On Fri, 23 Jan 2026 23:04:43 GMT, Mohamed Issa <[email protected]> wrote:
>> Intel® AVX10 ISA [1] extensions added new floating point comparison >> instructions. They set the EFLAGS register so that relationships can be >> tested independently to avoid extra checks when one of the inputs is NaN. >> >> Most of the work is covered in the architecture definition (`x86.ad`) file. >> A new comparison operand was created to be used by new CMove and JMP >> definitions with the APX specific portions of the CMove section being >> updated to rely on the new instructions because both sets of instructions >> are always expected to be available on the same platform. New floating point >> comparison definitions were also added. >> >> This change uses the new AVX10.2 (UCOMXSS or UCOMXSD) instructions on >> supported platforms to avoid the extra handling required with existing >> (UCOMISS or UCOMISD) instructions. To make sure no new failures were >> introduced, tier1, tier2, and tier3 tests were run on builds with and >> without the changes. Additionally, the JTREG tests listed below were used to >> verify correctness with `-XX:-UseAPX` / `-XX:+UseAPX` options. The baseline >> build used is [OpenJDK >> v26-b26](https://github.com/openjdk/jdk/releases/tag/jdk-26%2B26). >> >> 1. `jtreg:test/hotspot/jtreg/compiler/c2/irTests/CMoveLConstants.java` >> 2. `jtreg:test/hotspot/jtreg/compiler/c2/irTests/TestFPComparison.java` >> 3. >> `jtreg:test/hotspot/jtreg/compiler/intrinsics/math/TestSignumIntrinsic.java` >> 4. `jtreg:test/hotspot/jtreg/compiler/vectorization/TestSignumVector.java` >> >> Finally, the JMH micro-benchmark listed below was updated to separately >> exercise CMove and JMP code paths. >> >> 1. `micro:test/micro/org/openjdk/bench/java/lang/FPComparison.java` >> >> [1] >> https://www.intel.com/content/www/us/en/content-details/856721/intel-advanced-vector-extensions-10-2-intel-avx10-2-architecture-specification.html?wapkw=AVX10 > > Mohamed Issa has updated the pull request with a new target base due to a > merge or a rebase. The pull request now contains 25 commits: > > - Merge branch 'master' into user/missa-prime/avx10_2 > - Merge branch 'master' into user/missa-prime/avx10_2 > - Change function names and extend IR encoding rules for CMove tests > - Merge remote-tracking branch 'origin/master' into user/missa-prime/avx10_2 > - Remove unnecessary CMOV blocks and adjust predicates involving APX and > AVX10.2 > - Remove extra jump instruction in signum floating point and unify three-way > floating point comparison logic in x86.ad > - Also update copyright year in IREncodingPrinter.java > - Include apx_f in list of verified CPU features for IR encoding > - Fix CMove IR tests to account for APX presence > - Merge branch 'master' into user/missa-prime/avx10_2 > - ... and 15 more: https://git.openjdk.org/jdk/compare/44b74e16...09d1e44d test/hotspot/jtreg/compiler/c2/irTests/CMoveLConstants.java line 65: > 63: applyIfPlatform = {"x64", "true"}, > 64: applyIfCPUFeatureAnd = {"apx_f", "true", "avx10_2", "true"}, > 65: phase = CompilePhase.FINAL_CODE) These are all `CMove` cases. I see you added branch/call cases in the benchmark. Would it make sense to have some similar cases here? And: this here is a test for `float`. Where do we cover `double`, which you are also making VM changes for? ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/28337#discussion_r2730715729
