linaro-toolchain

linaro-toolchain@lists.linaro.org

16 participants
5599 discussions

by Peter Maydell

Progress: * UM-2 [QEMU upstream maintainership] - I'm now back on merging duty for the 8.0 release cycle, so some time spent on pull request processing - Sent pull requests with accumulated arm and reset-refactoring patches from the freeze period - Trying to cut down my code review backlog before the holidays -- PMM

2 years, 6 months

[ACTIVITY] Report for week #50

by Thiago Jung Bauermann

# [GNU-767] Support changing SVE vector length in remote debugging - Patches to gdbserver to support changing the SVE vector length: About halfway through implementing the new approach of sending new XML target descriptions through the wire. # Misc - Experimented with using a GDB wrapper to run the testsuite. Came up with a small patch that fixes the tests that fail when using the wrapper. -- Thiago

2 years, 6 months

[TCWG CI] Failure after basepoints/gcc-13-4618-g17ae956c0fa: AArch64: Support new tbranch optab.

by ci_notify＠linaro.org

Failure after basepoints/gcc-13-4618-g17ae956c0fa: AArch64: Support new tbranch optab.: Results changed to -10 # build_abe binutils: -9 # build_abe stage1: -5 # build_abe qemu: -2 # linux_n_obj: 7572 # First few build errors in logs: # 00:10:43 drivers/gpu/drm/v3d/v3d_perfmon.c:57:1: internal compiler error: in decompose, at rtl.h:2288 # 00:10:44 make[5]: *** [scripts/Makefile.build:250: drivers/gpu/drm/v3d/v3d_perfmon.o] Error 1 # 00:10:53 make[4]: *** [scripts/Makefile.build:502: drivers/gpu/drm/v3d] Error 2 # 00:13:52 drivers/media/mc/mc-device.c:198:1: internal compiler error: in decompose, at rtl.h:2288 # 00:13:53 make[4]: *** [scripts/Makefile.build:250: drivers/media/mc/mc-device.o] Error 1 # 00:13:57 make[3]: *** [scripts/Makefile.build:502: drivers/media/mc] Error 2 # 00:15:27 make[2]: *** [scripts/Makefile.build:502: drivers/media] Error 2 # 00:17:50 make[3]: *** [scripts/Makefile.build:502: drivers/gpu/drm] Error 2 # 00:17:50 make[2]: *** [scripts/Makefile.build:502: drivers/gpu] Error 2 # 00:17:50 make[1]: *** [scripts/Makefile.build:502: drivers] Error 2 from -10 # build_abe binutils: -9 # build_abe stage1: -5 # build_abe qemu: -2 # linux_n_obj: 8625 # linux build successful: all # linux boot successful: boot THIS IS THE END OF INTERESTING STUFF. BELOW ARE LINKS TO BUILDS, REPRODUCTION INSTRUCTIONS, AND THE RAW COMMIT. For latest status see comments in https://linaro.atlassian.net/browse/GNU-680 . Status of basepoints/gcc-13-4618-g17ae956c0fa commit for tcwg_kernel: commit 17ae956c0fa6baac3d22764019d5dd5ebf5c2b11 Author: Tamar Christina <tamar.christina(a)arm.com> Date: Mon Dec 12 15:18:56 2022 +0000 AArch64: Support new tbranch optab. This implements the new tbranch optab for AArch64. we cannot emit one big RTL for the final instruction immediately. The reason that all comparisons in the AArch64 backend expand to separate CC compares, and separate testing of the operands is for ifcvt. The separate CC compare is needed so ifcvt can produce csel, cset etc from the compares. Unlike say combine, ifcvt can not do recog on a parallel with a clobber. Should we emit the instruction directly then ifcvt will not be able to say, make a csel, because we have no patterns which handle zero_extract and compare. (unlike combine ifcvt cannot transform the extract into an AND). While you could provide various patterns for this (and I did try) you end up with broken patterns because you can't add the clobber to the CC register. If you do, ifcvt recog fails. i.e. int f1 (int x) { if (x & 1) return 1; return x; } We lose csel here. Secondly the reason the compare with an explicit CC mode is needed is so that ifcvt can transform the operation into a version that doesn't require the flags to be set. But it only does so if it know the explicit usage of the CC reg. For instance int foo (int a, int b) { return ((a & (1 << 25)) ? 5 : 4); } Doesn't require a comparison, the optimal form is: foo(int, int): ubfx x0, x0, 25, 1 add w0, w0, 4 ret and no compare is actually needed. If you represent the instruction using an ANDS instead of a zero_extract then you get close, but you end up with an ands followed by an add, which is a slower operation. gcc/ChangeLog: * config/aarch64/aarch64.md (*tb<optab><mode>1): Rename to... (*tb<optab><ALLI:mode><GPI:mode>1): ... this. (tbranch_<code><mode>4): New. * config/aarch64/iterators.md(ZEROM, zerom): New. gcc/testsuite/ChangeLog: * gcc.target/aarch64/tbz_1.c: New test. * gnu-master-aarch64-mainline-defconfig ** Failure after basepoints/gcc-13-4618-g17ae956c0fa: AArch64: Support new tbranch optab.: ** https://ci.linaro.org/job/tcwg_kernel-gnu-build-gnu-master-aarch64-mainline… Bad build: https://ci.linaro.org/job/tcwg_kernel-gnu-build-gnu-master-aarch64-mainline… Good build: https://ci.linaro.org/job/tcwg_kernel-gnu-build-gnu-master-aarch64-mainline… Reproduce current build: <cut> mkdir -p investigate-gcc-17ae956c0fa6baac3d22764019d5dd5ebf5c2b11 cd investigate-gcc-17ae956c0fa6baac3d22764019d5dd5ebf5c2b11 # Fetch scripts git clone https://git.linaro.org/toolchain/jenkins-scripts # Fetch manifests for bad and good builds mkdir -p bad/artifacts good/artifacts curl -o bad/artifacts/manifest.sh https://ci.linaro.org/job/tcwg_kernel-gnu-build-gnu-master-aarch64-mainline… --fail curl -o good/artifacts/manifest.sh https://ci.linaro.org/job/tcwg_kernel-gnu-build-gnu-master-aarch64-mainline… --fail # Reproduce bad build (cd bad; ../jenkins-scripts/tcwg_kernel-build.sh ^^ true %%rr[top_artifacts] artifacts) # Reproduce good build (cd good; ../jenkins-scripts/tcwg_kernel-build.sh ^^ true %%rr[top_artifacts] artifacts) </cut> Full commit (up to 1000 lines): <cut> commit 17ae956c0fa6baac3d22764019d5dd5ebf5c2b11 Author: Tamar Christina <tamar.christina(a)arm.com> Date: Mon Dec 12 15:18:56 2022 +0000 AArch64: Support new tbranch optab. This implements the new tbranch optab for AArch64. we cannot emit one big RTL for the final instruction immediately. The reason that all comparisons in the AArch64 backend expand to separate CC compares, and separate testing of the operands is for ifcvt. The separate CC compare is needed so ifcvt can produce csel, cset etc from the compares. Unlike say combine, ifcvt can not do recog on a parallel with a clobber. Should we emit the instruction directly then ifcvt will not be able to say, make a csel, because we have no patterns which handle zero_extract and compare. (unlike combine ifcvt cannot transform the extract into an AND). While you could provide various patterns for this (and I did try) you end up with broken patterns because you can't add the clobber to the CC register. If you do, ifcvt recog fails. i.e. int f1 (int x) { if (x & 1) return 1; return x; } We lose csel here. Secondly the reason the compare with an explicit CC mode is needed is so that ifcvt can transform the operation into a version that doesn't require the flags to be set. But it only does so if it know the explicit usage of the CC reg. For instance int foo (int a, int b) { return ((a & (1 << 25)) ? 5 : 4); } Doesn't require a comparison, the optimal form is: foo(int, int): ubfx x0, x0, 25, 1 add w0, w0, 4 ret and no compare is actually needed. If you represent the instruction using an ANDS instead of a zero_extract then you get close, but you end up with an ands followed by an add, which is a slower operation. gcc/ChangeLog: * config/aarch64/aarch64.md (*tb<optab><mode>1): Rename to... (*tb<optab><ALLI:mode><GPI:mode>1): ... this. (tbranch_<code><mode>4): New. * config/aarch64/iterators.md(ZEROM, zerom): New. gcc/testsuite/ChangeLog: * gcc.target/aarch64/tbz_1.c: New test. --- gcc/config/aarch64/aarch64.md | 33 ++++++++--- gcc/config/aarch64/iterators.md | 2 + gcc/testsuite/gcc.target/aarch64/tbz_1.c | 95 ++++++++++++++++++++++++++++++++ 3 files changed, 122 insertions(+), 8 deletions(-) diff --git a/gcc/config/aarch64/aarch64.md b/gcc/config/aarch64/aarch64.md index 896b6a8ac79..d749c98eef6 100644 --- a/gcc/config/aarch64/aarch64.md +++ b/gcc/config/aarch64/aarch64.md @@ -947,12 +947,29 @@ (const_int 1)))] ) -(define_insn "*tb<optab><mode>1" +(define_expand "tbranch_<code><mode>3" [(set (pc) (if_then_else - (EQL (zero_extract:DI (match_operand:GPI 0 "register_operand" "r") - (const_int 1) - (match_operand 1 - "aarch64_simd_shift_imm_<mode>" "n")) + (EQL (match_operand:ALLI 0 "register_operand") + (match_operand 1 "aarch64_simd_shift_imm_<mode>")) + (label_ref (match_operand 2 "")) + (pc)))] + "" +{ + rtx bitvalue = gen_reg_rtx (<ZEROM>mode); + rtx reg = gen_lowpart (<ZEROM>mode, operands[0]); + rtx val = GEN_INT (1UL << UINTVAL (operands[1])); + emit_insn (gen_and<zerom>3 (bitvalue, reg, val)); + operands[1] = const0_rtx; + operands[0] = aarch64_gen_compare_reg (<CODE>, bitvalue, + operands[1]); +}) + +(define_insn "*tb<optab><ALLI:mode><GPI:mode>1" + [(set (pc) (if_then_else + (EQL (zero_extract:GPI (match_operand:ALLI 0 "register_operand" "r") + (const_int 1) + (match_operand 1 + "aarch64_simd_shift_imm_<ALLI:mode>" "n")) (const_int 0)) (label_ref (match_operand 2 "" "")) (pc))) @@ -963,15 +980,15 @@ { if (get_attr_far_branch (insn) == 1) return aarch64_gen_far_branch (operands, 2, "Ltb", - "<inv_tb>\\t%<w>0, %1, "); + "<inv_tb>\\t%<ALLI:w>0, %1, "); else { operands[1] = GEN_INT (HOST_WIDE_INT_1U << UINTVAL (operands[1])); - return "tst\t%<w>0, %1\;<bcond>\t%l2"; + return "tst\t%<ALLI:w>0, %1\;<bcond>\t%l2"; } } else - return "<tbz>\t%<w>0, %1, %l2"; + return "<tbz>\t%<ALLI:w>0, %1, %l2"; } [(set_attr "type" "branch") (set (attr "length") diff --git a/gcc/config/aarch64/iterators.md b/gcc/config/aarch64/iterators.md index d10cf93572e..a521dbde1ec 100644 --- a/gcc/config/aarch64/iterators.md +++ b/gcc/config/aarch64/iterators.md @@ -1107,6 +1107,8 @@ ;; Give the number of bits in the mode (define_mode_attr sizen [(QI "8") (HI "16") (SI "32") (DI "64")]) +(define_mode_attr ZEROM [(QI "SI") (HI "SI") (SI "SI") (DI "DI")]) +(define_mode_attr zerom [(QI "si") (HI "si") (SI "si") (DI "di")]) ;; Give the ordinal of the MSB in the mode (define_mode_attr sizem1 [(QI "#7") (HI "#15") (SI "#31") (DI "#63") diff --git a/gcc/testsuite/gcc.target/aarch64/tbz_1.c b/gcc/testsuite/gcc.target/aarch64/tbz_1.c new file mode 100644 index 00000000000..39deb58e278 --- /dev/null +++ b/gcc/testsuite/gcc.target/aarch64/tbz_1.c @@ -0,0 +1,95 @@ +/* { dg-do compile } */ +/* { dg-additional-options "-O2 -std=c99 -fno-unwind-tables -fno-asynchronous-unwind-tables" } */ +/* { dg-final { check-function-bodies "**" "" "" { target { le } } } } */ + +#include <stdbool.h> + +void h(void); + +/* +** g1: +** tbnz w[0-9]+, #?0, .L([0-9]+) +** ret +** ... +*/ +void g1(bool x) +{ + if (__builtin_expect (x, 0)) + h (); +} + +/* +** g2: +** tbz w[0-9]+, #?0, .L([0-9]+) +** b h +** ... +*/ +void g2(bool x) +{ + if (__builtin_expect (x, 1)) + h (); +} + +/* +** g3_ge: +** tbnz w[0-9]+, #?31, .L[0-9]+ +** b h +** ... +*/ +void g3_ge(int x) +{ + if (__builtin_expect (x >= 0, 1)) + h (); +} + +/* +** g3_gt: +** cmp w[0-9]+, 0 +** ble .L[0-9]+ +** b h +** ... +*/ +void g3_gt(int x) +{ + if (__builtin_expect (x > 0, 1)) + h (); +} + +/* +** g3_lt: +** tbz w[0-9]+, #?31, .L[0-9]+ +** b h +** ... +*/ +void g3_lt(int x) +{ + if (__builtin_expect (x < 0, 1)) + h (); +} + +/* +** g3_le: +** cmp w[0-9]+, 0 +** bgt .L[0-9]+ +** b h +** ... +*/ +void g3_le(int x) +{ + if (__builtin_expect (x <= 0, 1)) + h (); +} + +/* +** g5: +** mov w[0-9]+, 65279 +** tst w[0-9]+, w[0-9]+ +** beq .L[0-9]+ +** b h +** ... +*/ +void g5(int x) +{ + if (__builtin_expect (x & 0xfeff, 1)) + h (); +} </cut>

2 years, 6 months

[ACTIVITY] Report for week #49

by Thiago Jung Bauermann

# [GNU-767] Support changing SVE vector length in remote debugging - Patches to gdbserver to support changing the SVE vector length: There was an upstream discussion about whether changing the implementation from relying on expedited registers to relying on sending target descriptions over the wire was a better approach. Simon Marchi detailed his idea on how to do that and it does seem better. - Started implementing Simon's approach of sending target descriptions over the wire for each thread. # Misc - Sent and later committed a couple of patches¹ fixing whitespace issues in a Python script that generates a GDB source file. -- Thiago ¹ https://inbox.sourceware.org/gdb-patches/20221202192200.405379-1-thiago.bau…

2 years, 6 months

[ACTIVITY] Report for week #48

by Thiago Jung Bauermann

Hello, # [GNU-767] Support changing SVE vector length in remote debugging - v2 of the gdbserver patches to support changing the SVE vector length was quickly reviewed by both Luis and Simon Marchi. I applied their review suggestions and I'm now working on fixing a bug with multi-threaded programs that they spotted. - Submitted a couple of small patches¹ fixing tab vs spaces issues in the gdbarch.py script that generates some source code in GDB. # Misc - Fixed problem in the tcwg-dev/start.sh script where asking docker to expose /dev/kvm to a dev container on a host which doesn't have KVM support causes docker to error out (reported by David Spickett). Sent Gerrit change request “42669: tcwg-dev: Add heuristic to check for KVM support on the host” to fix it. David reviewed and merged it. Thanks! -- Thiago ¹ https://inbox.sourceware.org/gdb-patches/20221202192200.405379-1-thiago.bau…

2 years, 6 months

[ACTIVITY] Report for week #47

by Thiago Jung Bauermann

Hello, # [GNU-767] Support changing SVE vector length in remote debugging - Worked on v2 of the gdbserver patches improving SVE support. Found a couple of simplifications that could be made to the code. Rebased on current master branch and finished regression testing and patch preparation. Wrote the cover letter. - Finally posted the patch series upstream¹. -- Thiago ¹ https://inbox.sourceware.org/gdb-patches/20221126020452.1686509-1-thiago.ba…

2 years, 7 months

[ACTIVITY] report week ending 25 Nov

by Peter Maydell

Progress: * UM-2 [QEMU upstream maintainership] - More conversions of devices to 3-phase reset in pursuit of the interim goal of getting rid of device_class_set_parent_reset() - Investigated a regression in a TF-A workload: this was due to recent pagetable walk refactoring breaking debug accesses. - code review, etc; mostly this has been of for-8.0 material, as (other than the above mentioned ptw regression) there haven't been many release-worthy issues this freeze cycle. -- PMM

2 years, 7 months

[TCWG CI] 403.gcc failed to run after working-3971-g5f7f484ee54e: [AArch64] Add GPR rr instructions to isAssociativeAndCommutative

by Maxim Kuvyrkov

Hi David, Our CI flagged your commit; it seems it miscompiles 403.gcc from SPEC CPU2006 at -O3 -flto for aarch64-linux-gnu. Would you please investigate? Let me know if you need any assistance in reproducing this. Thanks! === After working-3971-g5f7f484ee54e commit 5f7f484ee54ebbf702ee4c5fe9852502dc237121 Author: David Green <david.green(a)arm.com> [AArch64] Add GPR rr instructions to isAssociativeAndCommutative the following benchmarks slowed down by more than 3%: - 403.gcc failed to run Configuration: - Benchmark: SPEC CPU2006 - Toolchain: Clang + Glibc + LLVM Linker - Version: all components were built from their tip of trunk - Target: aarch64-linux-gnu - Compiler flags: -O3 -flto - Hardware: NVidia TX1 4x Cortex-A57 -- Maxim Kuvyrkov https://www.linaro.org

2 years, 7 months

[ACTIVITY] Report for week #46

by Thiago Jung Bauermann

Hello, # [GNU-767] Support changing SVE vector length in remote debugging - Working on v2 of the gdbserver patches improving SVE support. Found and fixed a memory leak in my code while preparing patches for upstream submission. Continuing preparing the patches. # [GNU-796] Stabilise GDB testsuite results in the CI - Analysed failures of “fast_check_gdb” CI jobs over the weekend. They were caused by unrelated build issues in GCC and libstdc++. -- Thiago

2 years, 7 months

[ACTIVITY] week ending Nov. 20 2022

by Alex Bennée

Project Stratos =============== - finished [blog post for virtio-camera] - now live [on web site] - did a review of [virtio-loopback] - started reviewing [PATCH v9 0/8] KVM: mm: fd-based approach for supporting KVM Message-Id: <20221025151344.3784230-1-chao.p.peng(a)linux.intel.com> - trying to assess if user-space facing solution for memory sharing [blog post for virtio-camera] <https://linaro.atlassian.net/jira/core/projects/LBO/board?selectedIssue=LBO…> [on web site] <https://www.linaro.org/blog/the-challenges-of-abstracting-virtio/> [virtio-loopback] <https://git.virtualopensystems.com/virtio-loopback> vhost-device maintainer effort ([UM-196]) - debugged regression in virtio-vsock and QEMU - should have some error message patches to post [UM-196] <https://linaro.atlassian.net/browse/UM-196> Single Binary ([QEMU-487]) ========================== - posted [PATCH for 8.0 v5 00/20] use MemTxAttrs to avoid current_cpu in hw/ Message-Id: <20221111182535.64844-1-alex.bennee(a)linaro.org> [QEMU-487] <https://linaro.atlassian.net/browse/QEMU-487> QEMU Upstream Work ([UM-2]) =========================== - posted [PATCH for 7.2-rc1 v2 00/12] testing, docs, plugins, arm pre-PR Message-Id: <20221111145529.4020801-1-alex.bennee(a)linaro.org> - posted [PULL for 7.2 00/10] testing and doc updates Message-Id: <20221115133439.2348929-1-alex.bennee(a)linaro.org> - did a bunch of follow-up testing to shake out other console bugs - posted [PATCH for 7.2-rc1 v2 00/12] testing, docs, plugins, arm pre-PR Message-Id: <20221111145529.4020801-1-alex.bennee(a)linaro.org> - posted [PATCH for 7.2? v1 0/2] Arm GICv2 patches Message-Id: <20221115134048.2352715-1-alex.bennee(a)linaro.org> [UM-2] <https://linaro.atlassian.net/browse/UM-2> Other ===== - hackbox2 upgrade follow-up - health check content Completed Reviews [1/1] ======================= [PATCH] ci: replace x86_64 macos-11 with aarch64 macos-12 Message-Id: <20221116175023.80627-1-berrange(a)redhat.com> Absences ======== Current Review Queue ==================== TODO [PATCH v6 00/21] Drivers for gunyah hypervisor Message-Id: <20221026185846.3983888-1-quic_eberman(a)quicinc.com> ================================================================================================================== TODO [PATCH v2 0/1] tcg: add perfmap and jitdump Message-Id: <20221114161321.3364875-1-iii(a)linux.ibm.com> ======================================================================================================== TODO [PATCH for-8.0 v3 00/45] tcg: Support for Int128 with helpers Message-Id: <20221111074101.2069454-1-richard.henderson(a)linaro.org> ===================================================================================================================================== -- Alex Bennée

2 years, 7 months

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

linaro-toolchain