summaryrefslogtreecommitdiff
path: root/compiler/optimizing/intrinsics_riscv64.cc
AgeCommit message (Collapse)Author
2025-03-05riscv64: handle invoke-virtual and invoke-direct in invokeExact Anton Romanov
Test: art/test/testrunner/testrunner.py --target --64 Change-Id: I65b8a2e9f9cc361c044476a322218b6ec85647d3
2025-02-28Remove `HInstruction::GetAllocator()`. Vladimir Marko
And clean up some uses of instruction->GetBlock()->GetGraph()->GetAllocator() Test: m test-art-host-gtest Test: testrunner.py --host --optimizing Change-Id: Iae218056495a0b9cf94d2a3c1cebd6c8eb909096
2025-01-28Remove unused dex_pc from InvokeRuntime Santiago Aboy Solanes
Bug: 392802982 Test: art/test/testrunner/testrunner.py --host --64 -b --optimizing Test: m test-art-host-gtest Change-Id: I6e40215a5b1b18223c5f17e9e0ac70e05515fa94
2025-01-28Remove explicit dex_pc from RecordPcInfo Santiago Aboy Solanes
Special cases considered: * Frame entry (hardcoded to be 0) or block entry. * Native debuggable + slow paths, which is the only case where we use the instruction's dex_pc. Test: m test-art-host-gtest Test: art/test/testrunner/testrunner.py --host --64 -b --optimizing Change-Id: Ic5e0a6b5106395b891a9a45ea48da39dfb44a0a5
2025-01-24riscv64: handle invoke-static in invokeExact intrinsic. Anton Romanov
Test: art/test/testrunner/testrunner.py --target --64 Change-Id: I987b29b7d78e2913eb728003fbc85bdfe1fbceff
2025-01-02Rename `GetBootImageVarHandleField()`, drop "Boot". Vladimir Marko
This is a minor cleanup after https://android-review.googlesource.com/3048514 . Test: m test-art-host-gtest Test: testrunner.py --host --optimizing --speed-profile Bug: 38313278 Change-Id: Ic47063231fd48656b612ede3ec100ceb8a379050
2025-01-02Rename Unsafe.putOrdered* intrinsic identifiers. Vladimir Marko
Test: m test-art-host-gtest Test: testrunner.py --host --optimizing Change-Id: I1658a61953aaf109d68a2df7a534c3972679c291
2024-11-13riscv: Remove bad DCHECK Santiago Aboy Solanes
update_old_value_slow_path_ is different than nullptr and that DCHECK should always fail. Test: LUCI run https://ci.chromium.org/b/8731455182373771345 and note no `update_old_value_slow_path_ == nullptr` failure. Change-Id: Ia9dd93fae84bb9758d8b80f297f3819ca4003c17
2024-11-12Add missing Location::kNoOutputOverlap Santiago Aboy Solanes
This can save some ParallelMove instructions. For x86(_64) all FPToFP intrinsics can add it. For RISC-V, MathSqrt can add it but the ones that call GenDoubleRound can't. Also, we can add it for MathMultiplyHigh. Test: art/test/testrunner/testrunner.py --host --64 -b --optimizing Test: LUCI run https://ci.chromium.org/b/8731845964396026257 Change-Id: I28e13caf84cd850566538efbd285c0264ce80a1a
2024-10-21Remove unused variables. Christopher Ferris
Test: Builds with -Wunused-variable Change-Id: Ie308b9f1d7a092c67ab2d86d3f877fe5b59fdac4
2024-10-17Revert^2 "Add intrinsics for the absolute forms of unsafe.{get,put}Int" Shai Barack
This reverts commit c1bd4376b18ff46e5156cc597bd3311dd86fabd9. Reason for revert: fixed original root cause for breakage Change-Id: Ic9c214ebd691abf7f163b6399e8a94fc6f277ea0
2024-10-11Revert "Add intrinsics for the absolute forms of unsafe.{get,put}Int" Santiago Aboy Solanes
This reverts commit bcb5c19e5e200607fe76294aeb5273ddac5f04ae. Bug: 370098695 Reason for revert: Breaks LUCI builds for arm https://ci.chromium.org/ui/p/art/builders/ci/angler-armv7-non-gen-cc/4506/overview Change-Id: Ia72938c65c45db0ccd14f8901e715a9ec2930087
2024-10-11Add intrinsics for the absolute forms of unsafe.{get,put}Int Shai Barack
Praise the Absolute! 🦑 Bug: 370098695 Change-Id: Icd0435c33a7a3705674eb2e3b027056a6c1298ba Tested: UnsafeTest Tested: manually inspected output disassembly
2024-10-07Address comments from aosp/3282234 Santiago Aboy Solanes
Follow up to aosp/3282234 regarding comments after submission. Bug: 362916226 Test: art/test/testrunner/testrunner.py --host --64 --optimizing -b Change-Id: Id572dd42e5aa877f4ae4a20cd43ad7a778e92815
2024-09-25Revert^2 "Add VarHandle implementations for void getAndUpdate methods" Santiago Aboy Solanes
This reverts commit 8567fcdf972ade85785b689a15e3593a7d3bff9d. Reason for revert: RISC-V fix Test: LUCI run for RISC-V Change-Id: I6392bfa045e7737b5c45d78da4f7ecb8fbd3b89a
2024-09-25Revert "Add VarHandle implementations for void getAndUpdate methods" Ulya Trofimovich
This reverts commit 9ff2b617341bfe574e2d8706553f0cd65a1a2fc8. Reason for revert: failed ART test 712-varhandle-invocations on riscv64 `F dex2oat64: intrinsics_riscv64.cc:4821 Check failed: locations->GetTempCount() == 2u (locations->GetTempCount()=3, 2u=2)` Change-Id: I68a3830ccd730afa66e9cff277fe2a06829ddabe
2024-09-24Add VarHandle implementations for void getAndUpdate methods Santiago Aboy Solanes
We were missing to intrinsify getAndUpdate methods whose value was unused. By adding this support, they get a considerable runtime improvement. Bug: 362916226 Test: art/test/testrunner/testrunner.py --host --64 --optimizing -b Test: art/test/testrunner/testrunner.py --host --32 --optimizing -b Test: art/test/testrunner/testrunner.py --target --64 --optimizing Test: art/test/testrunner/testrunner.py --target --32 --optimizing Change-Id: I63d5a897ac9ea3c133167ac186afd8595f75bb13
2024-08-28cleanup: Use AddRegisterTemps where appropriate Santiago Aboy Solanes
Test: art/test/testrunner/testrunner.py --host --64 --optimizing -b Change-Id: I2fc3cd539ee1f64d1757c3b2a0a532ff6e7366f7
2024-08-12riscv64: implement signum{float|double} and copySign{float|double} intrinsics Olga Mikhaltsova
Performance improvement: copySign(double) -29% copySign(float) -33% signum(double) -20% signum(float) -22% Test: testrunner.py --target --64 --ndebug --optimizing Change-Id: I3b69a4d35a0b37e5debed9f16c3a757a4204387e
2024-06-20riscv64: SystemArrayCopyChar/Int/Byte intrinsic Aleksandr Soldatov
Implemented SystemArrayCopy for primitive types (Char, Int, Byte). Used unrolling by 16 bytes for optimization. Also add additional tests for ArrayCopy intrinsic Test: testrunner.py --target --no-prebuild --jit-on-first-use --64 Change-Id: I60f59b8b6eb164c9daf0a61c669de1606712ad09
2024-06-12Implement {Integer|Long}.remainderUnsigned intrinsic Roman Artemev
Implement intrinsic for RISC-V target Integer.remainderUnsigned: +13.71% Long.remainderUnsigned: +1623.19% Test: testrunner.py --target --64 --ndebug --optimizing Change-Id: I4c3c47bad8fcbdde723e68d19aad45be0d4ff8bb
2024-05-15Fix bugprone-macro-parentheses clang-tidy issues Stefano Cianciulli
Bug: 264654008 Test: m tidy-art Change-Id: I6a96bee1510449760634dbef5c1fac21a7e1c80e
2024-05-15Fix bugprone-argument-comment clang-tidy issues Stefano Cianciulli
Bug: 264654008 Test: m tidy-art Change-Id: I7cbe86ab1cc9de2ca2e80583203e8d9e7abcf44b
2024-05-06riscv64: StringGetCharsNoCheck intrinsic. Mark
Implemented StringGetCharsNoCheck intrinsic. Added new checks in the 536-checker-intrinsic-optimization test in order to increase the quality of test coverage for the case of uncompressed strings. Test: testrunner.py --target --64 --ndebug --no-image --jit-on-first-use Change-Id: I08bc075440eef1750a057bad5fe6de45f792091f
2024-04-22Optimizing: Treat app image objects as non-movable. Vladimir Marko
Treat app image objects similar to boot image objects and avoid unnecessary read barriers for app image `HLoadClass` and `HInstanceOf` checks with app image `HLoadClass` input. Extend other optimizations to treat app image classes the same way as boot image classes even though this remains mostly dormant because we currently do not initialize app image classes with class initializers; the experimental flag `--initialize-app-image-classes` is false by default. Test: m test-art-host-gtest Test: testrunner.py --host --optimizing --speed-profile Bug: 38313278 Change-Id: I359dd8897f6d128213602f5731d40edace298ab8
2024-03-28riscv64: Implement StringCompareTo instrinsic denis.tomashev
Test: art/test.py --target -r --64 --optimizing --ndebug Change-Id: I9b68cb2994f1feda0f0cffb8a5d38e44f5cd807e
2024-03-19riscv64: Implemented StringEquals intrinsic denis.tomashev
Bug: 283082089 Test: art/test/testrunner/testrunner.py --target --64 --ndebug --no-image --jit-on-first-use Change-Id: Ifb4076d7f12cbad174049995980f0f08db41357a Signed-off-by: denis.tomashev <denis.tomashev@syntacore.com>
2024-03-12riscv64: Fix `VarHandle.getAndAdd(..., 0)`. Vladimir Marko
Test: testrunner.py --target --64 --optimizing -t 712 Bug: 328420806 Change-Id: I6c12e66e5d0602b23af157016ba423aae7425031
2024-02-29riscv64: Add extension restrictions to assembler. Vladimir Marko
Use them to select different implementations of the `Zext*` and `Sext*` macros and ensure that LR/SC sequences contain only permitted instructions. Test: m test-art-host-gtest Test: run-gtests.sh Test: testrunner.py --target --optimizing Bug: 283082089 Change-Id: I3c306eecc123e5f10a0181942f1cd3dad7ec9a4c
2024-01-29Revert^7 "Disable write-barrier elimination pass" Santiago Aboy Solanes
This reverts commit 1ba3516e8c3e2b86c73084893dd297f468469181. Reason for revert: PS1 is reland as-is PS2 has two related fixes (for RISC-V and arm64) taking into account that when we store zero, we use a special register. Bug: 301833859 Bug: 310755375 Bug: 260843353 Test: lunch cf_riscv64_wear-trunk_staging-userdebug && m Test: lunch starnix_wear_yukawa-trunk_staging-userdebug && m Change-Id: I5e69890fd56404ddde56ebc457179241363d9243
2024-01-24Revert^6 "Disable write-barrier elimination pass" Matthew Reynolds
This reverts commit 1be176f5a78750e2f0e32470f8c83e3d1643954d. Reason for revert: Potential cause of build breakage for cf_riscv64_wear-trunk_staging-userdebug build 11353124 Change-Id: I5db1c9fba1edd4ab1eef30e2b547bb9649af5c10
2024-01-24Revert^5 "Disable write-barrier elimination pass" Santiago Aboy Solanes
This reverts commit 31b949bc4a76e5c6d00a8e18c346f123b5321a1c. Reason for revert: PS1 is reland as-is PS2 has two fixes: * Missed poisoning heap references in a code path * Removed incorrect DCHECK Change-Id: I81b317ddc704dbd8a173f5d5c624dbc69e2d9e60 Test: art/test/testrunner/testrunner.py --host --64 --optimizing -b Test: art/test/testrunner/testrunner.py --target --64 --optimizing -b Both commnands with `export ART_HEAP_POISONING=true` Bug: 301833859 Bug: 310755375 Bug: 260843353
2024-01-22Revert^4 "Disable write-barrier elimination pass" Santiago Aboy Solanes
This reverts commit b5b98b9bb31acb2deffb692c50d0fbc71476663b. Reason for revert: Breaks tests in arm64 + heap poison configurations Bug: 310755375 Bug: 260843353 Change-Id: I682c74987a365497e0dbe47eba26a9ccf0513561
2024-01-22Revert^3 "Disable write-barrier elimination pass" Santiago Aboy Solanes
This reverts commit 9f8df195b7ff2ce47eec4e9b193ff3214ebed19c. Reason for revert: Fix for x86_64 with heap poison enabled This case uses a temp with index `1` in the regular FieldSet case. This is done like this due to GenerateVarHandleSet also calling HandleFieldSet. The bug was that we were allocating only one temp in the regular FieldSet case and therefore not having the temp with index `1` available. PS1 is the revert as-is. PS2 contains the fix. Test: art/test/testrunner/testrunner.py --host --64 --optimizing -b Test: Same command with heap poison enabled too Bug: 301833859 Bug: 310755375 Bug: 260843353 Change-Id: Ie2740b4c443158c4e72810ce1d8268353c5f0055
2024-01-19Revert^2 "Disable write-barrier elimination pass" Santiago Aboy Solanes
This reverts commit 7c1dd6e2d1893f288214413c4b97273980f3aa4a. Reason for revert: build breakages, using a different number of temps vs the expected (crashing in https://cs.android.com/android/platform/superproject/main/+/main:art/compiler/optimizing/code_generator_x86_64.cc;l=5488;drc=7c1dd6e2d1893f288214413c4b97273980f3aa4a) Change-Id: I843c039394dd666776ea5bcb5b10b1f47df12d53
2024-01-19Revert "Disable write-barrier elimination pass" Santiago Aboy Solanes
This reverts commit 5a3271d7caafefd10a20f5a5db09d2c178838b76. Reason for revert: This CL has two fixes (codegen not doing a null check if a write barrier is being relied on, and codegen not recomputing skipping write barriers), regression tests, a new runtime check which runs in debug mode for the CC GC to ensure that the card table is set correctly for skipped write barriers, and new compile time (graph checker) tests to ensure graph consistency. This patchset updates the WriteBarrierKind to be {emit being relied on, emit not being relied on, dont emit}, which leaves the null check skip implementation to codegen. Test 2247- is removed from knownfailures.json but still skipped in MTS due to SLO requirements. Test: art/test/testrunner/testrunner.py --host --64 --optimizing -b Bug: 301833859 Bug: 310755375 Bug: 260843353 Change-Id: I025597e284b2765986e2091538680ee629fb5ae7
2024-01-16riscv64: Implement `StringNewStringFrom*` intrinsics. Vladimir Marko
Test: testrunner.py --target --64 --ndebug --optimizing Bug: 283082089 Change-Id: Iba6321ed2b5aa2249eae3ec5b47d5c5364542748
2024-01-15riscv64: Skip WB for zero-length SystemArrayCopy. Vladimir Marko
Skip the unnecessary write barrier for zero-length arraycopy as we do on x86/x86-64 since https://android-review.git.corp.google.com/2893223 . Test: testrunner.py --target --64 --ndebug --optimizing Bug: 283082089 Change-Id: Ie30dc9e6ee66bee884c6f61141b14b7e02b404e3
2024-01-12riscv64: Implement Thread.Interrupted() instrinsic Lifang Xia
Test: testrunner.py --target --64 --ndebug --optimizing --jit -t 050-sync-test -t 1941-dispose-stress -t 988-method-trace Bug: 283082089 Change-Id: I113c90e399dd603c152241e0b89c7f19a5105c43
2024-01-05riscv64: Fix heap poisoning. Vladimir Marko
Fix heap poisoning for stubs, nterp, codegen and some intrinsics and disable a few intrinsics broken for the heap poisoning configuration. Note that this fixes heap poisoning only for the CC GC. The heap poisoning is pobably still broken for CMC GC. Test: Build with `ART_HEAP_POISONING=true`, then testrunner.py --target --64 --ndebug --optimizing --interpreter Bug: 283082089 Change-Id: I386235f6379722e1b2355150ab7d14f2ffb0da85
2023-12-13riscv64: Clean up the `SystemArrayCopy` intrinsic. Vladimir Marko
Define a new optimization flag for source and destination position match. Use it to avoid the forward-copy check (where the assembler optimized away a BLT instruction, so we had just a useless BNE to the next instruction) and one position sign check. Avoid checking that the position is inside the array. The subsequent subtraction cannot underflow an `int32_t` and the following BLT shall go to the slow path for negative values anyway. Rewrite the array type check to avoid unnecessary checks and read barriers. Use an allocated temporary instead of scratch register for the marking in the read barrier slow path. Simplify the gray bit check and the fake dependency. Use constant position and length locations for small constant values. (It was probably an oversight that we used it only for large constant values.) Emit threshold check when the length equals source or destination length. The old code allowed the intrinsic to process array copy of an arbirary length. Use `ShNAdd()` for faster array address calculations. Use helper functions and lambdas to simplify the code. Pass registers and locations by value. Prefer load/store macro instructions over raw load/store instructions. Use a bare conditional branch to assert the `TMP` shall not be clobbered. Test: testrunner.py --target --64 --ndebug --optimizing Bug: 283082089 Change-Id: I3f697b4a74497d6d712a92450a6a45e772430662
2023-12-12riscv64: Implement SystemArrayCopy intrinsic Aleksandr Soldatov
Test: testrunner.py --target --ndebug --optimizing --jit --64 Test: testrunner.py --target --no-prebuild --jit-on-first-use --64 Bug: 283082089 Signed-off-by: Aleksandr Soldatov <aleksandr.soldatov@syntacore.com> Change-Id: I8f9aa02299c08922b4edfe25d6ded368b41d1927
2023-12-11riscv64: Implement `Reference` intrinsics. Vladimir Marko
Test: testrunner.py --target --64 --ndebug --optimizing Bug: 283082089 Change-Id: I81c730d71169f0302c1ce32ce4dac50b0e3850cd
2023-12-06riscv64: Implement `Integer/Long.reverse()` intrinsics. Vladimir Marko
Also move `divideUnsigned()` intrinsic implementations together with other integral intrinsic implementations. Test: testrunner.py --target --64 --ndebug --optimizing Bug: 283082089 Change-Id: I57786789ce6114ee951f46b7614ce965bdce9617
2023-12-04riscv64: Implement Unsafe.getAndAdd/-Set intrinsics. Vladimir Marko
Test: testrunner.py --target --64 --ndebug --optimizing Bug: 283082089 Change-Id: I51bb208f2d5e43f4eab60136491bbd62ec52c812
2023-12-01riscv64: Implement Unsafe CAS intrinsics. Vladimir Marko
Test: testrunner.py --target --64 --ndebug --optimizing Bug: 283082089 Change-Id: I63a30dbf54e5eba0da08e0e99daef660e7660db7
2023-12-01riscv64: Fix VarHandle GetAndUpdate intrinsic. Vladimir Marko
Fix the byte-swap CAS case for narrow types (short, char). Note that these cases are not currently supported by libcore and we therefore take the slow path during the access mode check. However, the ART VarHandle implementation is ready to handle them in case this arbitrary libcore (or upstream?) decision changes and it generates code for them. This fix avoids a `DCHECK` failure in `GenerateByteSwapAndExtract()` by doing the `DCHECK` in two of its three callers and generates the correct (and smaller) code in all cases. Extend the run-test 712 to test offsets that would lead to non-zero shift in `GenerateByteSwapAndExtract()`. To test this code path, one needs to modify the libcore method `VarHandle.unalignedAccessModesBitMask()` to permit the `NUMERIC_ATOMIC_UPDATE_ACCESS_MODES_BIT_MASK` for `short` (and maybe `char` but that makes less sense than `short`). Test: testrunner.py --target --64 --ndebug --optimizing Bug: 283082089 Change-Id: I23926bbafc9eff21e094b00f36357d629cc550a9
2023-11-30riscv64: Fix VarHandle.compareAndSet intrinsics. Vladimir Marko
And enable a test that exposes the bug even for `--ndebug`. Test: testrunner.py --target --64 --ndebug --optimizing Bug: 283082089 Bug: 224733324 Change-Id: I34ba41cdf905c2da703f77739758450d8d305bf1
2023-11-30riscv64: Fix reference load in VarHandle checks. Vladimir Marko
Test: builbot-build.sh --target Bug: 283082089 Change-Id: I99c57cb1305c1f2bc5cd8310c104006febad1d3b
2023-11-29riscv64: Pass codegen through to GenerateReverseBytes Samuel Holland
This allows GenerateReverseBytes to check for instruction set features. Test: Boot AOSP on HiFive Unmatched Change-Id: I7bb880a24358d5712cd1ebd3a915e98eef1f40f5
generated by cgit v1.2.3-59-g8ed1b (git 2.39.0) at 2025-10-19 23:17:49 +0000