| ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 | |
| ; RUN: llc -global-isel=0 -mtriple=amdgpu12.00 < %s | FileCheck -check-prefixes=GFX12,GFX12-SDAG %s | |
| ; RUN: llc -global-isel=1 -mtriple=amdgpu12.00 < %s | FileCheck -check-prefixes=GFX12,GFX12-GISEL %s | |
| ; RUN: llc -global-isel=0 -mtriple=amdgpu13.10 < %s | FileCheck -check-prefixes=GFX13,GFX13-SDAG %s | |
| ; RUN: llc -global-isel=1 -mtriple=amdgpu13.10 < %s | FileCheck -check-prefixes=GFX13,GFX13-GISEL %s | |
| @bar = internal addrspace(15) global target("amdgcn.named.barrier", 0) poison | |
| @bar2 = internal addrspace(15) global target("amdgcn.named.barrier", 0) poison | |
| @bar3 = internal addrspace(15) global target("amdgcn.named.barrier", 0) poison | |
| ; Test using the workgroup barrier with the GV. | |
| @wgbarr = internal addrspace(15) global target("amdgcn.named.barrier", 0) poison, !absolute_symbol !0 | |
| define void @func1() { | |
| ; GFX12-SDAG-LABEL: func1: | |
| ; GFX12-SDAG: ; %bb.0: | |
| ; GFX12-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_mov_b32 m0, 3 | |
| ; GFX12-SDAG-NEXT: s_barrier_join m0 | |
| ; GFX12-SDAG-NEXT: s_mov_b32 m0, 0x70003 | |
| ; GFX12-SDAG-NEXT: s_barrier_signal m0 | |
| ; GFX12-SDAG-NEXT: s_barrier_wait 1 | |
| ; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-GISEL-LABEL: func1: | |
| ; GFX12-GISEL: ; %bb.0: | |
| ; GFX12-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_mov_b32 m0, 0x70003 | |
| ; GFX12-GISEL-NEXT: s_barrier_join 3 | |
| ; GFX12-GISEL-NEXT: s_barrier_signal m0 | |
| ; GFX12-GISEL-NEXT: s_barrier_wait 1 | |
| ; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX13-SDAG-LABEL: func1: | |
| ; GFX13-SDAG: ; %bb.0: | |
| ; GFX13-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_wait_expcnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_mov_b32 m0, 3 | |
| ; GFX13-SDAG-NEXT: s_barrier_join m0 | |
| ; GFX13-SDAG-NEXT: s_mov_b32 m0, 0x70003 | |
| ; GFX13-SDAG-NEXT: s_barrier_signal m0 | |
| ; GFX13-SDAG-NEXT: s_barrier_wait 1 | |
| ; GFX13-SDAG-NEXT: s_set_pc_i64 s[30:31] | |
| ; | |
| ; GFX13-GISEL-LABEL: func1: | |
| ; GFX13-GISEL: ; %bb.0: | |
| ; GFX13-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_wait_expcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_mov_b32 m0, 0x70003 | |
| ; GFX13-GISEL-NEXT: s_barrier_join 3 | |
| ; GFX13-GISEL-NEXT: s_barrier_signal m0 | |
| ; GFX13-GISEL-NEXT: s_barrier_wait 1 | |
| ; GFX13-GISEL-NEXT: s_set_pc_i64 s[30:31] | |
| call void @llvm.amdgcn.s.barrier.join(ptr addrspace(15) @bar3) | |
| call void @llvm.amdgcn.s.barrier.signal.var(ptr addrspace(15) @bar3, i32 7) | |
| call void @llvm.amdgcn.s.barrier.wait(i16 1) | |
| ret void | |
| } | |
| define void @func2() { | |
| ; GFX12-SDAG-LABEL: func2: | |
| ; GFX12-SDAG: ; %bb.0: | |
| ; GFX12-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_mov_b32 m0, 2 | |
| ; GFX12-SDAG-NEXT: s_barrier_join m0 | |
| ; GFX12-SDAG-NEXT: s_mov_b32 m0, 0x70002 | |
| ; GFX12-SDAG-NEXT: s_barrier_signal m0 | |
| ; GFX12-SDAG-NEXT: s_barrier_wait 1 | |
| ; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-GISEL-LABEL: func2: | |
| ; GFX12-GISEL: ; %bb.0: | |
| ; GFX12-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_mov_b32 m0, 0x70002 | |
| ; GFX12-GISEL-NEXT: s_barrier_join 2 | |
| ; GFX12-GISEL-NEXT: s_barrier_signal m0 | |
| ; GFX12-GISEL-NEXT: s_barrier_wait 1 | |
| ; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX13-SDAG-LABEL: func2: | |
| ; GFX13-SDAG: ; %bb.0: | |
| ; GFX13-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_wait_expcnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_mov_b32 m0, 2 | |
| ; GFX13-SDAG-NEXT: s_barrier_join m0 | |
| ; GFX13-SDAG-NEXT: s_mov_b32 m0, 0x70002 | |
| ; GFX13-SDAG-NEXT: s_barrier_signal m0 | |
| ; GFX13-SDAG-NEXT: s_barrier_wait 1 | |
| ; GFX13-SDAG-NEXT: s_set_pc_i64 s[30:31] | |
| ; | |
| ; GFX13-GISEL-LABEL: func2: | |
| ; GFX13-GISEL: ; %bb.0: | |
| ; GFX13-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_wait_expcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_mov_b32 m0, 0x70002 | |
| ; GFX13-GISEL-NEXT: s_barrier_join 2 | |
| ; GFX13-GISEL-NEXT: s_barrier_signal m0 | |
| ; GFX13-GISEL-NEXT: s_barrier_wait 1 | |
| ; GFX13-GISEL-NEXT: s_set_pc_i64 s[30:31] | |
| call void @llvm.amdgcn.s.barrier.join(ptr addrspace(15) @bar2) | |
| call void @llvm.amdgcn.s.barrier.signal.var(ptr addrspace(15) @bar2, i32 7) | |
| call void @llvm.amdgcn.s.barrier.wait(i16 1) | |
| ret void | |
| } | |
| define amdgpu_kernel void @kernel1(ptr addrspace(1) %out, ptr addrspace(15) %in) #0 { | |
| ; GFX12-SDAG-LABEL: kernel1: | |
| ; GFX12-SDAG: ; %bb.0: | |
| ; GFX12-SDAG-NEXT: s_mov_b64 s[10:11], s[6:7] | |
| ; GFX12-SDAG-NEXT: s_mov_b64 s[6:7], s[2:3] | |
| ; GFX12-SDAG-NEXT: s_load_b32 s2, s[4:5], 0x2c | |
| ; GFX12-SDAG-NEXT: s_mov_b32 m0, 0xc0001 | |
| ; GFX12-SDAG-NEXT: v_mov_b32_e32 v31, v0 | |
| ; GFX12-SDAG-NEXT: s_barrier_init m0 | |
| ; GFX12-SDAG-NEXT: s_add_nc_u64 s[8:9], s[4:5], 48 | |
| ; GFX12-SDAG-NEXT: s_mov_b64 s[4:5], s[0:1] | |
| ; GFX12-SDAG-NEXT: s_mov_b32 s32, 0 | |
| ; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_and_b32 s2, s2, 63 | |
| ; GFX12-SDAG-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | |
| ; GFX12-SDAG-NEXT: s_or_b32 s3, s2, 0x90000 | |
| ; GFX12-SDAG-NEXT: s_cmp_eq_u32 0, 0 | |
| ; GFX12-SDAG-NEXT: s_mov_b32 m0, s3 | |
| ; GFX12-SDAG-NEXT: s_barrier_init m0 | |
| ; GFX12-SDAG-NEXT: s_mov_b32 m0, 0xc0001 | |
| ; GFX12-SDAG-NEXT: s_barrier_signal m0 | |
| ; GFX12-SDAG-NEXT: s_mov_b32 m0, s3 | |
| ; GFX12-SDAG-NEXT: s_barrier_signal m0 | |
| ; GFX12-SDAG-NEXT: s_mov_b32 m0, s2 | |
| ; GFX12-SDAG-NEXT: s_barrier_signal -1 | |
| ; GFX12-SDAG-NEXT: s_barrier_join m0 | |
| ; GFX12-SDAG-NEXT: s_barrier_signal_isfirst -1 | |
| ; GFX12-SDAG-NEXT: s_mov_b32 m0, 1 | |
| ; GFX12-SDAG-NEXT: s_barrier_wait 1 | |
| ; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_barrier_leave | |
| ; GFX12-SDAG-NEXT: s_get_barrier_state s3, m0 | |
| ; GFX12-SDAG-NEXT: s_mov_b32 m0, s2 | |
| ; GFX12-SDAG-NEXT: s_get_barrier_state s2, m0 | |
| ; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_getpc_b64 s[2:3] | |
| ; GFX12-SDAG-NEXT: s_sext_i32_i16 s3, s3 | |
| ; GFX12-SDAG-NEXT: s_add_co_u32 s2, s2, func1@gotpcrel32@lo+8 | |
| ; GFX12-SDAG-NEXT: s_add_co_ci_u32 s3, s3, func1@gotpcrel32@hi+16 | |
| ; GFX12-SDAG-NEXT: s_barrier_signal -1 | |
| ; GFX12-SDAG-NEXT: s_load_b64 s[2:3], s[2:3], 0x0 | |
| ; GFX12-SDAG-NEXT: s_barrier_wait -1 | |
| ; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_swappc_b64 s[30:31], s[2:3] | |
| ; GFX12-SDAG-NEXT: s_getpc_b64 s[2:3] | |
| ; GFX12-SDAG-NEXT: s_wait_alu depctr_sa_sdst(0) | |
| ; GFX12-SDAG-NEXT: s_sext_i32_i16 s3, s3 | |
| ; GFX12-SDAG-NEXT: s_add_co_u32 s2, s2, func2@gotpcrel32@lo+12 | |
| ; GFX12-SDAG-NEXT: s_wait_alu depctr_sa_sdst(0) | |
| ; GFX12-SDAG-NEXT: s_add_co_ci_u32 s3, s3, func2@gotpcrel32@hi+24 | |
| ; GFX12-SDAG-NEXT: s_load_b64 s[2:3], s[2:3], 0x0 | |
| ; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_swappc_b64 s[30:31], s[2:3] | |
| ; GFX12-SDAG-NEXT: s_get_barrier_state s0, -1 | |
| ; GFX12-SDAG-NEXT: s_endpgm | |
| ; | |
| ; GFX12-GISEL-LABEL: kernel1: | |
| ; GFX12-GISEL: ; %bb.0: | |
| ; GFX12-GISEL-NEXT: s_mov_b64 s[12:13], s[4:5] | |
| ; GFX12-GISEL-NEXT: s_mov_b64 s[4:5], s[0:1] | |
| ; GFX12-GISEL-NEXT: s_load_b32 s0, s[12:13], 0x2c | |
| ; GFX12-GISEL-NEXT: s_mov_b32 m0, 0xc0001 | |
| ; GFX12-GISEL-NEXT: v_mov_b32_e32 v31, v0 | |
| ; GFX12-GISEL-NEXT: s_barrier_init m0 | |
| ; GFX12-GISEL-NEXT: s_mov_b64 s[10:11], s[6:7] | |
| ; GFX12-GISEL-NEXT: s_mov_b64 s[6:7], s[2:3] | |
| ; GFX12-GISEL-NEXT: s_mov_b32 s32, 0 | |
| ; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_and_b32 s0, s0, 63 | |
| ; GFX12-GISEL-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | |
| ; GFX12-GISEL-NEXT: s_or_b32 s1, s0, 0x90000 | |
| ; GFX12-GISEL-NEXT: s_cmp_eq_u32 0, 0 | |
| ; GFX12-GISEL-NEXT: s_mov_b32 m0, s1 | |
| ; GFX12-GISEL-NEXT: s_barrier_init m0 | |
| ; GFX12-GISEL-NEXT: s_mov_b32 m0, 0xc0001 | |
| ; GFX12-GISEL-NEXT: s_barrier_signal m0 | |
| ; GFX12-GISEL-NEXT: s_mov_b32 m0, s1 | |
| ; GFX12-GISEL-NEXT: s_barrier_signal m0 | |
| ; GFX12-GISEL-NEXT: s_mov_b32 m0, s0 | |
| ; GFX12-GISEL-NEXT: s_barrier_signal -1 | |
| ; GFX12-GISEL-NEXT: s_barrier_join m0 | |
| ; GFX12-GISEL-NEXT: s_barrier_signal_isfirst -1 | |
| ; GFX12-GISEL-NEXT: s_barrier_wait 1 | |
| ; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_barrier_leave | |
| ; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_add_co_u32 s8, s12, 48 | |
| ; GFX12-GISEL-NEXT: s_get_barrier_state s0, 1 | |
| ; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_get_barrier_state s0, m0 | |
| ; GFX12-GISEL-NEXT: s_add_co_ci_u32 s9, s13, 0 | |
| ; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_getpc_b64 s[0:1] | |
| ; GFX12-GISEL-NEXT: s_sext_i32_i16 s1, s1 | |
| ; GFX12-GISEL-NEXT: s_add_co_u32 s0, s0, func1@gotpcrel32@lo+8 | |
| ; GFX12-GISEL-NEXT: s_add_co_ci_u32 s1, s1, func1@gotpcrel32@hi+16 | |
| ; GFX12-GISEL-NEXT: s_barrier_signal -1 | |
| ; GFX12-GISEL-NEXT: s_load_b64 s[0:1], s[0:1], 0x0 | |
| ; GFX12-GISEL-NEXT: s_barrier_wait -1 | |
| ; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_swappc_b64 s[30:31], s[0:1] | |
| ; GFX12-GISEL-NEXT: s_add_co_u32 s8, s12, 48 | |
| ; GFX12-GISEL-NEXT: s_add_co_ci_u32 s9, s13, 0 | |
| ; GFX12-GISEL-NEXT: s_getpc_b64 s[0:1] | |
| ; GFX12-GISEL-NEXT: s_wait_alu depctr_sa_sdst(0) | |
| ; GFX12-GISEL-NEXT: s_sext_i32_i16 s1, s1 | |
| ; GFX12-GISEL-NEXT: s_add_co_u32 s0, s0, func2@gotpcrel32@lo+12 | |
| ; GFX12-GISEL-NEXT: s_wait_alu depctr_sa_sdst(0) | |
| ; GFX12-GISEL-NEXT: s_add_co_ci_u32 s1, s1, func2@gotpcrel32@hi+24 | |
| ; GFX12-GISEL-NEXT: s_load_b64 s[0:1], s[0:1], 0x0 | |
| ; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_swappc_b64 s[30:31], s[0:1] | |
| ; GFX12-GISEL-NEXT: s_get_barrier_state s0, -1 | |
| ; GFX12-GISEL-NEXT: s_endpgm | |
| ; | |
| ; GFX13-SDAG-LABEL: kernel1: | |
| ; GFX13-SDAG: ; %bb.0: | |
| ; GFX13-SDAG-NEXT: s_mov_b64 s[10:11], s[6:7] | |
| ; GFX13-SDAG-NEXT: s_mov_b64 s[6:7], s[2:3] | |
| ; GFX13-SDAG-NEXT: s_load_b32 s2, s[4:5], 0x2c nv | |
| ; GFX13-SDAG-NEXT: s_mov_b32 m0, 0xc0001 | |
| ; GFX13-SDAG-NEXT: v_mov_b32_e32 v31, v0 | |
| ; GFX13-SDAG-NEXT: s_barrier_init m0 | |
| ; GFX13-SDAG-NEXT: s_add_nc_u64 s[8:9], s[4:5], 48 | |
| ; GFX13-SDAG-NEXT: s_mov_b64 s[4:5], s[0:1] | |
| ; GFX13-SDAG-NEXT: s_mov_b32 s32, 0 | |
| ; GFX13-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_and_b32 s2, s2, 63 | |
| ; GFX13-SDAG-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | |
| ; GFX13-SDAG-NEXT: s_or_b32 s3, s2, 0x90000 | |
| ; GFX13-SDAG-NEXT: s_cmp_eq_u32 0, 0 | |
| ; GFX13-SDAG-NEXT: s_mov_b32 m0, s3 | |
| ; GFX13-SDAG-NEXT: s_barrier_init m0 | |
| ; GFX13-SDAG-NEXT: s_mov_b32 m0, 0xc0001 | |
| ; GFX13-SDAG-NEXT: s_barrier_signal m0 | |
| ; GFX13-SDAG-NEXT: s_mov_b32 m0, s3 | |
| ; GFX13-SDAG-NEXT: s_barrier_signal m0 | |
| ; GFX13-SDAG-NEXT: s_mov_b32 m0, s2 | |
| ; GFX13-SDAG-NEXT: s_barrier_signal -1 | |
| ; GFX13-SDAG-NEXT: s_barrier_join m0 | |
| ; GFX13-SDAG-NEXT: s_barrier_signal_isfirst -1 | |
| ; GFX13-SDAG-NEXT: s_mov_b32 m0, 1 | |
| ; GFX13-SDAG-NEXT: s_barrier_wait 1 | |
| ; GFX13-SDAG-NEXT: s_barrier_leave 1 | |
| ; GFX13-SDAG-NEXT: s_get_barrier_state s3, m0 | |
| ; GFX13-SDAG-NEXT: s_mov_b32 m0, s2 | |
| ; GFX13-SDAG-NEXT: s_get_barrier_state s2, m0 | |
| ; GFX13-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_get_pc_i64 s[2:3] | |
| ; GFX13-SDAG-NEXT: s_add_nc_u64 s[2:3], s[2:3], func1@gotpcrel+4 | |
| ; GFX13-SDAG-NEXT: s_barrier_signal -1 | |
| ; GFX13-SDAG-NEXT: s_load_b64 s[2:3], s[2:3], 0x0 nv | |
| ; GFX13-SDAG-NEXT: s_barrier_wait -1 | |
| ; GFX13-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_swap_pc_i64 s[30:31], s[2:3] | |
| ; GFX13-SDAG-NEXT: s_get_pc_i64 s[2:3] | |
| ; GFX13-SDAG-NEXT: s_add_nc_u64 s[2:3], s[2:3], func2@gotpcrel+4 | |
| ; GFX13-SDAG-NEXT: s_load_b64 s[2:3], s[2:3], 0x0 nv | |
| ; GFX13-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_swap_pc_i64 s[30:31], s[2:3] | |
| ; GFX13-SDAG-NEXT: s_get_barrier_state s0, -1 | |
| ; GFX13-SDAG-NEXT: s_endpgm | |
| ; | |
| ; GFX13-GISEL-LABEL: kernel1: | |
| ; GFX13-GISEL: ; %bb.0: | |
| ; GFX13-GISEL-NEXT: s_mov_b64 s[12:13], s[4:5] | |
| ; GFX13-GISEL-NEXT: s_mov_b64 s[4:5], s[0:1] | |
| ; GFX13-GISEL-NEXT: s_load_b32 s0, s[12:13], 0x2c nv | |
| ; GFX13-GISEL-NEXT: s_mov_b32 m0, 0xc0001 | |
| ; GFX13-GISEL-NEXT: v_mov_b32_e32 v31, v0 | |
| ; GFX13-GISEL-NEXT: s_barrier_init m0 | |
| ; GFX13-GISEL-NEXT: s_mov_b64 s[10:11], s[6:7] | |
| ; GFX13-GISEL-NEXT: s_mov_b64 s[6:7], s[2:3] | |
| ; GFX13-GISEL-NEXT: s_mov_b32 s32, 0 | |
| ; GFX13-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_and_b32 s0, s0, 63 | |
| ; GFX13-GISEL-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | |
| ; GFX13-GISEL-NEXT: s_or_b32 s1, s0, 0x90000 | |
| ; GFX13-GISEL-NEXT: s_cmp_eq_u32 0, 0 | |
| ; GFX13-GISEL-NEXT: s_mov_b32 m0, s1 | |
| ; GFX13-GISEL-NEXT: s_barrier_init m0 | |
| ; GFX13-GISEL-NEXT: s_mov_b32 m0, 0xc0001 | |
| ; GFX13-GISEL-NEXT: s_barrier_signal m0 | |
| ; GFX13-GISEL-NEXT: s_mov_b32 m0, s1 | |
| ; GFX13-GISEL-NEXT: s_barrier_signal m0 | |
| ; GFX13-GISEL-NEXT: s_mov_b32 m0, s0 | |
| ; GFX13-GISEL-NEXT: s_barrier_signal -1 | |
| ; GFX13-GISEL-NEXT: s_barrier_join m0 | |
| ; GFX13-GISEL-NEXT: s_barrier_signal_isfirst -1 | |
| ; GFX13-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_add_co_u32 s8, s12, 48 | |
| ; GFX13-GISEL-NEXT: s_add_co_ci_u32 s9, s13, 0 | |
| ; GFX13-GISEL-NEXT: s_barrier_wait 1 | |
| ; GFX13-GISEL-NEXT: s_barrier_leave 1 | |
| ; GFX13-GISEL-NEXT: s_get_barrier_state s0, 1 | |
| ; GFX13-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_get_barrier_state s0, m0 | |
| ; GFX13-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_get_pc_i64 s[0:1] | |
| ; GFX13-GISEL-NEXT: s_add_nc_u64 s[0:1], s[0:1], func1@gotpcrel+4 | |
| ; GFX13-GISEL-NEXT: s_barrier_signal -1 | |
| ; GFX13-GISEL-NEXT: s_load_b64 s[0:1], s[0:1], 0x0 nv | |
| ; GFX13-GISEL-NEXT: s_barrier_wait -1 | |
| ; GFX13-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_swap_pc_i64 s[30:31], s[0:1] | |
| ; GFX13-GISEL-NEXT: s_get_pc_i64 s[0:1] | |
| ; GFX13-GISEL-NEXT: s_add_nc_u64 s[0:1], s[0:1], func2@gotpcrel+4 | |
| ; GFX13-GISEL-NEXT: s_add_co_u32 s8, s12, 48 | |
| ; GFX13-GISEL-NEXT: s_load_b64 s[0:1], s[0:1], 0x0 nv | |
| ; GFX13-GISEL-NEXT: s_add_co_ci_u32 s9, s13, 0 | |
| ; GFX13-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_swap_pc_i64 s[30:31], s[0:1] | |
| ; GFX13-GISEL-NEXT: s_get_barrier_state s0, -1 | |
| ; GFX13-GISEL-NEXT: s_endpgm | |
| call void @llvm.amdgcn.s.barrier.init(ptr addrspace(15) @bar, i32 12) | |
| call void @llvm.amdgcn.s.barrier.init(ptr addrspace(15) %in, i32 9) | |
| call void @llvm.amdgcn.s.barrier.signal.var(ptr addrspace(15) @bar, i32 12) | |
| call void @llvm.amdgcn.s.barrier.signal.var(ptr addrspace(15) %in, i32 9) | |
| call void @llvm.amdgcn.s.barrier.signal(i32 -1) | |
| call void @llvm.amdgcn.s.barrier.join(ptr addrspace(15) %in) | |
| %isfirst = call i1 @llvm.amdgcn.s.barrier.signal.isfirst(i32 -1) | |
| call void @llvm.amdgcn.s.barrier.wait(i16 1) | |
| call void @llvm.amdgcn.s.barrier.leave(i16 1) | |
| %state = call i32 @llvm.amdgcn.s.get.named.barrier.state(ptr addrspace(15) @bar) | |
| %state2 = call i32 @llvm.amdgcn.s.get.named.barrier.state(ptr addrspace(15) %in) | |
| call void @llvm.amdgcn.s.barrier() | |
| call void @func1() | |
| call void @func2() | |
| %state3 = call i32 @llvm.amdgcn.s.get.barrier.state(i32 -1) | |
| ret void | |
| } | |
| define amdgpu_kernel void @kernel2(ptr addrspace(1) %out, ptr addrspace(15) %in) #0 { | |
| ; GFX12-SDAG-LABEL: kernel2: | |
| ; GFX12-SDAG: ; %bb.0: | |
| ; GFX12-SDAG-NEXT: s_mov_b64 s[10:11], s[6:7] | |
| ; GFX12-SDAG-NEXT: s_getpc_b64 s[6:7] | |
| ; GFX12-SDAG-NEXT: s_sext_i32_i16 s7, s7 | |
| ; GFX12-SDAG-NEXT: s_add_co_u32 s6, s6, func2@gotpcrel32@lo+8 | |
| ; GFX12-SDAG-NEXT: s_add_co_ci_u32 s7, s7, func2@gotpcrel32@hi+16 | |
| ; GFX12-SDAG-NEXT: v_mov_b32_e32 v31, v0 | |
| ; GFX12-SDAG-NEXT: s_load_b64 s[12:13], s[6:7], 0x0 | |
| ; GFX12-SDAG-NEXT: s_mov_b32 m0, 0x70001 | |
| ; GFX12-SDAG-NEXT: s_add_nc_u64 s[8:9], s[4:5], 48 | |
| ; GFX12-SDAG-NEXT: s_barrier_signal m0 | |
| ; GFX12-SDAG-NEXT: s_mov_b32 m0, 1 | |
| ; GFX12-SDAG-NEXT: s_mov_b64 s[4:5], s[0:1] | |
| ; GFX12-SDAG-NEXT: s_mov_b64 s[6:7], s[2:3] | |
| ; GFX12-SDAG-NEXT: s_mov_b32 s32, 0 | |
| ; GFX12-SDAG-NEXT: s_barrier_join m0 | |
| ; GFX12-SDAG-NEXT: s_barrier_wait 1 | |
| ; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_swappc_b64 s[30:31], s[12:13] | |
| ; GFX12-SDAG-NEXT: s_endpgm | |
| ; | |
| ; GFX12-GISEL-LABEL: kernel2: | |
| ; GFX12-GISEL: ; %bb.0: | |
| ; GFX12-GISEL-NEXT: s_add_co_u32 s8, s4, 48 | |
| ; GFX12-GISEL-NEXT: s_add_co_ci_u32 s9, s5, 0 | |
| ; GFX12-GISEL-NEXT: s_getpc_b64 s[4:5] | |
| ; GFX12-GISEL-NEXT: s_sext_i32_i16 s5, s5 | |
| ; GFX12-GISEL-NEXT: s_add_co_u32 s4, s4, func2@gotpcrel32@lo+8 | |
| ; GFX12-GISEL-NEXT: s_add_co_ci_u32 s5, s5, func2@gotpcrel32@hi+16 | |
| ; GFX12-GISEL-NEXT: v_mov_b32_e32 v31, v0 | |
| ; GFX12-GISEL-NEXT: s_load_b64 s[12:13], s[4:5], 0x0 | |
| ; GFX12-GISEL-NEXT: s_mov_b64 s[10:11], s[6:7] | |
| ; GFX12-GISEL-NEXT: s_mov_b32 m0, 0x70001 | |
| ; GFX12-GISEL-NEXT: s_mov_b64 s[4:5], s[0:1] | |
| ; GFX12-GISEL-NEXT: s_mov_b64 s[6:7], s[2:3] | |
| ; GFX12-GISEL-NEXT: s_mov_b32 s32, 0 | |
| ; GFX12-GISEL-NEXT: s_barrier_signal m0 | |
| ; GFX12-GISEL-NEXT: s_barrier_join 1 | |
| ; GFX12-GISEL-NEXT: s_barrier_wait 1 | |
| ; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_swappc_b64 s[30:31], s[12:13] | |
| ; GFX12-GISEL-NEXT: s_endpgm | |
| ; | |
| ; GFX13-SDAG-LABEL: kernel2: | |
| ; GFX13-SDAG: ; %bb.0: | |
| ; GFX13-SDAG-NEXT: s_mov_b64 s[10:11], s[6:7] | |
| ; GFX13-SDAG-NEXT: s_get_pc_i64 s[6:7] | |
| ; GFX13-SDAG-NEXT: s_add_nc_u64 s[6:7], s[6:7], func2@gotpcrel+4 | |
| ; GFX13-SDAG-NEXT: v_mov_b32_e32 v31, v0 | |
| ; GFX13-SDAG-NEXT: s_load_b64 s[12:13], s[6:7], 0x0 nv | |
| ; GFX13-SDAG-NEXT: s_mov_b32 m0, 0x70001 | |
| ; GFX13-SDAG-NEXT: s_add_nc_u64 s[8:9], s[4:5], 48 | |
| ; GFX13-SDAG-NEXT: s_barrier_signal m0 | |
| ; GFX13-SDAG-NEXT: s_mov_b32 m0, 1 | |
| ; GFX13-SDAG-NEXT: s_mov_b64 s[4:5], s[0:1] | |
| ; GFX13-SDAG-NEXT: s_mov_b64 s[6:7], s[2:3] | |
| ; GFX13-SDAG-NEXT: s_mov_b32 s32, 0 | |
| ; GFX13-SDAG-NEXT: s_barrier_join m0 | |
| ; GFX13-SDAG-NEXT: s_barrier_wait 1 | |
| ; GFX13-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_swap_pc_i64 s[30:31], s[12:13] | |
| ; GFX13-SDAG-NEXT: s_endpgm | |
| ; | |
| ; GFX13-GISEL-LABEL: kernel2: | |
| ; GFX13-GISEL: ; %bb.0: | |
| ; GFX13-GISEL-NEXT: s_mov_b64 s[10:11], s[6:7] | |
| ; GFX13-GISEL-NEXT: s_get_pc_i64 s[6:7] | |
| ; GFX13-GISEL-NEXT: s_add_nc_u64 s[6:7], s[6:7], func2@gotpcrel+4 | |
| ; GFX13-GISEL-NEXT: v_mov_b32_e32 v31, v0 | |
| ; GFX13-GISEL-NEXT: s_load_b64 s[12:13], s[6:7], 0x0 nv | |
| ; GFX13-GISEL-NEXT: s_add_co_u32 s8, s4, 48 | |
| ; GFX13-GISEL-NEXT: s_mov_b32 m0, 0x70001 | |
| ; GFX13-GISEL-NEXT: s_add_co_ci_u32 s9, s5, 0 | |
| ; GFX13-GISEL-NEXT: s_mov_b64 s[4:5], s[0:1] | |
| ; GFX13-GISEL-NEXT: s_mov_b64 s[6:7], s[2:3] | |
| ; GFX13-GISEL-NEXT: s_mov_b32 s32, 0 | |
| ; GFX13-GISEL-NEXT: s_barrier_signal m0 | |
| ; GFX13-GISEL-NEXT: s_barrier_join 1 | |
| ; GFX13-GISEL-NEXT: s_barrier_wait 1 | |
| ; GFX13-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_swap_pc_i64 s[30:31], s[12:13] | |
| ; GFX13-GISEL-NEXT: s_endpgm | |
| call void @llvm.amdgcn.s.barrier.signal.var(ptr addrspace(15) @bar, i32 7) | |
| call void @llvm.amdgcn.s.barrier.join(ptr addrspace(15) @bar) | |
| call void @llvm.amdgcn.s.barrier.wait(i16 1) | |
| call void @func2() | |
| ret void | |
| } | |
| define void @signal_var_cnt0_const_bar() { | |
| ; GFX12-LABEL: signal_var_cnt0_const_bar: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_barrier_signal 1 | |
| ; GFX12-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX13-LABEL: signal_var_cnt0_const_bar: | |
| ; GFX13: ; %bb.0: | |
| ; GFX13-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX13-NEXT: s_wait_expcnt 0x0 | |
| ; GFX13-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX13-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX13-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-NEXT: s_barrier_signal 1 | |
| ; GFX13-NEXT: s_set_pc_i64 s[30:31] | |
| call void @llvm.amdgcn.s.barrier.signal.var(ptr addrspace(15) @bar, i32 0) | |
| ret void | |
| } | |
| define void @signal_var_cnt0_dynamic_bar(ptr addrspace(15) inreg %bar) { | |
| ; GFX12-LABEL: signal_var_cnt0_dynamic_bar: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_and_b32 m0, s0, 63 | |
| ; GFX12-NEXT: s_barrier_signal m0 | |
| ; GFX12-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX13-LABEL: signal_var_cnt0_dynamic_bar: | |
| ; GFX13: ; %bb.0: | |
| ; GFX13-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX13-NEXT: s_wait_expcnt 0x0 | |
| ; GFX13-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX13-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX13-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-NEXT: s_and_b32 m0, s0, 63 | |
| ; GFX13-NEXT: s_barrier_signal m0 | |
| ; GFX13-NEXT: s_set_pc_i64 s[30:31] | |
| call void @llvm.amdgcn.s.barrier.signal.var(ptr addrspace(15) %bar, i32 0) | |
| ret void | |
| } | |
| define void @barrier_init_dynamic_cnt(ptr addrspace(15) inreg %bar, i32 inreg %cnt) { | |
| ; GFX12-SDAG-LABEL: barrier_init_dynamic_cnt: | |
| ; GFX12-SDAG: ; %bb.0: | |
| ; GFX12-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_and_b32 s1, s1, 63 | |
| ; GFX12-SDAG-NEXT: s_and_b32 s0, s0, 63 | |
| ; GFX12-SDAG-NEXT: s_wait_alu depctr_sa_sdst(0) | |
| ; GFX12-SDAG-NEXT: s_lshl_b32 s1, s1, 16 | |
| ; GFX12-SDAG-NEXT: s_wait_alu depctr_sa_sdst(0) | |
| ; GFX12-SDAG-NEXT: s_or_b32 m0, s1, s0 | |
| ; GFX12-SDAG-NEXT: s_barrier_init m0 | |
| ; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-GISEL-LABEL: barrier_init_dynamic_cnt: | |
| ; GFX12-GISEL: ; %bb.0: | |
| ; GFX12-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_and_b32 s1, s1, 63 | |
| ; GFX12-GISEL-NEXT: s_and_b32 s0, s0, 63 | |
| ; GFX12-GISEL-NEXT: s_wait_alu depctr_sa_sdst(0) | |
| ; GFX12-GISEL-NEXT: s_lshl_b32 s1, s1, 16 | |
| ; GFX12-GISEL-NEXT: s_wait_alu depctr_sa_sdst(0) | |
| ; GFX12-GISEL-NEXT: s_or_b32 m0, s0, s1 | |
| ; GFX12-GISEL-NEXT: s_barrier_init m0 | |
| ; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX13-SDAG-LABEL: barrier_init_dynamic_cnt: | |
| ; GFX13-SDAG: ; %bb.0: | |
| ; GFX13-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_wait_expcnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_and_b32 s1, s1, 0xff | |
| ; GFX13-SDAG-NEXT: s_and_b32 s0, s0, 63 | |
| ; GFX13-SDAG-NEXT: s_lshl_b32 s1, s1, 16 | |
| ; GFX13-SDAG-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | |
| ; GFX13-SDAG-NEXT: s_or_b32 m0, s1, s0 | |
| ; GFX13-SDAG-NEXT: s_barrier_init m0 | |
| ; GFX13-SDAG-NEXT: s_set_pc_i64 s[30:31] | |
| ; | |
| ; GFX13-GISEL-LABEL: barrier_init_dynamic_cnt: | |
| ; GFX13-GISEL: ; %bb.0: | |
| ; GFX13-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_wait_expcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_and_b32 s1, s1, 0xff | |
| ; GFX13-GISEL-NEXT: s_and_b32 s0, s0, 63 | |
| ; GFX13-GISEL-NEXT: s_lshl_b32 s1, s1, 16 | |
| ; GFX13-GISEL-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | |
| ; GFX13-GISEL-NEXT: s_or_b32 m0, s0, s1 | |
| ; GFX13-GISEL-NEXT: s_barrier_init m0 | |
| ; GFX13-GISEL-NEXT: s_set_pc_i64 s[30:31] | |
| call void @llvm.amdgcn.s.barrier.init(ptr addrspace(15) %bar, i32 %cnt) | |
| ret void | |
| } | |
| define void @signal_var_dynamic_cnt(ptr addrspace(15) inreg %bar, i32 inreg %cnt) { | |
| ; GFX12-SDAG-LABEL: signal_var_dynamic_cnt: | |
| ; GFX12-SDAG: ; %bb.0: | |
| ; GFX12-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-SDAG-NEXT: s_and_b32 s1, s1, 63 | |
| ; GFX12-SDAG-NEXT: s_and_b32 s0, s0, 63 | |
| ; GFX12-SDAG-NEXT: s_wait_alu depctr_sa_sdst(0) | |
| ; GFX12-SDAG-NEXT: s_lshl_b32 s1, s1, 16 | |
| ; GFX12-SDAG-NEXT: s_wait_alu depctr_sa_sdst(0) | |
| ; GFX12-SDAG-NEXT: s_or_b32 m0, s1, s0 | |
| ; GFX12-SDAG-NEXT: s_barrier_signal m0 | |
| ; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-GISEL-LABEL: signal_var_dynamic_cnt: | |
| ; GFX12-GISEL: ; %bb.0: | |
| ; GFX12-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-GISEL-NEXT: s_and_b32 s1, s1, 63 | |
| ; GFX12-GISEL-NEXT: s_and_b32 s0, s0, 63 | |
| ; GFX12-GISEL-NEXT: s_wait_alu depctr_sa_sdst(0) | |
| ; GFX12-GISEL-NEXT: s_lshl_b32 s1, s1, 16 | |
| ; GFX12-GISEL-NEXT: s_wait_alu depctr_sa_sdst(0) | |
| ; GFX12-GISEL-NEXT: s_or_b32 m0, s0, s1 | |
| ; GFX12-GISEL-NEXT: s_barrier_signal m0 | |
| ; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX13-SDAG-LABEL: signal_var_dynamic_cnt: | |
| ; GFX13-SDAG: ; %bb.0: | |
| ; GFX13-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_wait_expcnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-SDAG-NEXT: s_and_b32 s1, s1, 0xff | |
| ; GFX13-SDAG-NEXT: s_and_b32 s0, s0, 63 | |
| ; GFX13-SDAG-NEXT: s_lshl_b32 s1, s1, 16 | |
| ; GFX13-SDAG-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | |
| ; GFX13-SDAG-NEXT: s_or_b32 m0, s1, s0 | |
| ; GFX13-SDAG-NEXT: s_barrier_signal m0 | |
| ; GFX13-SDAG-NEXT: s_set_pc_i64 s[30:31] | |
| ; | |
| ; GFX13-GISEL-LABEL: signal_var_dynamic_cnt: | |
| ; GFX13-GISEL: ; %bb.0: | |
| ; GFX13-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_wait_expcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX13-GISEL-NEXT: s_and_b32 s1, s1, 0xff | |
| ; GFX13-GISEL-NEXT: s_and_b32 s0, s0, 63 | |
| ; GFX13-GISEL-NEXT: s_lshl_b32 s1, s1, 16 | |
| ; GFX13-GISEL-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | |
| ; GFX13-GISEL-NEXT: s_or_b32 m0, s0, s1 | |
| ; GFX13-GISEL-NEXT: s_barrier_signal m0 | |
| ; GFX13-GISEL-NEXT: s_set_pc_i64 s[30:31] | |
| call void @llvm.amdgcn.s.barrier.signal.var(ptr addrspace(15) %bar, i32 %cnt) | |
| ret void | |
| } | |
| define amdgpu_ps void @test_barrier_leave_write_to_scc(i32 inreg %val, ptr addrspace(1) %out) { | |
| ; GFX12-LABEL: test_barrier_leave_write_to_scc: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_barrier_leave | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_cmp_lg_u32 s0, 0 | |
| ; GFX12-NEXT: s_movk_i32 s0, 0x7b | |
| ; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1) | |
| ; GFX12-NEXT: s_cselect_b32 s0, s0, 0x1c8 | |
| ; GFX12-NEXT: v_mov_b32_e32 v2, s0 | |
| ; GFX12-NEXT: global_store_b32 v[0:1], v2, off | |
| ; GFX12-NEXT: s_endpgm | |
| ; | |
| ; GFX13-LABEL: test_barrier_leave_write_to_scc: | |
| ; GFX13: ; %bb.0: | |
| ; GFX13-NEXT: s_cmp_lg_u32 s0, 0 | |
| ; GFX13-NEXT: s_movk_i32 s0, 0x7b | |
| ; GFX13-NEXT: s_barrier_leave 1 | |
| ; GFX13-NEXT: s_cselect_b32 s0, s0, 0x1c8 | |
| ; GFX13-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | |
| ; GFX13-NEXT: v_mov_b32_e32 v2, s0 | |
| ; GFX13-NEXT: global_store_b32 v[0:1], v2, off | |
| ; GFX13-NEXT: s_endpgm | |
| call void @llvm.amdgcn.s.barrier.leave(i16 1) | |
| %cmp = icmp ne i32 %val, 0 | |
| %ret = select i1 %cmp, i32 123, i32 456 | |
| store i32 %ret, ptr addrspace(1) %out | |
| ret void | |
| } | |
| define amdgpu_kernel void @wgbarr_as_gv() { | |
| ; GFX12-LABEL: wgbarr_as_gv: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_mov_b32 m0, 0x7003f | |
| ; GFX12-NEXT: s_barrier_signal m0 | |
| ; GFX12-NEXT: s_barrier_wait -1 | |
| ; GFX12-NEXT: s_endpgm | |
| ; | |
| ; GFX13-LABEL: wgbarr_as_gv: | |
| ; GFX13: ; %bb.0: | |
| ; GFX13-NEXT: s_mov_b32 m0, 0x7003f | |
| ; GFX13-NEXT: s_barrier_signal m0 | |
| ; GFX13-NEXT: s_barrier_wait -1 | |
| ; GFX13-NEXT: s_endpgm | |
| call void @llvm.amdgcn.s.barrier.signal.var(ptr addrspace(15) @wgbarr, i32 7) | |
| call void @llvm.amdgcn.s.barrier.wait(i16 -1) | |
| ret void | |
| } | |
| define amdgpu_kernel void @null_barrier() { | |
| ; GFX12-LABEL: null_barrier: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_barrier_join 0 | |
| ; GFX12-NEXT: s_endpgm | |
| ; | |
| ; GFX13-LABEL: null_barrier: | |
| ; GFX13: ; %bb.0: | |
| ; GFX13-NEXT: s_barrier_join 0 | |
| ; GFX13-NEXT: s_endpgm | |
| call void @llvm.amdgcn.s.barrier.join(ptr addrspace(15) null) | |
| ret void | |
| } | |
| declare void @llvm.amdgcn.s.barrier() #1 | |
| declare void @llvm.amdgcn.s.barrier.wait(i16) #1 | |
| declare void @llvm.amdgcn.s.barrier.signal(i32) #1 | |
| declare void @llvm.amdgcn.s.barrier.signal.var(ptr addrspace(15), i32) #1 | |
| declare i1 @llvm.amdgcn.s.barrier.signal.isfirst(i32) #1 | |
| declare void @llvm.amdgcn.s.barrier.init(ptr addrspace(15), i32) #1 | |
| declare void @llvm.amdgcn.s.barrier.join(ptr addrspace(15)) #1 | |
| declare void @llvm.amdgcn.s.barrier.leave(i16) #1 | |
| declare i32 @llvm.amdgcn.s.get.barrier.state(i32) #1 | |
| declare i32 @llvm.amdgcn.s.get.named.barrier.state(ptr addrspace(15)) #1 | |
| attributes #0 = { nounwind } | |
| attributes #1 = { convergent nounwind } | |
| attributes #2 = { nounwind readnone } | |
| !0 = !{i32 -1, i32 0} |