Corrected copying back derived type device members
17812 11.5 C/C++ pgcc reports an error for the gcc __attribute__
Allow the __attribute__ to appear after a struct or union keyword
17835 11.5 64-bit PGI Accelerator Windows Fortran Code containing COMPLEX operations fails
Corrected the handling of the divide of COMPLEX data within accelerator regions
16766 11.4 PGI Accelerator C Allow C99 math functions within accelerator code
Added the support for C99 math functions appearing in accelerator regions
17222 11.4 PGI Accelerator User requests support for erf from accelerator region
Added the support for erf appearing in accelerator regions
17637 11.4 CUDA Fortran Error reported, "Internal compiler error. Errors in Lowering", host to device copying of user-defined types
Corrected the host to device copying of the components of a user-defined type
17668 11.4 PGI Accelerator User requests functions to retrieve profiling information
Added functions to enable/disable timers for accelerator regions and kernels
17673 11.4 Fortran Generic subroutine calls fail within a "select type" construct
Corrected the handling of associated variables within select type
17675 11.4 CUDA Fortran Missing cudaMemGetInfo
Added cudaMemGetInfo to the cudafor module
17701 11.4 64-bit C/C++ Inline C99 functions cimag, creal and conj
Enabled the support in pgcc to inline these functions
17704 11.4 Fortran Wrong ubound when compiling with -i8
Corrected the computation of ubound and lbound with a dim argument in the presence of -i8
17707 11.4 Fortran Passing complex by value fails
Corrected passing complex arguments by value on the stack
17715 11.4 Fortran Example causes compiler to hang
Corrected the automatic deallocation of allocatable components when the derived type also contains type bound procedures
17720 11.4 Fortran Erroneous POINTER assignment rank mismatch error
Corrected the processing of procedure pointer assignments when the target is an array-valued function
17727 11.4 Fortran Error reported, "Using structure constructor for type with private components"
Corrected the processing of structure constructors of derived types containing type bound procedures
17728 11.4 Fortran Error reported, "ALLOCATE Type-Specification is incompatible with type of object"
Corrected the type compatibility checks of unlimited polymorphic entities
17737 11.4 Fortran FLOAT intrinsic does not handle i8 argument
Corrected the evaluation of the FLOAT intrinsic of a 64-bit integer
17741 11.4 CUDA Fortran cudamemcpy2d error when using c_devptr
Corrected passing a c_devptr object to memcpy2d
17750 11.4 Windows Fortran pgfortran mssing 3F functions derf and derfc on Windows
Added the routines, derf and derfc.
17751 11.4 C/C++ pgcpp compilation fails for an empty namespace
Corrected processing of empry namespaces
17616 11.3 CUDA Fortran Error reported, "Internal compiler error. unsupported procedure, pgf90_alloc"
Report a compilation error if an allocation of a temporary for an argument occurs in a global or device routine
17636 11.3 Profiler pgprof-Error-option -text is not available in this release
Corrected the support for text mode
17647 11.3 PVF Add -Mbounds property to PVF
Added the "Runtime" property page which contains properties for checking array bounds, checking pointers, and checking the stack
17669 11.3 CUDA Fortran Error reported, "Internal compiler error. bad sincos operation"
Corrected the compiler's processing of sin and cos when they have the same argument
17677 11.3 Fortran pgfortran fails to compute the correct result for sum
Corrected the assignment of a sum with dim=1 and rank >= 2 to an array section where its bounds are not the same as the bounds in the 2nd dimension of the argument
17683 11.3 CUDA Fortran Use of device memory declarations of complex(8) causes the error "identifier "dcmplx2" is undefined
Corrected the processing of complex(4) and complex(8) declarations
17687 11.3 C/C++ Missing _mm_cvtss_f32 SSE intrinsic
Added support for _mm_cvtss_f32
17689 11.3 CUDA Fortran Incorrect answers when using ISHFT
Corrected the evaluation of ISHFT when the shift count is negative
17690 11.3 C/C++ C++ OpenMP program gets incorrect answers
Corrected the processing of the omp for variable
17693 11.3 Fortran pgfortran reports a severe error, "Derived type argument cannot be INTENT(IN)"
Corrected the checking of the INTENT of a derived type dummy argument when its pointer component appears in a pointer assignment statement
17695 11.3 Fortran pgfortran format string '(-16PF7.4)' fails during execution
Corrected 'P' editing for the 'F' edit descriptor
17681 11.2 32-bit Fortran Link error "unresolved external symbol"
Corrected the generation of device function names
17686 11.2 PGI Accelerator Call to cuMemcpy2D returns invalid value error
Corrected arguments to CUDA 3.2 driver version of cuMemcpy2D
16166 11.2 Fortran pgcc -Mnostdlib suppresses -pgf90libs
Enhanced the driver to keep -pgf90libs when -Mnostdlib is specified.
17200 11.2 PGI Accelerator Presence of contained procedures cause spurious 'live-out' messages in the host
Enhanced the analysis to find host subprogram variables used in any contained procedures.
17227 11.2 CUDA Fortran Request to add the elemental functions that manipulate reals
Added support of the intrinsics, exponent, fraction, nearest, rrspacing, scale, set_exponent, and spacing.
17251 11.2 32-bit Fortran User code causes pgfortran to seg fault.
Corrected the processing of type bound procedures and generic type procedures/operators.
17487 11.2 Fortran Inconsistent driver behavior for '-I dir ' versus '-Idir' when -Mipa is present
Corrected the handling of spaces between -I and the directory name.
17502 11.2 Fortran PRIVATE access not being applied
Report errors when processing USE ONLY statements referring to PRIVATE entities.
17536 11.2 Fortran Error reported "Internal compiler error size_of:bad dtype 'and Scale_Of: bad dtype"
Corrected the handling an elemental function returning adjustable length character.
17576 11.2 Fortran Error reported "Internal compiler error. export_dt: illegal dtype"
Corrected processing of an interface appearing in a procedure declaration within a module.
17584 11.2 Fortran Array with the pointer attribute gives incorrect results with SHAPE, LBOUND, UBOUND
Corrected a pointer assignment where the left-hand side contains bounds specifiers.
17598 11.2 Fortran Error reported "Internal compiler error. getsname: bad stype for 0"
Corrected the processing of type bound procedures and generic type procedures/operators.
17599 11.2 Fortran pgfortran reports error "Type-Specification is incompatible with type of object"
Corrected the processing of type bound procedures and generic type procedures/operators.
17600 11.2 Fortran Error reported, "Mismatched data type for member shape"
Corrected the processing of a structure constructor of an extended type.
17601 11.2 Fortran pgfortran namelist read does not recover from when the input file does not exist
Corrected the namelist run-time to properly set IOSTAT when the file does not exist.
17605 11.2 OSX CUDA
Fortran
cufft library is not installed into the CUDA area
Corrected the install packages for OSX.
17606 11.2 Fortran pgfortran error with -12p edit descriptor
Corrected the handling of illegal scale factors.
17609 11.2 64-bit Fortran -Mimsl incorrect for 64-bit systems
Corrected the driver to include libimslscalar.lib when -Mimsl is present.
17611 11.2 Windows Debugger pgdbg fails on Windows when invoked from the start menu
Corrected launching the GUI when invoking pgdbg from the start menu.
17617 11.2 CUDA Fortran GPU error "class "" has no member "m2560"" is reported when calling device routine
Corrected the handling of descriptor arrays in a common block.
17619 11.2 Fortran Error reported, "Internal compiler error. Interf:new_symbol, symbol not found"
Corrected the processing of type bound procedures and generic type procedures/operators.
17620 11.2 32-bit Fortran Incorrect warning messages when using a complicated array declaration
Corrected the processing of an array declaration where a dimension specifier contains an array constructor.
17628 11.2 Fortran F2003 get_command_argument() intrinsic fails to provide accurate error status
Corrected the run-time to return the error value when the length of the command argument exceeds the length of the 'value' argument.
17630 11.2 64-bit C++ Debug information for member functions missing
Corrected the generation of debug information of objects declared in include files.
15067 11.1 Fortran/C/C++ Driver enhancement, ability to create default -tp settings including unified binary
Added the PREOPTIONS variable which can be set with -tp and a list of values in the .mypgirc or siterc file
15979 11.1 Fortran pgfortran resports the error "Internal compiler error. sym_is_refd:bad sty 297" for an illegal context
Corrected the compiler to emit a better message when a variable appears in a specification expression and is later declared as a PARAMETER
17117 11.1 Fortran FORM='BINARY' for the OPEN statement not documented
Updated the pgfortran manual to include a description of 'BINARY' for the FORM specifier
17275 11.1 Fortran Use of COMPLEX by value gives the error "Internal compiler error. unrec dtype in cp_iregarg 11"
Corrected passing complex arguments by value
17287 11.1 PGI Accelerator Feature request for acc_set_device num be a no-op when a context has already been created
Added the enhancement
17580 11.1 C/C++ pgcpp reports false array bounds checking error
Corrected the subscript bounds calculations for -Mbounds
17581 11.1 64-bit Fortran mpif90 does not add -lpthread
Corrected mpif90 to add -lpthread to the link step
17582 11.1 PVF Values of COMPLEX array elements are not displayed correctly
Corrected the display of values in COMPLEX arrays
17585 11.1 PGI Accelerator pgaccelinfo requires explicit paths for shared libs
Corrected the build of pgaccelinfo
17586 11.1 Fortran/C/C++ -Bstatic does not create a static executable
Corrected the linking step for -Bstatic
17593 11.1 OSX Installer Installation incorrect in the 2011 directory
Corrected the installer to add links from the 2011 area to the 11.0 area
15018 11.0 C/C++ pgCC test fails when compiled -Mnoautoinline
Corrected inlining the same function in several instances of a for statement
15667 11.0 PGI Accelerator Support do loops whose loop counts depend on values computed in the region
Enhanced the accelerator analysis to allow these types of loops
16208 11.0 Fortran Request support for the f2003 PROTECTED attribute
Implemented the PROTECTED attribute
16222 11.0 64-bit Fortran Failure with automatic arrays in VASP 5.2
Corrected the size computation of certain automatic arrays where their extents included the use of the max intrinsic
16241 11.0 Cygwin Request support for Cygwin paths
Updated the driver to allow Cygwin path names
16273 11.0 64-bit Fortran Compilation error with -Mipa=fast
Corrected IPA's handling of arguments from certain subprograms
16439 11.0 Fortran User requests flag that would allow setting a Fortran suffix
Added the option -fsuffix to allow users to set the suffix of fortran source files
16691 11.0 64-bit CUDA Fortran Errors "argument mismatch" or "unsupported procedure"
Corrected the error messages for character variables appearing in device routines
16767 11.0 CUDA Fortran Error reported, "copyout Symbol Memcpy FAILED" when passing a module variable as an argument
Corrected the translation of the host address of a module variable to its device address
16790 11.0 CUDA Fortran Cudafor val test fails
Corrected the translation of the host address of a module variable to its device address
16832 11.0 Fortran f2003 serialization.f90 code fails to compile
Added support for f2003 type bound procedures
16873 11.0 64-bit CUDA Fortran Error reported, "unsupported procedure 330"
Corrected the error message when an array assignment requires an temp to be dynamically allocated
16898 11.0 CUDA Fortran Type mismatch error when declaring device variables in non-entity declarations
Corrected the handling of the device attribute in "attribute-oriented" declarations
16926 11.0 64-bit Fortran Error reported, "Internal compiler error. Subscript for array is out-of-bounds"
Corrected the subscript checking within an implied do expression used as an argument to the reshape intrinsic
17051 11.0 64-bit PGI Accelerator Fortran Error reported, "Internal compiler error. no matching ref for symbol"
Corrected the processing of an accelerator region with a sequentially scheduled loop
17115 11.0 Fortran Error reported, "Illegal use of symbol" for a variable with the same name as a module
Corrected the handling of PRIVATE for the name of a module subprogram
17141 11.0 64-bit CUDA Fortran Error reported, "call to cuMemcpyDtoH returned error 700: Launch failed"
Corrected the compiler's analysis of induction expressions within a accelerator region
17156 11.0 32-bit PGI Accelerator 32-bit PGI Unified Binary fails to execute on host without CUDA installed.
Added an error check to the driver to report than unified binaries are not supported on 32-bit systems
17163 11.0 Fortran Error reported, "'internal compiler error. rewrite_block_where: no outer endwhere"
Corrected the processing of a single-statement or block containing a call to an array-valued function
17166 11.0 32-bit CUDA Fortran Incorrect gpu code generated for character constants
Corrected the compiler's code generation of character expressions in global routines
17167 11.0 32-bit CUDA Fortran Type mismatch error when declaring device variables in non-entity declarations
Corrected the handling of the device attribute in "attribute-oriented" declarations
17176 11.0 Fortran Error reported, "Internal compiler error. Errors in Lowering" when an array section is used in as a write's format
Corrected the compiler to report that the array section is an illegal format specifier
17177 11.0 Fortran Error reported, "Internal compiler error. mkexprt1" for the REDIMENSION statement
Corrected the compiler to report a "syntax error" since REDIMENSION is not a valid pgfortran extension
17181 11.0 C/C++ pgcc reports the warning "No va_start() seen"
Corrected the compiler to report the warning only if va_arg is not present
17183 11.0 64-bit Linux Debugger Add X11 warning to linux users for pgdbg
Removed the dependency on 32-bit X libs for the 64-bit version of pgdbg
17184 11.0 PGI Accelerator Fortran Logicals not supported in accelerator regions
Added support for logicals within accelerator regions
17186 11.0 C/C++ pgcc reports the warning "Pragma ignored - string expected after #pragma ident"
Allow macro replacement to occur within #pragma ident
17197 11.0 Fortran pgfortran lbound and ubound test fails
Corrected the compiler's processing of a pointer assignment when the target is a pointer component of an assumed-shape array
17204 11.0 64-bit CUDA Fortran Error reported, "too few arguments in function call"
Corrected the handling of a global subroutine which has several device arguments
17220 11.0 Fortran Error reported, linking an OpenMP example "undefined ref to TPptr_p_"
Corrected the code generation for referencing host-associated threadprivate variables from within an internal procedure
17221 11.0 Linux Installer Install errors on Gentoo Linux
Corrected the installer to properly configure localrc for Gentoo Linux
17223 11.0 CUDA Fortran Error reported, "Internal compiler error. unsupported operation"
Corrected the copying of a derived type variable
17229 11.0 64-bit C/C++ Error reported, "Internal compiler error. copy_elem, unexp. dt"
Corrected assigning structures containing members of type __m128d
17231 11.0 CUDA Fortran Device code does not generate sincos calls
Added the sincos optimization
17245 11.0 Fortran pgfortran f2003 example fails
Corrected the processing of a procedure pointer declared in a module whose interface is a contained procedure in the same module
17250 11.0 CUDA Fortran Missing double precision complex exp and log support
Added support for the single and double complex versions of the exp, log, and sqrt intrinsics
17251 11.0 32-bit Fortran pgfortran f20o3 code causes a seg fault during compilation
Corrected the compiler's processing of the CLASS attribute
17255 11.0 C/C++ pgcc fails for a structure with the attribute __packed__
Corrected the initialization of a struct with the packed attribute which contains pointer members
17256 11.0 Fortran Error reported, "Illegal procedure interface"
Corrected the compiler's handling of a procedure declaration whose interface is a module-contained procedure
17258 11.0 32-bit Fortran pgfortran error for Windows routine globalmemorystatus
Corrected the interface to block the routine
17275 11.0 CUDA Fortran Complex passed by value gives the error "Internal compiler error. unrec dtype in cp_iregarg"
Corrected the processing of a complex argument which has the value attribute
17283 11.0 64-bit Windows Fortran A read of a direct access file larger than 2GB fails on Windows
Corrected the fortran runtime when performing a seek operation in a large file on Windows
17297 11.0 Fortran pgfortran fails to detect an OpenMP error
Corrected the compiler to report an error if a CYCLE statement does not refer to the innermost associated OpenMP loop
17300 11.0 PGI Accelerator Fortran Error reported, "cuMemcpyHtoD returned error 700"
Corrected the handling of multiple sum intrinsics in an accelerator region
17301 11.0 PGI Accelerator Incorrect answers for a loop in an accelerator region
Corrected the handling of multiple sum intrinsics in an accelerator region
17334 11.0 PGI Accelerator Fortran Error reported, "Internal compiler error. pgnvd job exited with nonzero status code"
Improved the cache tuning of large arrays in an accelerator region
17335 11.0 64-bit Fortran Incorrect answers with -Mvect=sse
Corrected the vectorizer's peel-and-shuffle optimization
17399 11.0 32-bit Fortran Bogus subscript bounds errors for zero-size arrays
Corrected the handling of zero-sized arrays in the presence of -C/-Mbounds
17412 11.0 Fortran pgfortran NAMELIST fails when -i8
Corrected the runtime for the IOSTAT variable when compiled -i8
17431 11.0 Fortran pgfortran incorrect result for the iostat= variable when -i8 used
Corrected the runtime for the IOSTAT variable when compiled -i8
17432 11.0 C/C++ The include file sched.h lacks guards for nested includes
Added the appropriate #define and #ifndef directives to sched.h
17433 11.0 32-bit C/C++ The 32-bit pgcc preprocessor does not allow 'ULL' suffixed-constants
Updated the preprocessor to 64-bit numbers
17436 11.0 Fortran pgfortran example issues the runtime error "0: RESHAPE: result type != SOURCE type"
Corrected the compiler's processing of the RESHAPE intrinsic when its argument is an allocatable array
17487 11.0 Fortran Incorrect behavior for '-I dir ' versus '-Idir' when -Mipa is present
Corrected IPA to allow for space between -I and the directory name
17492 11.0 Windows Fortran/C/C++ Runtime ignores MP_BIND and always sets the affinity
Corrected the runtime on Windows to set affinity only if MP_BIND is present and has the value yes
17503 11.0 32-bit Fortran Error reported, "bld_use, ast nyd 53"
Corrected the compiler's processing of character functions whose length is adjustable
17516 11.0 64-bit Fortran CAM code from CCSM fails when compiled -O2
Corrected the optimization of an optional argument with the pointer argument
17533 11.0 64-bit C/C++ Error reported, "Internal compiler error. gen_aili unrec. ili opcode" using -fastsse
Corrected the compiler's handling of __m128 variables in extended asm statements
Click me