SPMI: per-context tpdiff examples

[AndyAyersMS]

Add a "top method regressions / improvements" section to the tpdiff markdown summary, mirroring how asmdiffs surfaces specific contexts. The per-context data already lives in the details CSV; aggregate it once and emit FullOpts/MinOpts top-N tables by PDIFF %.

Release JITs don't currently report MethodFullName, so the table also shows the SPMI context number for fallback lookup via mcs -dumpMap or superpmi -c N.

Fixes #85755.

[AndyAyersMS]

@jakobbotsch PTAL
fyi @dotnet/jit-contrib

---

Sample output

Existing pivot section (unchanged):

Overall (-4.84%)

Collection	PDIFF
benchmarks.run.windows.x64.checked.mch	-4.84%

FullOpts (-4.82%)

Collection	PDIFF
benchmarks.run.windows.x64.checked.mch	-4.82%

New per-context section (trimmed here to top 5 of each; the actual report shows top 20):

Top method regressions (FullOpts, by PDIFF %)

Collection	Context	Method	Base	Diff	PDIFF
benchmarks.run.windows.x64.checked.mch	224	<no name reported by JIT>	77,320	122,551	+58.50%
benchmarks.run.windows.x64.checked.mch	32723	<no name reported by JIT>	78,626	123,097	+56.56%
benchmarks.run.windows.x64.checked.mch	42681	<no name reported by JIT>	91,558	143,311	+56.52%
benchmarks.run.windows.x64.checked.mch	8751	<no name reported by JIT>	57,133	87,846	+53.76%
benchmarks.run.windows.x64.checked.mch	1631	<no name reported by JIT>	57,165	87,876	+53.72%

Top method improvements (FullOpts, by PDIFF %)

Collection	Context	Method	Base	Diff	PDIFF
benchmarks.run.windows.x64.checked.mch	12166	<no name reported by JIT>	1,392,193	1,170,242	-15.94%
benchmarks.run.windows.x64.checked.mch	52652	<no name reported by JIT>	165,001	138,917	-15.81%
benchmarks.run.windows.x64.checked.mch	18230	<no name reported by JIT>	156,170	132,310	-15.28%
benchmarks.run.windows.x64.checked.mch	33041	<no name reported by JIT>	1,351,314	1,146,456	-15.16%
benchmarks.run.windows.x64.checked.mch	9415	<no name reported by JIT>	219,937	187,271	-14.85%

The Method column is empty here because Release JITs (what tpdiff uses) don't currently
emit MethodFullName via reportMetadata — see compiler.cpp:311-322, which gates the
report on DEBUG || LATE_DISASM || DUMP_FLOWGRAPHS. The Context number is included so
the method can still be resolved via e.g. DOTNET_JitDisasm=* superpmi -c <N> <jit> <mch>
(verified: context 224 above is System.Net.SocketAddress:.cctor()). A small follow-up
to emit MethodFullName unconditionally would let this column populate directly.

[Copilot]

Pull request overview

Adds per-context throughput “examples” to SuperPMI tpdiff markdown summaries by extracting per-context instruction deltas from the existing details CSV and emitting top-N regression/improvement tables (split by FullOpts/MinOpts). Also updates the tuple shape flowing through aggregate_diff_metrics/tpdiff reporting and makes the markdown writer tolerant of older 3-tuple JSON summaries.

Changes:

• Extend aggregate_diff_metrics() to collect per-context throughput diffs (base vs diff instruction counts) and return them to callers.
• Plumb per-context throughput data through replay_with_throughput_diff() and into markdown summary generation.
• Add a new markdown section that lists top per-context regressions/improvements (FullOpts and MinOpts) by PDIFF%.

[AndyAyersMS]

@tannergooding @jakobbotsch can you re-review?

[tannergooding]

Changes still LGTM, functionally speaking. Jakob is much more familiar with the long term implementation and goals of the script

[Copilot]

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated 4 comments.

[jakobbotsch]

LGTM!