Repository navigation
refactor(cuda): simplify benchmark session setup (#10885) - #10885
Conversation
|
@claude review this pr in depth |
|
Claude finished @0ax1's task in 1m 38s —— View job Review of #10885Verdict: LGTM. The change preserves behaviour: the new sessions act the same as the old ones. I found nothing blocking, only two optional cleanups.
Why the sessions behave the same
Performance noteThe PR description is accurate:
Compile and format checks (by reading the code)
Optional cleanups (not blocking)
A follow-up outside this PR's scope: in five of the six benches (all but ChecksI didn't build, lint or run the benches. Following |
Merging this PR will not alter performance1 warning
Comparing |
0b63c70 to
f6ee1d6
Compare
that register `ArraySession`, `KernelSession`, `DTypeSession`, `ScalarFnSession`, `AggregateFnSession` and `MemorySession` up front, although `VortexSession::empty()` creates each of them with the same defaults on first use. Register only the `CudaSession`. Signed-off-by: Alexander Droste <alexander.droste@protonmail.com>
Signed-off-by: Alexander Droste <alexander.droste@protonmail.com>
d2a402f to
336e223
Compare
Summary
Simplifies session setup in seven CUDA benchmarks by removing explicit
registration of default host session variables.
VortexSession already initializes these variables with the same defaults
on first access. Register only CudaSession explicitly, preserving the
custom FSST export-layout configuration.
Changes
ScalarFnSession, AggregateFnSession, and MemorySession registrations.
This is benchmark setup cleanup, not a decoding optimization. Host
session defaults are initialized on first use rather than eagerly.
Production code is unaffected.
No builds, tests, or benchmarks were run for this change.