[PATCH] D112406: [Driver][AArch64]Add driver support for neoverse-512tvb target

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rG2186b011e966: [Driver][AArch64]Add driver support for neoverse-512tvb target (authored by CarolineConcatto). Changed prior to commit:

[PATCH] D112406: [Driver][AArch64]Add driver support for neoverse-512tvb target

dmgreen added a comment. > The total vector bandwidth includes unrolling so currently having > `VScaleForTuning=1` and `MaxInterleaveFactor=4` implies 512 tvb. If the > target has >128bit vectors then vector loops will likely have more work than > they can handle in parallel but as long as

[PATCH] D112406: [Driver][AArch64]Add driver support for neoverse-512tvb target

sdesmalen accepted this revision. sdesmalen added a comment. This revision is now accepted and ready to land. LGTM with nit addressed. Comment at: llvm/lib/Target/AArch64/AArch64.td:840 +def TuneNeoverse512TVB : SubtargetFeature<"neoverse512tvb", "ARMProcFamily",

[PATCH] D112406: [Driver][AArch64]Add driver support for neoverse-512tvb target

paulwalker-arm added a comment. In D112406#3087191 , @dmgreen wrote: > Thanks. If the cpu has a 512 bit total vector bandwidth, should the > VScaleForTuning be 1 or 2 (or higher)? llvm doesn't usually deal with total > bandwidth a lot, perhaps not as

[PATCH] D112406: [Driver][AArch64]Add driver support for neoverse-512tvb target

dmgreen added a comment. Thanks. If the cpu has a 512 bit total vector bandwidth, should the VScaleForTuning be 1 or 2 (or higher)? llvm doesn't usually deal with total bandwidth a lot, perhaps not as much as it should. @david-arm any thoughts? Repository: rG LLVM Github Monorepo CHANGES

[PATCH] D112406: [Driver][AArch64]Add driver support for neoverse-512tvb target

CarolineConcatto marked an inline comment as done. CarolineConcatto added a comment. Thank you for your review @dmgreen and @tschuett. I rebase the patch, now VScaleForTuning is being set. And I removed support for neoverse-512tvb from Arm 32 bits. @paulwalker-arm pointed me that

[PATCH] D112406: [Driver][AArch64]Add driver support for neoverse-512tvb target

CarolineConcatto updated this revision to Diff 382273. CarolineConcatto added a comment. - Rebase and remove support on for ARM 32 bits Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D112406/new/ https://reviews.llvm.org/D112406 Files:

[PATCH] D112406: [Driver][AArch64]Add driver support for neoverse-512tvb target

tschuett added a comment. This one might get a `VScaleForTuning`: https://reviews.llvm.org/D112459 Do you need this as well? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D112406/new/ https://reviews.llvm.org/D112406

[PATCH] D112406: [Driver][AArch64]Add driver support for neoverse-512tvb target

dmgreen added a comment. This is a bit of a shame. I was hoping we wouldn't need the same hacks as GCC. The llvm cost modelling can work quite differently at times to GCC and I didn't think we were close enough to optimal code to need to worry about these kinds of differences. I guess having

[PATCH] D112406: [Driver][AArch64]Add driver support for neoverse-512tvb target

CarolineConcatto created this revision. Herald added subscribers: dexonsmith, hiraditya, kristof.beyls. CarolineConcatto requested review of this revision. Herald added projects: clang, LLVM. Herald added subscribers: llvm-commits, cfe-commits. The support for neoverse-512tvb mirrors the same