; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1200 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX12,GFX12-SDAG %s
; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1200 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX12,GFX12-GISEL %s
define amdgpu_kernel void @test1_s_barrier_signal(ptr addrspace(1) %out) #0 {
; GFX12-SDAG-LABEL: test1_s_barrier_signal:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-SDAG-NEXT: v_mul_u32_u24_e32 v1, v0, v0
; GFX12-SDAG-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-SDAG-NEXT: s_wait_storecnt 0x0
; GFX12-SDAG-NEXT: s_barrier_signal -1
; GFX12-SDAG-NEXT: s_barrier_wait -1
; GFX12-SDAG-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test1_s_barrier_signal:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v0, v0
; GFX12-GISEL-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-GISEL-NEXT: s_wait_storecnt 0x0
; GFX12-GISEL-NEXT: s_barrier_signal -1
; GFX12-GISEL-NEXT: s_barrier_wait -1
; GFX12-GISEL-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
call void @llvm.amdgcn.s.barrier.signal(i32 -1)
call void @llvm.amdgcn.s.barrier.wait(i16 -1)
%tmp3 = mul i32 %tmp, %tmp
%tmp4 = sub i32 %tmp3, %tmp
store i32 %tmp4, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test2_s_barrier_signal(ptr addrspace(1) %out) #0 {
; GFX12-SDAG-LABEL: test2_s_barrier_signal:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-SDAG-NEXT: v_mul_u32_u24_e32 v1, v0, v0
; GFX12-SDAG-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-SDAG-NEXT: s_wait_storecnt 0x0
; GFX12-SDAG-NEXT: s_barrier_signal 1
; GFX12-SDAG-NEXT: s_barrier_wait 1
; GFX12-SDAG-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test2_s_barrier_signal:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v0, v0
; GFX12-GISEL-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-GISEL-NEXT: s_wait_storecnt 0x0
; GFX12-GISEL-NEXT: s_barrier_signal 1
; GFX12-GISEL-NEXT: s_barrier_wait 1
; GFX12-GISEL-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
call void @llvm.amdgcn.s.barrier.signal(i32 1)
call void @llvm.amdgcn.s.barrier.wait(i16 1)
%tmp3 = mul i32 %tmp, %tmp
%tmp4 = sub i32 %tmp3, %tmp
store i32 %tmp4, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test3_s_barrier_signal(ptr addrspace(1) %out) #0 {
; GFX12-SDAG-LABEL: test3_s_barrier_signal:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-SDAG-NEXT: v_mul_u32_u24_e32 v1, v0, v0
; GFX12-SDAG-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-SDAG-NEXT: s_wait_storecnt 0x0
; GFX12-SDAG-NEXT: s_barrier_signal 0
; GFX12-SDAG-NEXT: s_barrier_wait 0
; GFX12-SDAG-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test3_s_barrier_signal:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v0, v0
; GFX12-GISEL-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-GISEL-NEXT: s_wait_storecnt 0x0
; GFX12-GISEL-NEXT: s_barrier_signal 0
; GFX12-GISEL-NEXT: s_barrier_wait 0
; GFX12-GISEL-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
call void @llvm.amdgcn.s.barrier.signal(i32 0)
call void @llvm.amdgcn.s.barrier.wait(i16 0)
%tmp3 = mul i32 %tmp, %tmp
%tmp4 = sub i32 %tmp3, %tmp
store i32 %tmp4, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test1_s_barrier_signal_var(ptr addrspace(1) %out) #0 {
; GFX12-SDAG-LABEL: test1_s_barrier_signal_var:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_mov_b32 m0, 1
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX12-SDAG-NEXT: v_mul_u32_u24_e32 v2, v0, v0
; GFX12-SDAG-NEXT: v_lshlrev_b32_e32 v3, 2, v0
; GFX12-SDAG-NEXT: v_sub_nc_u32_e32 v0, v2, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: global_store_b32 v3, v1, s[0:1]
; GFX12-SDAG-NEXT: s_wait_storecnt 0x0
; GFX12-SDAG-NEXT: s_barrier_signal m0
; GFX12-SDAG-NEXT: s_barrier_wait 1
; GFX12-SDAG-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test1_s_barrier_signal_var:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: v_mov_b32_e32 v2, 0
; GFX12-GISEL-NEXT: s_mov_b32 m0, 1
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_lshlrev_b32_e32 v3, 2, v0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v0, v0
; GFX12-GISEL-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-GISEL-NEXT: s_wait_storecnt 0x0
; GFX12-GISEL-NEXT: s_barrier_signal m0
; GFX12-GISEL-NEXT: s_barrier_wait 1
; GFX12-GISEL-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
call void @llvm.amdgcn.s.barrier.signal.var(i32 1)
call void @llvm.amdgcn.s.barrier.wait(i16 1)
%tmp3 = mul i32 %tmp, %tmp
%tmp4 = sub i32 %tmp3, %tmp
store i32 %tmp4, ptr addrspace(1) %tmp1
ret void
}
define void @test2_s_barrier_signal_var(i32 %arg) {
; GFX12-SDAG-LABEL: test2_s_barrier_signal_var:
; 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: v_readfirstlane_b32 s0, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-SDAG-NEXT: s_mov_b32 m0, s0
; GFX12-SDAG-NEXT: s_wait_storecnt 0x0
; GFX12-SDAG-NEXT: s_barrier_signal m0
; GFX12-SDAG-NEXT: s_wait_alu 0xfffe
; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX12-GISEL-LABEL: test2_s_barrier_signal_var:
; 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: v_readfirstlane_b32 m0, v0
; GFX12-GISEL-NEXT: s_wait_storecnt 0x0
; GFX12-GISEL-NEXT: s_barrier_signal m0
; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31]
call void @llvm.amdgcn.s.barrier.signal.var(i32 %arg)
ret void
}
define amdgpu_kernel void @test1_s_barrier_signal_isfirst(ptr addrspace(1) %a, ptr addrspace(1) %b, ptr addrspace(1) %c, ptr addrspace(1) %out) #0 {
; GFX12-SDAG-LABEL: test1_s_barrier_signal_isfirst:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b256 s[0:7], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-SDAG-NEXT: s_wait_storecnt 0x0
; GFX12-SDAG-NEXT: s_barrier_signal_isfirst -1
; GFX12-SDAG-NEXT: s_cselect_b32 s3, s3, s5
; GFX12-SDAG-NEXT: s_cselect_b32 s2, s2, s4
; GFX12-SDAG-NEXT: s_clause 0x1
; GFX12-SDAG-NEXT: global_load_b32 v2, v1, s[0:1]
; GFX12-SDAG-NEXT: global_load_b32 v1, v1, s[2:3]
; GFX12-SDAG-NEXT: s_wait_loadcnt 0x0
; GFX12-SDAG-NEXT: v_mul_lo_u32 v1, v1, v2
; GFX12-SDAG-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test1_s_barrier_signal_isfirst:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b256 s[0:7], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-GISEL-NEXT: s_wait_storecnt 0x0
; GFX12-GISEL-NEXT: s_barrier_signal_isfirst -1
; GFX12-GISEL-NEXT: s_cselect_b32 s8, 1, 0
; GFX12-GISEL-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX12-GISEL-NEXT: s_and_b32 s8, s8, 1
; GFX12-GISEL-NEXT: s_cmp_lg_u32 s8, 0
; GFX12-GISEL-NEXT: s_cselect_b64 s[2:3], s[2:3], s[4:5]
; GFX12-GISEL-NEXT: s_clause 0x1
; GFX12-GISEL-NEXT: global_load_b32 v2, v1, s[0:1]
; GFX12-GISEL-NEXT: global_load_b32 v1, v1, s[2:3]
; GFX12-GISEL-NEXT: s_wait_loadcnt 0x0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v1, v2
; GFX12-GISEL-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
%isfirst = call i1 @llvm.amdgcn.s.barrier.signal.isfirst(i32 -1)
%0 = load i32, ptr addrspace(1) %a, align 4
%b.c = select i1 %isfirst, ptr addrspace(1) %b, ptr addrspace(1) %c
%1 = load i32, ptr addrspace(1) %b.c, align 4
%mul1 = mul nsw i32 %1, %0
store i32 %mul1, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test2_s_barrier_signal_isfirst(ptr addrspace(1) %a, ptr addrspace(1) %b, ptr addrspace(1) %c, ptr addrspace(1) %out) #0 {
; GFX12-SDAG-LABEL: test2_s_barrier_signal_isfirst:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b256 s[0:7], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-SDAG-NEXT: s_wait_storecnt 0x0
; GFX12-SDAG-NEXT: s_barrier_signal_isfirst 1
; GFX12-SDAG-NEXT: s_cselect_b32 s3, s3, s5
; GFX12-SDAG-NEXT: s_cselect_b32 s2, s2, s4
; GFX12-SDAG-NEXT: s_clause 0x1
; GFX12-SDAG-NEXT: global_load_b32 v2, v1, s[0:1]
; GFX12-SDAG-NEXT: global_load_b32 v1, v1, s[2:3]
; GFX12-SDAG-NEXT: s_wait_loadcnt 0x0
; GFX12-SDAG-NEXT: v_mul_lo_u32 v1, v1, v2
; GFX12-SDAG-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test2_s_barrier_signal_isfirst:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b256 s[0:7], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-GISEL-NEXT: s_wait_storecnt 0x0
; GFX12-GISEL-NEXT: s_barrier_signal_isfirst 1
; GFX12-GISEL-NEXT: s_cselect_b32 s8, 1, 0
; GFX12-GISEL-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX12-GISEL-NEXT: s_and_b32 s8, s8, 1
; GFX12-GISEL-NEXT: s_cmp_lg_u32 s8, 0
; GFX12-GISEL-NEXT: s_cselect_b64 s[2:3], s[2:3], s[4:5]
; GFX12-GISEL-NEXT: s_clause 0x1
; GFX12-GISEL-NEXT: global_load_b32 v2, v1, s[0:1]
; GFX12-GISEL-NEXT: global_load_b32 v1, v1, s[2:3]
; GFX12-GISEL-NEXT: s_wait_loadcnt 0x0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v1, v2
; GFX12-GISEL-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
%isfirst = call i1 @llvm.amdgcn.s.barrier.signal.isfirst(i32 1)
%0 = load i32, ptr addrspace(1) %a, align 4
%b.c = select i1 %isfirst, ptr addrspace(1) %b, ptr addrspace(1) %c
%1 = load i32, ptr addrspace(1) %b.c, align 4
%mul1 = mul nsw i32 %1, %0
store i32 %mul1, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test3_s_barrier_signal_isfirst(ptr addrspace(1) %a, ptr addrspace(1) %b, ptr addrspace(1) %c, ptr addrspace(1) %out) #0 {
; GFX12-SDAG-LABEL: test3_s_barrier_signal_isfirst:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b256 s[0:7], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-SDAG-NEXT: s_wait_storecnt 0x0
; GFX12-SDAG-NEXT: s_barrier_signal_isfirst 1
; GFX12-SDAG-NEXT: s_cselect_b32 s3, s3, s5
; GFX12-SDAG-NEXT: s_cselect_b32 s2, s2, s4
; GFX12-SDAG-NEXT: s_clause 0x1
; GFX12-SDAG-NEXT: global_load_b32 v2, v1, s[0:1]
; GFX12-SDAG-NEXT: global_load_b32 v1, v1, s[2:3]
; GFX12-SDAG-NEXT: s_wait_loadcnt 0x0
; GFX12-SDAG-NEXT: v_mul_lo_u32 v1, v1, v2
; GFX12-SDAG-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test3_s_barrier_signal_isfirst:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b256 s[0:7], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-GISEL-NEXT: s_wait_storecnt 0x0
; GFX12-GISEL-NEXT: s_barrier_signal_isfirst 1
; GFX12-GISEL-NEXT: s_cselect_b32 s8, 1, 0
; GFX12-GISEL-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX12-GISEL-NEXT: s_and_b32 s8, s8, 1
; GFX12-GISEL-NEXT: s_cmp_lg_u32 s8, 0
; GFX12-GISEL-NEXT: s_cselect_b64 s[2:3], s[2:3], s[4:5]
; GFX12-GISEL-NEXT: s_clause 0x1
; GFX12-GISEL-NEXT: global_load_b32 v2, v1, s[0:1]
; GFX12-GISEL-NEXT: global_load_b32 v1, v1, s[2:3]
; GFX12-GISEL-NEXT: s_wait_loadcnt 0x0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v1, v2
; GFX12-GISEL-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
%isfirst = call i1 @llvm.amdgcn.s.barrier.signal.isfirst(i32 1)
%0 = load i32, ptr addrspace(1) %a, align 4
%b.c = select i1 %isfirst, ptr addrspace(1) %b, ptr addrspace(1) %c
%1 = load i32, ptr addrspace(1) %b.c, align 4
%mul1 = mul nsw i32 %1, %0
store i32 %mul1, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test1_s_barrier_signal_isfirst_var(ptr addrspace(1) %a, ptr addrspace(1) %b, ptr addrspace(1) %c, ptr addrspace(1) %out) #0 {
; GFX12-SDAG-LABEL: test1_s_barrier_signal_isfirst_var:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b256 s[0:7], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_mov_b32 m0, 1
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-SDAG-NEXT: s_wait_storecnt 0x0
; GFX12-SDAG-NEXT: s_barrier_signal_isfirst m0
; GFX12-SDAG-NEXT: s_cselect_b32 s3, s3, s5
; GFX12-SDAG-NEXT: s_cselect_b32 s2, s2, s4
; GFX12-SDAG-NEXT: s_clause 0x1
; GFX12-SDAG-NEXT: global_load_b32 v2, v1, s[0:1]
; GFX12-SDAG-NEXT: global_load_b32 v1, v1, s[2:3]
; GFX12-SDAG-NEXT: s_wait_loadcnt 0x0
; GFX12-SDAG-NEXT: v_mul_lo_u32 v1, v1, v2
; GFX12-SDAG-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test1_s_barrier_signal_isfirst_var:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b256 s[0:7], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_mov_b32 m0, 1
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-GISEL-NEXT: s_wait_storecnt 0x0
; GFX12-GISEL-NEXT: s_barrier_signal_isfirst m0
; GFX12-GISEL-NEXT: s_cselect_b32 s8, 1, 0
; GFX12-GISEL-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX12-GISEL-NEXT: s_and_b32 s8, s8, 1
; GFX12-GISEL-NEXT: s_cmp_lg_u32 s8, 0
; GFX12-GISEL-NEXT: s_cselect_b64 s[2:3], s[2:3], s[4:5]
; GFX12-GISEL-NEXT: s_clause 0x1
; GFX12-GISEL-NEXT: global_load_b32 v2, v1, s[0:1]
; GFX12-GISEL-NEXT: global_load_b32 v1, v1, s[2:3]
; GFX12-GISEL-NEXT: s_wait_loadcnt 0x0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v1, v2
; GFX12-GISEL-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
%isfirst = call i1 @llvm.amdgcn.s.barrier.signal.isfirst.var(i32 1)
%0 = load i32, ptr addrspace(1) %a, align 4
%b.c = select i1 %isfirst, ptr addrspace(1) %b, ptr addrspace(1) %c
%1 = load i32, ptr addrspace(1) %b.c, align 4
%mul1 = mul nsw i32 %1, %0
store i32 %mul1, ptr addrspace(1) %tmp1
ret void
}
define void @test2_s_barrier_signal_isfirst_var(ptr addrspace(1) %a, ptr addrspace(1) %b, ptr addrspace(1) %c, i32 %arg, ptr addrspace(1) %out) {
; GFX12-SDAG-LABEL: test2_s_barrier_signal_isfirst_var:
; 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: v_dual_mov_b32 v10, 0 :: v_dual_and_b32 v9, 0x3ff, v31
; GFX12-SDAG-NEXT: v_readfirstlane_b32 s0, v6
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX12-SDAG-NEXT: v_lshlrev_b32_e32 v9, 2, v9
; GFX12-SDAG-NEXT: s_mov_b32 m0, s0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_add_co_u32 v7, vcc_lo, v7, v9
; GFX12-SDAG-NEXT: v_add_co_ci_u32_e32 v8, vcc_lo, 0, v8, vcc_lo
; GFX12-SDAG-NEXT: global_store_b32 v[7:8], v10, off
; GFX12-SDAG-NEXT: s_wait_storecnt 0x0
; GFX12-SDAG-NEXT: s_barrier_signal_isfirst m0
; GFX12-SDAG-NEXT: s_cselect_b32 vcc_lo, -1, 0
; GFX12-SDAG-NEXT: s_wait_alu 0xfffe
; GFX12-SDAG-NEXT: v_dual_cndmask_b32 v2, v4, v2 :: v_dual_cndmask_b32 v3, v5, v3
; GFX12-SDAG-NEXT: global_load_b32 v0, v[0:1], off
; GFX12-SDAG-NEXT: global_load_b32 v1, v[2:3], off
; GFX12-SDAG-NEXT: s_wait_loadcnt 0x0
; GFX12-SDAG-NEXT: v_mul_lo_u32 v0, v1, v0
; GFX12-SDAG-NEXT: global_store_b32 v[7:8], v0, off
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX12-GISEL-LABEL: test2_s_barrier_signal_isfirst_var:
; 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: v_and_b32_e32 v9, 0x3ff, v31
; GFX12-GISEL-NEXT: v_readfirstlane_b32 m0, v6
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_lshlrev_b32_e32 v9, 2, v9
; GFX12-GISEL-NEXT: v_add_co_u32 v7, vcc_lo, v7, v9
; GFX12-GISEL-NEXT: v_add_co_ci_u32_e32 v8, vcc_lo, 0, v8, vcc_lo
; GFX12-GISEL-NEXT: v_mov_b32_e32 v9, 0
; GFX12-GISEL-NEXT: global_store_b32 v[7:8], v9, off
; GFX12-GISEL-NEXT: s_wait_storecnt 0x0
; GFX12-GISEL-NEXT: s_barrier_signal_isfirst m0
; GFX12-GISEL-NEXT: s_cselect_b32 s0, 1, 0
; GFX12-GISEL-NEXT: s_wait_alu 0xfffe
; GFX12-GISEL-NEXT: s_and_b32 s0, 1, s0
; GFX12-GISEL-NEXT: s_wait_alu 0xfffe
; GFX12-GISEL-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0
; GFX12-GISEL-NEXT: v_dual_cndmask_b32 v2, v4, v2 :: v_dual_cndmask_b32 v3, v5, v3
; GFX12-GISEL-NEXT: global_load_b32 v0, v[0:1], off
; GFX12-GISEL-NEXT: global_load_b32 v1, v[2:3], off
; GFX12-GISEL-NEXT: s_wait_loadcnt 0x0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v0, v1, v0
; GFX12-GISEL-NEXT: global_store_b32 v[7:8], v0, off
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31]
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
%isfirst = call i1 @llvm.amdgcn.s.barrier.signal.isfirst.var(i32 %arg)
%1 = load i32, ptr addrspace(1) %a, align 4
%b.c = select i1 %isfirst, ptr addrspace(1) %b, ptr addrspace(1) %c
%2 = load i32, ptr addrspace(1) %b.c, align 4
%mul1 = mul nsw i32 %2, %1
store i32 %mul1, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test1_s_barrier_init(ptr addrspace(1) %out, i32 %mbrCnt) #0 {
; GFX12-SDAG-LABEL: test1_s_barrier_init:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b96 s[0:2], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-SDAG-NEXT: v_mul_u32_u24_e32 v1, v0, v0
; GFX12-SDAG-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: s_lshl_b32 s2, s2, 16
; GFX12-SDAG-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-SDAG-NEXT: s_mov_b32 m0, s2
; GFX12-SDAG-NEXT: s_barrier_init -1
; GFX12-SDAG-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test1_s_barrier_init:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b96 s[0:2], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v0, v0
; GFX12-GISEL-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: s_lshl_b32 m0, 16, s2
; GFX12-GISEL-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-GISEL-NEXT: s_barrier_init -1
; GFX12-GISEL-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
call void @llvm.amdgcn.s.barrier.init(i32 -1, i32 %mbrCnt)
%tmp2 = mul i32 %tmp, %tmp
%tmp3 = sub i32 %tmp2, %tmp
store i32 %tmp3, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test2_s_barrier_init(ptr addrspace(1) %out, i32 %mbrCnt) #0 {
; GFX12-SDAG-LABEL: test2_s_barrier_init:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b96 s[0:2], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-SDAG-NEXT: v_mul_u32_u24_e32 v1, v0, v0
; GFX12-SDAG-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: s_lshl_b32 s2, s2, 16
; GFX12-SDAG-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-SDAG-NEXT: s_mov_b32 m0, s2
; GFX12-SDAG-NEXT: s_barrier_init 1
; GFX12-SDAG-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test2_s_barrier_init:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b96 s[0:2], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v0, v0
; GFX12-GISEL-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: s_lshl_b32 m0, 16, s2
; GFX12-GISEL-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-GISEL-NEXT: s_barrier_init 1
; GFX12-GISEL-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
call void @llvm.amdgcn.s.barrier.init(i32 1, i32 %mbrCnt)
%tmp3 = mul i32 %tmp, %tmp
%tmp4 = sub i32 %tmp3, %tmp
store i32 %tmp4, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test3_s_barrier_init(ptr addrspace(1) %out, i32 %mbrCnt) #0 {
; GFX12-SDAG-LABEL: test3_s_barrier_init:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b96 s[0:2], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-SDAG-NEXT: v_mul_u32_u24_e32 v1, v0, v0
; GFX12-SDAG-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: s_lshl_b32 s2, s2, 16
; GFX12-SDAG-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-SDAG-NEXT: s_mov_b32 m0, s2
; GFX12-SDAG-NEXT: s_barrier_init 0
; GFX12-SDAG-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test3_s_barrier_init:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b96 s[0:2], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v0, v0
; GFX12-GISEL-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: s_lshl_b32 m0, 16, s2
; GFX12-GISEL-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-GISEL-NEXT: s_barrier_init 0
; GFX12-GISEL-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
call void @llvm.amdgcn.s.barrier.init(i32 0, i32 %mbrCnt)
%tmp3 = mul i32 %tmp, %tmp
%tmp4 = sub i32 %tmp3, %tmp
store i32 %tmp4, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test4_s_barrier_init(ptr addrspace(1) %out, i32 %bar, i32 %mbrCnt) #0 {
; GFX12-SDAG-LABEL: test4_s_barrier_init:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b128 s[0:3], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-SDAG-NEXT: v_mul_u32_u24_e32 v1, v0, v0
; GFX12-SDAG-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: s_lshl_b32 s3, s3, 16
; GFX12-SDAG-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-SDAG-NEXT: s_or_b32 s2, s2, s3
; GFX12-SDAG-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX12-SDAG-NEXT: s_mov_b32 m0, s2
; GFX12-SDAG-NEXT: s_barrier_init m0
; GFX12-SDAG-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test4_s_barrier_init:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b128 s[0:3], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v0, v0
; GFX12-GISEL-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: s_lshl_b32 s3, 16, s3
; GFX12-GISEL-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-GISEL-NEXT: s_or_b32 m0, s2, s3
; GFX12-GISEL-NEXT: s_barrier_init m0
; GFX12-GISEL-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
call void @llvm.amdgcn.s.barrier.init(i32 %bar, i32 %mbrCnt)
%tmp3 = mul i32 %tmp, %tmp
%tmp4 = sub i32 %tmp3, %tmp
store i32 %tmp4, ptr addrspace(1) %tmp1
ret void
}
define void @test5_s_barrier_init_m0(i32 %arg1 ,i32 %arg2) {
; GFX12-SDAG-LABEL: test5_s_barrier_init_m0:
; 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: v_lshlrev_b32_e32 v1, 16, v1
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_or_b32_e32 v0, v0, v1
; GFX12-SDAG-NEXT: v_readfirstlane_b32 s0, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-SDAG-NEXT: s_mov_b32 m0, s0
; GFX12-SDAG-NEXT: s_barrier_init m0
; GFX12-SDAG-NEXT: s_wait_alu 0xfffe
; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX12-GISEL-LABEL: test5_s_barrier_init_m0:
; 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: v_readfirstlane_b32 s0, v1
; GFX12-GISEL-NEXT: v_readfirstlane_b32 s1, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: s_lshl_b32 s0, 16, s0
; GFX12-GISEL-NEXT: s_wait_alu 0xfffe
; GFX12-GISEL-NEXT: s_or_b32 m0, s1, s0
; GFX12-GISEL-NEXT: s_barrier_init m0
; GFX12-GISEL-NEXT: s_wait_alu 0xfffe
; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31]
call void @llvm.amdgcn.s.barrier.init(i32 %arg1, i32 %arg2)
ret void
}
define amdgpu_kernel void @test1_s_barrier_join(ptr addrspace(1) %out) #0 {
;
; GFX12-SDAG-LABEL: test1_s_barrier_join:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-SDAG-NEXT: v_mul_u32_u24_e32 v1, v0, v0
; GFX12-SDAG-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-SDAG-NEXT: s_barrier_join -1
; GFX12-SDAG-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test1_s_barrier_join:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v0, v0
; GFX12-GISEL-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-GISEL-NEXT: s_barrier_join -1
; GFX12-GISEL-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
call void @llvm.amdgcn.s.barrier.join(i32 -1)
%tmp3 = mul i32 %tmp, %tmp
%tmp4 = sub i32 %tmp3, %tmp
store i32 %tmp4, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test2_s_barrier_join(ptr addrspace(1) %out) #0 {
;
; GFX12-SDAG-LABEL: test2_s_barrier_join:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-SDAG-NEXT: v_mul_u32_u24_e32 v1, v0, v0
; GFX12-SDAG-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-SDAG-NEXT: s_barrier_join 1
; GFX12-SDAG-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test2_s_barrier_join:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v0, v0
; GFX12-GISEL-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-GISEL-NEXT: s_barrier_join 1
; GFX12-GISEL-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
call void @llvm.amdgcn.s.barrier.join(i32 1)
%tmp3 = mul i32 %tmp, %tmp
%tmp4 = sub i32 %tmp3, %tmp
store i32 %tmp4, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test3_s_barrier_join(ptr addrspace(1) %out) #0 {
;
; GFX12-SDAG-LABEL: test3_s_barrier_join:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-SDAG-NEXT: v_mul_u32_u24_e32 v1, v0, v0
; GFX12-SDAG-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-SDAG-NEXT: s_barrier_join 0
; GFX12-SDAG-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test3_s_barrier_join:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v0, v0
; GFX12-GISEL-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-GISEL-NEXT: s_barrier_join 0
; GFX12-GISEL-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
call void @llvm.amdgcn.s.barrier.join(i32 0)
%tmp3 = mul i32 %tmp, %tmp
%tmp4 = sub i32 %tmp3, %tmp
store i32 %tmp4, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test4_s_barrier_join_m0(ptr addrspace(1) %out, i32 %bar) #0 {
; GFX12-SDAG-LABEL: test4_s_barrier_join_m0:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b96 s[0:2], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX12-SDAG-NEXT: v_mul_u32_u24_e32 v2, v0, v0
; GFX12-SDAG-NEXT: v_lshlrev_b32_e32 v3, 2, v0
; GFX12-SDAG-NEXT: v_sub_nc_u32_e32 v0, v2, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: s_mov_b32 m0, s2
; GFX12-SDAG-NEXT: global_store_b32 v3, v1, s[0:1]
; GFX12-SDAG-NEXT: s_barrier_join m0
; GFX12-SDAG-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test4_s_barrier_join_m0:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b96 s[0:2], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v0, v0
; GFX12-GISEL-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: s_mov_b32 m0, s2
; GFX12-GISEL-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-GISEL-NEXT: s_barrier_join m0
; GFX12-GISEL-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
call void @llvm.amdgcn.s.barrier.join(i32 %bar)
%tmp3 = mul i32 %tmp, %tmp
%tmp4 = sub i32 %tmp3, %tmp
store i32 %tmp4, ptr addrspace(1) %tmp1
ret void
}
define void @test5_s_barrier_join_m0(i32 %arg) {
; GFX12-SDAG-LABEL: test5_s_barrier_join_m0:
; 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: v_readfirstlane_b32 s0, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-SDAG-NEXT: s_mov_b32 m0, s0
; GFX12-SDAG-NEXT: s_barrier_join m0
; GFX12-SDAG-NEXT: s_wait_alu 0xfffe
; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX12-GISEL-LABEL: test5_s_barrier_join_m0:
; 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: v_readfirstlane_b32 m0, v0
; GFX12-GISEL-NEXT: s_barrier_join m0
; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31]
call void @llvm.amdgcn.s.barrier.join(i32 %arg)
ret void
}
define void @test6_s_barrier_join_0() {
; GFX12-LABEL: test6_s_barrier_join_0:
; 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_join 0
; GFX12-NEXT: s_setpc_b64 s[30:31]
call void @llvm.amdgcn.s.barrier.join(i32 0)
ret void
}
define amdgpu_kernel void @test1_s_barrier_leave(ptr addrspace(1) %a, ptr addrspace(1) %b, ptr addrspace(1) %c, ptr addrspace(1) %out) #0 {
; GFX12-SDAG-LABEL: test1_s_barrier_leave:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b256 s[0:7], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-SDAG-NEXT: s_barrier_leave
; GFX12-SDAG-NEXT: s_cselect_b32 s3, s3, s5
; GFX12-SDAG-NEXT: s_cselect_b32 s2, s2, s4
; GFX12-SDAG-NEXT: s_clause 0x1
; GFX12-SDAG-NEXT: global_load_b32 v2, v1, s[0:1]
; GFX12-SDAG-NEXT: global_load_b32 v1, v1, s[2:3]
; GFX12-SDAG-NEXT: s_wait_loadcnt 0x0
; GFX12-SDAG-NEXT: v_mul_lo_u32 v1, v1, v2
; GFX12-SDAG-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test1_s_barrier_leave:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b256 s[0:7], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
; GFX12-GISEL-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-GISEL-NEXT: s_barrier_leave
; GFX12-GISEL-NEXT: s_cselect_b32 s8, 1, 0
; GFX12-GISEL-NEXT: s_and_b32 s8, s8, 1
; GFX12-GISEL-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX12-GISEL-NEXT: s_cmp_lg_u32 s8, 0
; GFX12-GISEL-NEXT: s_cselect_b64 s[2:3], s[2:3], s[4:5]
; GFX12-GISEL-NEXT: s_clause 0x1
; GFX12-GISEL-NEXT: global_load_b32 v2, v1, s[0:1]
; GFX12-GISEL-NEXT: global_load_b32 v1, v1, s[2:3]
; GFX12-GISEL-NEXT: s_wait_loadcnt 0x0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v1, v2
; GFX12-GISEL-NEXT: global_store_b32 v0, v1, s[6:7]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
%0 = call i1 @llvm.amdgcn.s.barrier.leave()
%1 = load i32, ptr addrspace(1) %a, align 4
%b.c = select i1 %0, ptr addrspace(1) %b, ptr addrspace(1) %c
%2 = load i32, ptr addrspace(1) %b.c, align 4
%mul1 = mul nsw i32 %2, %1
store i32 %mul1, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test1_s_wakeup_barrier(ptr addrspace(1) %out) #0 {
;
; GFX12-SDAG-LABEL: test1_s_wakeup_barrier:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-SDAG-NEXT: v_mul_u32_u24_e32 v1, v0, v0
; GFX12-SDAG-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-SDAG-NEXT: s_wakeup_barrier -1
; GFX12-SDAG-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test1_s_wakeup_barrier:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v0, v0
; GFX12-GISEL-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-GISEL-NEXT: s_wakeup_barrier -1
; GFX12-GISEL-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
call void @llvm.amdgcn.s.wakeup.barrier(i32 -1)
%tmp3 = mul i32 %tmp, %tmp
%tmp4 = sub i32 %tmp3, %tmp
store i32 %tmp4, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test2_s_wakeup_barrier(ptr addrspace(1) %out) #0 {
;
; GFX12-SDAG-LABEL: test2_s_wakeup_barrier:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-SDAG-NEXT: v_mul_u32_u24_e32 v1, v0, v0
; GFX12-SDAG-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-SDAG-NEXT: s_wakeup_barrier 1
; GFX12-SDAG-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test2_s_wakeup_barrier:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v0, v0
; GFX12-GISEL-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-GISEL-NEXT: s_wakeup_barrier 1
; GFX12-GISEL-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
call void @llvm.amdgcn.s.wakeup.barrier(i32 1)
%tmp3 = mul i32 %tmp, %tmp
%tmp4 = sub i32 %tmp3, %tmp
store i32 %tmp4, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test3_s_wakeup_barrier(ptr addrspace(1) %out) #0 {
;
; GFX12-SDAG-LABEL: test3_s_wakeup_barrier:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-SDAG-NEXT: v_mul_u32_u24_e32 v1, v0, v0
; GFX12-SDAG-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-SDAG-NEXT: s_wakeup_barrier 0
; GFX12-SDAG-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test3_s_wakeup_barrier:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v0, v0
; GFX12-GISEL-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-GISEL-NEXT: s_wakeup_barrier 0
; GFX12-GISEL-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
call void @llvm.amdgcn.s.wakeup.barrier(i32 0)
%tmp3 = mul i32 %tmp, %tmp
%tmp4 = sub i32 %tmp3, %tmp
store i32 %tmp4, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test4_s_wakeup_barrier_m0(ptr addrspace(1) %out, i32 %bar) #0 {
; GFX12-SDAG-LABEL: test4_s_wakeup_barrier_m0:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b96 s[0:2], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX12-SDAG-NEXT: v_mul_u32_u24_e32 v2, v0, v0
; GFX12-SDAG-NEXT: v_lshlrev_b32_e32 v3, 2, v0
; GFX12-SDAG-NEXT: v_sub_nc_u32_e32 v0, v2, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: s_mov_b32 m0, s2
; GFX12-SDAG-NEXT: global_store_b32 v3, v1, s[0:1]
; GFX12-SDAG-NEXT: s_wakeup_barrier m0
; GFX12-SDAG-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test4_s_wakeup_barrier_m0:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b96 s[0:2], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v0, v0
; GFX12-GISEL-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: s_mov_b32 m0, s2
; GFX12-GISEL-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-GISEL-NEXT: s_wakeup_barrier m0
; GFX12-GISEL-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
call void @llvm.amdgcn.s.wakeup.barrier(i32 %bar)
%tmp3 = mul i32 %tmp, %tmp
%tmp4 = sub i32 %tmp3, %tmp
store i32 %tmp4, ptr addrspace(1) %tmp1
ret void
}
define void @test5_s_wakeup_barrier_m0(i32 %arg) {
; GFX12-SDAG-LABEL: test5_s_wakeup_barrier_m0:
; 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: v_readfirstlane_b32 s0, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-SDAG-NEXT: s_mov_b32 m0, s0
; GFX12-SDAG-NEXT: s_wakeup_barrier m0
; GFX12-SDAG-NEXT: s_wait_alu 0xfffe
; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX12-GISEL-LABEL: test5_s_wakeup_barrier_m0:
; 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: v_readfirstlane_b32 m0, v0
; GFX12-GISEL-NEXT: s_wakeup_barrier m0
; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31]
call void @llvm.amdgcn.s.wakeup.barrier(i32 %arg)
ret void
}
define amdgpu_kernel void @test1_s_get_barrier_state(ptr addrspace(1) %out) #0 {
; GFX12-LABEL: test1_s_get_barrier_state:
; GFX12: ; %bb.0: ; %entry
; GFX12-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_2)
; GFX12-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: global_store_b32 v0, v1, s[0:1]
; GFX12-NEXT: s_get_barrier_state s2, -1
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: v_mov_b32_e32 v1, s2
; GFX12-NEXT: global_store_b32 v0, v1, s[0:1]
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
%state = call i32 @llvm.amdgcn.s.get.barrier.state(i32 -1)
store i32 %state, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test2_s_get_barrier_state(ptr addrspace(1) %out) #0 {
; GFX12-LABEL: test2_s_get_barrier_state:
; GFX12: ; %bb.0: ; %entry
; GFX12-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_2)
; GFX12-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: global_store_b32 v0, v1, s[0:1]
; GFX12-NEXT: s_get_barrier_state s2, 1
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: v_mov_b32_e32 v1, s2
; GFX12-NEXT: global_store_b32 v0, v1, s[0:1]
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
%state = call i32 @llvm.amdgcn.s.get.barrier.state(i32 1)
store i32 %state, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test3_s_get_barrier_state(ptr addrspace(1) %out) #0 {
; GFX12-LABEL: test3_s_get_barrier_state:
; GFX12: ; %bb.0: ; %entry
; GFX12-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_2)
; GFX12-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: global_store_b32 v0, v1, s[0:1]
; GFX12-NEXT: s_get_barrier_state s2, 0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: v_mov_b32_e32 v1, s2
; GFX12-NEXT: global_store_b32 v0, v1, s[0:1]
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
%state = call i32 @llvm.amdgcn.s.get.barrier.state(i32 0)
store i32 %state, ptr addrspace(1) %tmp1
ret void
}
define amdgpu_kernel void @test4_s_get_barrier_state_m0(ptr addrspace(1) %out, i32 %bar) #0 {
; GFX12-LABEL: test4_s_get_barrier_state_m0:
; GFX12: ; %bb.0: ; %entry
; GFX12-NEXT: s_load_b96 s[0:2], s[2:3], 0x24
; GFX12-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0x3ff, v0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-NEXT: v_lshlrev_b32_e32 v0, 2, v0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: s_mov_b32 m0, s2
; GFX12-NEXT: global_store_b32 v0, v1, s[0:1]
; GFX12-NEXT: s_get_barrier_state s2, m0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_2)
; GFX12-NEXT: v_mov_b32_e32 v1, s2
; GFX12-NEXT: global_store_b32 v0, v1, s[0:1]
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
%state = call i32 @llvm.amdgcn.s.get.barrier.state(i32 %bar)
store i32 %state, ptr addrspace(1) %tmp1
ret void
}
define i32 @test5_s_get_barrier_state_m0(i32 %arg) {
; GFX12-SDAG-LABEL: test5_s_get_barrier_state_m0:
; 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: v_readfirstlane_b32 s0, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(SALU_CYCLE_1)
; GFX12-SDAG-NEXT: s_mov_b32 m0, s0
; GFX12-SDAG-NEXT: s_get_barrier_state s0, m0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: s_wait_alu 0xfffe
; GFX12-SDAG-NEXT: v_mov_b32_e32 v0, s0
; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31]
;
; GFX12-GISEL-LABEL: test5_s_get_barrier_state_m0:
; 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: v_readfirstlane_b32 m0, v0
; GFX12-GISEL-NEXT: s_get_barrier_state s0, m0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: s_wait_alu 0xfffe
; GFX12-GISEL-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX12-GISEL-NEXT: v_mov_b32_e32 v0, s0
; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31]
%state = call i32 @llvm.amdgcn.s.get.barrier.state(i32 %arg)
ret i32 %state
}
define i32 @test6_s_get_barrier_state_0() {
; GFX12-LABEL: test6_s_get_barrier_state_0:
; 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_get_barrier_state s0, 0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: s_wait_alu 0xfffe
; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX12-NEXT: v_mov_b32_e32 v0, s0
; GFX12-NEXT: s_setpc_b64 s[30:31]
%state = call i32 @llvm.amdgcn.s.get.barrier.state(i32 0)
ret i32 %state
}
define amdgpu_kernel void @test_barrier_convert(ptr addrspace(1) %out) #0 {
; GFX12-SDAG-LABEL: test_barrier_convert:
; GFX12-SDAG: ; %bb.0: ; %entry
; GFX12-SDAG-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-SDAG-NEXT: v_mul_u32_u24_e32 v1, v0, v0
; GFX12-SDAG-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-SDAG-NEXT: s_wait_storecnt 0x0
; GFX12-SDAG-NEXT: s_barrier_signal -1
; GFX12-SDAG-NEXT: s_barrier_wait -1
; GFX12-SDAG-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-SDAG-NEXT: s_nop 0
; GFX12-SDAG-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-SDAG-NEXT: s_endpgm
;
; GFX12-GISEL-LABEL: test_barrier_convert:
; GFX12-GISEL: ; %bb.0: ; %entry
; GFX12-GISEL-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
; GFX12-GISEL-NEXT: v_and_b32_e32 v0, 0x3ff, v0
; GFX12-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_lshlrev_b32 v3, 2, v0
; GFX12-GISEL-NEXT: v_mul_lo_u32 v1, v0, v0
; GFX12-GISEL-NEXT: v_sub_nc_u32_e32 v0, v1, v0
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT: global_store_b32 v3, v2, s[0:1]
; GFX12-GISEL-NEXT: s_wait_storecnt 0x0
; GFX12-GISEL-NEXT: s_barrier_signal -1
; GFX12-GISEL-NEXT: s_barrier_wait -1
; GFX12-GISEL-NEXT: global_store_b32 v3, v0, s[0:1]
; GFX12-GISEL-NEXT: s_nop 0
; GFX12-GISEL-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-GISEL-NEXT: s_endpgm
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = getelementptr i32, ptr addrspace(1) %out, i32 %tmp
store i32 0, ptr addrspace(1) %tmp1
call void @llvm.amdgcn.s.barrier()
%tmp3 = mul i32 %tmp, %tmp
%tmp4 = sub i32 %tmp3, %tmp
store i32 %tmp4, ptr addrspace(1) %tmp1
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(i32) #1
declare i1 @llvm.amdgcn.s.barrier.signal.isfirst(i32) #1
declare i1 @llvm.amdgcn.s.barrier.signal.isfirst.var(i32) #1
declare void @llvm.amdgcn.s.barrier.init(i32, i32) #1
declare void @llvm.amdgcn.s.barrier.join(i32) #1
declare i1 @llvm.amdgcn.s.barrier.leave() #1
declare void @llvm.amdgcn.s.wakeup.barrier(i32) #1
declare i32 @llvm.amdgcn.s.get.barrier.state(i32) #1
declare i32 @llvm.amdgcn.s.get.barrier.state.var(i32) #1
declare i32 @llvm.amdgcn.workitem.id.x() #2
attributes #0 = { nounwind }
attributes #1 = { convergent nounwind }
attributes #2 = { nounwind readnone }
Codebase Browser
llvm