[llvm] [AMDGPU] Add s_delay_alu in relocated PC-relative symbol sequence (PR #71061)

Jay Foad via llvm-commits llvm-commits at lists.llvm.org
Thu Nov 2 06:51:22 PDT 2023


================
@@ -124,8 +124,9 @@ define amdgpu_kernel void @test_call_external_void_func_i1_imm() #0 {
 ; GFX11-NEXT:    s_mov_b64 s[6:7], s[0:1]
 ; GFX11-NEXT:    s_mov_b32 s32, 0
 ; GFX11-NEXT:    s_getpc_b64 s[2:3]
-; GFX11-NEXT:    s_add_u32 s2, s2, external_void_func_i1 at rel32@lo+4
-; GFX11-NEXT:    s_addc_u32 s3, s3, external_void_func_i1 at rel32@hi+12
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
----------------
jayfoad wrote:

Open question: should `AMDGPUInsertDelayAlu` have updated this `s_delay_alu` instruction (which is inside a bundle) instead of adding the new `s_delay_alu` instruction on line 130 right after the bundle?

https://github.com/llvm/llvm-project/pull/71061


More information about the llvm-commits mailing list