Skip to content

feat(LoadQueueReplay): add support of fast wakeup for C_MA & C_FF#6092

Open
weidingliu wants to merge 1 commit into
kunminghu-v3from
feat-fast-wakeup-replay
Open

feat(LoadQueueReplay): add support of fast wakeup for C_MA & C_FF#6092
weidingliu wants to merge 1 commit into
kunminghu-v3from
feat-fast-wakeup-replay

Conversation

@weidingliu

Copy link
Copy Markdown
Member

This PR add the support of fast wakeup for load of mdp hit && forward fail.

The timing of Sta wakeup:

image

The timing of Std wakeup:

image

The timing of Cancel

image

Note: The Std pipeline only has 1 stage. Here, a virtual pipeline stage is drawn for easier comparison.

Performance:

image

@github-actions github-actions Bot added module: backend Decode, Rename, Issue, Dispatch, Rob, Alu, Csr, fudian, yunsuan module: memory Memblock, DCache, TLB, Prefetcher, coupledL2, huancun topic: functionality To introduce new function, e.g. new isa extensions, new components, bug fixes... module: top XSTop, XSTile, XSParameters, configs labels Jun 13, 2026
@github-actions

github-actions Bot commented Jun 13, 2026

Copy link
Copy Markdown

Emu - Performance Summary

Metadata

- SHA Run ID
Base 96c3f56 27999991129
Current 80dca63 28007422326

IPC Report

Testcase Current Base Diff
GemsFDTD_49458 1.492659 1.544234 -3.34%
astar_rivers_8728 2.343521 2.331856 0.50%
bwaves_30350 1.042206 1.041898 0.03%
bzip2_liberty_739 7.384723 7.385661 -0.01%
cactusADM_61235 4.203873 4.215580 -0.28%
calculix_53456 5.369883 5.401027 -0.58%
dealII_13061 4.149530 4.124533 0.61%
gamess_gradient_36450 4.982815 4.976041 0.14%
gcc_s04_2772 2.383243 2.385674 -0.10%
gobmk_nngs_453 0.753839 0.764295 -1.37%
gromacs_2907 2.071699 2.060931 0.52%
h264ref_foreman.main_10053 4.371775 4.339641 0.74%
hmmer_nph3_33214 5.816273 5.816291 -0.00%
lbm_31064 2.562498 2.562498 0.00%
legacy-astar 2.132444 2.135677 -0.15%
legacy-gcc 1.542792 1.532376 0.68%
legacy-gromacs 3.048744 3.044628 0.14%
legacy-lbm 2.649839 2.649839 0.00%
legacy-mcf 1.203734 1.189050 1.23%
legacy-milc 1.823334 1.823334 0.00%
legacy-namd 3.700641 3.691252 0.25%
legacy-wrf 2.758577 2.739936 0.68%
legacy-xalancbmk 4.017809 3.997902 0.50%
leslie3d_37552 2.085383 2.082479 0.14%
libquantum_81539 5.198913 5.198913 0.00%
mcf_6388 0.471749 0.470476 0.27%
milc_7124 2.662981 2.540540 4.82%
namd_75757 3.397575 3.427555 -0.87%
omnetpp_14042 0.911805 0.907783 0.44%
perlbench_splitmail_3995 4.281096 4.289633 -0.20%
povray_8362 2.784029 2.775264 0.32%
sjeng_64284 2.152629 2.147114 0.26%
soplex_ref_10774 4.375660 4.301526 1.72%
sphinx3_141036 3.874010 3.914751 -1.04%
tonto_69015 4.274239 4.254901 0.45%
wrf_112496 2.467694 2.451723 0.65%
xalancbmk_9874 3.236085 3.236085 0.00%
zeusmp_45598 3.272749 3.370671 -2.91%
GEOMEAN 2.664686397549851 2.6619046263822277 0.10%

@weidingliu weidingliu force-pushed the feat-fast-wakeup-replay branch 2 times, most recently from 3175c14 to 72881e2 Compare June 13, 2026 09:30
This Commit add support of fast wakeup from issueQueue of sta && std to LRQ.

We will send sqIdx at og0 stage of issueQueue, latch 1 cycle. We will use scoreBoard to trace the wakeup and detect wakeup cancel.

the source of wakeup cancel: og1Cancel, s0Cancel, s1Cancel.
@weidingliu weidingliu force-pushed the feat-fast-wakeup-replay branch from 72881e2 to 80dca63 Compare June 23, 2026 06:39
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

module: backend Decode, Rename, Issue, Dispatch, Rob, Alu, Csr, fudian, yunsuan module: memory Memblock, DCache, TLB, Prefetcher, coupledL2, huancun module: top XSTop, XSTile, XSParameters, configs topic: functionality To introduce new function, e.g. new isa extensions, new components, bug fixes...

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant