RISC-V Daily Intelligence Report (2026-11-11)
Research window: 2026-11-10 05:00 ~ 2026-11-11 05:00 (Beijing time, 24 hours) Sources: Google News aggregation (24 query sets in Chinese and English, via proxy), Hacker News, Phoronix, CNX Software, LinuxGizmos, RISC-V International official site, openRuyi official site and main repo, RISC-V member company website sweep (391 companies), riscv100.top leaderboard sweep, plus commit and release searches on GitHub across organizations and key repositories including XiangShan/NEMU/difftest/XSCache, CORE-V CVW/CVA6/cvfpu, DAMO XuanTie, ESWIN, Sophgo, Nuclei, Espressif, SpacemiT and spacemit-robotics, Sipeed, revyos, openRuyi-Project, and Radxa Linkr.
Issue Index
- Today’s Highlights: Automotive-grade RISC-V passes the baton to toolchains, as HighTec and TASKING announce AURIX RC1 support on the same day (11-10)
- I. Open Source Community and Software Ecosystem
- RevyOS kernel lands six commits at once: K3, K1, DP1000, and EIC7700 platform configs merged into the tree (11-10)
- openRuyi main repo moves into a security-fix batch: policycoreutils bumped to 3.11, cups/lxc/libmtp backport upstream patches (11-10)
- NEMU fixes SIP write latching timer interrupts on riscv64, AME switches to CRC32C for matrix completion hashing (11-10)
- II. Product and Ecosystem Developments
- riscv100.top leaderboard page redesigned: full table replaced with “architecture cards + overview ranking” (11-10)
- RISC-V Day Tokyo 2026 Autumn opens today at the University of Tokyo (11-11)
- Jamestown Foundation assesses China’s RISC-V path: strategy precedes capability (11-10)
- III. Deep Dive on Domestic Teams
- XiangShan Kunminghu V3, five commits: VBTB switches to taken-first selection, NewCSR gates breakpoint refetch (11-10)
- SpacemiT llama.cpp: restores IME1 dense Q4 path, SSM convolution fusion keeps inputs alive (11-10)
- T-Head IPO progress: Hong Kong’s Hong Kong Economic Times reports confidential filing with HKEX (11-10)
- Sipeed on two fronts: TangMega-138K fixes PCIe example and delivers Gen3 conclusion, MaixPy fixes packaging (11-10)
- IV. Embodied Intelligence and Robotics Special Section
- SpacemiT robotics stack: Unitree G1 and humanoid_common align on SONIC tracking and scheduling (11-10)
- New repo ros2_spacemit_vslam makes initial commit, SpacemiT completes the SLAM link (11-10)
- V. Summary and Trend Observations
- Appendix: Full Source List and Source Verification Table
Today’s Focus: Automotive RISC-V Passes the Baton to Toolchains, HighTec and TASKING Announce AURIX RC1 Support on the Same Day (11-10)
After Infineon’s AURIX RC1 was released on November 4 and confirmed on November 9 to use RISC-V cores derived from MIPS IP, November 10 brought the third phase: toolchains. Two European automotive-grade software tool vendors issued support announcements on the same day, landing exactly on the opening day of electronica 2026 in Munich (November 10–13).
HighTec EDV-Systeme’s announcement is specific: its RISC-V-oriented Embedded AI, Rust, and C/C++ development platform now supports RC1. The approach folds RC1’s device-specific compiler options into a single multi-architecture compiler and, in collaboration with Infineon, adds optimizations targeting pipeline and code size. What sets this route apart is language coverage—it offers both safety- and security-certified C/C++ and Rust toolchains, with an Embedded AI development platform planned for RC1; support also extends from AURIX to TRAVEO and PSOC, i.e., the entire MCU family formed after Infineon merged its two former product lines. HighTec is also a participant in the DRIVECORE RISC-V virtual prototype bundle, which includes the Synopsys VDK, Lauterbach TRACE32, HighTec’s RISC-V C/C++ platform and peripheral drivers, plus a FreeRTOS implementation and co-simulation examples—software running before silicon arrives, the usual cadence for automotive projects.
TASKING’s framing is to encompass the entire lifecycle in one toolset: VX-toolset for compilation, winIDEA and BlueBox for debugging, LDRA for verification, and Agentic AI built into the toolchain. The timeline is explicit: the VX-toolset for RISC-V with RC1 support will be available in December 2026, with toolchain certification planned for 2027. The AI component lands in two places: the December release will connect TASKING’s toolchain knowledge (builds, diagnostics, map file and LSL analysis, startup code) into the VS Code environment via MCP, while the certified tools themselves remain unchanged; on the debug side, the winIDEA AI Kit has already shown at least a 50% reduction in effort across three use cases. For teams in transition, the selling point is “change architecture, not toolchain”—those migrating from TriCore keep their existing CI/CD and expertise, while those migrating from Arm use the same toolset to cover both architectures.
Stringing the three days of news together, the progression of automotive RISC-V follows IP, silicon, toolchain, certification, with clear external signals at each of the four steps: 11-09 was the IP source (MIPS, GlobalFoundries manufacturing), 11-04 was product definition (up to 12 cores, 40MB memory, ASIL D, sampling in the first half of 2027), and 11-10 was toolchain availability (compiler in December, certification in 2027). The certification timing given by the two tool vendors aligns with Infineon’s sampling timing, indicating that this schedule is not each party speaking for itself. For domestic teams working on automotive RISC-V, what deserves attention in this round is not that there are two more compiler vendors, but that “certification timelines” are becoming the standard format for customer questions.
Date: 2026-11-10 Source: HighTec press release, TASKING press release (Business Wire)
I. Open-Source Community and Software Ecosystem
RevyOS Kernel Lands Six Commits at Once: K3, K1, DP1000, and EIC7700 Platform Configs Merged (11-10)
The RevyOS linux repository (revyos/7.3.y branch) landed six configuration commits in one batch at the end of the window: enabling SpacemiT K3 support, enabling SpacemiT K1 support, enabling UltraRISC DP1000 support, enabling ESWIN EIC7700 support, plus a linuxboot configuration and CI cross/native build jobs. All six commits share the same author and timestamp—a batch consolidation of the platform matrix rather than incremental development.
The significance of merging these configs is turning “it can boot” into “it compiles by default.” The four platforms span three vendors: SpacemiT’s K1 and next-generation K3, UltraRISC’s DP1000, and ESWIN’s EIC7700—together covering the SoCs most commonly used on current domestic RISC-V development boards and industrial control boards. The linuxboot config and dual CI builds landing together show the distribution is moving toward reproducible builds rather than just maintaining one machine that runs.
Date: 2026-11-10 Source: revyos/linux
openRuyi Main Repo Shifts to Security Fix Batch: policycoreutils Bumped to 3.11, cups/lxc/libmtp Backport Upstream Patches (11-10)
Within the window, the openRuyi main repository shifted its focus from version upgrades to security: policycoreutils was bumped to 3.11 with a backported fixfiles fix, while lxc, libmtp, and cups each received backported upstream security patches (all commits tagged security-fixes), libssh was bumped to 0.12.2, and llvm-snapshot tracked the 20261109 snapshot.
This contrasts with the previous day’s 20 “routine rolling upgrades” in the same repo: one day for feature versions, one day for security backports—a typical rhythm shift for a rolling distribution. cups and lxc sit on the two common paths of desktop and containers, while libmtp affects mobile device mounting; the backports for these three packages matter more to actual users than version number changes—without security fixes in the mainline, a distribution has no claim to being “production-ready.”
Date: 2026-11-10 Source: openRuyi-Project/openRuyi
NEMU Fixes riscv64 SIP Write Latching Timer Interrupts; AME Switches to CRC32C for Matrix Completion Hashing (11-10)
XiangShan’s NEMU emulator merged two commits in opposite directions within the window. One is a correctness fix: on the riscv64 side, writes to SIP (supervisor-level interrupt pending register) are prevented from latching timer interrupts (#1242)—a subtle bug in the interrupt virtualization path, since writing SIP should not be treated as a source of software-triggered interrupts. The other is a performance optimization: AME (matrix extension) now uses CRC32C to compute matrix completion hashes (#1280), complementing the earlier hosttlb shared matrix access optimization in #1275.
The simultaneous appearance of these two commit types shows the matrix extension software stack is being seriously polished: fixing privileged-mode semantics on one hand, reducing overhead in verification and cache paths on the other. For teams evaluating RISC-V matrix/vector extensions, NEMU’s AME implementation is one of the few open-source references capable of running a complete software stack—commits like these say more about maturity than benchmarks do.
Date: 2026-11-10 Source: OpenXiangShan/NEMU
II. Product and Ecosystem Updates
riscv100.top Leaderboard Page Redesigned: Full Table Replaced by “Architecture Cards + Overall Ranking” (11-10)
The leaderboard page on riscv100.top was redesigned within the window: the table that previously listed all results row by row no longer appears in the page source, replaced by a structure of “one top-result card per architecture + an Overall Ranking overview list.” Each card gives the platform, compiler, Ratio/GHz, Run rank, Tests, and Run date, while the overview list retains only the best row for each platform. Taking cross-architecture SPEC CPU2006 as an example, the cards correspond to AWS EC2 Graviton5 (GCC 15.1.0, 24.91), XiangShan KMHv3 RTL (XSCC 2.0, 22.80, marked NEW, record date 2026-10-17), and AMD Ryzen 9 9950X (GCC 15.1.0, 17.72); on the RISC-V side, the overview separately lists SpacemiT K3 X100 (XSCC 1.1, 8.54) and Yitian 710 (XSCC 1.1, 17.69). The homepage’s What’s New now has six entries, including SpaceMiT K3 X100 XSCC 1.1 SPEC CPU2006, SpaceMiT K3 X100 Version Progress Now Available, and Run Comparison Now Available.
The direct impact of the redesign is that “full historical results” are no longer directly retrievable from the page, and the daily briefing’s original table parsing consequently no longer works. This issue does not use it to determine any new entries or delistings—the site refreshed both the page structure and the snapshot baseline in the same redesign, and such changes cannot distinguish “score changes” from “display contraction.” Practical advice for readers: when specific scores are needed, read them from the cards and the overview; when historical trends are needed, switch to the site’s Version Progress and Run Comparison sections.
Date: 2026-11-10 Source: riscv100.top cross-arch SPEC CPU2006, riscv100.top homepage
RISC-V Day Tokyo 2026 Autumn Opens Today at the University of Tokyo (11-11)
RISC-V Day Tokyo 2026 Autumn, hosted by the RISC-V Association (Japan), is held on November 11 at the University of Tokyo’s Hongo Campus, with an expected attendance of about 400 people. The morning agenda features a chip design tutorial, while the afternoon consists of technical talks and booths: Tenstorrent speaks on physical AI and the Open Chiplet Atlas, the National Institute of Advanced Industrial Science and Technology (AIST) speaks on 2nm edge AI, and Andes Technology gives a live demonstration; the venue also hosts a special Chip Museum exhibit showcasing the hardware specifications and software overview of Google TPU v3.
This is the only offline event node in this issue, and the next stop for regional events after last week’s RISC-V China Summit. The division of labor between the two events is already fairly clear: the China event leans toward industrial deployment and domestic substitution, while the Japan event leans toward design methodology and education. Worth noting are Tenstorrent’s chiplet stance and AIST’s advanced-process edge AI direction—both will influence where the “RISC-V servers and edge computing” narrative heads next year.
Date: 2026-11-11 (event) Source: RISC-V Day Tokyo 2026 Autumn
Jamestown Foundation Assesses China’s RISC-V Path: Strategy Ahead of Capability (11-10)
The U.S. think tank Jamestown Foundation published an article on November 10 assessing China’s RISC-V progress, authored by Sunny Cheung. The article offers several verifiable factual threads: SpacemiT and the XiangShan team released the 46-core server chip V100 at the 6th RISC-V China Summit on October 19, claiming it to be the first general-purpose chip fully compliant with the RISC-V server standard; the “15th Five-Year Plan for the Electronic Information Manufacturing Industry” released in September proposed advancing RISC-V R&D and industrialization and developing RISC-V chips for artificial intelligence, whereas previously national-level policy documents rarely named the instruction set directly.
The article’s core judgment is “strategy over capability”: expanded use and influence over international organizations are clear, but performance still lags proprietary chips; the article cites Chinese analysts’ concerns about a “glass ceiling” and states that the substitution share of RISC-V chips in the national IT substitution program is “negligible.” This kind of external assessment is worth reading as a contrast—it tests the domestically oft-repeated claim that “the ecosystem is already in place” against a different set of metrics, and the gap between these two sets of metrics is precisely the question that domestic RISC-V must answer in the next stage.
Date: 2026-11-10 Source: Jamestown Foundation
III. In-depth look at domestic teams
XiangShan Kunminghu V3, five commits: VBTB switches to taken-first selection, NewCSR breakpoint-gating re-fetch (11-10)
The main XiangShan repo (kunminghu-v3 branch) merged five substantive commits in the window: changing the VBTB prediction selection logic in MBTB to taken-first (previously a different priority scheme), fixing NewCSR not re-fetching when a CSR write changes breakpoint gating (#6862), preserving ROB pending-slot reset during redirect cleanup, aligning CBO invalidate events to cache line granularity (#6864), and one codeowners maintenance commit. In the same window, NEMU, difftest, and XSCache each had commits of their own; the difftest side centered on Verilator compile options, gsim assertion registration, and FPGA-side XDMA group draining.
The most telling of the five are VBTB taken-first and the CBO invalidate alignment. The former is a priority trade-off inside the branch predictor that directly affects fetch bandwidth utilization; the latter is alignment between the cache coherence protocol and the verification model — getting it wrong won’t crash immediately, but it will produce hard-to-reproduce deviations in long-running regressions. XiangShan’s recent commit style has consistently alternated between “microarchitectural tuning” and “verification infrastructure,” and this window saw both at once.
Date: 2026-11-10 Source: OpenXiangShan/XiangShan, OpenXiangShan/difftest
SpacemiT llama.cpp: restores IME1 dense Q4 path, SSM convolution fusion keeps inputs alive (11-10)
The spacemit-com llama.cpp fork (mtmd-backend branch) merged three commits directly related to the K3 backend in the window: restoring the IME1 dense Q4 path (#60), ensuring concat inputs are not released early during ssm_conv fusion (#58), and strip-mining the gated delta network by head size (#59; the same direction was covered the previous day, this issue marks the merge landing). A further #61 merges the upstream sync (sync/upstream-20261110), with the server switching to upstream’s video capability detection logic.
Each of the three lines points to a specific problem: IME1 dense Q4 is the main path for quantized inference on K3, and restoring it means an earlier rollback has been reverted; keeping ssm_conv inputs alive is lifecycle management for a fused operator, a class of bug that only surfaces under specific graph structures; and expanding by head size is about fully using RVV’s vector width. Upstream sync and K3-specific optimizations advancing in parallel is the pattern this fork has always maintained — never diverging from upstream, never abandoning the local backend.
Date: 2026-11-10 Source: spacemit-com/llama.cpp
T-Head IPO progress: Hong Kong’s Hong Kong Economic Times says confidential filing submitted to HKEX (11-10)
Following the November 9 report that T-Head plans a Hong Kong IPO in 2027 raising US$2–3 billion, the Hong Kong Economic Times reported further on November 10 that T-Head has confidentially submitted a listing application to HKEX; Alibaba and T-Head had not responded as of publication. Previously disclosed information includes: a valuation range of US$25–62 billion; cumulative shipments of over 470,000 Zhenwu-series AI chips (PPU) and annualized revenue above RMB 10 billion, serving more than 650 enterprise customers as of June 2026 with external customers accounting for over 60%; and the Zhenwu V900 planned for mass production in Q1 2027.
The relationship to RISC-V needs to be stated clearly: the entity being listed is T-Head, whose products are the Zhenwu PPU (AI accelerator card) and Yitian (Arm server CPU); the XuanTie RISC-V processor IP was spun off from T-Head in 2024 and transferred to DAMO Academy. Media often fold “cumulative XuanTie shipments exceeding 6 billion units” into retrospectives of T-Head’s product line — that is a historical-scope figure and should not be read as meaning this IPO’s assets include the RISC-V IP business. This item is useful for judging “the capitalization path of domestic RISC-V,” but it does not constitute a financing event for the RISC-V industry itself.
Date: 2026-11-10 Source: 10jqka reprint of Caiwen report
Sipeed on two fronts: TangMega-138K fixes PCIe examples and states Gen3 conclusion, MaixPy fixes packaging (11-10)
Sipeed wrapped up two product lines in the window. TangMega-138K-example produced five documentation and example fixes in a row: removing a duplicate rstn and an invalid write to 0x808760 in the PCIe example, applying the same fixes to the Gen3 (8G) project, correcting the Raspberry Pi-side PCIe notes and documenting the REFCLK fix, adding NextTang Gen3 test results, and explicitly stating two conclusions: “Gen3 long-run stability testing is unreliable; Gen2 is still recommended” and “DRP replay is not built in.” On the MaixPy side, image packaging issues were fixed and the CMake version used for releases was pinned.
The two are different in nature but both amount to “stating things clearly”: the PCIe Gen3 conclusion deliberately lowers expectations so users don’t buy based on marketing claims, while MaixPy’s packaging and version pinning make release artifacts reproducible. For a RISC-V ecosystem whose main form factor is development boards, this kind of cleanup is often worth more than shipping another new board.
Date: 2026-11-10 Source: sipeed/TangMega-138K-example, sipeed/MaixPy
IV. Special section: embodied intelligence and robotics
SpacemiT robotics stack: Unitree G1 and humanoid_common align SONIC tracking and scheduling (11-10)
spacemit-robotics made a round of consistency fixes to its bipedal humanoid stack in the window: humanoid_unitree_g1 fixed G1’s SONIC tracking, motion presets, and startup-readiness determination; humanoid_common aligned SONIC scheduling while maintaining tracking response; and control_whole_body separated feedback position clamping from command boundary handling. The three commits share an author and close timestamps — a cross-repo coordinated fix.
The technical point is that “tracking” and “scheduling” are now described separately. Earlier implementations coupled the planner’s CPU scheduling with the tracking loop (the 11-04 #14 commit began the split), and this round further decouples startup readiness, motion presets, and tracking-error determination. For teams running humanoid control on K3, the significance is that jitter sources in the real-time loop become localizable — when something goes wrong, you can tell whether it’s planner queuing or the tracking itself exceeding tolerance.
Date: 2026-11-10 Source: humanoid_unitree_g1, control_whole_body
New repo ros2_spacemit_vslam makes initial commit, SpacemiT fills in the SLAM piece (11-10)
spacemit-robotics created the repo ros2_spacemit_vslam and completed its initial commit; judging by the name, it wraps visual SLAM as a ROS 2 node for delivery on the SpacemiT platform. The organization already had perception (packaging the inference runtime into las2), unitree-go2, the humanoid series, and quadruped ports; SLAM was the previously missing piece in its robotics software stack.
With SLAM in place, SpacemiT’s robotics stack now covers four layers at the software level: perception, localization and mapping (vslam), whole-body control (control_whole_body), and full-robot adaptation (unitree_g1/go2). This combination looks more like “completing a reproducible full-robot reference implementation” than releasing individual algorithms one by one — for teams wanting to put K3 into a robot body, the completeness of the reference implementation is more useful than any single benchmark.
Date: 2026-11-10 Source: ros2_spacemit_vslam
V. Summary and trend observations
1. The competitive front for automotive-grade has moved to toolchains and certification timing. The three steps of AURIX RC1 over three days — product definition on 11-04, IP sourcing on 11-09, toolchain support on 11-10 — lay out the automotive-grade RISC-V sequence clearly: chip vendors define the platform, IP vendors supply the core, tool vendors provide compilation and debugging, and certification comes last. HighTec and TASKING’s “tools available in December, certification in 2027” dovetails precisely with Infineon’s “sampling in H1 2027,” indicating the timetable was coordinated. The question domestic automotive-grade teams will likely face next is not “do you have a RISC-V core” but “when will your toolchain pass certification.”
2. What the distro side did this day was “deliverability,” not “new features.” RevyOS added four platform configs (K3, K1, DP1000, EIC7700) to the tree in one go and added linuxboot and dual-build CI, while openRuyi moved into a security backport batch — both answering the same question: can this system be delivered to someone else as a deliverable? Configs in the tree determine “can others build it”; security backports determine “dare we put it into production.” By contrast, new feature releases were absent this day.
3. Domestic teams’ work focused on completing the chain, not raising the numbers. XiangShan changed branch prediction and verification alignment; SpacemiT restored the main quantized path in llama.cpp and fixed fused-operator lifecycle; the robotics side decoupled SONIC scheduling from tracking and added the SLAM repo. What these changes share is “making existing capabilities stable and reproducible,” alternating in rhythm with the previous day’s Nuclei NI1000 launch (new IP, new benchmarks).
4. The comparability of leaderboard-type sources needs re-confirmation. The riscv100.top redesign removed full results from the page, leaving only per-architecture top entries and an overview list. For day-to-day tracking, this means the “score change” dimension is temporarily not continuously observable; our approach this issue is not to judge new entries or drop-offs, and to rebuild the baseline once the site structure stabilizes. Any statement drawing conclusions from a single leaderboard should include a note on methodology during the redesign period.
5. Offshore perspectives are starting to assess China’s RISC-V by “capability” rather than “scale.” The Jamestown Foundation article acknowledges the intensity of policy and ecosystem push, but also points out that the performance gap and substitution share are “negligible.” Such assessments need not be accepted wholesale, but they suggest one thing: while the domestic narrative centers on shipment volumes, summit scale, and standards compliance, external observers look at single-core performance, software maturity, and actual substitution rate — the gap between these two sets of metrics is exactly the homework for the next stage.
Appendix
Source Verification Table
| Source | Verification Result |
|---|---|
| Google News aggregation (24 Chinese and English queries, 24-hour window) | 4 substantive new items within the window: HighTec support for AURIX RC1, TASKING unified toolchain, Jamestown Foundation commentary, MIPS Actus reprint (already included on 09-02, not counted again); the T-Head IPO and MIPS IP items were already included in the 11-10 report, so only follow-up developments are covered this issue; stock market pieces, fund NAV pieces, and sports/betting SEO pieces were removed in bulk |
| Hacker News / Phoronix / CNX Software / LinuxGizmos | No new RISC-V-related items within the window; Phoronix content for the period was unrelated to RISC-V, and CNX and LinuxGizmos had no corresponding entries |
| GitHub organization and key repository checks (30+ organizations and repositories) | 13 organizations pushed within the window: spacemit-com 9 repos, sipeed 8 repos, openRuyi-Project 7 repos, spacemit-robotics 4 repos, openhwfoundation 3 repos, revyos 1 repo; on the XiangShan side, five repos — XiangShan / NEMU / difftest / XSCache / GEM5 — had commits; radxa-linkr and DAMO XuanTie (XUANTIE-RV) had zero pushes, and ESWIN, Sophgo, Nuclei, and Espressif had no new commits within the window |
| riscv100.top inspection | The site’s leaderboard page was redesigned into “architecture cards + overview ranking,” breaking the original table parsing; this issue does not determine new entries or drop-offs; the body text is based on manually reading the cards and What’s New, and the baseline will be rebuilt once the site structure stabilizes |
| RISC-V member website inspection (391 sites) | 322 sites fetched successfully, of which 8 content changes were all homepage rotations at universities and research institutions (MIT, KTH, imec, METU, etc.), with no substantive product activity, so not included; 69 sites failed to fetch (including ISCAS, Tencent, Ingenic, MediaTek, DENSO); failures do not block |
| RISC-V Day Tokyo 2026 Autumn official site | Agenda and speakers unchanged; the event opens today and is included as an event entry |
Full Source List
- [1] HighTec support for AURIX RC1 press release — https://www.presseagentur.com/hightec/detail.php?pr_id=7860&lang=en
- [2] TASKING unified toolchain supports AURIX RC1 (Business Wire) — https://www.businesswire.com/news/home/20261110215415/en/TASKINGs-Unified-Toolchain-Supports-Infineons-New-AURIX-RC1-RISC-V-based-Automotive-MCU-Family-with-Industrys-First-AI-Enabled-End-to-End-Workflow
- [3] revyos/linux repository — https://github.com/revyos/linux
- [4] openRuyi-Project/openRuyi main repository — https://github.com/openRuyi-Project/openRuyi
- [5] OpenXiangShan/NEMU — https://github.com/OpenXiangShan/NEMU
- [6] riscv100.top cross-architecture SPEC CPU2006 leaderboard — https://riscv100.top/cross-arch/spec06/
- [7] riscv100.top homepage — https://riscv100.top/
- [8] RISC-V Day Tokyo 2026 Autumn — https://riscv.or.jp/en/risc-v-day-tokyo-2026-autumn-2
- [9] Jamestown Foundation commentary — https://jamestown.org/program/risc-v-open-source-chip-push-puts-strategy-over-capability/
- [10] OpenXiangShan/XiangShan — https://github.com/OpenXiangShan/XiangShan
- [11] OpenXiangShan/difftest — https://github.com/OpenXiangShan/difftest
- [12] spacemit-com/llama.cpp — https://github.com/spacemit-com/llama.cpp
- [13] T-Head IPO progress report (10jqka reprint of Caiwen) — https://news.10jqka.com.cn/20261110/c681642866.shtml
- [14] sipeed/TangMega-138K-example — https://github.com/sipeed/TangMega-138K-example
- [15] sipeed/MaixPy — https://github.com/sipeed/MaixPy
- [16] spacemit-robotics/humanoid_unitree_g1 — https://github.com/spacemit-robotics/humanoid_unitree_g1
- [17] spacemit-robotics/control_whole_body — https://github.com/spacemit-robotics/control_whole_body
- [18] spacemit-robotics/ros2_spacemit_vslam — https://github.com/spacemit-robotics/ros2_spacemit_vslam
Limitations: This issue covers a 24-hour window from 05:00 on November 10 to 05:00 on November 11. On the Google News side, this issue is dominated by toolchain follow-ups for Infineon AURIX RC1, with a gap in chip-launch news; the T-Head IPO and MIPS IP items were already included as entries in the November 10 report, so this issue covers only new developments and does not count them again. riscv100.top was redesigned within the window, breaking parsing of the original leaderboard table, so this issue does not report score changes or drop-offs. Of the 391 member websites inspected, 69 failed to fetch (including ISCAS, Tencent, Ingenic, MediaTek, etc.); the 8 judged to have content changes were all university homepage rotations with no substantive product releases. On the Chinese-language side, queries such as “XuanTie,” “Sophgo,” “Espressif,” and “ESWIN” were still dominated by stock market pieces and unrelated content, which were removed based on headline semantics.