Age | Commit message (Collapse) | Author | Files | Lines | |
---|---|---|---|---|---|
2023-05-22 | Add vec-sgemm | Jerry Zhao | 1 | -0/+16 | |
2023-05-22 | Enable VS for benchmarks | Jerry Zhao | 1 | -2/+2 | |
2023-05-22 | Improve performance of syscalls/memcpy | Jerry Zhao | 1 | -1/+15 | |
2020-04-17 | The HTIF device must live in its own page since it is (generally) a ↵ | Adrian Harris | 1 | -0/+1 | |
bus/hardware device (#274) | |||||
2020-02-08 | Solves https://github.com/riscv/riscv-tests/issues/241 : Each mhartid has ↵ | Sho Nakatani | 1 | -2/+2 | |
the same size of stack & TLS. (#242) | |||||
2019-12-10 | benchmarks: Simplify TLS initialisation (#224) | James Clarke | 3 | -19/+5 | |
The symbols used to query the size of .tdata and .tbss need not be thread-local themselves; instead, make them linker script-provided non-thread-local symbols. | |||||
2017-08-04 | Improve RVC test | Andrew Waterman | 1 | -1/+2 | |
Make the page-crossing instruction non-idempotent to detect erroneously executing the first 16 bits of the instruction with garbage MSBs. | |||||
2017-05-01 | Set ELF entry point correctly | Andrew Waterman | 1 | -0/+1 | |
2017-04-26 | Add abort() for benefit of benchmark code | Andrew Waterman | 1 | -0/+6 | |
2017-03-29 | Prohibit relaxing the initial gp generation | Palmer Dabbelt | 1 | -0/+3 | |
I've added an additional linker relaxation that relaxes two instruction pc-relative sequences to one instruction gp relative sequences when possible. This sequence now optimizes the initial gp generation to mv gp, gp which is obviously bogus. The fix is to disable relaxations when setting up gp, preventing the linker from relaxing away this setup code. | |||||
2017-03-29 | Change the global pointer symbol to __global_pointer$ | Palmer Dabbelt | 2 | -2/+2 | |
This changed a while ago in binutils. | |||||
2017-03-25 | Fix FPU initialization code | Andrew Waterman | 1 | -6/+8 | |
2017-03-24 | Like tests, pass the benchmarks if XLEN disagrees | Andrew Waterman | 1 | -5/+6 | |
2017-03-22 | Clean up benchmarks build | Andrew Waterman | 2 | -33/+10 | |
2017-03-21 | Clean up bmarks build | Andrew Waterman | 2 | -5/+5 | |
2017-03-21 | Remove smips/host-debugging cruft | Andrew Waterman | 1 | -50/+0 | |
2017-01-20 | Use correct macro to guard floating-point initialization | Andrew Waterman | 1 | -1/+1 | |
2017-01-20 | Fix build with glibc | Andrew Waterman | 1 | -0/+2 | |
2016-12-06 | avoid non-standard predefined macros | Andrew Waterman | 1 | -2/+2 | |
2016-08-29 | On RV32, zero-extend pointers for HTIF | Andrew Waterman | 1 | -12/+12 | |
2016-07-11 | Remove deprecated uarch counters; support RVC for benchmarks | Andrew Waterman | 2 | -27/+13 | |
2016-07-07 | Don't use FPU in benchmarks that don't need to use the FPU | Andrew Waterman | 3 | -20/+12 | |
2016-05-25 | Keep tohost/fromhost at deterministic address | Andrew Waterman | 3 | -7/+15 | |
2016-05-02 | Stop using tohost/fromhost registers | Andrew Waterman | 1 | -3/+11 | |
2016-04-30 | ERET -> xRET; new memory map | Andrew Waterman | 3 | -10/+96 | |
For now, we no longer build hex files, because the programs don't start at address 0. This decision will likely be revisited. | |||||
2016-04-06 | Don't use stats register; refer to uarch counters by number | Andrew Waterman | 1 | -10/+5 | |
2016-03-14 | Rework benchmarks to run in M-mode | Andrew Waterman | 4 | -26/+18 | |
This broadens their utility and lets them use the M-mode counters directly. | |||||
2016-03-03 | WIP on priv spec v1.9 | Andrew Waterman | 1 | -3/+3 | |
2015-07-02 | Fix RV32 handling of syscall arguments | Andrew Waterman | 2 | -63/+65 | |
2015-07-01 | Add _hard_float check in crt.S. | Christopher Celio | 1 | -0/+3 | |
2015-05-09 | Update to privileged architecture version 1.7 | Andrew Waterman | 3 | -18/+8 | |
2015-05-01 | Make dhrystone report correct-ish numbers | Andrew Waterman | 2 | -1/+16 | |
2015-04-21 | Don't rely on reset values of MSTATUS_UA/SA | Andrew Waterman | 2 | -3/+18 | |
2015-04-21 | Fix benchmark compilation/execution on RV32 | Andrew Waterman | 1 | -1/+1 | |
2015-03-17 | Merge [shm]call into ecall, [shm]ret into eret | Andrew Waterman | 2 | -3/+3 | |
2015-03-12 | Update to new privileged spec | Andrew Waterman | 3 | -38/+40 | |
2015-01-09 | Add LICENSE | Andrew Waterman | 3 | -0/+6 | |
2014-12-15 | Remove dependence on machine/syscall.h | Andrew Waterman | 1 | -1/+2 | |
2014-12-13 | Initialize static TLS for the benchmarks | Andrew Waterman | 3 | -12/+33 | |
2014-12-12 | Add more entropy to matrix multiplication input | Andrew Waterman | 1 | -0/+8 | |
This will exercise the floating-point units more thoroughly. | |||||
2014-12-12 | Use user stack in supervisor mode | Andrew Waterman | 1 | -14/+1 | |
This fixes a race condition, as there was only one kernel stack. | |||||
2014-12-03 | Use new toolchain and calling convention | Andrew Waterman | 2 | -27/+19 | |
2014-11-07 | Clean up canonical mt benchmarks and reorganize extra versions in /mt. All ↵ | Henry Cook | 1 | -0/+11 | |
versions support support at least 1/2/4 threads. | |||||
2014-11-06 | Fix build with riscv-gcc version 4.9 | Andrew Waterman | 1 | -0/+1 | |
2014-11-06 | Fix TLS in benchmarks | Andrew Waterman | 1 | -0/+3 | |
Linker relaxations were screwing up loading the thread pointer. | |||||
2014-09-10 | Enable interrupts in bmarks | Christopher Celio | 1 | -6/+5 | |
2014-04-14 | commit high-performance mm (scalar and vector versions) | Yunsup Lee | 3 | -18/+14 | |
2014-04-07 | Add radix sort benchmark | Andrew Waterman | 2 | -2/+3 | |
2014-04-03 | setStats in benchmarks now should set and unset the stats register. Also, ↵ | Stephen Twigg | 2 | -5/+14 | |
removed legacy SET_STATS macro. | |||||
2014-03-25 | Make qsort benchmark more meaningful | Andrew Waterman | 1 | -2/+2 | |
Before, we were sorting a sorted array :\ |