Intel® Software Development Products - Intel® Software Network

	*Intel® Itanium® /Linux Software Instrumentation** Changes in kit 137
	Release Notes for Updates to the Annotations Libraries None - this is the initial version
	Release Notes for Updates to the Network Applications Performance Analyzer None - this is the initial version
	Release Notes for Updates to the Intel® C++ ABI Test Suite version 1.0 Release Notes for Updates to the Itanium® Architecture Floating-Point IEEE Filter Version 1.0 Release Notes for Updates to the Itanium® Architecture Knobs API Version 5.0 Version 4.0 Version 3.0 Version 2.0 Release Notes for Updates to the Itanium® Architecture Dependency Violation Checking Library Version 2.0 Release Notes for Updates to the Itanium® Architecture Assembler and Disassembler Version 6.0 Version 5.0 Version 4.0 Version 3.0 Version 2.0 Release Notes for the Highly Optimized Mathematical Functions for the Itanium® Architecture (Itanium Libm) Version 2.2 Version 2.1 Version 2.0 Version 1.7 Version 1.6 Version 1.5 Version 1.4 Version 1.3 Version 1.2 Version 1.1 Version 1.0 Release Notes for Divide, Square Root, and Remainder Algorithms for the Itanium® Architecture Version 1.2 Version 1.1 Version 1.0 Release Notes for the Non-IEEE Division, Square Root, Reciprocal, and Reciprocal Square Root Algorithms for the Intel® Itanium™ Architecture. Version 1.0 Release Notes for Updates to the Optimized String and Memory Functions Version 1.2 Version 1.1 Release Notes for Updates to the Stack Unwind Library for Intel® Itanium® ; Architecture Version 2.0 released on 8/12/2002

	*Intel® Itanium® /Linux Software Instrumentation** Changes in Kit 137 Set return value for sigaction correctly. Add -ldl to fix build problem for pinprof. Back to top

	Release Notes for Updates to the Annotations Libraries None. This is the initial version. Back to top

	Release Notes for Updates to the Network Applications Performance Analyzer None. This is the initial version. Back to top

	Release Notes for Updates to the Intel® C++ ABI Test Suite None. This is the initial version. Back to top

	Release Notes for Updates to the IEEE Filter Version 1.0 of the Floating-Point IEEE Filter for the Itanium® Architecture is the initial release. Separate versions are provided for the Linux, Windows, HP-UX, and VMS operating systems. Back to top

	Release Notes for Updates to the Itanium® Architecture Knobs API Itanium Architecture Knobs API Version 5.0 (10/30/02) has made the following improvements to Version 4.0: Removed support in BRL instruction type Removed a debug printf statement Itanium Architecture Knobs API Version 4.0 (05/25/01) has made the following improvements to Version 3.0: Fixed a bug related to latency in v26-itanium-41-eternal.knb Itanium Architecture Knobs API Version 3.0(01/17/01) has made the following improvements to Version 2.0: Fixed problems which occured during the initialization of the kapi_ia64. Fixed segmentation fault which occured during kapi initialization on Microsoft Windows* 2000. The reason for this fault was an incorrect number of instruction operands were counted which resulted in an allocation of memory that was too small. Updated the Programmer's Guide. Itanium Architecture Knobs API Version 2.0(09/11/00) has made the following improvement to Version 1.0: L2 cache has 256 lines and 64 bytes per line, as specified in the Itanium® Architecture Reference. This was corrected in the V26-itanium-41-external.knb file. Back to top

	Release Notes for the Itanium® Architecture Dependency Violation Checking Library Version 2.0 of the Itanium® ; Architecture Dependency Violation Checking Library made the following improvements to Version 1.0: Implemented the new dependence semantic "stop" (DCO 511): a branch predicated with p63 and a subsequent mod-sched-br must be separated by a 'stop' . Two bugs were fixed - relevant to abstract mode only! Both are in the API DVLoc_validateAbstrInstr, used to check the correctness of an abstract instruction: The first bug incorrectly produced an DVLOC_ERR_WRONG_OPERANDS error on floating point load pair instructions (ldfp...). The second bug incorrectly produced an DVLOC_ERR_WRONG_OPERANDS error on the following instructions: mov pr = r2,mask17 mov pr.rot = imm44 mov r1 = pr mov itr.d dtr[r3] = r2 Back to top

	Release Notes for the Highly Optimized Mathematical Functions for the Itanium® Architecture (Itanium Libm) Version 2.2 of the Itanium Libm made the following improvements to Version 2.1: Added the following functions: isgreater, isgreaterequal, isless, islessequal, islessgreater, isunordered Improved the speed of cbrtl, cos, fmodl, ldexp, ldexpf, ldexpl, llrint, llrintf, llrintl, llround, llroundf, llroundl, lrint, lrintf, lrintl, lround, lroundf, lroundl, nearbyint, nearbyintf, nearbyintl, powf, round, roundf, roundl, scalb, scalbf, scalbl, scalbln, scalblnf, scalblnl, scalbn, scalbnf, scalbnl, sin, sincos Version 2.1 of the Itanium Libm made the following improvements to Version 2.0: Added the following functions: sincosd, sincosdf, sincosdl Improved the speed of acosh, acoshf, asinh, asinhf, atanh, atanhf, cbrtf, ceil, ceilf, ceill, cosdf, cotd, cotdf, expl, exp10l, exp2l, expm1l, fdim, fdimf, fdiml, floor, floorf, floorl, ilogb, ilogbf, ilogbl, logb, logbf, logbl, rint, rintf, rintl, round, roundf, roundl, sindf, tand, tandf, trunc, truncf, truncl Improved the accuracy of cbrt, cbrtf, expl Version 2.0 of the Itanium Libm made the following improvements to Version 1.7: Added the following functions: acosd, acosdf, acosdl, annuity, annuityf, annuityl, asind, asindf, asindl, atan2d, atan2df, atan2dl, atand, atandf, atandl, compound, compoundf, compoundl, cosd, cosdf, cosdl, cot, cotf, cotl, cotd, cotdf, cotdl, sincos, sincosf, sincosl, sind, sindf, sindl, sinhcosh, sinhcoshf, sinhcoshl, tand, tandf, tandl Improved the speed of acos, asin, cosh, coshf, exp10, exp10f, exp2, exp2f, expm1, expm1f, logf, log10, log10f, log1p, log1pf, powl, sinh, sinhf, tanf Improved the speed on large arguments for cos, cosf, cosl, sin, sinf, sinl, tan, tanf, tanl Improved the accuracy of atan, expf, exp10f, exp2f, log, log10 Version 1.7 of the Itanium Libm made the following improvements to Version 1.6: Added the following functions: acoshl, asinhl, atanhl, erf, erff, erfl, erfc, erfcf, erfcl, lgamma, lgammaf, lgammal, llrint, llrintf, llrintl, llround, llroundf, llroundl, lrint, lrintf, lrintl, lround, lroundf, lroundl, nexttoward, nexttowardf, nexttowardl, scalbln, scalblnf, scalblnl, tanhl, tgamma, tgammaf, tgammal Improved the speed of acosl, asinl, atanl, atan2l, cosl, expl, expm1, expm1f, expm1l, log1p, log1pf, log1pl, sinl, tanh, tanl Improved the accuracy of acosl, asinl, asinh Version 1.6 of the Itanium Libm made the following improvements to Version 1.5: Added the following functions: acosh, acoshf, asinh, asinhf, atanh, atanhf, fdim, fdimf, fdiml, fma, fmaf, fmal, fmax, fmaxf, fmaxl, fmin, fminf, fminl, remquo, remquof, remquol, tanh, tanhf Improved the speed of atan2, cosh, coshf, coshl, logl, log10l, pow, powf, sinh, sinhf, sinhl Improved accuracy of log2 Corrected monotonicity errors in expf, powf, and pow that occurred in directed rounding modes Version 1.5 of the Itanium Libm made the following improvements to Version 1.4: Added the following functions: invsqrt, invsqrtf, invsqrtl, log2, log2f, log2l, nearbyint, nearbyintf, nearbyintl, round, roundf, roundl Improved the speed of atan2f, cos, cosf, exp, ldexp, ldexpf, ldexpl, log, logf, log10, log10f, modf, modff, modfl, pow, powf, scalb, scalbf, scalbl, scalbn, scalbnf, scalbnl, sin, sinf, tan, tanf Changed one sincos table entry to insure symmetry Fixed bug in ilogb, ilogbf, ilogbl for x=0 Fixed bug in nextafterl for smallest normal input producing largest denormal result Fixed ldexp, ldexpf, ldexpl, scalb, scalbf, scalbl, scalbn, scalbnf, scalbnl to correctly generate tiny results Fixed powl(1,inf), (1,nan), and (snan,0) to conform to C99 standard Corrected flag settings in the following functions: acos, acosf, acosl, asin, asinf, asinl, atan2, atan2f, cos, cosf, coshl, modf, log, logf, log10, log10f, modff, modfl, nextafter, nextafterf, nextafterl, pow, powf, powl, sin, sinf, sinh, sinhf, sinhl, tan, tanf, tanl Corrected error handling for the following functions: exp2, exp2f, exp2l, fmod, fmodf, fmodl, ldexp, ldexpf, ldexpl, powl, remainder, remainderf, remainderl, scalb, scalbf, scalbl, scalbn, scalbnf, scalbnl, sinhl Version 1.4 of the Itanium Libm made the following improvements to Version 1.3: Added the following functions: exp10, exp10f, exp10l, exp2, exp2f, exp2l, trunc, truncf, truncl Improved the speed of acos, acosf, asin, asinf, atan, cos, cosf, expf, log, log10, pow, powf, sin, sinf Updated sincos tables to insure symmetry Fixed bug in remainder, remainderf, remainderl at midpoints with quotient exponent=23 Corrected problem in routines calling libm_error where pointer to result could potentially be overwritten To conform to the C99 standard, now pow(+-1,+-inf)=1 and pow(+1,nan)=1 Version 1.3 of the Itanium Libm made the following improvements to Version 1.2: Added the following double-extended precision function: cbrtl Improved the speed of atanf, atan2f, cbrt, cbrtf, ceil, ceilf, ceill, cosf, floor, floorf, floorl, hypot, hypotf, hypotl, ldexp, ldexpf, ldexpl, logf, log10f, scalb, scalbf, scalbl, scalbn, scalbnf, scalbnl, significand, significandf, significandl, sinf Improved the accuracy of hypotl Updated sincos tables to insure symmetry; log tables are now correctly rounded Fixed nextafter, nextafterf, nextafterl to not take software assistance Fixed exception flag settings in several routines Fixed several routines to correctly handle double-extended denormals Fixed pow(inf,0) to return +1.0 Fixed rintl(2^63-0.5) Fixed an accuracy problem in sinhl Back to top Version 1.2 of the Itanium Libm made the following improvements to Version 1.1: Added the following double-extended precision functions: fmodl, frexpl, nextafterl, remainderl Added the following floating-point environment functions: fpgetmask, fpgetround, fpgetsticky, fpsetmask, fpsetround, fpsetsticky Added stack unwind directives Improved the speed of expf, floor, fmod, fmodf, frexp, frexpf, modf, modff, modfl, nextafter, nextafterf, remainder, remainderf Improved the accuracy of log10l Fixed uninitialized variables in atan, atanf, and tanl Several bugs in pow and powf were fixed. The invalid flag was incorrectly set. Error support was called when it should not have been. pow(+-1, inf) did not return NaN. Fixed several problems in libm_error_support Version 1.1 of the Itanium Libm made the following improvements to Version 1.0: Added the following double-extended precision functions: coshl, ldexpl, logbl, modfl, scalbl, scalbnl, sinhl Improved the accuracy of log1pl Modified logbf and logb to adhere to C9X handling of denormals and 0.0 input Fixed coshf to send correct tag to error support Fixed expm1l to avoid segmentation violation Version 1.0 of the Itanium Libm was the initial release, and included 98 highly optimized mathematical functions. Back to top

	Release Notes for Updates to the Itanium® Architecture Assembler and Disassambler Itanium™ Architecture Assembler Version 6.0 (dated Jan 24, 2005) includes the following improvements and new features: Added support to TLS section type. Added new instructions, ld16, st16, cmp8xchg16, tf and hint. Minor bug fixes. Itanium™ Architecture Assembler Version 5.0 (dated Jan 29, 2003) includes the following improvements and new features: Support build on RedHat 7.1 machine. Some more minor bug were fixed. Itanium™ Architecture Assembler Version 4.0 (dated March 15, 2001) includes the following improvements and new features: Fix bug of implicit nop in mlx bundle. Support the directive .restore sp (ELF only). Change the directive .spillreg @restore to .restorereg (ELF only). Add the .hidden and .protected symbol visibility property (ELF only). Some more minor bug were fixed. Itanium® Architecture Assembler Version 3.0 (dated June 15, 2000) includes the following improvements and new features: Automatic generation of unwind info (under the command line flag -X unwind), please see Itanium® Architecture Assembler Users Guide for details). Temporary labels are converted to new symbol with generated name (COFF object files only) All relocation entries based on section start and offset are converted to relocation based on new symbol. (COFF object files only) Created special relocations for labels difference. (COFF object files only) Extended support for long-branch instructions used for optimization. (COFF object files only) Generation of new unwind records (X type spill records). (ELF64 object files only) The March 6th update to Version 2.0 provided the following improvements: The DVLoc library whose sources are in /src/DVLoc can now be built independently of the assembler. The DVLoc library now additionally supports the detection of implied dependency violations. For more details refer to the RELEASE-notes.txt file in the package. The first release of the Version 2.0 Itanium® Architecture Assembler and Disassembler was dated February 15th, 2000. Back to top

	Release Notes for the Divide, Square Root, and Remainder Algorithms for the Itanium® Architecture Version 1.2 of the Divide, Square Root, and Remainder Algorithms for the Intel ® Itanium® Architecture made the following improvements to Version 1.1: Updated latency and throughput data for Itanium 2 Unrolled loops in latency-optimized 8-bit integer division and remainder algorithms; eliminated redundant calculations of the quotient Eliminated the latency-optimized 16-bit integer division and remainder algorithms (no longer better for the Itanium 2 processors than the former throughput-optimized algorithms) In the integer division and remainder implementations, added code to handle division by zero Changed certain section titles to better reflect contents Version 1.1 of the Divide, Square Root, and Remainder Algorithms for the Itanium® Architecture made the following improvements to Version 1.0: Added faster minimum latency algorithms for 8-bit integer divide, 16-bit integer divide, 8-bit integer remainder, and 16-bit integer remainder Improved the maximum throughput double precision square root algorithm Replaced the latency optimized SIMD square root algorithm to avoid unnecessary software assistance requests Improved the latency of unpacked code for SIMD algorithms (executed when pseudo-software assistance requests are raised) Version 1.0 of the Divide, Square Root, and Remainder Algorithms for Itanium® Archiecture was the initial release. Back to top

	Release Notes for the Non-IEEE Division, Square Root, Reciprocal, and Reciprocal Square Root Algorithms for the Intel® Itanium™ Architecture. Version 1.0 is the initial release of the Non-IEEE Division, Square Root, Reciprocal, and Reciprocal Square Root Algorithms for the Intel® Itanium™ Architecture. Back to top

	Release Notes for Updates to the Optimized String and Memory Functions Version 1.2 of the Optimized String and Memory Functions made the following improvements to Version 1.1: Fixed recovery code in strcmp and strlen Fixed memset Version 1.1 of the Optimized String and Memory Functions made the following improvements to Version 1.0: Improved the speed of memset, strcat, strcpy and strcmp Fixed memcpy, strcat, strcmp and strlen

	Release Notes for Updates to the Stack Unwind Library for Intel® Itanium® ; Architecture Version 1.0 released on Aug 3, 2001.