Feature fall stall prevention in pipeline CPUs #13
Conversation
|
I defer to @eastWillow for confirmation. |
This comment was marked as duplicate.
This comment was marked as duplicate.
Sorry, something went wrong.
jgw0915
force-pushed
the
Control-Hazard-Logic
branch
2 times, most recently
from
7e14aa8 to
878a0ce
Compare
|
Hi @eastWillow, I have made an update on this PR. Please check it out. |
jserv
left a comment
jserv
left a comment
There was a problem hiding this comment.
Enforce the rules specified in https://chris.beams.io/git-commit carefully.
jgw0915
force-pushed
the
Control-Hazard-Logic
branch
2 times, most recently
from
9d648f6 to
3c78cba
Compare
jgw0915
changed the title
|
Hi @eastWillow, I have made an update on this PR. Please check it out. |
jgw0915
force-pushed
the
Control-Hazard-Logic
branch
from
3c78cba to
429ca3d
Compare
jgw0915
changed the title
| val memory_read_enable_ex = Input(Bool()) // id2ex.io.output_memory_read_enable | ||
| val rd_ex = Input(UInt(Parameters.PhysicalRegisterAddrWidth)) // id2ex.io.output_regs_write_address | ||
| val uses_rs1_id = Input(Bool()) // true only if current ID instruction really reads rs1 | ||
| val uses_rs2_id = Input(Bool()) |
There was a problem hiding this comment.
This should match the comment in fivestage_stall/Control.scala.
| val uses_rs2_id = Input(Bool()) | |
| val uses_rs2_id = Input(Bool()) // true only if current ID instruction really reads rs2 |
|
I noticed that 3-pipeline/csrc/hazard_extended.S was updated in PR #8, the corresponding hazard_extended.asmbin was not.
|
Add uses_rs1_id/uses_rs2_id from decode and feed them into stall/forward CPU's control unit so hazard checks know which operands are real. Without gating, stall logic can misread a lw’s immediate bits as an rs2 match, causing false load-use stalls/flushes. We now guard comparisons with uses_rs1_id/uses_rs2_id to avoid that. Unused operand read addresses are driven to x0 to keep dependency matching consistent and prevent accidental collisions.
jgw0915
force-pushed
the
Control-Hazard-Logic
branch
from
ed47301 to
986db74
Compare
There was a problem hiding this comment.
I locally rebased my solution with this Pull Request.
Running sbt "project pipeline" test passed successfully.
[success] Total time: 47 s, completed Jan 31, 2026 6:25:19 AM
Thanks for the great work! I don't have any further suggestions.
Thanks @jgw0915 for contributing!
|
Thank @jgw0915 for contributing! |
3 participants
Solving issue mentioned by @eastWillow