[PATCH] D149893: Rewrite LSV to handle longer chains.

[Artem Belevich via Phabricator via llvm-commits]

tra added inline comments.


================
Comment at: llvm/test/CodeGen/AMDGPU/GlobalISel/sdivrem.ll:1581
 ; GFX8-NEXT:    v_subb_u32_e32 v7, vcc, v7, v8, vcc
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v9, s5
----------------
This looks like some sort of barrier which was probably important. It would be great if someone familiar with AMDGPU backend could double check if the removal of this instruction is OK.


================
Comment at: llvm/test/CodeGen/AMDGPU/GlobalISel/sdivrem.ll:1890
 ; GFX9-NEXT:    v_subb_co_u32_e32 v8, vcc, v8, v9, vcc
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    global_store_dwordx4 v0, v[1:4], s[4:5]
----------------
ditto.


================
Comment at: llvm/test/CodeGen/AMDGPU/GlobalISel/sdivrem.ll:2190
 ; GFX10-NEXT:    v_subrev_co_ci_u32_e32 v7, vcc_lo, s12, v7, vcc_lo
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    global_store_dwordx4 v16, v[0:3], s[4:5]
----------------
ditto.


================
Comment at: llvm/test/CodeGen/AMDGPU/GlobalISel/udivrem.ll:1513
 ; GFX9-NEXT:    v_cndmask_b32_e64 v8, v11, v9, s[0:1]
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    global_store_dwordx4 v0, v[1:4], s[4:5]
----------------
ditto, though in this case it may have moved to line 1328 in the new version.


================
Comment at: llvm/test/CodeGen/AMDGPU/GlobalISel/udivrem.ll:1775
 ; GFX10-NEXT:    v_cndmask_b32_e64 v7, v13, v14, s1
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    global_store_dwordx4 v9, v[0:3], s[4:5]
----------------
ditto.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D149893/new/

https://reviews.llvm.org/D149893