For Debug, make sure we use right register selection heuristics to match with Release

[kunalspathak]

If a method needs consecutive register, we do not use *minimal() register allocation strategy added in #96386. The reason being that in some cases, the *minimal might allocate registers that will sometime obstruct us from allocating consecutive registers. Also, to handle consecutive registers scenarios in *minimal() methods eats up the massive TP gain that we see in Arm64 for MinOpts because of the checks we need to add if refPosition needs consecutive register or not. I forgot to add a condition for consecutive register while deciding the selection heuristics.

Fixes: #97638

[ghost]

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

Issue Details

---

If a method needs consecutive register, we do not use *minimal() register allocation strategy added in #96386. The reason being that in some cases, the *minimal might allocate registers that will sometime obstruct us from allocating consecutive registers. Also, to handle consecutive registers scenarios in *minimal() methods eats up the massive TP gain that we see in Arm64 for MinOpts because of the checks we need to add if refPosition needs consecutive register or not. I forgot to add a condition for consecutive register while deciding the selection heuristics.

Fixes: #97638

Author:	kunalspathak
Assignees:	-
Labels:	area-CodeGen-coreclr
Milestone:	-

[ryujit-bot]

Diff results for #97657

Assembly diffs

Assembly diffs for osx/arm64 ran on linux/x64

Diffs are based on 2,029,386 contexts (927,368 MinOpts, 1,102,018 FullOpts).

MISSED contexts: 109 (0.01%)

Overall (-41,352 bytes)

Collection	Base size (bytes)	Diff size (bytes)
benchmarks.run_pgo.osx.arm64.checked.mch	34,533,064	-372
benchmarks.run_tiered.osx.arm64.checked.mch	15,508,464	-220
coreclr_tests.run.osx.arm64.checked.mch	483,586,020	-40,760

MinOpts (-41,352 bytes)

Collection	Base size (bytes)	Diff size (bytes)
benchmarks.run_pgo.osx.arm64.checked.mch	16,373,348	-372
benchmarks.run_tiered.osx.arm64.checked.mch	11,504,272	-220
coreclr_tests.run.osx.arm64.checked.mch	330,172,768	-40,760

Assembly diffs for windows/arm64 ran on linux/x64

Diffs are based on 2,070,850 contexts (937,853 MinOpts, 1,132,997 FullOpts).

MISSED contexts: 139 (0.01%)

Overall (-41,352 bytes)

Collection	Base size (bytes)	Diff size (bytes)
benchmarks.run_pgo.windows.arm64.checked.mch	46,609,220	-372
benchmarks.run_tiered.windows.arm64.checked.mch	15,506,140	-220
coreclr_tests.run.windows.arm64.checked.mch	496,298,628	-40,760

MinOpts (-41,352 bytes)

Collection	Base size (bytes)	Diff size (bytes)
benchmarks.run_pgo.windows.arm64.checked.mch	16,258,192	-372
benchmarks.run_tiered.windows.arm64.checked.mch	11,177,748	-220
coreclr_tests.run.windows.arm64.checked.mch	339,674,404	-40,760

Details here

---

Throughput diffs

Throughput diffs for osx/arm64 ran on windows/x64

MinOpts (-0.01% to +0.00%)

Collection	PDIFF
libraries.pmi.osx.arm64.checked.mch	-0.01%

Throughput diffs for windows/arm64 ran on windows/x64

MinOpts (-0.00% to +0.01%)

Collection	PDIFF
libraries.pmi.windows.arm64.checked.mch	+0.01%

Details here

---

[ryujit-bot]

Diff results for #97657

Assembly diffs

Assembly diffs for linux/arm64 ran on windows/x64

Diffs are based on 2,259,470 contexts (1,008,044 MinOpts, 1,251,426 FullOpts).

MISSED contexts: 159 (0.01%)

Overall (-41,352 bytes)

Collection	Base size (bytes)	Diff size (bytes)
benchmarks.run_pgo.linux.arm64.checked.mch	79,903,244	-372
benchmarks.run_tiered.linux.arm64.checked.mch	22,276,872	-220
coreclr_tests.run.linux.arm64.checked.mch	509,740,232	-40,760

MinOpts (-41,352 bytes)

Collection	Base size (bytes)	Diff size (bytes)
benchmarks.run_pgo.linux.arm64.checked.mch	25,548,372	-372
benchmarks.run_tiered.linux.arm64.checked.mch	17,338,964	-220
coreclr_tests.run.linux.arm64.checked.mch	348,907,856	-40,760

Details here

---

[kunalspathak]

@dotnet/jit-contrib