On Tue, 12 Aug 2025 04:33:04 GMT, Galder Zamarreño <gal...@openjdk.org> wrote:
>> I've done some testing on x86_64 and aarch64 and the tests pass. >> >> I also made sure that the test output demonstrated execution of the expected >> IR rule as per the requirements of each platform. >> >> ## `c7gn.2xlarge` Graviton3 >> >> >> ============================== >> Test summary >> ============================== >> TEST TOTAL PASS FAIL ERROR >> SKIP >> >> jtreg:test/hotspot/jtreg/compiler/c2/irTests/ConvF2HFIdealizationTests.java >> 1 1 0 0 >> 0 >> >> jtreg:test/hotspot/jtreg/compiler/c2/irTests/TestFloat16ScalarOperations.java >> 1 1 0 0 >> 0 >> >> jtreg:test/hotspot/jtreg/compiler/loopopts/superword/TestCompatibleUseDefTypeSize.java >> 1 1 0 0 >> 0 >> ============================== >> TEST SUCCESS >> >> $ tail ConvF2HFIdealizationTests.jtr >> Messages from Test VM >> --------------------- >> [IREncodingPrinter] Disabling IR matching for rule 1 of 2 in test1: None of >> the feature constraints met (applyIfCPUFeatureOr): avx512_fp16, true, zfh, >> true >> >> >> ----------System.err:(3/35)---------- >> >> JavaTest Message: Test complete. >> >> result: Passed. Execution successful >> >> >> test result: Passed. Execution successful >> >> $ tail TestFloat16ScalarOperations.jtr >> Messages from Test VM >> --------------------- >> [IREncodingPrinter] Disabling IR matching for rule 1 of 2 in testDivByPOT: >> None of the feature constraints met (applyIfCPUFeatureOr): avx512_fp16, >> true, zfh, true >> [IREncodingPrinter] Disabling IR matching for rule 1 of 2 in testMulByTWO: >> None of the feature constraints met (applyIfCPUFeatureOr): avx512_fp16, >> true, zfh, true >> [IREncodingPrinter] Disabling IR matching for rule 1 of 2 in >> testInexactFP16ConstantPatterns: None of the feature constraints met >> (applyIfCPUFeatureOr): avx512_fp16, true, zfh, true >> [IREncodingPrinter] Disabling IR matching for rule 1 of 2 in >> testSNaNFP16ConstantPatterns: None of the feature constraints met >> (applyIfCPUFeatureOr): avx512_fp16, true, zfh, true >> [IREncodingPrinter] Disabling IR matching for rule 1 of 2 in >> testQNaNFP16ConstantPatterns: None of the feature constraints met >> (applyIfCPUFeatureOr): avx512_fp16, true, zfh, true >> [IREncodingPrinter] Disabling IR matching for rule 1 of 2 in >> testExactFP16ConstantPatterns: None of the feature constraints met >> (applyIfCPUFeat... > > Btw, I've noticed that `TestFloat16ScalarOperations` does not have `package` > definition. Is that an oversight? It runs fine in spite of not having it Hi, as you mostly touched the auto-vectorization part of c2, could you please run these float16 tests as well (most of these enable auto-vectorization for Float16) - `compiler/vectorization/TestFloat16VectorOperations.java` `compiler/vectorization/TestFloatConversionsVectorNaN.java` `compiler/vectorization/TestFloatConversionsVector.java` `compiler/vectorization/TestFloat16ToFloatConv.java` `compiler/vectorization/TestFloat16VectorConvChain.java` `compiler/intrinsics/float16/*` ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/26457#discussion_r2269084605
