maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[AMDGPU] Re-arrange ds_read/ds_write ISel pattern for better readability. 

Reviewed By: foad

Differential Revision: https://reviews.llvm.org/D100773
This commit is contained in:
hsmahesha 2021-04-20 16:17:05 +05:30
parent 319c9f6e58
commit 840c4e4e90
1 changed files with 4 additions and 16 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
llvm/lib/Target/AMDGPU/DSInstructions.td
View File

@ -842,9 +842,6 @@ let AddedComplexity = 100 in {
foreach vt = VReg_64.RegTypes in {
defm : DSReadPat_mc <DS_READ_B64, vt, "load_align8_local">;
}
foreach vt = VReg_64.RegTypes in {
defm : DSWritePat_mc <DS_WRITE_B64, vt, "store_align8_local">;
}
@ -852,34 +849,25 @@ let SubtargetPredicate = isGFX7Plus in {
foreach vt = VReg_96.RegTypes in {
defm : DSReadPat_mc <DS_READ_B96, vt, "load_align16_local">;
}
foreach vt = VReg_96.RegTypes in {
defm : DSWritePat_mc <DS_WRITE_B96, vt, "store_align16_local">;
}
foreach vt = VReg_128.RegTypes in {
defm : DSReadPat_mc <DS_READ_B128, vt, "load_align16_local">;
}
foreach vt = VReg_128.RegTypes in {
defm : DSWritePat_mc <DS_WRITE_B128, vt, "store_align16_local">;
}
let SubtargetPredicate = HasUnalignedAccessMode in {
// FIXME: Is ds_read_b96/ds_write_b96 better choice in unaligned-access-mode?
// FIXME: From performance point of view, is ds_read_b96/ds_write_b96 better choice
// for unaligned accesses?
foreach vt = VReg_96.RegTypes in {
defm : DSReadPat_mc <DS_READ_B96, vt, "load_local">;
}
foreach vt = VReg_96.RegTypes in {
defm : DSWritePat_mc <DS_WRITE_B96, vt, "store_local">;
}
// For performance reasons, *do not* select ds_read_b128/ds_write_b128 in
// unaligned-access-mode. At lower alignments ds_read2_b64/ds_write2_b64 is
// always a better choice.
// For performance reasons, *do not* select ds_read_b128/ds_write_b128 for unaligned
// accesses.
} // End SubtargetPredicate = HasUnalignedAccessMode