tree 5a43d424c32a19ec36c79b6068bf39e25c13dcf1
parent 4c2834dc79a7b8702c16938a4c6d094540424c11
author Shilei Tian <i@tianshilei.me> 1776703799 -0400
committer GitHub <noreply@github.com> 1776703799 +0000
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsFcBAABCAAQBQJp5lk3CRC1aQ7uu5UhlAAArAYQAK3MlR/5wo5nycuFWPzkEswj
 XzjLsHaYbPayUrb99MVerrUIOVl4svhIlXlfwqVfCfU5OucRRUn0EC+Fh7D8gmrV
 n8EjAFkgDiRAISFFisOTP2x4KUG1zPkwG+bmBdicpf8yPnHDPADjmn9FU2zycupA
 2q377Hw7zfrUarDO+Z8CpNLZw4ZIUQ1ST2Xqeyoapl1Nf5yRS0InPxVvOny09UVT
 tY/5cjYxUIf4nso5Zi64qWNIRHTYomq1PIALD9y1R7Iwn4Ik7LrrhCbqyNONNYpK
 vLUtQryeqR7i6B249k+BUDvXxhHcS/arzQzDIzsWmnk5+iJlU6z7tO12AMcHWpde
 QH3s54jhwrz8FCXloz8jfRRuMgqzI+guPJFEiPsWAcldpmUzGaifGbugh+GgpNK9
 dl3m4BupnKL6Wjg+xGwtFBxsPo0xKSoX/uWWFvqJneWI0yDq6S6vExrpbhWQ1O9+
 GeLQ03CkDKnjaZ7lSwluA/Kp93569NnR7gr1ycJzL6sPQTngYaodMzN8fc39BLDU
 hfxcYgOHVP7kE6OkEnA3cJ8mrC11oNJhWTqbUNWPlBg+eQNk5vz7B92zTyd1szli
 u5pcUDAF0cI+gqxf0CQEtJk95pLKAfnAtULROLJHsqC7V1fUpvBSBISJrxjPIqIl
 BFD5ifzwAjZrt+UJIbm7
 =Oq6N
 -----END PGP SIGNATURE-----
 

[AMDGPU] Report only local per-function resource usage when object linking is enabled (#192594)

With object linking the linker aggregates resource usage across TUs, so
compile-time pessimism and call-graph propagation duplicate the linker's
work or pollute its inputs.

In this mode, skip the per-callsite conservative bumps in
`AMDGPUResourceUsageAnalysis` and assign each resource symbol in
`AMDGPUMCResourceInfo` a concrete local constant instead of building
call-graph max/or expressions.