; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -amdgpu-atomic-optimizer-strategy=None < %s | FileCheck -check-prefix=GFX6 %s
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=hawaii -amdgpu-atomic-optimizer-strategy=None < %s | FileCheck -check-prefix=GFX7 %s
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -amdgpu-atomic-optimizer-strategy=None < %s | FileCheck -check-prefix=GFX12 %s
; Test end to end matching of addressing modes when MUBUF is used for
; global memory.
define amdgpu_ps void @mubuf_store_sgpr_ptr(ptr addrspace(1) inreg %ptr) {
; GFX6-LABEL: mubuf_store_sgpr_ptr:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: v_mov_b32_e32 v0, 0
; GFX6-NEXT: s_mov_b32 s2, -1
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX6-NEXT: s_endpgm
;
; GFX7-LABEL: mubuf_store_sgpr_ptr:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: v_mov_b32_e32 v0, 0
; GFX7-NEXT: s_mov_b32 s2, -1
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX7-NEXT: s_endpgm
;
; GFX12-LABEL: mubuf_store_sgpr_ptr:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_mov_b32_e32 v0, 0
; GFX12-NEXT: global_store_b32 v0, v0, s[2:3]
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
store i32 0, ptr addrspace(1) %ptr
ret void
}
define amdgpu_ps void @mubuf_store_sgpr_ptr_offset4095(ptr addrspace(1) inreg %ptr) {
; GFX6-LABEL: mubuf_store_sgpr_ptr_offset4095:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: v_mov_b32_e32 v0, 0
; GFX6-NEXT: s_mov_b32 s2, -1
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: s_movk_i32 s4, 0x3ffc
; GFX6-NEXT: buffer_store_dword v0, off, s[0:3], s4
; GFX6-NEXT: s_endpgm
;
; GFX7-LABEL: mubuf_store_sgpr_ptr_offset4095:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: v_mov_b32_e32 v0, 0
; GFX7-NEXT: s_mov_b32 s2, -1
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: s_movk_i32 s4, 0x3ffc
; GFX7-NEXT: buffer_store_dword v0, off, s[0:3], s4
; GFX7-NEXT: s_endpgm
;
; GFX12-LABEL: mubuf_store_sgpr_ptr_offset4095:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_mov_b32_e32 v0, 0
; GFX12-NEXT: global_store_b32 v0, v0, s[2:3] offset:16380
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
%gep = getelementptr i32, ptr addrspace(1) %ptr, i64 4095
store i32 0, ptr addrspace(1) %gep
ret void
}
define amdgpu_ps void @mubuf_store_sgpr_ptr_offset4294967296(ptr addrspace(1) inreg %ptr) {
; GFX6-LABEL: mubuf_store_sgpr_ptr_offset4294967296:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_mov_b32_e32 v0, 0
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: v_mov_b32_e32 v2, 0
; GFX6-NEXT: v_mov_b32_e32 v1, 4
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; GFX6-NEXT: s_endpgm
;
; GFX7-LABEL: mubuf_store_sgpr_ptr_offset4294967296:
; GFX7: ; %bb.0:
; GFX7-NEXT: v_mov_b32_e32 v0, 0
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: v_mov_b32_e32 v2, 0
; GFX7-NEXT: v_mov_b32_e32 v1, 4
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; GFX7-NEXT: s_endpgm
;
; GFX12-LABEL: mubuf_store_sgpr_ptr_offset4294967296:
; GFX12: ; %bb.0:
; GFX12-NEXT: s_add_co_u32 s0, s2, 0
; GFX12-NEXT: s_add_co_ci_u32 s1, s3, 4
; GFX12-NEXT: v_mov_b32_e32 v0, s0
; GFX12-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s1
; GFX12-NEXT: global_store_b32 v[0:1], v2, off
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
%gep = getelementptr i32, ptr addrspace(1) %ptr, i64 4294967296
store i32 0, ptr addrspace(1) %gep
ret void
}
define amdgpu_ps void @mubuf_store_sgpr_ptr_offset4294967297(ptr addrspace(1) inreg %ptr) {
; GFX6-LABEL: mubuf_store_sgpr_ptr_offset4294967297:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_mov_b32_e32 v0, 4
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: v_mov_b32_e32 v2, 0
; GFX6-NEXT: v_mov_b32_e32 v1, 4
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; GFX6-NEXT: s_endpgm
;
; GFX7-LABEL: mubuf_store_sgpr_ptr_offset4294967297:
; GFX7: ; %bb.0:
; GFX7-NEXT: v_mov_b32_e32 v0, 4
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: v_mov_b32_e32 v2, 0
; GFX7-NEXT: v_mov_b32_e32 v1, 4
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; GFX7-NEXT: s_endpgm
;
; GFX12-LABEL: mubuf_store_sgpr_ptr_offset4294967297:
; GFX12: ; %bb.0:
; GFX12-NEXT: s_add_co_u32 s0, s2, 4
; GFX12-NEXT: s_add_co_ci_u32 s1, s3, 4
; GFX12-NEXT: v_mov_b32_e32 v0, s0
; GFX12-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s1
; GFX12-NEXT: global_store_b32 v[0:1], v2, off
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
%gep = getelementptr i32, ptr addrspace(1) %ptr, i64 4294967297
store i32 0, ptr addrspace(1) %gep
ret void
}
define amdgpu_ps void @mubuf_store_sgpr_ptr_offset4096(ptr addrspace(1) inreg %ptr) {
; GFX6-LABEL: mubuf_store_sgpr_ptr_offset4096:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: v_mov_b32_e32 v0, 0
; GFX6-NEXT: s_mov_b32 s2, -1
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: s_movk_i32 s4, 0x4000
; GFX6-NEXT: buffer_store_dword v0, off, s[0:3], s4
; GFX6-NEXT: s_endpgm
;
; GFX7-LABEL: mubuf_store_sgpr_ptr_offset4096:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: v_mov_b32_e32 v0, 0
; GFX7-NEXT: s_mov_b32 s2, -1
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: s_movk_i32 s4, 0x4000
; GFX7-NEXT: buffer_store_dword v0, off, s[0:3], s4
; GFX7-NEXT: s_endpgm
;
; GFX12-LABEL: mubuf_store_sgpr_ptr_offset4096:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_mov_b32_e32 v0, 0
; GFX12-NEXT: global_store_b32 v0, v0, s[2:3] offset:16384
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
%gep = getelementptr i32, ptr addrspace(1) %ptr, i64 4096
store i32 0, ptr addrspace(1) %gep
ret void
}
define amdgpu_ps void @mubuf_store_vgpr_ptr_offset4095(ptr addrspace(1) %ptr) {
; GFX6-LABEL: mubuf_store_vgpr_ptr_offset4095:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: v_mov_b32_e32 v2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: s_mov_b64 s[0:1], 0
; GFX6-NEXT: s_movk_i32 s4, 0x3ffc
; GFX6-NEXT: buffer_store_dword v2, v[0:1], s[0:3], s4 addr64
; GFX6-NEXT: s_endpgm
;
; GFX7-LABEL: mubuf_store_vgpr_ptr_offset4095:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: v_mov_b32_e32 v2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: s_mov_b64 s[0:1], 0
; GFX7-NEXT: s_movk_i32 s4, 0x3ffc
; GFX7-NEXT: buffer_store_dword v2, v[0:1], s[0:3], s4 addr64
; GFX7-NEXT: s_endpgm
;
; GFX12-LABEL: mubuf_store_vgpr_ptr_offset4095:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_mov_b32_e32 v2, 0
; GFX12-NEXT: global_store_b32 v[0:1], v2, off offset:16380
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
%gep = getelementptr i32, ptr addrspace(1) %ptr, i64 4095
store i32 0, ptr addrspace(1) %gep
ret void
}
define amdgpu_ps void @mubuf_store_vgpr_ptr_offset4294967296(ptr addrspace(1) %ptr) {
; GFX6-LABEL: mubuf_store_vgpr_ptr_offset4294967296:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s0, 0
; GFX6-NEXT: s_mov_b32 s1, 4
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: v_mov_b32_e32 v2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; GFX6-NEXT: s_endpgm
;
; GFX7-LABEL: mubuf_store_vgpr_ptr_offset4294967296:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s0, 0
; GFX7-NEXT: s_mov_b32 s1, 4
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: v_mov_b32_e32 v2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; GFX7-NEXT: s_endpgm
;
; GFX12-LABEL: mubuf_store_vgpr_ptr_offset4294967296:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v0, 0
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, 4, v1, vcc_lo
; GFX12-NEXT: v_mov_b32_e32 v2, 0
; GFX12-NEXT: global_store_b32 v[0:1], v2, off
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
%gep = getelementptr i32, ptr addrspace(1) %ptr, i64 4294967296
store i32 0, ptr addrspace(1) %gep
ret void
}
define amdgpu_ps void @mubuf_store_vgpr_ptr_offset4294967297(ptr addrspace(1) %ptr) {
; GFX6-LABEL: mubuf_store_vgpr_ptr_offset4294967297:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s0, 4
; GFX6-NEXT: s_mov_b32 s1, 4
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: v_mov_b32_e32 v2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; GFX6-NEXT: s_endpgm
;
; GFX7-LABEL: mubuf_store_vgpr_ptr_offset4294967297:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s0, 4
; GFX7-NEXT: s_mov_b32 s1, 4
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: v_mov_b32_e32 v2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; GFX7-NEXT: s_endpgm
;
; GFX12-LABEL: mubuf_store_vgpr_ptr_offset4294967297:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v0, 4
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, 4, v1, vcc_lo
; GFX12-NEXT: v_mov_b32_e32 v2, 0
; GFX12-NEXT: global_store_b32 v[0:1], v2, off
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
%gep = getelementptr i32, ptr addrspace(1) %ptr, i64 4294967297
store i32 0, ptr addrspace(1) %gep
ret void
}
define amdgpu_ps void @mubuf_store_vgpr_ptr_offset4096(ptr addrspace(1) %ptr) {
; GFX6-LABEL: mubuf_store_vgpr_ptr_offset4096:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: v_mov_b32_e32 v2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: s_mov_b64 s[0:1], 0
; GFX6-NEXT: s_movk_i32 s4, 0x4000
; GFX6-NEXT: buffer_store_dword v2, v[0:1], s[0:3], s4 addr64
; GFX6-NEXT: s_endpgm
;
; GFX7-LABEL: mubuf_store_vgpr_ptr_offset4096:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: v_mov_b32_e32 v2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: s_mov_b64 s[0:1], 0
; GFX7-NEXT: s_movk_i32 s4, 0x4000
; GFX7-NEXT: buffer_store_dword v2, v[0:1], s[0:3], s4 addr64
; GFX7-NEXT: s_endpgm
;
; GFX12-LABEL: mubuf_store_vgpr_ptr_offset4096:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_mov_b32_e32 v2, 0
; GFX12-NEXT: global_store_b32 v[0:1], v2, off offset:16384
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
%gep = getelementptr i32, ptr addrspace(1) %ptr, i64 4096
store i32 0, ptr addrspace(1) %gep
ret void
}
define amdgpu_ps void @mubuf_store_sgpr_ptr_sgpr_offset(ptr addrspace(1) inreg %ptr, i32 inreg %soffset) {
; GFX6-LABEL: mubuf_store_sgpr_ptr_sgpr_offset:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_ashr_i32 s5, s4, 31
; GFX6-NEXT: s_lshl_b64 s[4:5], s[4:5], 2
; GFX6-NEXT: v_mov_b32_e32 v0, s4
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: v_mov_b32_e32 v2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: v_mov_b32_e32 v1, s5
; GFX6-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; GFX6-NEXT: s_endpgm
;
; GFX7-LABEL: mubuf_store_sgpr_ptr_sgpr_offset:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_ashr_i32 s5, s4, 31
; GFX7-NEXT: s_lshl_b64 s[4:5], s[4:5], 2
; GFX7-NEXT: v_mov_b32_e32 v0, s4
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: v_mov_b32_e32 v2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: v_mov_b32_e32 v1, s5
; GFX7-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; GFX7-NEXT: s_endpgm
;
; GFX12-LABEL: mubuf_store_sgpr_ptr_sgpr_offset:
; GFX12: ; %bb.0:
; GFX12-NEXT: s_ashr_i32 s5, s4, 31
; GFX12-NEXT: v_mov_b32_e32 v0, 0
; GFX12-NEXT: s_lshl_b64 s[0:1], s[4:5], 2
; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX12-NEXT: s_add_co_u32 s0, s2, s0
; GFX12-NEXT: s_add_co_ci_u32 s1, s3, s1
; GFX12-NEXT: global_store_b32 v0, v0, s[0:1]
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
%gep = getelementptr i32, ptr addrspace(1) %ptr, i32 %soffset
store i32 0, ptr addrspace(1) %gep
ret void
}
define amdgpu_ps void @mubuf_store_vgpr_ptr_sgpr_offset(ptr addrspace(1) %ptr, i32 inreg %soffset) {
; GFX6-LABEL: mubuf_store_vgpr_ptr_sgpr_offset:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_ashr_i32 s3, s2, 31
; GFX6-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: v_mov_b32_e32 v2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; GFX6-NEXT: s_endpgm
;
; GFX7-LABEL: mubuf_store_vgpr_ptr_sgpr_offset:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_ashr_i32 s3, s2, 31
; GFX7-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: v_mov_b32_e32 v2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; GFX7-NEXT: s_endpgm
;
; GFX12-LABEL: mubuf_store_vgpr_ptr_sgpr_offset:
; GFX12: ; %bb.0:
; GFX12-NEXT: s_ashr_i32 s3, s2, 31
; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX12-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX12-NEXT: v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo
; GFX12-NEXT: v_mov_b32_e32 v2, 0
; GFX12-NEXT: global_store_b32 v[0:1], v2, off
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
%gep = getelementptr i32, ptr addrspace(1) %ptr, i32 %soffset
store i32 0, ptr addrspace(1) %gep
ret void
}
define amdgpu_ps void @mubuf_store_vgpr_ptr_sgpr_offset_offset256(ptr addrspace(1) %ptr, i32 inreg %soffset) {
; GFX6-LABEL: mubuf_store_vgpr_ptr_sgpr_offset_offset256:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_ashr_i32 s3, s2, 31
; GFX6-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: v_mov_b32_e32 v2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64 offset:1024
; GFX6-NEXT: s_endpgm
;
; GFX7-LABEL: mubuf_store_vgpr_ptr_sgpr_offset_offset256:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_ashr_i32 s3, s2, 31
; GFX7-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: v_mov_b32_e32 v2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64 offset:1024
; GFX7-NEXT: s_endpgm
;
; GFX12-LABEL: mubuf_store_vgpr_ptr_sgpr_offset_offset256:
; GFX12: ; %bb.0:
; GFX12-NEXT: s_ashr_i32 s3, s2, 31
; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX12-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX12-NEXT: v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo
; GFX12-NEXT: v_mov_b32_e32 v2, 0
; GFX12-NEXT: global_store_b32 v[0:1], v2, off offset:1024
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
%gep0 = getelementptr i32, ptr addrspace(1) %ptr, i32 %soffset
%gep1 = getelementptr i32, ptr addrspace(1) %gep0, i32 256
store i32 0, ptr addrspace(1) %gep1
ret void
}
define amdgpu_ps void @mubuf_store_vgpr_ptr_sgpr_offset256_offset(ptr addrspace(1) %ptr, i32 inreg %soffset) {
; GFX6-LABEL: mubuf_store_vgpr_ptr_sgpr_offset256_offset:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_ashr_i32 s3, s2, 31
; GFX6-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: v_mov_b32_e32 v2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64 offset:1024
; GFX6-NEXT: s_endpgm
;
; GFX7-LABEL: mubuf_store_vgpr_ptr_sgpr_offset256_offset:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_ashr_i32 s3, s2, 31
; GFX7-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: v_mov_b32_e32 v2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64 offset:1024
; GFX7-NEXT: s_endpgm
;
; GFX12-LABEL: mubuf_store_vgpr_ptr_sgpr_offset256_offset:
; GFX12: ; %bb.0:
; GFX12-NEXT: s_ashr_i32 s3, s2, 31
; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX12-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX12-NEXT: v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo
; GFX12-NEXT: v_mov_b32_e32 v2, 0
; GFX12-NEXT: global_store_b32 v[0:1], v2, off offset:1024
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
%gep0 = getelementptr i32, ptr addrspace(1) %ptr, i32 256
%gep1 = getelementptr i32, ptr addrspace(1) %gep0, i32 %soffset
store i32 0, ptr addrspace(1) %gep1
ret void
}
define amdgpu_ps void @mubuf_store_sgpr_ptr_vgpr_offset(ptr addrspace(1) inreg %ptr, i32 %voffset) {
; GFX6-LABEL: mubuf_store_sgpr_ptr_vgpr_offset:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], 2
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: v_mov_b32_e32 v2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; GFX6-NEXT: s_endpgm
;
; GFX7-LABEL: mubuf_store_sgpr_ptr_vgpr_offset:
; GFX7: ; %bb.0:
; GFX7-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX7-NEXT: v_lshl_b64 v[0:1], v[0:1], 2
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: v_mov_b32_e32 v2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_store_dword v2, v[0:1], s[0:3], 0 addr64
; GFX7-NEXT: s_endpgm
;
; GFX12-LABEL: mubuf_store_sgpr_ptr_vgpr_offset:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX12-NEXT: v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-NEXT: v_lshlrev_b64_e32 v[0:1], 2, v[0:1]
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v2, v0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2)
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v3, v1, vcc_lo
; GFX12-NEXT: v_mov_b32_e32 v2, 0
; GFX12-NEXT: global_store_b32 v[0:1], v2, off
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
%gep = getelementptr i32, ptr addrspace(1) %ptr, i32 %voffset
store i32 0, ptr addrspace(1) %gep
ret void
}
define amdgpu_ps void @mubuf_store_sgpr_ptr_vgpr_offset_offset4095(ptr addrspace(1) inreg %ptr, i32 %voffset) {
; GFX6-LABEL: mubuf_store_sgpr_ptr_vgpr_offset_offset4095:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], 2
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: v_mov_b32_e32 v2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: s_movk_i32 s4, 0x3ffc
; GFX6-NEXT: buffer_store_dword v2, v[0:1], s[0:3], s4 addr64
; GFX6-NEXT: s_endpgm
;
; GFX7-LABEL: mubuf_store_sgpr_ptr_vgpr_offset_offset4095:
; GFX7: ; %bb.0:
; GFX7-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX7-NEXT: v_lshl_b64 v[0:1], v[0:1], 2
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: v_mov_b32_e32 v2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: s_movk_i32 s4, 0x3ffc
; GFX7-NEXT: buffer_store_dword v2, v[0:1], s[0:3], s4 addr64
; GFX7-NEXT: s_endpgm
;
; GFX12-LABEL: mubuf_store_sgpr_ptr_vgpr_offset_offset4095:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX12-NEXT: v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-NEXT: v_lshlrev_b64_e32 v[0:1], 2, v[0:1]
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v2, v0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2)
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v3, v1, vcc_lo
; GFX12-NEXT: v_mov_b32_e32 v2, 0
; GFX12-NEXT: global_store_b32 v[0:1], v2, off offset:16380
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
%gep0 = getelementptr i32, ptr addrspace(1) %ptr, i32 %voffset
%gep1 = getelementptr i32, ptr addrspace(1) %gep0, i32 4095
store i32 0, ptr addrspace(1) %gep1
ret void
}
define amdgpu_ps void @mubuf_store_sgpr_ptr_offset4095_vgpr_offset(ptr addrspace(1) inreg %ptr, i32 %voffset) {
; GFX6-LABEL: mubuf_store_sgpr_ptr_offset4095_vgpr_offset:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], 2
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: v_mov_b32_e32 v2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: s_movk_i32 s4, 0x3ffc
; GFX6-NEXT: buffer_store_dword v2, v[0:1], s[0:3], s4 addr64
; GFX6-NEXT: s_endpgm
;
; GFX7-LABEL: mubuf_store_sgpr_ptr_offset4095_vgpr_offset:
; GFX7: ; %bb.0:
; GFX7-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX7-NEXT: v_lshl_b64 v[0:1], v[0:1], 2
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: v_mov_b32_e32 v2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: s_movk_i32 s4, 0x3ffc
; GFX7-NEXT: buffer_store_dword v2, v[0:1], s[0:3], s4 addr64
; GFX7-NEXT: s_endpgm
;
; GFX12-LABEL: mubuf_store_sgpr_ptr_offset4095_vgpr_offset:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX12-NEXT: v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-NEXT: v_lshlrev_b64_e32 v[0:1], 2, v[0:1]
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v2, v0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2)
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v3, v1, vcc_lo
; GFX12-NEXT: v_mov_b32_e32 v2, 0
; GFX12-NEXT: global_store_b32 v[0:1], v2, off offset:16380
; GFX12-NEXT: s_nop 0
; GFX12-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX12-NEXT: s_endpgm
%gep0 = getelementptr i32, ptr addrspace(1) %ptr, i32 4095
%gep1 = getelementptr i32, ptr addrspace(1) %gep0, i32 %voffset
store i32 0, ptr addrspace(1) %gep1
ret void
}
define amdgpu_ps float @mubuf_load_sgpr_ptr(ptr addrspace(1) inreg %ptr) {
; GFX6-LABEL: mubuf_load_sgpr_ptr:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: s_mov_b32 s2, -1
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_load_sgpr_ptr:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: s_mov_b32 s2, -1
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_load_sgpr_ptr:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_mov_b32_e32 v0, 0
; GFX12-NEXT: global_load_b32 v0, v0, s[2:3] scope:SCOPE_SYS
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: ; return to shader part epilog
%val = load volatile float, ptr addrspace(1) %ptr
ret float %val
}
define amdgpu_ps float @mubuf_load_sgpr_ptr_offset4095(ptr addrspace(1) inreg %ptr) {
; GFX6-LABEL: mubuf_load_sgpr_ptr_offset4095:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: s_mov_b32 s2, -1
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: s_movk_i32 s4, 0x3ffc
; GFX6-NEXT: buffer_load_dword v0, off, s[0:3], s4 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_load_sgpr_ptr_offset4095:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: s_mov_b32 s2, -1
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: s_movk_i32 s4, 0x3ffc
; GFX7-NEXT: buffer_load_dword v0, off, s[0:3], s4 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_load_sgpr_ptr_offset4095:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_mov_b32_e32 v0, 0
; GFX12-NEXT: global_load_b32 v0, v0, s[2:3] offset:16380 scope:SCOPE_SYS
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr float, ptr addrspace(1) %ptr, i64 4095
%val = load volatile float, ptr addrspace(1) %gep
ret float %val
}
define amdgpu_ps float @mubuf_load_sgpr_ptr_offset4294967296(ptr addrspace(1) inreg %ptr) {
; GFX6-LABEL: mubuf_load_sgpr_ptr_offset4294967296:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_mov_b32_e32 v0, 0
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: v_mov_b32_e32 v1, 4
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_load_sgpr_ptr_offset4294967296:
; GFX7: ; %bb.0:
; GFX7-NEXT: v_mov_b32_e32 v0, 0
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: v_mov_b32_e32 v1, 4
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_load_sgpr_ptr_offset4294967296:
; GFX12: ; %bb.0:
; GFX12-NEXT: s_add_co_u32 s0, s2, 0
; GFX12-NEXT: s_add_co_ci_u32 s1, s3, 4
; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX12-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
; GFX12-NEXT: global_load_b32 v0, v[0:1], off scope:SCOPE_SYS
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr float, ptr addrspace(1) %ptr, i64 4294967296
%val = load volatile float, ptr addrspace(1) %gep
ret float %val
}
define amdgpu_ps float @mubuf_load_sgpr_ptr_offset4294967297(ptr addrspace(1) inreg %ptr) {
; GFX6-LABEL: mubuf_load_sgpr_ptr_offset4294967297:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_mov_b32_e32 v0, 4
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: v_mov_b32_e32 v1, 4
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_load_sgpr_ptr_offset4294967297:
; GFX7: ; %bb.0:
; GFX7-NEXT: v_mov_b32_e32 v0, 4
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: v_mov_b32_e32 v1, 4
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_load_sgpr_ptr_offset4294967297:
; GFX12: ; %bb.0:
; GFX12-NEXT: s_add_co_u32 s0, s2, 4
; GFX12-NEXT: s_add_co_ci_u32 s1, s3, 4
; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX12-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
; GFX12-NEXT: global_load_b32 v0, v[0:1], off scope:SCOPE_SYS
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr float, ptr addrspace(1) %ptr, i64 4294967297
%val = load volatile float, ptr addrspace(1) %gep
ret float %val
}
define amdgpu_ps float @mubuf_load_sgpr_ptr_offset4096(ptr addrspace(1) inreg %ptr) {
; GFX6-LABEL: mubuf_load_sgpr_ptr_offset4096:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: s_mov_b32 s2, -1
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: s_movk_i32 s4, 0x4000
; GFX6-NEXT: buffer_load_dword v0, off, s[0:3], s4 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_load_sgpr_ptr_offset4096:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: s_mov_b32 s2, -1
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: s_movk_i32 s4, 0x4000
; GFX7-NEXT: buffer_load_dword v0, off, s[0:3], s4 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_load_sgpr_ptr_offset4096:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_mov_b32_e32 v0, 0
; GFX12-NEXT: global_load_b32 v0, v0, s[2:3] offset:16384 scope:SCOPE_SYS
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr float, ptr addrspace(1) %ptr, i64 4096
%val = load volatile float, ptr addrspace(1) %gep
ret float %val
}
define amdgpu_ps float @mubuf_load_vgpr_ptr_offset4095(ptr addrspace(1) %ptr) {
; GFX6-LABEL: mubuf_load_vgpr_ptr_offset4095:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: s_mov_b64 s[0:1], 0
; GFX6-NEXT: s_movk_i32 s4, 0x3ffc
; GFX6-NEXT: buffer_load_dword v0, v[0:1], s[0:3], s4 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_load_vgpr_ptr_offset4095:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: s_mov_b64 s[0:1], 0
; GFX7-NEXT: s_movk_i32 s4, 0x3ffc
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[0:3], s4 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_load_vgpr_ptr_offset4095:
; GFX12: ; %bb.0:
; GFX12-NEXT: global_load_b32 v0, v[0:1], off offset:16380 scope:SCOPE_SYS
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr float, ptr addrspace(1) %ptr, i64 4095
%val = load volatile float, ptr addrspace(1) %gep
ret float %val
}
define amdgpu_ps float @mubuf_load_vgpr_ptr_offset4294967296(ptr addrspace(1) %ptr) {
; GFX6-LABEL: mubuf_load_vgpr_ptr_offset4294967296:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s0, 0
; GFX6-NEXT: s_mov_b32 s1, 4
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_load_vgpr_ptr_offset4294967296:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s0, 0
; GFX7-NEXT: s_mov_b32 s1, 4
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_load_vgpr_ptr_offset4294967296:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v0, 0
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, 4, v1, vcc_lo
; GFX12-NEXT: global_load_b32 v0, v[0:1], off scope:SCOPE_SYS
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr float, ptr addrspace(1) %ptr, i64 4294967296
%val = load volatile float, ptr addrspace(1) %gep
ret float %val
}
define amdgpu_ps float @mubuf_load_vgpr_ptr_offset4294967297(ptr addrspace(1) %ptr) {
; GFX6-LABEL: mubuf_load_vgpr_ptr_offset4294967297:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s0, 4
; GFX6-NEXT: s_mov_b32 s1, 4
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_load_vgpr_ptr_offset4294967297:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s0, 4
; GFX7-NEXT: s_mov_b32 s1, 4
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_load_vgpr_ptr_offset4294967297:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v0, 4
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, 4, v1, vcc_lo
; GFX12-NEXT: global_load_b32 v0, v[0:1], off scope:SCOPE_SYS
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr float, ptr addrspace(1) %ptr, i64 4294967297
%val = load volatile float, ptr addrspace(1) %gep
ret float %val
}
define amdgpu_ps float @mubuf_load_vgpr_ptr_offset4096(ptr addrspace(1) %ptr) {
; GFX6-LABEL: mubuf_load_vgpr_ptr_offset4096:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: s_mov_b64 s[0:1], 0
; GFX6-NEXT: s_movk_i32 s4, 0x4000
; GFX6-NEXT: buffer_load_dword v0, v[0:1], s[0:3], s4 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_load_vgpr_ptr_offset4096:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: s_mov_b64 s[0:1], 0
; GFX7-NEXT: s_movk_i32 s4, 0x4000
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[0:3], s4 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_load_vgpr_ptr_offset4096:
; GFX12: ; %bb.0:
; GFX12-NEXT: global_load_b32 v0, v[0:1], off offset:16384 scope:SCOPE_SYS
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr float, ptr addrspace(1) %ptr, i64 4096
%val = load volatile float, ptr addrspace(1) %gep
ret float %val
}
define amdgpu_ps float @mubuf_load_sgpr_ptr_sgpr_offset(ptr addrspace(1) inreg %ptr, i32 inreg %soffset) {
; GFX6-LABEL: mubuf_load_sgpr_ptr_sgpr_offset:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_ashr_i32 s5, s4, 31
; GFX6-NEXT: s_lshl_b64 s[4:5], s[4:5], 2
; GFX6-NEXT: v_mov_b32_e32 v0, s4
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: v_mov_b32_e32 v1, s5
; GFX6-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_load_sgpr_ptr_sgpr_offset:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_ashr_i32 s5, s4, 31
; GFX7-NEXT: s_lshl_b64 s[4:5], s[4:5], 2
; GFX7-NEXT: v_mov_b32_e32 v0, s4
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: v_mov_b32_e32 v1, s5
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_load_sgpr_ptr_sgpr_offset:
; GFX12: ; %bb.0:
; GFX12-NEXT: s_ashr_i32 s5, s4, 31
; GFX12-NEXT: v_mov_b32_e32 v0, 0
; GFX12-NEXT: s_lshl_b64 s[0:1], s[4:5], 2
; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX12-NEXT: s_add_co_u32 s0, s2, s0
; GFX12-NEXT: s_add_co_ci_u32 s1, s3, s1
; GFX12-NEXT: global_load_b32 v0, v0, s[0:1] scope:SCOPE_SYS
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr float, ptr addrspace(1) %ptr, i32 %soffset
%val = load volatile float, ptr addrspace(1) %gep
ret float %val
}
define amdgpu_ps float @mubuf_load_vgpr_ptr_sgpr_offset(ptr addrspace(1) %ptr, i32 inreg %soffset) {
; GFX6-LABEL: mubuf_load_vgpr_ptr_sgpr_offset:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_ashr_i32 s3, s2, 31
; GFX6-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_load_vgpr_ptr_sgpr_offset:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_ashr_i32 s3, s2, 31
; GFX7-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_load_vgpr_ptr_sgpr_offset:
; GFX12: ; %bb.0:
; GFX12-NEXT: s_ashr_i32 s3, s2, 31
; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX12-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX12-NEXT: v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo
; GFX12-NEXT: global_load_b32 v0, v[0:1], off scope:SCOPE_SYS
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr float, ptr addrspace(1) %ptr, i32 %soffset
%val = load volatile float, ptr addrspace(1) %gep
ret float %val
}
define amdgpu_ps float @mubuf_load_vgpr_ptr_sgpr_offset_offset256(ptr addrspace(1) %ptr, i32 inreg %soffset) {
; GFX6-LABEL: mubuf_load_vgpr_ptr_sgpr_offset_offset256:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_ashr_i32 s3, s2, 31
; GFX6-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 offset:1024 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_load_vgpr_ptr_sgpr_offset_offset256:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_ashr_i32 s3, s2, 31
; GFX7-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 offset:1024 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_load_vgpr_ptr_sgpr_offset_offset256:
; GFX12: ; %bb.0:
; GFX12-NEXT: s_ashr_i32 s3, s2, 31
; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX12-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX12-NEXT: v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo
; GFX12-NEXT: global_load_b32 v0, v[0:1], off offset:1024 scope:SCOPE_SYS
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: ; return to shader part epilog
%gep0 = getelementptr float, ptr addrspace(1) %ptr, i32 %soffset
%gep1 = getelementptr float, ptr addrspace(1) %gep0, i32 256
%val = load volatile float, ptr addrspace(1) %gep1
ret float %val
}
define amdgpu_ps float @mubuf_load_vgpr_ptr_sgpr_offset256_offset(ptr addrspace(1) %ptr, i32 inreg %soffset) {
; GFX6-LABEL: mubuf_load_vgpr_ptr_sgpr_offset256_offset:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_ashr_i32 s3, s2, 31
; GFX6-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 offset:1024 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_load_vgpr_ptr_sgpr_offset256_offset:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_ashr_i32 s3, s2, 31
; GFX7-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 offset:1024 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_load_vgpr_ptr_sgpr_offset256_offset:
; GFX12: ; %bb.0:
; GFX12-NEXT: s_ashr_i32 s3, s2, 31
; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX12-NEXT: s_lshl_b64 s[0:1], s[2:3], 2
; GFX12-NEXT: v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo
; GFX12-NEXT: global_load_b32 v0, v[0:1], off offset:1024 scope:SCOPE_SYS
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: ; return to shader part epilog
%gep0 = getelementptr float, ptr addrspace(1) %ptr, i64 256
%gep1 = getelementptr float, ptr addrspace(1) %gep0, i32 %soffset
%val = load volatile float, ptr addrspace(1) %gep1
ret float %val
}
define amdgpu_ps float @mubuf_load_sgpr_ptr_vgpr_offset(ptr addrspace(1) inreg %ptr, i32 %voffset) {
; GFX6-LABEL: mubuf_load_sgpr_ptr_vgpr_offset:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], 2
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_load_sgpr_ptr_vgpr_offset:
; GFX7: ; %bb.0:
; GFX7-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX7-NEXT: v_lshl_b64 v[0:1], v[0:1], 2
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_load_sgpr_ptr_vgpr_offset:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX12-NEXT: v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-NEXT: v_lshlrev_b64_e32 v[0:1], 2, v[0:1]
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v2, v0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2)
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v3, v1, vcc_lo
; GFX12-NEXT: global_load_b32 v0, v[0:1], off scope:SCOPE_SYS
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr float, ptr addrspace(1) %ptr, i32 %voffset
%val = load volatile float, ptr addrspace(1) %gep
ret float %val
}
define amdgpu_ps float @mubuf_load_sgpr_ptr_vgpr_offset_offset4095(ptr addrspace(1) inreg %ptr, i32 %voffset) {
; GFX6-LABEL: mubuf_load_sgpr_ptr_vgpr_offset_offset4095:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], 2
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: s_movk_i32 s4, 0x3ffc
; GFX6-NEXT: buffer_load_dword v0, v[0:1], s[0:3], s4 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_load_sgpr_ptr_vgpr_offset_offset4095:
; GFX7: ; %bb.0:
; GFX7-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX7-NEXT: v_lshl_b64 v[0:1], v[0:1], 2
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: s_movk_i32 s4, 0x3ffc
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[0:3], s4 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_load_sgpr_ptr_vgpr_offset_offset4095:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX12-NEXT: v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-NEXT: v_lshlrev_b64_e32 v[0:1], 2, v[0:1]
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v2, v0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2)
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v3, v1, vcc_lo
; GFX12-NEXT: global_load_b32 v0, v[0:1], off offset:16380 scope:SCOPE_SYS
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: ; return to shader part epilog
%gep0 = getelementptr float, ptr addrspace(1) %ptr, i32 %voffset
%gep1 = getelementptr float, ptr addrspace(1) %gep0, i64 4095
%val = load volatile float, ptr addrspace(1) %gep1
ret float %val
}
define amdgpu_ps float @mubuf_load_sgpr_ptr_offset4095_vgpr_offset(ptr addrspace(1) inreg %ptr, i32 %voffset) {
; GFX6-LABEL: mubuf_load_sgpr_ptr_offset4095_vgpr_offset:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], 2
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: s_movk_i32 s4, 0x3ffc
; GFX6-NEXT: buffer_load_dword v0, v[0:1], s[0:3], s4 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_load_sgpr_ptr_offset4095_vgpr_offset:
; GFX7: ; %bb.0:
; GFX7-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX7-NEXT: v_lshl_b64 v[0:1], v[0:1], 2
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: s_movk_i32 s4, 0x3ffc
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[0:3], s4 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_load_sgpr_ptr_offset4095_vgpr_offset:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX12-NEXT: v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-NEXT: v_lshlrev_b64_e32 v[0:1], 2, v[0:1]
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v2, v0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2)
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v3, v1, vcc_lo
; GFX12-NEXT: global_load_b32 v0, v[0:1], off offset:16380 scope:SCOPE_SYS
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: ; return to shader part epilog
%gep0 = getelementptr float, ptr addrspace(1) %ptr, i64 4095
%gep1 = getelementptr float, ptr addrspace(1) %gep0, i32 %voffset
%val = load volatile float, ptr addrspace(1) %gep1
ret float %val
}
define amdgpu_ps float @mubuf_atomicrmw_sgpr_ptr_offset4095(ptr addrspace(1) inreg %ptr) {
; GFX6-LABEL: mubuf_atomicrmw_sgpr_ptr_offset4095:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: v_mov_b32_e32 v0, 2
; GFX6-NEXT: s_mov_b32 s2, -1
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: s_movk_i32 s4, 0x3ffc
; GFX6-NEXT: buffer_atomic_add v0, off, s[0:3], s4 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_atomicrmw_sgpr_ptr_offset4095:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: v_mov_b32_e32 v0, 2
; GFX7-NEXT: s_mov_b32 s2, -1
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: s_movk_i32 s4, 0x3ffc
; GFX7-NEXT: buffer_atomic_add v0, off, s[0:3], s4 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_atomicrmw_sgpr_ptr_offset4095:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_dual_mov_b32 v0, 2 :: v_dual_mov_b32 v1, 0
; GFX12-NEXT: global_wb scope:SCOPE_DEV
; GFX12-NEXT: global_atomic_add_u32 v0, v1, v0, s[2:3] offset:16380 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_DEV
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr i32, ptr addrspace(1) %ptr, i64 4095
%result = atomicrmw add ptr addrspace(1) %gep, i32 2 syncscope("agent") seq_cst
%cast = bitcast i32 %result to float
ret float %cast
}
define amdgpu_ps float @mubuf_atomicrmw_sgpr_ptr_offset4294967296(ptr addrspace(1) inreg %ptr) {
; GFX6-LABEL: mubuf_atomicrmw_sgpr_ptr_offset4294967296:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_mov_b32_e32 v1, 0
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: v_mov_b32_e32 v0, 2
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: v_mov_b32_e32 v2, 4
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_atomic_add v0, v[1:2], s[0:3], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_atomicrmw_sgpr_ptr_offset4294967296:
; GFX7: ; %bb.0:
; GFX7-NEXT: v_mov_b32_e32 v1, 0
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: v_mov_b32_e32 v0, 2
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: v_mov_b32_e32 v2, 4
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_atomic_add v0, v[1:2], s[0:3], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_atomicrmw_sgpr_ptr_offset4294967296:
; GFX12: ; %bb.0:
; GFX12-NEXT: s_add_co_u32 s0, s2, 0
; GFX12-NEXT: s_add_co_ci_u32 s1, s3, 4
; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX12-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
; GFX12-NEXT: v_mov_b32_e32 v2, 2
; GFX12-NEXT: global_wb scope:SCOPE_DEV
; GFX12-NEXT: global_atomic_add_u32 v0, v[0:1], v2, off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_DEV
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr i32, ptr addrspace(1) %ptr, i64 4294967296
%result = atomicrmw add ptr addrspace(1) %gep, i32 2 syncscope("agent") seq_cst
%cast = bitcast i32 %result to float
ret float %cast
}
define amdgpu_ps float @mubuf_atomicrmw_vgpr_ptr_offset4095(ptr addrspace(1) %ptr) {
; GFX6-LABEL: mubuf_atomicrmw_vgpr_ptr_offset4095:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_mov_b32_e32 v2, 2
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: s_mov_b64 s[0:1], 0
; GFX6-NEXT: s_movk_i32 s4, 0x3ffc
; GFX6-NEXT: buffer_atomic_add v2, v[0:1], s[0:3], s4 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
; GFX6-NEXT: v_mov_b32_e32 v0, v2
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_atomicrmw_vgpr_ptr_offset4095:
; GFX7: ; %bb.0:
; GFX7-NEXT: v_mov_b32_e32 v2, 2
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: s_mov_b64 s[0:1], 0
; GFX7-NEXT: s_movk_i32 s4, 0x3ffc
; GFX7-NEXT: buffer_atomic_add v2, v[0:1], s[0:3], s4 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
; GFX7-NEXT: v_mov_b32_e32 v0, v2
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_atomicrmw_vgpr_ptr_offset4095:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_mov_b32_e32 v2, 2
; GFX12-NEXT: global_wb scope:SCOPE_DEV
; GFX12-NEXT: global_atomic_add_u32 v0, v[0:1], v2, off offset:16380 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_DEV
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr i32, ptr addrspace(1) %ptr, i64 4095
%result = atomicrmw add ptr addrspace(1) %gep, i32 2 syncscope("agent") seq_cst
%cast = bitcast i32 %result to float
ret float %cast
}
define amdgpu_ps float @mubuf_atomicrmw_vgpr_ptr_offset4294967296(ptr addrspace(1) %ptr) {
; GFX6-LABEL: mubuf_atomicrmw_vgpr_ptr_offset4294967296:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s0, 0
; GFX6-NEXT: s_mov_b32 s1, 4
; GFX6-NEXT: v_mov_b32_e32 v2, 2
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_atomic_add v2, v[0:1], s[0:3], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
; GFX6-NEXT: v_mov_b32_e32 v0, v2
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_atomicrmw_vgpr_ptr_offset4294967296:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s0, 0
; GFX7-NEXT: s_mov_b32 s1, 4
; GFX7-NEXT: v_mov_b32_e32 v2, 2
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_atomic_add v2, v[0:1], s[0:3], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
; GFX7-NEXT: v_mov_b32_e32 v0, v2
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_atomicrmw_vgpr_ptr_offset4294967296:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v0, 0
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, 4, v1, vcc_lo
; GFX12-NEXT: v_mov_b32_e32 v2, 2
; GFX12-NEXT: global_wb scope:SCOPE_DEV
; GFX12-NEXT: global_atomic_add_u32 v0, v[0:1], v2, off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_DEV
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr i32, ptr addrspace(1) %ptr, i64 4294967296
%result = atomicrmw add ptr addrspace(1) %gep, i32 2 syncscope("agent") seq_cst
%cast = bitcast i32 %result to float
ret float %cast
}
define amdgpu_ps float @mubuf_atomicrmw_sgpr_ptr_vgpr_offset(ptr addrspace(1) inreg %ptr, i32 %voffset) {
; GFX6-LABEL: mubuf_atomicrmw_sgpr_ptr_vgpr_offset:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], 2
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: v_mov_b32_e32 v2, 2
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_atomic_add v2, v[0:1], s[0:3], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
; GFX6-NEXT: v_mov_b32_e32 v0, v2
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_atomicrmw_sgpr_ptr_vgpr_offset:
; GFX7: ; %bb.0:
; GFX7-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX7-NEXT: v_lshl_b64 v[0:1], v[0:1], 2
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: v_mov_b32_e32 v2, 2
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_atomic_add v2, v[0:1], s[0:3], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
; GFX7-NEXT: v_mov_b32_e32 v0, v2
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_atomicrmw_sgpr_ptr_vgpr_offset:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX12-NEXT: v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
; GFX12-NEXT: v_mov_b32_e32 v4, 2
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-NEXT: v_lshlrev_b64_e32 v[0:1], 2, v[0:1]
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v2, v0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2)
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v3, v1, vcc_lo
; GFX12-NEXT: global_wb scope:SCOPE_DEV
; GFX12-NEXT: global_atomic_add_u32 v0, v[0:1], v4, off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_DEV
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr i32, ptr addrspace(1) %ptr, i32 %voffset
%result = atomicrmw add ptr addrspace(1) %gep, i32 2 syncscope("agent") seq_cst
%cast = bitcast i32 %result to float
ret float %cast
}
define amdgpu_ps float @mubuf_cmpxchg_sgpr_ptr_offset4095(ptr addrspace(1) inreg %ptr, i32 %old, i32 %in) {
; GFX6-LABEL: mubuf_cmpxchg_sgpr_ptr_offset4095:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: v_mov_b32_e32 v2, v0
; GFX6-NEXT: s_mov_b32 s2, -1
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: s_movk_i32 s4, 0x3ffc
; GFX6-NEXT: buffer_atomic_cmpswap v[1:2], off, s[0:3], s4 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
; GFX6-NEXT: v_mov_b32_e32 v0, v1
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_cmpxchg_sgpr_ptr_offset4095:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: v_mov_b32_e32 v2, v0
; GFX7-NEXT: s_mov_b32 s2, -1
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: s_movk_i32 s4, 0x3ffc
; GFX7-NEXT: buffer_atomic_cmpswap v[1:2], off, s[0:3], s4 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
; GFX7-NEXT: v_mov_b32_e32 v0, v1
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_cmpxchg_sgpr_ptr_offset4095:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_mov_b32_e32 v2, v0
; GFX12-NEXT: v_mov_b32_e32 v0, 0
; GFX12-NEXT: global_wb scope:SCOPE_DEV
; GFX12-NEXT: global_atomic_cmpswap_b32 v0, v0, v[1:2], s[2:3] offset:16380 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_DEV
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr i32, ptr addrspace(1) %ptr, i64 4095
%result.struct = cmpxchg ptr addrspace(1) %gep, i32 %old, i32 %in syncscope("agent") seq_cst seq_cst
%result = extractvalue { i32, i1 } %result.struct, 0
%cast = bitcast i32 %result to float
ret float %cast
}
define amdgpu_ps float @mubuf_cmpxchg_sgpr_ptr_offset4294967296(ptr addrspace(1) inreg %ptr, i32 %old, i32 %in) {
; GFX6-LABEL: mubuf_cmpxchg_sgpr_ptr_offset4294967296:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_mov_b32_e32 v3, 0
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: v_mov_b32_e32 v2, v0
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: v_mov_b32_e32 v4, 4
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_atomic_cmpswap v[1:2], v[3:4], s[0:3], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
; GFX6-NEXT: v_mov_b32_e32 v0, v1
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_cmpxchg_sgpr_ptr_offset4294967296:
; GFX7: ; %bb.0:
; GFX7-NEXT: v_mov_b32_e32 v3, 0
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: v_mov_b32_e32 v2, v0
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: v_mov_b32_e32 v4, 4
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_atomic_cmpswap v[1:2], v[3:4], s[0:3], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
; GFX7-NEXT: v_mov_b32_e32 v0, v1
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_cmpxchg_sgpr_ptr_offset4294967296:
; GFX12: ; %bb.0:
; GFX12-NEXT: s_add_co_u32 s0, s2, 0
; GFX12-NEXT: s_add_co_ci_u32 s1, s3, 4
; GFX12-NEXT: v_mov_b32_e32 v2, v0
; GFX12-NEXT: v_dual_mov_b32 v4, s1 :: v_dual_mov_b32 v3, s0
; GFX12-NEXT: global_wb scope:SCOPE_DEV
; GFX12-NEXT: global_atomic_cmpswap_b32 v0, v[3:4], v[1:2], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_DEV
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr i32, ptr addrspace(1) %ptr, i64 4294967296
%result.struct = cmpxchg ptr addrspace(1) %gep, i32 %old, i32 %in syncscope("agent") seq_cst seq_cst
%result = extractvalue { i32, i1 } %result.struct, 0
%cast = bitcast i32 %result to float
ret float %cast
}
define amdgpu_ps float @mubuf_cmpxchg_vgpr_ptr_offset4095(ptr addrspace(1) %ptr, i32 %old, i32 %in) {
; GFX6-LABEL: mubuf_cmpxchg_vgpr_ptr_offset4095:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_mov_b32_e32 v4, v2
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: s_mov_b64 s[0:1], 0
; GFX6-NEXT: s_movk_i32 s4, 0x3ffc
; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[0:3], s4 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
; GFX6-NEXT: v_mov_b32_e32 v0, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_cmpxchg_vgpr_ptr_offset4095:
; GFX7: ; %bb.0:
; GFX7-NEXT: v_mov_b32_e32 v4, v2
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: s_mov_b64 s[0:1], 0
; GFX7-NEXT: s_movk_i32 s4, 0x3ffc
; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[0:3], s4 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
; GFX7-NEXT: v_mov_b32_e32 v0, v3
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_cmpxchg_vgpr_ptr_offset4095:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_mov_b32_e32 v4, v2
; GFX12-NEXT: global_wb scope:SCOPE_DEV
; GFX12-NEXT: global_atomic_cmpswap_b32 v0, v[0:1], v[3:4], off offset:16380 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_DEV
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr i32, ptr addrspace(1) %ptr, i64 4095
%result.struct = cmpxchg ptr addrspace(1) %gep, i32 %old, i32 %in syncscope("agent") seq_cst seq_cst
%result = extractvalue { i32, i1 } %result.struct, 0
%cast = bitcast i32 %result to float
ret float %cast
}
define amdgpu_ps float @mubuf_cmpxchg_vgpr_ptr_offset4294967296(ptr addrspace(1) %ptr, i32 %old, i32 %in) {
; GFX6-LABEL: mubuf_cmpxchg_vgpr_ptr_offset4294967296:
; GFX6: ; %bb.0:
; GFX6-NEXT: s_mov_b32 s0, 0
; GFX6-NEXT: v_mov_b32_e32 v4, v2
; GFX6-NEXT: s_mov_b32 s1, 4
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[0:3], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
; GFX6-NEXT: v_mov_b32_e32 v0, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_cmpxchg_vgpr_ptr_offset4294967296:
; GFX7: ; %bb.0:
; GFX7-NEXT: s_mov_b32 s0, 0
; GFX7-NEXT: v_mov_b32_e32 v4, v2
; GFX7-NEXT: s_mov_b32 s1, 4
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[0:3], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
; GFX7-NEXT: v_mov_b32_e32 v0, v3
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_cmpxchg_vgpr_ptr_offset4294967296:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v0, 0
; GFX12-NEXT: v_mov_b32_e32 v4, v2
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, 4, v1, vcc_lo
; GFX12-NEXT: global_wb scope:SCOPE_DEV
; GFX12-NEXT: global_atomic_cmpswap_b32 v0, v[0:1], v[3:4], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_DEV
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr i32, ptr addrspace(1) %ptr, i64 4294967296
%result.struct = cmpxchg ptr addrspace(1) %gep, i32 %old, i32 %in syncscope("agent") seq_cst seq_cst
%result = extractvalue { i32, i1 } %result.struct, 0
%cast = bitcast i32 %result to float
ret float %cast
}
define amdgpu_ps float @mubuf_cmpxchg_sgpr_ptr_vgpr_offset(ptr addrspace(1) inreg %ptr, i32 %voffset, i32 %old, i32 %in) {
; GFX6-LABEL: mubuf_cmpxchg_sgpr_ptr_vgpr_offset:
; GFX6: ; %bb.0:
; GFX6-NEXT: v_mov_b32_e32 v3, v1
; GFX6-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX6-NEXT: v_lshl_b64 v[0:1], v[0:1], 2
; GFX6-NEXT: s_mov_b32 s0, s2
; GFX6-NEXT: s_mov_b32 s1, s3
; GFX6-NEXT: s_mov_b32 s2, 0
; GFX6-NEXT: s_mov_b32 s3, 0xf000
; GFX6-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[0:3], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
; GFX6-NEXT: v_mov_b32_e32 v0, v2
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: ; return to shader part epilog
;
; GFX7-LABEL: mubuf_cmpxchg_sgpr_ptr_vgpr_offset:
; GFX7: ; %bb.0:
; GFX7-NEXT: v_mov_b32_e32 v3, v1
; GFX7-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX7-NEXT: v_lshl_b64 v[0:1], v[0:1], 2
; GFX7-NEXT: s_mov_b32 s0, s2
; GFX7-NEXT: s_mov_b32 s1, s3
; GFX7-NEXT: s_mov_b32 s2, 0
; GFX7-NEXT: s_mov_b32 s3, 0xf000
; GFX7-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[0:3], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
; GFX7-NEXT: v_mov_b32_e32 v0, v2
; GFX7-NEXT: ; return to shader part epilog
;
; GFX12-LABEL: mubuf_cmpxchg_sgpr_ptr_vgpr_offset:
; GFX12: ; %bb.0:
; GFX12-NEXT: v_mov_b32_e32 v3, v1
; GFX12-NEXT: v_ashrrev_i32_e32 v1, 31, v0
; GFX12-NEXT: v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-NEXT: v_lshlrev_b64_e32 v[0:1], 2, v[0:1]
; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v4, v0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2)
; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v5, v1, vcc_lo
; GFX12-NEXT: global_wb scope:SCOPE_DEV
; GFX12-NEXT: global_atomic_cmpswap_b32 v0, v[0:1], v[2:3], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_DEV
; GFX12-NEXT: ; return to shader part epilog
%gep = getelementptr i32, ptr addrspace(1) %ptr, i32 %voffset
%result.struct = cmpxchg ptr addrspace(1) %gep, i32 %old, i32 %in syncscope("agent") seq_cst seq_cst
%result = extractvalue { i32, i1 } %result.struct, 0
%cast = bitcast i32 %result to float
ret float %cast
}