tree 81d127f59c7838bb8e4540c76a6fabe2a4dd1d19
parent 6f1347d57bdaed75b73b2013a96a4a69c8969ebe
author Brox Chen <guochen2@amd.com> 1743624506 -0400
committer GitHub <noreply@github.com> 1743624506 -0400
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsFcBAABCAAQBQJn7Zk6CRC1aQ7uu5UhlAAAbi0QAKDnLfdJF+V/OjZ9SfV7vbFM
 SmjBxIn0lginp/NSWlW6A3nUTt09sNejznFcLx5RLsLsrFP3Gda2NEZdKWUP635q
 Jc2Jh/LTk2fQ6ruC0/+8Xa/guijgUD/NWVveexuwf9V+XfwMvZQGGP8GC9yMgAIT
 bcabc+STjTCmrEh4Jh7er16+Mn8mnpnQRanxQZS1cz75pPfOfz2TfnTWd9W9YUy+
 y/Q2jm/6GW+c5yDJOhjH4Vi4jnqsAEY5NLD22zkpujuoMkZ0g7QMaoxftQitMkm8
 +MrJKnSXxQpKMsbjSV8JEF//UUS40ZtbcTDrZcxhd+izE0ax8nJoelNC61LyXtsa
 t6PvWTI81lRbil2SrL1xVzV8tIRUbHD6abWNDy8ndsVreU+/XupSAcOIPKt5JTg9
 TV+pnoyYEWd+7oMUh3OZyyvLwI8f1zpHX9hGZXPeGhN5pEuS7ghDMgBbZ/nF3xSu
 WZhaT17D+36i0Bby4YhlbfJ8FOTi4D+XTVI1aehnSKaJx2Nx2MDvETeUzz6ctgf4
 iVjZMrmM7GzAGfTZbb9TsUacWyTTCLAUtgfV284pVcAeszkWFESpumH/RgmKXVro
 /2KcWbmPcTnzXraM4b32PTijQYYFmx2e/Zr8SQ+97W76XkQe4xJIgDzsuenSzLcm
 Dnb0QsNWB7rMK9qrc2Qu
 =BUZk
 -----END PGP SIGNATURE-----
 

[AMDGPU][True16][CodeGen] Implement sgpr folding in true16 (#128929)

We haven't implemented 16 bit SGPRs. Currently allow 32-bit SGPRs to be
folded into True16 bit instructions taking 16 bit values. Also use
sgpr_32 when Imm is copied to spgr_lo16 so it could be further folded.
This improves generated code quality.