PGI Release 2013 Technical Problem Reports
What problems have been fixed in PGI 2013?
Each release of the PGI compilers brings a number of technical corrections to problems discovered by PGI and by users. Submit problems found using our online technical 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 |
|---|---|---|---|
| 19614 | 13.10 | Fortran | pgfortran reports the error, "Internal compiler error, redundss:mark unexpected AST type 4" Corrected the handling of parameterized derived types containing lengths for arrays |
| 19599 | 13.10 | PGI Accelerator | Error with loop unrolling Corrected the processing of a DO index variable which is used after the loop |
| 19589 | 13.10 | Fortran | Code containing a derived type with an allocatable component of abstract type fails during executes Corrected source allocation of an object containing dynamic types |
| 19586 | 13.10 | PGI Accelerator | pgcc reports the error, "Internal compiler error. sym_is_refd: bad sc" Corrected the handling of host_data with VLAs |
| 19581 | 13.10 | Fortran | pgfortran reports the error, "Internal compiler error. size_of: bad dtype 0" Detect the use of c_loc() as an output item as an error |
| 19579 | 13.10 | PGI Accelerator | pgfortran reports the error, "Internal compiler error. gen_aili: unrec. ili opcode: 345" Corrected the handing of a divide by an invariant value within an accelerator region |
| 19573 | 13.10 | Fortran | Excessive compile-time when the power of ** is a large 64-bit integer constant Improved the compile-time evaluation of ** when the power is large |
| 19567 | 13.10 | Fortran | pgfortran reports the error, "Argument number 1 to work: type mismatch" Corrected the handling of a type bound procedure binding name |
| 19566 | 13.10 | Fortran | pgfortran reports the error "Too many type parameter specifiers" Corrected source allocation of an object containing dynamic types |
| 19552 | 13.10 | C/C++ | pgcc OpenMP example fails to terminate Corrected the handling of a ?: expression within a parallel region |
| 19546 | 13.10 | PGI Accelerator | OpenACC code reports "call to cuStreamSynchronize returned error 700: Launch failed" Corrected the test for the worker or vector loop which guards against running over the end of the loop |
| 19545 | 13.10 | 64-bit C/C++ | pgc++ reports the error, " Internal Error: read_memory_region …" Corrected the check of the memory region during zc_eh processing |
| 19544 | 13.10 | C/C++ | pgc++ generates incorrect names for templated classes for debug Account for name mangling when emitting out class names |
| 19532 | 13.10 | Fortran | program reports "_mp_pcpu_idle: more than 256 threads created" with- Mconcur on Windows Corrected the run-time to correctly swap context within a parallel region on Windows |
| 19524 | 13.10 | PGI Accelerator | pgfortran reports the error, "Internal compiler error. Unknown end of block for rewriting: 46" Correctly handle case statements appearing in an accelerator region |
| 19504 | 13.10 | Fortran | pgfortran fails to detect the error, "specified interface is not declared" Corrected the handling of ABSTRACT interfaces within deferred types |
| 19484 | 13.10 | PGI Accelerator | OpenACC: Launch error when using larger private data Corrected the allocation of private arrays within the kernels constructs |
| 19427 | 13.10 | PGI Accelerator | pgcc fails to compile code using a non-constant stride Corrected the handling of accelerator loops with non-constant strides |
| 19425 | 13.10 | PGI Accelerator | pgcpp emits the message, "Accelerator restriction: unsupported statement type" Correctly handle certain functions which are inlined into accelerator regions |
| 19370 | 13.10 | CUDA Fortran | CUDA Fortran reports the error, "too many initializer values" Corrected the data initialization of an array in element form |
| 19227 | 13.10 | C/C++ | pgCC -mp fails within JNI Corrected the run-time which determines the thread id in the presence of JNI or pthreads |
| 18952 | 13.10 | PGI Accelerator | OpenACC: "-acc -ta=host" does not include libacc1 Ensure that -lacc1 is added with just -acc -ta=host or -ta=host |
| 18349 | 13.10 | PGI Accelerator | ACC: Allow a char variable to be a reduction variable Allow a reduction to be type char |
| 18075 | 13.10 | CUDA Fortran | Code fails to execute correctly when compiled -fast Corrected the handling of stores to shared arrays in the presence of optimizations |
| 17898 | 13.10 | Fortran | Poor performance of list-directed read on Windows Improved the I/O run-time's handling of formatted and list-directed read on Windows |
| 19577 | 13.9 | Fortran | pgfortran reports the error, "Argument type mismatch " Corrected an argument check error when the name of an actual argument is also the name of a component |
| 19523 | 13.9 | Windows C/C++ | c++ program fails to execute on windows with -defaultlib:user32 Corrected an error in the exception handling within the c++ run-time |
| 19519 | 13.9 | PGI Accelerator | CUDA Fortran kernel gets GPU error "expression must have integral or enum type" Corrected the handling of several shared input arrays |
| 19517 | 13.9 | C/C++ | pgc++ fails to generate debug information for loop index variables Corrected the handling of -g in pgc++ |
| 19516 | 13.9 | Fortran | IEEE_ARITHMETIC == operator may be elemental Corrected the handling of an overloaded == when the operands are derived type |
| 19515 | 13.9 | Fortran | pgfortran reports the error, "Lowering Error: symbol buff$sd is an inconsistent array descriptor", with ASSOCIATE" Corrected the handling of the selector of an ASSOCIATE constructor |
| 19513 | 13.9 | 32-bit Fortran | pgfortran does not terminate an unformatted record suitable for BACKSPACE Corrected the run-time to update the trailing record length of an unformatted record |
| 19510 | 13.9 | Fortran | pgfortran reports an error when the SOURCE in ALLOCATE is scalar Corrected the use of a scalar SOURCE value in ALLOCATE |
| 19508 | 13.9 | Fortran | Pointer assignment fails when the left-hand side contains sections Corrected the handling of nested pointer assignments |
| 19502 | 13.9 | Fortran | pgfortran reports an error when the SOURCE in ALLOCATE is scalar Corrected the use of a scalar SOURCE value in ALLOCATE |
| 19494 | 13.9 | PGI Accelerator | OACC: Using acc_init reserves all devices on a system Corrected the handling of systems with multiple GPUs |
| 19492 | 13.9 | Fortran | pgfortran example containing a namelist read fails to execute Corrected the skipping items in a file when performing a namelist read |
| 19485 | 13.9 | CUDA Fortran | CUDA Fortran reports an error when a module device variable is initialized Corrected the handling of initialized variables |
| 19478 | 13.9 | 64-bit C/C++ | pgCC fails to compile use of Corrected the failing cases when the gnu version is 4.7 |
| 19473 | 13.9 | Fortran | pgfortran reports an error when the SOURCE in ALLOCATE is scalar Corrected the use of a scalar SOURCE value in ALLOCATE |
| 19472 | 13.9 | Fortran | pgfortran source allocation fails to execute Implemented the source allocation f2008 feature where the destination array inherits the bounds of the source |
| 19470 | 13.9 | Fortran | pgfortran zero-size array assignment fails to compile Corrected handling of an empty array constructor |
| 19468 | 13.9 | C/C++ | pgc++ fails to compile cmath example containing std::isnan Corrected the use of the cmath function isnan |
| 19462 | 13.9 | CUDA Fortran | CUDA Fortran fails to compile '**' when -i8 is present Corrected the handling of the 'x**I' when I is a 64-bit integer |
| 19461 | 13.9 | Fortran | pgfortran example containing a FORALL fails to compile Corrected the handling of a derived type member appearing in a FORALL assignment |
| 19455 | 13.9 | Fortran | pgfortran list-directed write of a denormal number results in a field overflow Corrected the list-directed output of a denormal when -Mdaz and -Mnoflushz are present |
| 19454 | 13.9 | Fortran | pgfortran random_number is not thread safe Corrected the run-time for random_number and random_seed |
| 19422 | 13.9 | Fortran | pgfortran reports the error, "PGF90-F-0487-Index variable j doesn't appear in a subscript on the lhs of FORALL" Corrected the handling of array and structure constructors within a FORALL |
| 19419 | 13.9 | PGI Accelerator | OpenACC tdot example fails to execute Corrected the processing of a copyin of an allocatable array with -ta= nvidia,host |
| 19414 | 13.9 | CUDA Fortran | CUDA Fortran pinned arrays fail when used in ACC kernels Corrected the handling of arrays which are already pinned |
| 19413 | 13.9 | 64-bit C/C++ | pgcc fails to compile code due to missing definition of __REDIRECT_NTH in cdefs.h Updated cdefs.h to be compatible with systems containing gnu version 4.7 |
| 19410 | 13.9 | All | OpenMP programs with tasks act differently than gnu and intel Improved the execute-immediate checks of a task enclosed within a parallel construct |
| 19326 | 13.9 | C/C++ | pgcc reports errors for designated initializers within a struct Corrected the processing of nested structs containing designated initializers |
| 19284 | 13.9 | PGI Accelerator | Need support for OpenACC F90 pointer within accelerator kernel region Implemented the use of F90 pointers from within kernel regions |
| 19280 | 13.9 | Debugger | files command appears to hang with OpenMPI debugging Improved the implementation of the files commands |
| 19237 | 13.9 | Installation | pgcc reports errors with types.h on systems containing gcc version 4.7.1 Adjusted various header files to be compatible with gcc version 4.7.1 |
| 19216 | 13.9 | C/C++ | pgcc builtin for pow does not raise FE_INVALID Added the appropriate checks to the pow run-time routine |
| 19041 | 13.9 | PGI Accelerator | Binary & operator fails in an OpenACC compute region Corrected the handling of the & operator |
| 18860 | 13.9 | Debugger | pgdbg fails to parse func "" Corrected handling of null function names |
| 18585 | 13.9 | C/C++ | pgCC reports the error, ''terminate called without an active exception'' Corrected the handling of exceptions for member functions |
| 18513 | 13.9 | Fortran | pgfortran should detect illegal declarations of arrays Updated the compiler to report an error if local variables are used to define array bounds |
| 18415 | 13.9 | Fortran | pgfortran array section assignment fails with -i8 Corrected the handling of array sections containing integer(4) expressions when -i8 is present |
| 18414 | 13.9 | Fortran | pgfortran problem with derived types in a FORALL statement Corrected the handling of structure constructors within a FORALL |
| 17524 | 13.9 | Fortran | Request the f2008 STOP feature Implemented the f2008 feature where the stop-code is used as the exit status |
| 16211 | 13.9 | Fortran | iso_c_binding with struct/derived type-valued function fail during execution Corrected the handling of BIND C functions returning derived type |
| 15583 | 13.9 | Fortran | pgfortran array section assignment fails with -i8 Corrected the handling of array sections containing integer(4) expressions when -i8 is present |
| 14746 | 13.9 | PVF | PVF does not handle .inc files correctly Added the .inc file extension to the list of files which are treated as source files |
| 17171 | 13.7 | Debugger Windows | "Windows - No Disk" error occurs with PGDBG Corrected the processing of source file paths in the presence of removable drivers without connected media |
| 17211 | 13.7 | 64-bit C/C++ | pgcpp program fails to catch exception when compiled -m64 -fPIC Corrected the processing of destructors called from a try block |
| 17688 | 13.7 | Fortran | OpenMP NBP 3.3.1 LU fails on multiple threads Corrected an optimization of variables appearing in the FLUSH directive |
| 17968 | 13.7 | Profiler | pgcollect -time with -Mprof=time fails Report as an error the use of pgcollect time with -Mprof=time |
| 18115 | 13.7 | CUDA Fortran | CUDA Fortran: cannot allocate 2GB or more on GPU Updated the compiler and run-time to allow large objects on the GPU |
| 18330 | 13.7 | 64-bit Fortran | Unformatted files with large records are not compatible with ifort Updated the representation of large unformatted records so that they are compatible with ifort |
| 18389 | 13.7 | Fortran | Incorrect result returned ieee_is_normal when using -i8 Updated the IEEE module support to be compatible with -i8 |
| 18656 | 13.7 | CUDA Fortran | CUDA Fortran reports errors for variables named the same as intrinsics Corrected the compiler's handling of name conflicts |
| 18881 | 13.7 | C/C++ | pgcpp reports the error, "internal error: insert_statement: cannot insert non-expr statement aaTable[0] =" Corrected an error with dynamic initializations |
| 19046 | 13.7 | Profiler | pgsampt fails when using pgcollect -time with -exe and a script argument Report as an error the use of pgcollect with -exe and -time |
| 19055 | 13.7 | PGI Accelerator | OpenACC example fails to compile Corrected the handling of a derived type argument |
| 19134 | 13.7 | Fortran | Incorrect results with pgfortran code compiled -Mvect=sse Corrected an error with vectorizing integer MIN and MAX intrinsics |
| 19163 | 13.7 | Fortran | pgfortran example containing OpenMP WORKSHARE fails to execute Corrected the handling of an array assignment within WORKSHARE when the right-hand side is first evaluated to a temporary |
| 19174 | 13.7 | PGI Accelerator Windows | PGI_ACC_TIME and PGI_ACC_NOTIFY fail Correct the install of the DLLs needed to aupport PGI_ACC_TIME and PGI_ACC_NOTIFY |
| 19236 | 13.7 | PGI Accelerator | MPI code containing OpenACC and CUDA reports the error, "call to cuMemAlloc returned error 709: Context is destroyed or not yet created" Corrected the run-time to ensure the CUDA context is created when mixing OpenACC and CUDA |
| 19254 | 13.7 | C/C++ | pgcc is missing the C99 TC2 feature, static may be a qualifier in an array declarator Allow static to appear as array declarator qualifier |
| 19328 | 13.7 | All | Options "-Bstatic -mp=numa" results in a dynamic executable being created Document that a statically linked executable cannot be created with -mp=numa |
| 19340 | 13.7 | Fortran | pghpf reports the error "Unrecognized command line switch: -tp" Add the -tp switch to pghpf |
| 19354 | 13.7 | 64-bit Fortran | pgcpp examples fails to compile with '-g -mp -fast' Corrected the dwarf generation of lexical blocks |
| 19362 | 13.7 | C/C++ | Add support for C_INCLUDE_PATH + other GNU environment flags Add support for CPATH and C_INCLUDE_PATH to pgcc and pgcpp; add support for CPLUS_INCLUDE_PATH to pgcpp. |
| 19365 | 13.7 | Fortran | pgcpp does not handle inline assembly correctly Account for variables appearing in input and output expressions in asm statements |
| 19368 | 13.7 | C/C++ | pgcpp reports the error, "Internal compiler error. sym_is_refd: bad sptr" Corrected processing inline functions and templates from iostream |
| 19369 | 13.7 | Fortran | pgfortran reports the error, "Lowering Error: symbol things$sd is an inconsistent array descriptor" Correct the handling of ASSOCIATE construct where the identifier represents an allocatable |
| 19390 | 13.7 | Fortran | pghpf fails when the -tp switch is present Add the -tp switch to pghpf |
| 19391 | 13.7 | 64-bit Fortran | pgfortran program produces incorrect answers with -O3 Corrected loop interchange where an outer loop contained an expandable scalar |
| 19392 | 13.7 | PGI Accelerator | pgfortran reports the error, "Compiler failed to translate accelerator region (see -Minfo messages): Unsupported operation" Allow for the sqrt operation generated by -Mfprelaxed |
| 19396 | 13.7 | Fortran | pgfortran reports the error, "Internal compiler error. size_of: bad dtype 0" for NAMELIST Allow a module NAMELIST to specify module allocatable variables |
| 19398 | 13.7 | Fortran | pgfortran: incorrect type bound procedure gets called Corrected the binding names for certain type bound procedures |
| 19402 | 13.7 | Fortran | The presence of SAVE for module variables inhibits inlining the module's contained procedures Remove the check of the SAVE attribute as an inhibitor for inlining |
| 19411 | 13.7 | PVF | MPI PVF always a process-number license check Removed the license check when "Start Without Debugging" is enabled |
| 19412 | 13.7 | Fortran | pgfortran reports the error, "Too few data constants in initialization statement" Corrected initializing nested derived type variables which contain type bound procedures |
| 19419 | 13.7 | PGI Accelerator | OpenACC code fails with -ta=nvidia,host Corrected generating present_or_copyin of certain arrays when -ta=nvidia,host is used |
| 19423 | 13.7 | C/C++ | pgcpp failure with OpenMP loops containing switch statements Corrected the break statements for a switch within an OpenMP c++ loop |
| 19434 | 13.7 | Fortran | pgfortran errors with overloaded unary operators Corrected the handling of generic unary operators within the context of type bound procedures |
| 19439 | 13.7 | C/C++ | Conflicting OpenMPI mipCC on Mac OS X Removed the name, mpiCC, from the installation |
| 19440 | 13.7 | Fortran | pgfortran reports the error, "Could not resolve generic procedure hdfget" Corrected generic resolution when defined in multiple modules |
| 19443 | 13.7 | Fortran | pgfortran reports the error, "Internal compiler error. interf:new_symbol, symbol not found 1174" Corrected pointer assignments with pointer derived components |
| 19445 | 13.7 | Fortran | pgfortran reports the error, "Internal compiler error. string_expr_length: ast not string op 13" Corrected calling a type bound procedure returning character |
| 19373 | 13.6 | All | Floating point exception occurs with -Ktrap=fp Corrected the processing of double precision exp when the input is NaN |
| 19360 | 13.6 | PGI Accelerator | Code containing an unmatched "!$acc end kernels" fails to compile Corrected the detection of an unmatched END KERNELS directive |
| 19353 | 13.6 | CUDA Fortran | CUDA Fortran memset routines return an incorrect value Return the correct value for certain CUDA memset routines |
| 19350 | 13.6 | Fortran | Invalid error, "Must specify an interface-name or a procedure-name, not both in type bound procedure" Corrected the handling of a type bound procedure with an interface whose name is the same as a deferred type bound procedure |
| 19349 | 13.6 | 64-bit Fortran | Alignment fault occurs in lapack routine, zlasr Corrected a code generation issue with a complex multiple on Nehalem processors |
| 19348 | 13.6 | Fortran | pgfortan error with sourced allocation Corrected the handling of allocate when the destination of the source allocation is polymorphic |
| 19339 | 13.6 | CUDA Fortran | CUDA Fortran code containing "ishift" fails to execute correctly Corrected the handling of an assignment when the LHS is a 1- or 2- byte assignment |
| 19336 | 13.6 | PGI Accelerator | cudaMemcpy2DAsync function on Windows returns an incorrect value Corrected passing an argument by value when the dummy has the IGNORE_TKR attribute |
| 19335 | 13.6 | Fortran | Build of FFTW 3.3.3 fails Corrected the processing of the SSE and AVX intrinsics |
| 19327 | 13.6 | All | Symbols in the run-time pollute the global namespace Removed the offending names from the run-time |
| 19317 | 13.6 | Fortran | pgfortran produces the error, "Internal compiler error. convert_subscript: must be array" Corrected source allocation of an unlimited polymorphic entity |
| 19312 | 13.6 | Fortran | pgfortran reports the error, "Internal compiler error. memsym_of_ast:unexp.ast" Corrected source allocation when the source argument is a reshape intrinsic call |
| 19305 | 13.6 | Fortran | pgfortran example produces the warnings "ieee_/invalid/overflow/inexact is signaling" Replaced the use of libm functions in the implementation of complex double exp |
| 19298 | 13.6 | PGI Accelerator | Code when using "unroll" fails to execute Corrected a code generation bug for loops with "gang unroll" or "parallel unroll" |
| 19293 | 13.6 | Fortran | pgfortran produces the error, "Lowering Error: bad ast optype in expression" Corrected source allocation when the destination is subscripted |
| 19279 | 13.6 | Fortran | pgfortran reports the error, "Internal compiler error. mk_mem_ptr_shape: lwb not subs" Corrected the handling of the ASSOCIATE statement |
| 19272 | 13.6 | Fortran | pgfortran reports the error" "Non-polymorphic selector in select type" Corrected the detection of "Non-polymorphic" errors in SELECT TYPE |
| 19262 | 13.6 | Fortran | pgfortran reports the error, "Invalid type extension" Corrected extending a use-associated type |
| 19253 | 13.6 | PGI Accelerator | pgfortran reports the warning, "Accelerator region loops nested too deeply" Increased the limit of the allowed nesting depth of loops |
| 19250 | 13.6 | C/C++ | pgcpp reports the error, "Internal compiler error. union_find_last_lp_per_handler:empty throw_bih" Updated the processing of throw and catch regions while inlining |
| 19217 | 13.6 | Fortran | pgfortran fails to compile code containing zero-size array arguments Corrected the handling of a zero-size array constructor appearing as an actual argument |
| 19176 | 13.6 | Windows | .Net 4.5 Windows prerequisite does not work if .Net 4.0 is installed Corrected the Windows installation to install .Net 4.5 when .Net 4.0 is present |
| 19172 | 13.6 | Fortran | ieee_is_finite returns incorrect value when -fpic is used Corrected how use-associated variables are referenced when -fpic is present |
| 19133 | 13.6 | C/C++ | Incorrect results in c++ code using limits STL Corrected an optimization error with an unsigned multiply with large inputs |
| 19117 | 13.6 | PGI Accelerator | Using "double" for loop bounds causes the error, "Load of NULL symbol" Improved the processing of "acc parallel" loops which have floating point loop tests |
| 19114 | 13.6 | PGI Accelerator | pgfortran -Mcuda driver attempts to link in non-existent libcuda Removed -lcuda from the link when -Mcuda is used |
| 19092 | 13.6 | C/C++ | pgcpp reports the error, "Illegal type conversion required" Corrected processing a type conversion within a conditional operator |
| 19059 | 13.6 | C/C++ | pgcc reports errors for VLA arguments Allow a VLA specifier for an argument to refer to another argument |
| 19014 | 13.6 | PGI Accelerator | OpenACC - allocated zero-size arrays are detected as not PRESENT Corrected the run-time to properly allocate an array whose size is zero |
| 18956 | 13.6 | CUDA Fortran | CUDA Fortran does not allow I**J in a kernel where I and J are integer Improved the handling of ** when the base and power are integers |
| 18833 | 13.6 | CUDA Fortran | CUDA Fortran reports the error, "Internal compiler error. Unsupported procedure" Improved the error detection of allocate/deallocate within device code |
| 18721 | 13.6 | 64-bit Fortran | -Macml option on Windows fails to use the correct linker flags Corrected PVF's handling of the "Use ACML" property |
| 18059 | 13.6 | CUDA Fortran | CUDA Fortran reports the error, "Internal compiler error. unsupported procedure", in the presence of -mp Improved the detection of unsupported procedures in device code |
| 17664 | 13.6 | CUDA Fortran | CUDA Fortran does not allow MINLOC and MAXLOC Added support for the use of the MINLOC and MAXLOC intrinsics in kernel regions |
| 17556 | 13.6 | Fortran | Link errors occur with ACML on Windows Support static linking of the ACML library |
| 17012 | 13.6 | Fortran | pgfortran fails to compile code containing the OpenMP COLLAPSE directives Corrected the processing of named DO loops and the presence of COLLAPSE clauses |
| 19318 | 13.5 | Fortran | complex(16) matmul not giving consistent answers Updated the run-time to use the correct precision when forming a complex value |
| 19314 | 13.5 | PGI Accelerator Fortran | pgfortran reports the error, "Compiler failed to translate PGI Accelerator region (see -Minfo messages): Unable to find flow graph successor edge to remove" Corrected the processing of kernel regions containing nested loops and sum reductions |
| 19303 | 13.5 | PGI Accelerator | -Mcuda code calling internal procedures fails Corrected the processing of an internal procedure whose host routine is a module-contained procedure |
| 19296 | 13.5 | PGI Accelerator | OpenACC reports the error, "Compiler failed to translate PGI Accelerator region (see -Minfo messages): illegal opcode" Corrected the processing of kernel regions containing multiple sum reduction variables |
| 19295 | 13.5 | Debugger | pgdbg fails on Windows for long command lines Corrected memory corruption issues in the command line processing of pgdbg |
| 19294 | 13.5 | Fortran | pgfortran code containing a 2D pointer reshape fails Corrected pointer reshape when the target is a multi-dimension type containing type bound procedures |
| 19285 | 13.5 | PGI Accelerator | Kernel launch fails when there are device arrays larger than 2GB Allow copying arrays whose size in bytes exceeds a 32-bit integer |
| 19274 | 13.5 | Fortran | pgfortran reports a syntax error for the ROUND write specifier Added support for the ROUND specifier in a WRITE statement |
| 19273 | 13.5 | Windows Fortran | msvcr110.dll is not compatible with Visual Studio Update 2 Removed the file mscvrt110.dll from the Windows installation packages |
| 19271 | 13.5 | PGI Accelerator | Code containing the acc if clause fails Added support for the acc if clause |
| 19269 | 13.5 | PGI Accelerator | Incorrect results occur when mixing the mirror, reflected, and update clauses Corrected processing combinations of mirror, reflected, and update |
| 19260 | 13.5 | Fortran | Compilation errors occur with mirror/update and USE rename Corrected the processing of variables in a mirror/update clause which have been renamed with USE association |
| 19259 | 13.5 | 32-bit Windows C/C++ | Compilation errors occur with winnt11.h on 32-bit Windows Corrected the syntax of the __asm statements in winnt11.h |
| 19251 | 13.5 | PGI Accelerator | Errors occur in OpenACC and CUDA Fortran with x**n, where n is integer(8) Detect and handle the exponentiation operator when the power is a 64-bit integer |
| 19249 | 13.5 | PGI Accelerator | cudaMemcpy2D fails with P2P enabled Corrected the interfaces through the layers of calls for certain P2P and memory copy operations |
| 19246 | 13.5 | CUDA Fortran | pgfortran -Mcuda reports the error, "Internal compiler error. Unsupported procedure 0" Added support for the minloc and maxloc intrinsics in CUDA fortran |
| 19243 | 13.5 | Fortran | Request for the f2008 NEWUNIT feature Added support for the NEWUNIT specifier in the OPEN statement |
| 19242 | 13.5 | Profiler | pgprof is not always able to generate CCFF information a PVF-built executable Corrected how file path names are processed on Windows by pgprof |
| 19241 | 13.5 | PGI Accelerator | Incorrect results for OpenACC code due to auto-caching Corrected the handling of acc register-cached subscripted value |
| 19239 | 13.5 | Fortran | pgfortran code containing a finalizable type fails Corrected the handling of an automatic array of a finalizable type |
| 19237 | 13.5 | C/C++ | pgcc code using types/bits.h fails to compile on OpenSuSE 12.3 Added a bits.h file to be used when the GNU version is 4.3 |
| 19232 | 13.5 | Fortran | pgfortran reports an error when a contained routine has an argument whose name is the same as the host routine Allow the name of a formal argument to an internal procedure to be the same as the name of the host function |
| 19230 | 13.5 | Fortran | Code using a base type containing a default initializer fails during execution Corrected finalization in the context of the automatic deallocation of allocatables |
| 19228 | 13.5 | C/C++ | pgc++ fails to link with -fPIC Corrected processing a catch clause when compiling -fPIC |
| 19226 | 13.5 | PGI Accelerator | OpenACC code calling CUDA C kernels fails Corrected computing the arguments being passed to the CUDA C routines |
| 19223 | 13.5 | All | Missing AVX intrinsics Added support for the intrinsics, _mm_cmp_ps and _mm_cmp_pd intrinsics, and the predicate, _CMP_EQ_OQ |
| 19220 | 13.5 | 64-bit Fortran | pgfortran examples fails to execute Corrected the use of PRESENT of an OPTIONAL fixed-size array argument when the actual argument is an assumed-shape array |
| 19213 | 13.5 | PGI Accelerator Fortran | pgfortran fails to compile OpenACC code Corrected the processing of nested loops in a kernel region |
| 19206 | 13.5 | PGI Accelerator Fortran | pgfortran produces the error, "Lowering Error: unexpected symbol type statement function(15)" Report the use of the name of a statement function appearing in a data clause as an error |
| 19196 | 13.5 | PGI Accelerator Fortran | pgfortran emits the warning, "Compiler failed to translate PGI Accelerator region (see -Minfo messages): illegal opcode" Corrected the handling expression containing a combination of sqrt and divide operations |
| 19190 | 13.5 | PGI Accelerator | OpenACC code containing a reduction clause fails to compile Corrected the processing of a reduction clause appearing before the outermost loop of a loop nest |
| 19175 | 13.5 | PGI Accelerator | OpenACC ignores the if clause on parallel and kernels clauses Added support for the acc if clause |
| 19173 | 13.5 | PGI Accelerator | pgfortran fails to compile OpenACC code Corrected the processing of nested loops in a kernel region |
| 19149 | 13.5 | PGI Accelerator | User-specified gang size is ignored during execution Corrected the processing of the gang size at the point of launching the kernel |
| 19141 | 13.5 | PGI Accelerator | pgfortran emits the warning, "Compiler failed to translate PGI Accelerator region (see -Minfo messages): Unknown variable reference" Corrected the handling of an assignment in a kernel region but not in any loop |
| 19130 | 13.5 | PGI Accelerator | pgcc emits the warning, "Compiler failed to translate PGI Accelerator region (see -Minfo messages): Unexpected load/store type" Improved the handling of the acc cache pragma |
| 19107 | 13.5 | Fortran | Allow the type C_PTR to be used in a function type specifier appearing before USE of iso_c_binding Allow the type C_PTR from the iso_c_binding module to appear in the function statement |
| 19103 | 13.5 | PGI Accelerator | pgfortran example fails to compile with -acc Corrected accessing certain compiler data structures created from PGI Accelerator regions |
| 19081 | 13.5 | PGI Accelerator | pgfortran reports the error, "Internal compiler error. cf_data_init: unexpected zeroes" Corrected the handling of -Mcuda=nofma |
| 18995 | 13.5 | PGI Accelerator | Various compilation errors occur with pgcc -acc -ta=nvidia:cc20,keepgpu Corrected various subscripting issues within PGI Accelerator regions |
| 18911 | 13.5 | PGI Accelerator | OpenACC COSMO fails during execution Corrected the processing of the update clauses for certain types of variables |
| 18889 | 13.5 | All | pgcc/pgcpp/pgfortran lack -Mmpi=openmpi switch on OS X and Linux Ensured that the documentation states that the Open MPI compiler wrappers must be used to build Open MPI programs |
| 18476 | 13.5 | Profiler | Profiling Open MPI applications fails for pgf77 and pgfortran applications Updated the Open MPI compiler wrapper instructions in the Profiler User's Guide |
| 18369 | 13.5 | Fortran | Unformatted direct access read fails on a file whose last record is incomplete Updated the run-time to allow a partial last record in an unformatted direct access file |
| 18286 | 13.5 | Profiler | Open MPI compiler wrapper instructions are obsolete Updated the Open MPI compiler wrapper instructions in the Profiler User's Guide |
| 18147 | 13.5 | PGI Accelerator | Timing individual loops fails with -Mcuda Corrected the compilers to time PGI Accelerator kernels the same way with -acc with or without -Mcuda |
| 17719 | 13.5 | Fortran | pgfortran reports the error, "Internal compiler error. getsname: bad stype" for namelist I/O with derived types Corrected the processing of namelist I/O of a derived type containing allocatable or pointer components |
| 17171 | 13.5 | Debugger | pgdbg user on Windows sees the error, "Windows - No Disk", and an exception Revised how pgdbg determines file and directory existence on Windows |
| 16712 | 13.5 | C/C++ | Compiler reports an error when the statement of the for loop is an OpenMP statement Corrected the compiler to allow an OpenMP statement in a for without being enclosed in braces |
| 3939 | 13.5 | Fortran | Enhancement Request - allow a USE-associated kind selector in a function type specifier Allow a type specifier or kind selector from a module to appear in the function statement |
| 19240 | 13.4 | All | Name of function in -Minline may not contain digits Corrected the driver to allow digits in a function name |
| 19221 | 13.4 | Fortran | FORTRAN type bound procedure call fails Corrected the processing of a pointer-valued function with a class argument of the same type |
| 19211 | 13.4 | PGI CDK | PGI CDK node connectivity check is excessive for installations with a large number of nodes Improved the installation script to optionally perform the connectivity check |
| 19210 | 13.4 | OS X CUDA Fortran | Compilation fails with the -Mcuda=rdc option on OS X Corrected the handling of the rdc option |
| 19207 | 13.4 | Fortran | pgfortran reports the error, "Internal compiler error. Macro recursion 0". Emit a proper error message when macro recursion is detected |
| 19203 | 13.4 | Fortran | Nested OpenMP program fails to execute Corrected the handling of task regions within nested parallel regions |
| 19180 | 13.4 | PVF | PVF fails to link due to missing libaccapimp.lib Added the missing library to the 64-bit Windows installs |
| 19171 | 13.4 | Fortran | pgfortran accepts illegal input for a number in a formatted read Corrected the checking of the syntax for real numbers |
| 19075 | 13.4 | 64-bit Fortran | pgfortran incorrectly reports an error for a missing .mod file Corrected the handling of nested USE statements within mixed PRIVATE and PUBLIC access specifiers |
| 18877 | 13.4 | All | Improper check of OMP_NUM_THREADs for a non-MP program. Defer the check of OMP_NUM_THREADs until a program enters a parallel region |
| 18501 | 13.4 | All | Poor performance of the STREAMs benchmark on Interlagos and SandyBridge Improved the compilers to perform the streaming when targeting AVX-enabled processor |
| 17633 | 13.4 | Fortran | pgfortran reports the error, "Could not resolve generic procedure" Corrected the processing of generics defined in multiple modules with mixed PUBLIC and PRIVATE access specifiers |
| 15413 | 13.4 | Fortran | pgfortran example containing EQUIVALENCE statements fails during execution Corrected the processing of EQUIVALENCE statements in a MODULE |
| 19150 | 13.3 | PGI Accelerator C/C++ Windows | pgcpp fails to recognize -acc on Windows Enabled accelerator support in pgcpp on Windows |
| 19140 | 13.3 | Fortran | pgfortran INQUIRE by SIZE fails Corrected the processing of the SIZE specifier of an INQUIRE statement |
| 19138 | 13.3 | C/C++ | pgcpp compiled code leads to the linker error, "undefined ref: `LAMMPS_NS::PairGranHertzHistory::__dt( (void))" Corrected the handling of out-of-class inline declarations |
| 19122 | 13.3 | C/C++ | pgcpp reports the error, "Internal compiler error. unexpected init in gen_init_expr 0" Corrected an issue with a discarded initializer |
| 19118 | 13.3 | Fortran | pgfortran reports the error, "Internal compiler error. mkexpr1: bad id 14" Corrected the processing of a user-defined assignment from generic type bound procedures |
| 19110 | 13.3 | Fortran | pgfortran reports the error, "Constant expression of wrong data type" Allow C_NULL_PTR to be an initializer for an array of TYPE(C_PTR) |
| 19109 | 13.3 | Fortran | pgfortran reports the error, "Empty structure constructor()" Correcting the handling of an empty structure constructor |
| 19106 | 13.3 | Fortran | pgfortran reports the error, "Internal compiler error. getsname: bad stype for 0" Corrected the processing of the deallocate statement of a member of a derived type which is finalizable |
| 19096 | 13.3 | All | Errors with creating static executables Corrected the handling of -Bstatic in the absence of numa libraries |
| 19094 | 13.3 | Fortran | pgfortran reports a syntax error when using TYPE IS Corrected the recognition of the name of a derived type in the TYPE IS statement |
| 19092 | 13.3 | C/C++ | pgcpp fails in type conversion in conditional operator Corrected the handling of m128 operand types in a conditional operator |
| 19090 | 13.3 | Fortran | pgfortran reports the error, "Derived type has not been declared - c_mpfr_type" Corrected the handling of the access attribute for an entity-oriented PARAMETER declaration |
| 19083 | 13.3 | Fortran | pgfortran incorrectly reports a finalization error Corrected the checks of the requirements for final procedures |
| 17716 | 13.3 | Fortran | pgfortran reports many errors for legal f2003 code Corrected the handling of parameterized derived types and structure constructors |
| 15105 | 13.3 | Debugger | pgdbg attach command fails Corrected an issue with attach immediately following a detach command |
| 19105 | 13.2 | Profiler | pgsampt attempts to write to a file in a non-existent directory Removed the erroneous writes from pgsampt. |
| 19104 | 13.2 | Fortran | pghpf reports the error, "ILM file line 21: expecting bit datacnst" Corrected the compiler |
| 19101 | 13.2 | C/C++ | pgc++ fails to recognize the option, -acc Added -acc to the pgc++ driver |
| 19095 | 13.2 | PGI Accelerator Fortran | pgfortran fails to compile the COSMO OpenACC code Corrected the handling of optional arguments for device routines |
| 19093 | 13.2 | 64-bit Fortran | pgfortran -i8 fails to compile Gaussian correctly Corrected the data initialization of a Hollerith constant in the presence of -i8 |
| 19084 | 13.2 | Fortran | pgfortran reports the error, "Internal compiler error. add_subscript_list: nonconstant subscript" Corrected the processing of array sections with negative strides in initialization expressions |
| 19061 | 13.2 | Fortran | pgfortran NAMELIST and type extension fails Corrected the processing of a NAMELIST group containing an object of extended type |
| 19056 | 13.2 | Fortran | pgfortran fails to report a redefinition error Improved the error detection of duplicate declarations of allocatable arrays |
| 18998 | 13.2 | Fortran | pgfortran reports the error, "Internal compiler error. rewrite_subroutine: unknown stmt found" Corrected an internal failure with allocatables in the absence of -Mallocatable=03 |
| 18991 | 13.2 | PVF Accelerator | PVF lacks support for CUDA emulation Added the CUDA emulation libraries to the PVF install package |
| 18986 | 13.2 | PVF | PVF fails to parse multiple .for files Corrected the detection of files with the .for suffix as fixed-form files |
| 18957 | 13.2 | PVF | 'Stop Debugging' action fails when debugging an MPI program Corrected the handling of the 'Stop Debugging' action in pgdebug and PVF |
| 18933 | 13.2 | OS X | Sandybridge not supported on OSX Upgraded the assembler and compilers to support Sandy Bridge processors on OS X |
| 18932 | 13.2 | PVF | .f77 unsupported file extension Added .f77 as a supported file extension in PVF |
| 18838 | 13.2 | C/C++ | pgcpp fails to compile code which includes "ppl.h" on Windows Corrected the support of new[] and delete[] |
| 18772 | 13.2 | C/C++ | Add support for -MT and -MQ to pgcc Implemented the -MT and -MQ features of cpp |
| 18624 | 13.2 | Fortran | F2003 get_environment_variable() test fails Corrected the implementation of get_environment_variable() |
| 17925 | 13.2 | Fortran | system_clock() fails on Windows Improved the run-time to allow for the poor clock resolution on Windows |
| 17862 | 13.2 | Fortran | Performance deficit between PGI- and Intel- compiled code Added optimizations for the f90 intrinsics matmul and reshape |
| 19070 | 13.1 | C/C++ | pgCC error with OpenMP flush Allow the OpenMP flush pragma to appear inside an empty while statement |
| 19069 | 13.1 | C/C++ | pgcc preprocessing _Pragma operator fails for OpenMP Change the behavior of the preprocessor to insert a newline after the expansion of the _Pragma operator |
| 19066 | 13.1 | CUDA Fortran | pgfortran -Mcuda fails to compile user's code. Correct the processing of the dynamic shared memory |
| 19062 | 13.1 | Fortran | Fortran OpenMP set lock routine overwrite data within a derived type Corrected the Fortran omp set lock routine when called from a serial region |
| 19054 | 13.1 | Fortran | pgfortran sourced allocation of a class fails at execution time Corrected the sourced allocation of a derived type allocatable member |
| 19051 | 13.1 | Fortran | Use of the ASSOCIATE construct fails during execution Corrected the handling of the ASSOCIATE construct |
| 19042 | 13.1 | 64-bit Fortran | character*1 value arguments are not passed correctly to FUNCTIONs Corrected passing character data by value |
| 19038 | 13.1 | Fortran | pgfortran reports the error, ""Derived type has not been declared - a_t" Corrected the use of an empty derived type in a class or type declaration |
| 19037 | 13.1 | Fortran | pgfortran array remapping gives incorrect results Corrected the pointer assignment of a 1D array to an equivalent 3D array |
| 19034 | 13.1 | Fortran | pgfortran reports the error, "ILM file: missing subtype for array datatype 77" Corrected the processing of certain array derived type members in the presence of type bound procedures |
| 19031 | 13.1 | C/C++ | pgCC reports the error, "Internal error: assertion failed: find_assoc_pragma: pragma not found" Corrected the scope processing of pragmas within case statements |
| 19021 | 13.1 | Fortran | False null pointer reported with -Mchkptr Inhibit -Mchkptr when performing an associated inquiry of a pointer |
| 19018 | 13.1 | Fortran | pgfortran code fails to execute with "glibc detected *** ./a.out: double free or corruption (fasttop):" Apply the allocatable assignment semantics for a pointer object of a derived type containing allocatable components |
| 19016 | 13.1 | Fortran | pgfortran reports the error, "Internal compiler error:mk_forall_sptr: not member or subscript" Corrected the processing of a FORALL statement when the index variables appears in a character substring reference |
| 19011 | 13.1 | C/C++ | pgcc incorrectly reports the error, "Redefinition of symbol" Corrected the processing of an argument in a function prototype when it also appears in a VLA expression |
| 19005 | 13.1 | 64-bit Fortran | pgfortran fails to compile code containing parameterized derived type objects with the PARAMETER attribute Corrected applying the PARAMETER attribute to parameterized derived types |
| 18979 | 13.1 | CUDA Fortran Documentation | Remove confusing thread ID description from the CUDA Fortran manual Updated the manual |
| 18977 | 13.1 | PGI Accelerator Documentation | Errors in the OpenACC chapter of the reference manual Corrected the documentation |
| 18973 | 13.1 | Fortran | pgfortran reports an error for a module procedure pointer whose interface is a module subprogram Corrected the procedure pointer assignment of a module subprogram defined in the same module |
| 18968 | 13.1 | Fortran | Calling through procedure pointer without () fails Corrected calling a procedure via a procedure component without () |
| 18965 | 13.1 | Fortran | pgfortran produces the error, "Lowering Error 'symbol ring$sd is an inconsistent array descriptor' Corrected the handling of an allocatable member in a call to move_alloc |
| 18964 | 13.1 | Fortran | pgfortran produces the error, "Internal compiler error. make_alnd: descriptor does not exist" Corrected calling a module function returning an allocatable array from within an internal procedure |
| 18962 | 13.1 | Fortran | pgfortran produces the error, "Illegal use of symbol- not public entity of module" Correct the processing of USE ONLY when mixed access types are present |
| 18953 | 13.1 | Fortran | pgfortran fails to report an error for a recursive function without a RESULT clause Corrected the compiler to report error when the name of function is not used to invoke the function |
| 18945 | 13.1 | 32-bit Fortran | pgcc reports the error, "Illegal use of symbol, __SIZE_TYPE__", when including glob.h Added the appropriate definition of __SIZE_TYPE__ |
| 18939 | 13.1 | Fortran | pgfortran -Minline function names should not be case sensitive The function names specified by -Minline =name and -Minline=except are no longer case sensitive when -Mupcase is absent. |
| 18921 | 13.1 | Fortran | Undefined reference to a name of a generic interface from a shared library with -g Corrected generating debug information of user generics |
| 18920 | 13.1 | C/C++ | Invalid results when optimizing unsigned char Corrected optimizations of small unsigned induction variables |
| 18911 | 13.1 | PGI Accelerator Fortran | OpenACC COSMO fails during execution Corrected the handling of arrays appearing in update clauses |
| 18862 | 13.1 | Fortran | pgfortran generates the error, "Internal compiler error. search_arr: not TY_ARRAY" Corrected the handling of generics specified in modules when mixed access types are present |
| 18856 | 13.1 | Fortran | pgfortran could not resolve the user generic procedure trace Corrected the handling of generics specified in modules when mixed access types are present |
| 18822 | 13.1 | Fortran | pgfortran kind argument mismatch in calls to omp_set_nested and omp_set_num_threads when compiled -i8 Added generic interfaces for various omp_ routines to the omp_lib module |
| 18793 | 13.1 | Installation | FLEXlm utilities cannot detect network device 'em1' which many newer versions of Linux use by default. Upgraded to Flexera's FNP version 11.11 |
| 18780 | 13.1 | ALL | Errors in modulefiles delivered with products Corrected the entry for CPP to refer to the preprocessor |
| 18707 | 13.1 | Fortran | pgfortran produces incorrect results for OpenMP COLLAPSE Corrected the handling of multiple levels of the COLLAPSE directive |
| 18702 | 13.1 | C/C++ | Improper handling of _Pragma in pgcc, pgprepro, Change the behavior of the prepocessor to insert a newline after the expansion of the _Pragma operator |
| 18692 | 13.1 | Fortran | pgfortran program using .not.allocated array in OpenMP Private clause fails during execution Corrected the check of the allocation status of shared allocatables upon entry to a parallel region |
| 18628 | 13.1 | C/C++ | pgcc emits warnings when wcchar.h is included Inhibited potential truncation warning messages when casting 64-bit unslgned values to int |
| 18543 | 13.1 | Fortran | pgfortran should not detect an error for passing a procedure dummy argument Distinguish between passing a subroutine argument which appeared in an EXTERNAL statement versus a procedure statement |
| 18413 | 13.1 | 64-bit C/C++ | pgcc -Bstatic -mp=numa causes execution failure on SLES 11 SP 1 Changed the default behavior of -mp to be -mp=nonuma |
| 18291 | 13.1 | C/C++ | pgcc code containing the OpenMP pragma with ordered clause fails during execution Corrected the handling of ordered loop interations |
| 18263 | 13.1 | 64-bit Fortran | pgfortran generates incorrect debug information for certain derived types Corrected the debug information for a derived type declared in a host routine and used to declare local variables in a contained routine |
| 18258 | 13.1 | Fortran | pgfortran reports the error, "Syntax error at or near WHERE", when a construct name is present Corrected the compiler to allow a construct name to appear in the WHERE, ELSEWHERE, and ENDWHERE statements |
| 18256 | 13.1 | Fortran | Program using OpenMP tasks fails at execution Corrected the handling of firstprivate and lastprivate semantics of f90 pointers in tasks |
| 18191 | 13.1 | Fortran | pgfortran produces warnings indicating truncation of identifiers to 31 characters Corrected the error message to reflect the f2003 maximum identifier of 63 characters |
| 18144 | 13.1 | Fortran | pgfortran produces the error, "Lowering Error unknown target type for conversion" when ISO_C_BINDING is used Corrected the handling of C_LOC appearing in an array constructor |
| 17841 | 13.1 | Fortran | pgfortran produces the error, "Argument missing for formal argument node$sd7" Corrected the checking of arguments when an internal procedure recursively calls its host subprogram |
| 17785 | 13.1 | Fortran | pgfortran code referencing OpenMP firstprivate allocatables fails at execution Corrected the allocatable semantics of OpenMP private allocatable arrays |
| 17780 | 13.1 | All | Need to add GCC 4.5 predefined macros to stdint.h Updated stdint.h to include the macros new to 4.5 |
| 17413 | 13.1 | Fortran | pgfortran code containing UPCASE(TRIM(ADJUSTL(HSTATU))) fails at execution Corrected the processing of the trim intrinsic when its argument is a call of the adjustl intrinsic |
| 17293 | 13.1 | C/C++ | pgcc OpenMP code containing omp for with lastprivate and static schedule fails Corrected the handling of the last iteration of the loop |
| 16816 | 13.1 | C/C++ | Improve the error message for an illegal OpenMP usage of taskwait Corrected the error message to indicate that the taskwait cannot follow a switch statement |
| 16568 | 13.1 | C/C++ | pgcc fails to detect illegal OpenMP #pragma usage Improved the error checking for the appearance of taskwait |
| 14852 | 13.1 | 64-bit C/C++ | Programs compiled with -fpic -shared -ql fail to link when -Mprof or -pg is present t Updated the driver to report an unsupported combination of options |
| 14601 | 13.1 | 64-bit Fortran | Enhance pgfortran to allow alignment of allocatables on page boundaries Added the specifier ALIGN to the ALLOCATE statement as an extension to effect aligning the object on the specified boundary |
