Intel JCC bug #263
Comments
nuald
added a commit
to nuald/benchmarks
that referenced
this issue
Aug 30, 2020
nuald
added a commit
that referenced
this issue
Aug 30, 2020
* Added workaround for Intel JCC bug (#263). * Maintenance update.
|
Fixed for GCC. Apparently, Clang has the similar option, but it's not available in the stable version yet. |
|
It should be in clang 10, |
|
Thanks, I couldn't find this option in their docs. Updated the flags, and provided new results (it's only one test now). |
nuald
added a commit
to nuald/benchmarks
that referenced
this issue
Sep 21, 2020
* Added workaround for Intel JCC bug (kostya#263). * Maintenance update.
nuald
added a commit
to nuald/benchmarks
that referenced
this issue
Oct 31, 2020
* Added workaround for Intel JCC bug (kostya#263). * Maintenance update.
nuald
added a commit
that referenced
this issue
Dec 3, 2020
* Re-run bf mandel tests (addressing #258). * Maintenance update (#261) * Re-run bf mandel tests (addressing #258). * Preparation for Perl 7. * Lua module conflicts have been resolved. * Update README.md Fixed UPDATE date. * Maintenance update (#265) * Added workaround for Intel JCC bug (#263). * Maintenance update. * Added JCC bugfix flag to clang. (#266) * Vala numbers have been updated. (#268) * Update test_dawjsonlink.cpp (#271) Updated code to reflect API changes in v2 of DAW JSON Link * Maintenance update with the renamed tests (#269). (#270) * Switch the simdjson code to On Demand (#272) * Updating simdjson to 0.6 API (on demand). * Making things safer by specifying the commit + removing unnecessary std:: qualifiers. * Minor changes following review * Maintenance update and simdjson test changes (#273) * Maintenance update * Maintenance update and simdjson test changes. * The JSON key ordering tests have been added. (#275) * Boost.JSON test has been added and the old test has been renamed. (#277) * Updated to use better numbers for floating point ops, changed Boost.JSON test to use doubles. (#279) * Improved precisions of the tests (#281) * Median cals have been added. * Try another formatting. * Changed memory calculations. * Printing has been moved out from JSON benchmarks. * Up-to-dated results. * TOC updated. * Typo * Added printer to Scheme langs. * Racket code has been fixed to use idiomatic struct instead of macros. * C++ and Kotlin tests have been updated. * D tests have been fixed to use proper initializations. * Go and Elixir tests have improved. * Nim microoptimization have been removed. * Julia test has been updated. * All haskell tests have been updated. * Lua tests have been improved. * OCaml tests have been improved. * SML tests have been improved. * Fsharp test has been improved. * Tcl tests have been improved. * Perl tests have been updated to use v5.32. * Other tests have been fixed. * Haskell networking has been fixed. * Fixed JS code. Co-authored-by: Darrell Wright <beached@users.noreply.github.com> Co-authored-by: Daniel Lemire <lemire@gmail.com>
nuald
added a commit
to nuald/benchmarks
that referenced
this issue
Dec 4, 2020
* Re-run bf mandel tests (addressing kostya#258). * Maintenance update (kostya#261) * Re-run bf mandel tests (addressing kostya#258). * Preparation for Perl 7. * Lua module conflicts have been resolved. * Update README.md Fixed UPDATE date. * Maintenance update (kostya#265) * Added workaround for Intel JCC bug (kostya#263). * Maintenance update. * Added JCC bugfix flag to clang. (kostya#266) * Vala numbers have been updated. (kostya#268) * Update test_dawjsonlink.cpp (kostya#271) Updated code to reflect API changes in v2 of DAW JSON Link * Maintenance update with the renamed tests (kostya#269). (kostya#270) * Switch the simdjson code to On Demand (kostya#272) * Updating simdjson to 0.6 API (on demand). * Making things safer by specifying the commit + removing unnecessary std:: qualifiers. * Minor changes following review * Maintenance update and simdjson test changes (kostya#273) * Maintenance update * Maintenance update and simdjson test changes. * The JSON key ordering tests have been added. (kostya#275) * Boost.JSON test has been added and the old test has been renamed. (kostya#277) * Updated to use better numbers for floating point ops, changed Boost.JSON test to use doubles. (kostya#279) * Improved precisions of the tests (kostya#281) * Median cals have been added. * Try another formatting. * Changed memory calculations. * Printing has been moved out from JSON benchmarks. * Up-to-dated results. * TOC updated. * Typo * Added printer to Scheme langs. * Racket code has been fixed to use idiomatic struct instead of macros. * C++ and Kotlin tests have been updated. * D tests have been fixed to use proper initializations. * Go and Elixir tests have improved. * Nim microoptimization have been removed. * Julia test has been updated. * All haskell tests have been updated. * Lua tests have been improved. * OCaml tests have been improved. * SML tests have been improved. * Fsharp test has been improved. * Tcl tests have been improved. * Perl tests have been updated to use v5.32. * Other tests have been fixed. * Haskell networking has been fixed. * Fixed JS code. Co-authored-by: Darrell Wright <beached@users.noreply.github.com> Co-authored-by: Daniel Lemire <lemire@gmail.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
gcc supports a work around for the Intel JCC bug, but it is opt in via
-Wa,-mbranches-within-32B-boundariescommand line option. There is a performance regression in most modern Intel chips when conditionals overlap or end on 32bit boundaries.Background: https://www.intel.com/content/dam/support/us/en/documents/processors/mitigations-jump-conditional-code-erratum.pdf
The text was updated successfully, but these errors were encountered: