; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GPRIDX %s
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,MOVREL %s
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10PLUS,GFX10 %s
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10PLUS,GFX11 %s
define float @dyn_extract_v8f32_const_s_v(i32 %sel) {
; GCN-LABEL: dyn_extract_v8f32_const_s_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; GCN-NEXT: v_mov_b32_e32 v1, 0x40400000
; GCN-NEXT: v_cndmask_b32_e64 v6, 1.0, 2.0, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v0
; GCN-NEXT: v_cndmask_b32_e32 v1, v6, v1, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v0
; GCN-NEXT: v_mov_b32_e32 v2, 0x40a00000
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, 4.0, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v0
; GCN-NEXT: v_mov_b32_e32 v3, 0x40c00000
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v0
; GCN-NEXT: v_mov_b32_e32 v4, 0x40e00000
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 6, v0
; GCN-NEXT: v_mov_b32_e32 v5, 0x41000000
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v4, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 7, v0
; GCN-NEXT: v_cndmask_b32_e32 v0, v1, v5, vcc
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: dyn_extract_v8f32_const_s_v:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, 1.0, 2.0, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 0x40400000, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 4.0, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 0x40a00000, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 0x40c00000, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 0x40e00000, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v0, v1, 0x41000000, vcc_lo
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
entry:
%ext = extractelement <8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, i32 %sel
ret float %ext
}
define amdgpu_ps float @dyn_extract_v8f32_const_s_s(i32 inreg %sel) {
; GPRIDX-LABEL: dyn_extract_v8f32_const_s_s:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_cmp_eq_u32 s2, 1
; GPRIDX-NEXT: s_cselect_b32 s0, 2.0, 1.0
; GPRIDX-NEXT: s_cmp_eq_u32 s2, 2
; GPRIDX-NEXT: s_cselect_b32 s0, 0x40400000, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s2, 3
; GPRIDX-NEXT: s_cselect_b32 s0, 4.0, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s2, 4
; GPRIDX-NEXT: s_cselect_b32 s0, 0x40a00000, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s2, 5
; GPRIDX-NEXT: s_cselect_b32 s0, 0x40c00000, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s2, 6
; GPRIDX-NEXT: s_cselect_b32 s0, 0x40e00000, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s2, 7
; GPRIDX-NEXT: s_cselect_b32 s0, 0x41000000, s0
; GPRIDX-NEXT: v_mov_b32_e32 v0, s0
; GPRIDX-NEXT: ; return to shader part epilog
;
; MOVREL-LABEL: dyn_extract_v8f32_const_s_s:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_mov_b32 s4, 1.0
; MOVREL-NEXT: s_mov_b32 m0, s2
; MOVREL-NEXT: s_mov_b32 s11, 0x41000000
; MOVREL-NEXT: s_mov_b32 s10, 0x40e00000
; MOVREL-NEXT: s_mov_b32 s9, 0x40c00000
; MOVREL-NEXT: s_mov_b32 s8, 0x40a00000
; MOVREL-NEXT: s_mov_b32 s7, 4.0
; MOVREL-NEXT: s_mov_b32 s6, 0x40400000
; MOVREL-NEXT: s_mov_b32 s5, 2.0
; MOVREL-NEXT: s_movrels_b32 s0, s4
; MOVREL-NEXT: v_mov_b32_e32 v0, s0
; MOVREL-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v8f32_const_s_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s4, 1.0
; GFX10PLUS-NEXT: s_mov_b32 m0, s2
; GFX10PLUS-NEXT: s_mov_b32 s11, 0x41000000
; GFX10PLUS-NEXT: s_mov_b32 s10, 0x40e00000
; GFX10PLUS-NEXT: s_mov_b32 s9, 0x40c00000
; GFX10PLUS-NEXT: s_mov_b32 s8, 0x40a00000
; GFX10PLUS-NEXT: s_mov_b32 s7, 4.0
; GFX10PLUS-NEXT: s_mov_b32 s6, 0x40400000
; GFX10PLUS-NEXT: s_mov_b32 s5, 2.0
; GFX10PLUS-NEXT: s_movrels_b32 s0, s4
; GFX10PLUS-NEXT: v_mov_b32_e32 v0, s0
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, i32 %sel
ret float %ext
}
define amdgpu_ps float @dyn_extract_v8f32_s_v(<8 x float> inreg %vec, i32 %sel) {
; GCN-LABEL: dyn_extract_v8f32_s_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 s1, s3
; GCN-NEXT: s_mov_b32 s2, s4
; GCN-NEXT: v_mov_b32_e32 v1, s0
; GCN-NEXT: v_mov_b32_e32 v2, s1
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; GCN-NEXT: s_mov_b32 s3, s5
; GCN-NEXT: v_mov_b32_e32 v3, s2
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v0
; GCN-NEXT: v_mov_b32_e32 v4, s3
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v0
; GCN-NEXT: v_mov_b32_e32 v5, s6
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v4, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v0
; GCN-NEXT: v_mov_b32_e32 v6, s7
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v0
; GCN-NEXT: v_mov_b32_e32 v7, s8
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v6, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 6, v0
; GCN-NEXT: v_mov_b32_e32 v8, s9
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 7, v0
; GCN-NEXT: v_cndmask_b32_e32 v0, v1, v8, vcc
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v8f32_s_v:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX10PLUS-NEXT: v_mov_b32_e32 v1, s1
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v1, s0, v1, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: s_mov_b32 s6, s8
; GFX10PLUS-NEXT: s_mov_b32 s7, s9
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s3, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s5, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s6, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v0, v1, s7, vcc_lo
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <8 x float> %vec, i32 %sel
ret float %ext
}
define float @dyn_extract_v8f32_v_v(<8 x float> %vec, i32 %sel) {
; GCN-LABEL: dyn_extract_v8f32_v_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v8
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v8
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v8
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v8
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v8
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 6, v8
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 7, v8
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: dyn_extract_v8f32_v_v:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc_lo
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
entry:
%ext = extractelement <8 x float> %vec, i32 %sel
ret float %ext
}
define amdgpu_ps float @dyn_extract_v8f32_v_s(<8 x float> %vec, i32 inreg %sel) {
; GPRIDX-LABEL: dyn_extract_v8f32_v_s:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: v_cmp_eq_u32_e64 vcc, s2, 1
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e64 vcc, s2, 2
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e64 vcc, s2, 3
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e64 vcc, s2, 4
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e64 vcc, s2, 5
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e64 vcc, s2, 6
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e64 vcc, s2, 7
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc
; GPRIDX-NEXT: ; return to shader part epilog
;
; MOVREL-LABEL: dyn_extract_v8f32_v_s:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_mov_b32 m0, s2
; MOVREL-NEXT: v_movrels_b32_e32 v0, v0
; MOVREL-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v8f32_v_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 m0, s2
; GFX10PLUS-NEXT: v_movrels_b32_e32 v0, v0
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <8 x float> %vec, i32 %sel
ret float %ext
}
define amdgpu_ps float @dyn_extract_v8f32_s_s(<8 x float> inreg %vec, i32 inreg %sel) {
; GPRIDX-LABEL: dyn_extract_v8f32_s_s:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 1
; GPRIDX-NEXT: s_cselect_b32 s0, s3, s2
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 2
; GPRIDX-NEXT: s_cselect_b32 s0, s4, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 3
; GPRIDX-NEXT: s_cselect_b32 s0, s5, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 4
; GPRIDX-NEXT: s_cselect_b32 s0, s6, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 5
; GPRIDX-NEXT: s_cselect_b32 s0, s7, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 6
; GPRIDX-NEXT: s_cselect_b32 s0, s8, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 7
; GPRIDX-NEXT: s_cselect_b32 s0, s9, s0
; GPRIDX-NEXT: v_mov_b32_e32 v0, s0
; GPRIDX-NEXT: ; return to shader part epilog
;
; MOVREL-LABEL: dyn_extract_v8f32_s_s:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_mov_b32 s0, s2
; MOVREL-NEXT: s_mov_b32 m0, s10
; MOVREL-NEXT: s_mov_b32 s1, s3
; MOVREL-NEXT: s_mov_b32 s2, s4
; MOVREL-NEXT: s_mov_b32 s3, s5
; MOVREL-NEXT: s_mov_b32 s4, s6
; MOVREL-NEXT: s_mov_b32 s5, s7
; MOVREL-NEXT: s_mov_b32 s6, s8
; MOVREL-NEXT: s_mov_b32 s7, s9
; MOVREL-NEXT: s_movrels_b32 s0, s0
; MOVREL-NEXT: v_mov_b32_e32 v0, s0
; MOVREL-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v8f32_s_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 m0, s10
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: s_mov_b32 s6, s8
; GFX10PLUS-NEXT: s_mov_b32 s7, s9
; GFX10PLUS-NEXT: s_movrels_b32 s0, s0
; GFX10PLUS-NEXT: v_mov_b32_e32 v0, s0
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <8 x float> %vec, i32 %sel
ret float %ext
}
define i64 @dyn_extract_v8i64_const_s_v(i32 %sel) {
; GCN-LABEL: dyn_extract_v8i64_const_s_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: s_mov_b64 s[4:5], 1
; GCN-NEXT: s_mov_b64 s[6:7], 2
; GCN-NEXT: v_mov_b32_e32 v1, s4
; GCN-NEXT: v_mov_b32_e32 v2, s5
; GCN-NEXT: v_mov_b32_e32 v3, s6
; GCN-NEXT: v_mov_b32_e32 v4, s7
; GCN-NEXT: s_mov_b64 s[8:9], 3
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; GCN-NEXT: v_mov_b32_e32 v5, s8
; GCN-NEXT: v_mov_b32_e32 v6, s9
; GCN-NEXT: s_mov_b64 s[10:11], 4
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GCN-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v0
; GCN-NEXT: v_mov_b32_e32 v7, s10
; GCN-NEXT: v_mov_b32_e32 v8, s11
; GCN-NEXT: s_mov_b64 s[12:13], 5
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; GCN-NEXT: v_cndmask_b32_e32 v2, v2, v6, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v0
; GCN-NEXT: s_mov_b64 s[14:15], 6
; GCN-NEXT: v_mov_b32_e32 v9, s12
; GCN-NEXT: v_mov_b32_e32 v10, s13
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc
; GCN-NEXT: v_cndmask_b32_e32 v2, v2, v8, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v0
; GCN-NEXT: s_mov_b64 s[16:17], 7
; GCN-NEXT: v_mov_b32_e32 v11, s14
; GCN-NEXT: v_mov_b32_e32 v12, s15
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc
; GCN-NEXT: v_cndmask_b32_e32 v2, v2, v10, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v0
; GCN-NEXT: s_mov_b64 s[18:19], 8
; GCN-NEXT: v_mov_b32_e32 v13, s16
; GCN-NEXT: v_mov_b32_e32 v14, s17
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc
; GCN-NEXT: v_cndmask_b32_e32 v2, v2, v12, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 6, v0
; GCN-NEXT: v_mov_b32_e32 v15, s18
; GCN-NEXT: v_mov_b32_e32 v16, s19
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc
; GCN-NEXT: v_cndmask_b32_e32 v2, v2, v14, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 7, v0
; GCN-NEXT: v_cndmask_b32_e32 v0, v1, v15, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v2, v16, vcc
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: dyn_extract_v8i64_const_s_v:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: s_mov_b64 s[6:7], 2
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX10-NEXT: v_mov_b32_e32 v1, s6
; GFX10-NEXT: v_mov_b32_e32 v2, s7
; GFX10-NEXT: s_mov_b64 s[4:5], 1
; GFX10-NEXT: s_mov_b64 s[8:9], 3
; GFX10-NEXT: s_mov_b64 s[10:11], 4
; GFX10-NEXT: v_cndmask_b32_e32 v1, s4, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v2, s5, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10-NEXT: s_mov_b64 s[12:13], 5
; GFX10-NEXT: s_mov_b64 s[14:15], 6
; GFX10-NEXT: s_mov_b64 s[16:17], 7
; GFX10-NEXT: s_mov_b64 s[18:19], 8
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s13, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s14, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s15, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s16, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s17, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, s18, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v1, v2, s19, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: dyn_extract_v8i64_const_s_v:
; GFX11: ; %bb.0: ; %entry
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-NEXT: s_mov_b64 s[2:3], 2
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX11-NEXT: v_dual_mov_b32 v1, s2 :: v_dual_mov_b32 v2, s3
; GFX11-NEXT: s_mov_b64 s[0:1], 1
; GFX11-NEXT: s_mov_b64 s[4:5], 3
; GFX11-NEXT: s_mov_b64 s[6:7], 4
; GFX11-NEXT: v_cndmask_b32_e32 v1, s0, v1, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e32 v2, s1, v2, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX11-NEXT: s_mov_b64 s[8:9], 5
; GFX11-NEXT: s_mov_b64 s[10:11], 6
; GFX11-NEXT: s_mov_b64 s[12:13], 7
; GFX11-NEXT: s_mov_b64 s[14:15], 8
; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s5, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s6, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s7, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s9, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s10, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s11, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0
; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s12, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s13, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v0
; GFX11-NEXT: v_cndmask_b32_e64 v0, v1, s14, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v1, v2, s15, vcc_lo
; GFX11-NEXT: s_setpc_b64 s[30:31]
entry:
%ext = extractelement <8 x i64> <i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7, i64 8>, i32 %sel
ret i64 %ext
}
define amdgpu_ps void @dyn_extract_v8i64_const_s_s(i32 inreg %sel) {
; GPRIDX-LABEL: dyn_extract_v8i64_const_s_s:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_mov_b64 s[4:5], 1
; GPRIDX-NEXT: s_mov_b32 m0, s2
; GPRIDX-NEXT: s_mov_b64 s[18:19], 8
; GPRIDX-NEXT: s_mov_b64 s[16:17], 7
; GPRIDX-NEXT: s_mov_b64 s[14:15], 6
; GPRIDX-NEXT: s_mov_b64 s[12:13], 5
; GPRIDX-NEXT: s_mov_b64 s[10:11], 4
; GPRIDX-NEXT: s_mov_b64 s[8:9], 3
; GPRIDX-NEXT: s_mov_b64 s[6:7], 2
; GPRIDX-NEXT: s_movrels_b64 s[0:1], s[4:5]
; GPRIDX-NEXT: v_mov_b32_e32 v0, s0
; GPRIDX-NEXT: v_mov_b32_e32 v1, s1
; GPRIDX-NEXT: global_store_dwordx2 v[0:1], v[0:1], off
; GPRIDX-NEXT: s_endpgm
;
; MOVREL-LABEL: dyn_extract_v8i64_const_s_s:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_mov_b64 s[4:5], 1
; MOVREL-NEXT: s_mov_b32 m0, s2
; MOVREL-NEXT: s_mov_b64 s[18:19], 8
; MOVREL-NEXT: s_mov_b64 s[16:17], 7
; MOVREL-NEXT: s_mov_b64 s[14:15], 6
; MOVREL-NEXT: s_mov_b64 s[12:13], 5
; MOVREL-NEXT: s_mov_b64 s[10:11], 4
; MOVREL-NEXT: s_mov_b64 s[8:9], 3
; MOVREL-NEXT: s_mov_b64 s[6:7], 2
; MOVREL-NEXT: s_movrels_b64 s[0:1], s[4:5]
; MOVREL-NEXT: v_mov_b32_e32 v0, s0
; MOVREL-NEXT: v_mov_b32_e32 v1, s1
; MOVREL-NEXT: flat_store_dwordx2 v[0:1], v[0:1]
; MOVREL-NEXT: s_endpgm
;
; GFX10-LABEL: dyn_extract_v8i64_const_s_s:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_mov_b64 s[4:5], 1
; GFX10-NEXT: s_mov_b32 m0, s2
; GFX10-NEXT: s_mov_b64 s[18:19], 8
; GFX10-NEXT: s_mov_b64 s[16:17], 7
; GFX10-NEXT: s_mov_b64 s[14:15], 6
; GFX10-NEXT: s_mov_b64 s[12:13], 5
; GFX10-NEXT: s_mov_b64 s[10:11], 4
; GFX10-NEXT: s_mov_b64 s[8:9], 3
; GFX10-NEXT: s_mov_b64 s[6:7], 2
; GFX10-NEXT: s_movrels_b64 s[0:1], s[4:5]
; GFX10-NEXT: v_mov_b32_e32 v0, s0
; GFX10-NEXT: v_mov_b32_e32 v1, s1
; GFX10-NEXT: global_store_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_endpgm
;
; GFX11-LABEL: dyn_extract_v8i64_const_s_s:
; GFX11: ; %bb.0: ; %entry
; GFX11-NEXT: s_mov_b64 s[4:5], 1
; GFX11-NEXT: s_mov_b32 m0, s2
; GFX11-NEXT: s_mov_b64 s[18:19], 8
; GFX11-NEXT: s_mov_b64 s[16:17], 7
; GFX11-NEXT: s_mov_b64 s[14:15], 6
; GFX11-NEXT: s_mov_b64 s[12:13], 5
; GFX11-NEXT: s_mov_b64 s[10:11], 4
; GFX11-NEXT: s_mov_b64 s[8:9], 3
; GFX11-NEXT: s_mov_b64 s[6:7], 2
; GFX11-NEXT: s_movrels_b64 s[0:1], s[4:5]
; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
; GFX11-NEXT: global_store_b64 v[0:1], v[0:1], off
; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX11-NEXT: s_endpgm
entry:
%ext = extractelement <8 x i64> <i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7, i64 8>, i32 %sel
store i64 %ext, i64 addrspace(1)* undef
ret void
}
define amdgpu_ps void @dyn_extract_v8i64_s_v(<8 x i64> inreg %vec, i32 %sel) {
; GPRIDX-LABEL: dyn_extract_v8i64_s_v:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_mov_b32 s0, s2
; GPRIDX-NEXT: s_mov_b32 s1, s3
; GPRIDX-NEXT: s_mov_b32 s2, s4
; GPRIDX-NEXT: s_mov_b32 s3, s5
; GPRIDX-NEXT: s_mov_b32 s4, s6
; GPRIDX-NEXT: s_mov_b32 s5, s7
; GPRIDX-NEXT: v_mov_b32_e32 v1, s0
; GPRIDX-NEXT: v_mov_b32_e32 v2, s1
; GPRIDX-NEXT: v_mov_b32_e32 v3, s2
; GPRIDX-NEXT: v_mov_b32_e32 v4, s3
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; GPRIDX-NEXT: s_mov_b32 s6, s8
; GPRIDX-NEXT: s_mov_b32 s7, s9
; GPRIDX-NEXT: v_mov_b32_e32 v5, s4
; GPRIDX-NEXT: v_mov_b32_e32 v6, s5
; GPRIDX-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GPRIDX-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 2, v0
; GPRIDX-NEXT: s_mov_b32 s8, s10
; GPRIDX-NEXT: s_mov_b32 s9, s11
; GPRIDX-NEXT: v_mov_b32_e32 v7, s6
; GPRIDX-NEXT: v_mov_b32_e32 v8, s7
; GPRIDX-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; GPRIDX-NEXT: v_cndmask_b32_e32 v2, v2, v6, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 3, v0
; GPRIDX-NEXT: s_mov_b32 s10, s12
; GPRIDX-NEXT: s_mov_b32 s11, s13
; GPRIDX-NEXT: v_mov_b32_e32 v9, s8
; GPRIDX-NEXT: v_mov_b32_e32 v10, s9
; GPRIDX-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc
; GPRIDX-NEXT: v_cndmask_b32_e32 v2, v2, v8, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 4, v0
; GPRIDX-NEXT: v_mov_b32_e32 v11, s10
; GPRIDX-NEXT: v_mov_b32_e32 v12, s11
; GPRIDX-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc
; GPRIDX-NEXT: v_cndmask_b32_e32 v2, v2, v10, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 5, v0
; GPRIDX-NEXT: v_mov_b32_e32 v13, s14
; GPRIDX-NEXT: v_mov_b32_e32 v14, s15
; GPRIDX-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc
; GPRIDX-NEXT: v_cndmask_b32_e32 v2, v2, v12, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 6, v0
; GPRIDX-NEXT: v_mov_b32_e32 v15, s16
; GPRIDX-NEXT: v_mov_b32_e32 v16, s17
; GPRIDX-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc
; GPRIDX-NEXT: v_cndmask_b32_e32 v2, v2, v14, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 7, v0
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v1, v15, vcc
; GPRIDX-NEXT: v_cndmask_b32_e32 v1, v2, v16, vcc
; GPRIDX-NEXT: global_store_dwordx2 v[0:1], v[0:1], off
; GPRIDX-NEXT: s_endpgm
;
; MOVREL-LABEL: dyn_extract_v8i64_s_v:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_mov_b32 s0, s2
; MOVREL-NEXT: s_mov_b32 s1, s3
; MOVREL-NEXT: s_mov_b32 s2, s4
; MOVREL-NEXT: s_mov_b32 s3, s5
; MOVREL-NEXT: s_mov_b32 s4, s6
; MOVREL-NEXT: s_mov_b32 s5, s7
; MOVREL-NEXT: v_mov_b32_e32 v1, s0
; MOVREL-NEXT: v_mov_b32_e32 v2, s1
; MOVREL-NEXT: v_mov_b32_e32 v3, s2
; MOVREL-NEXT: v_mov_b32_e32 v4, s3
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; MOVREL-NEXT: s_mov_b32 s6, s8
; MOVREL-NEXT: s_mov_b32 s7, s9
; MOVREL-NEXT: v_mov_b32_e32 v5, s4
; MOVREL-NEXT: v_mov_b32_e32 v6, s5
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; MOVREL-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 2, v0
; MOVREL-NEXT: s_mov_b32 s8, s10
; MOVREL-NEXT: s_mov_b32 s9, s11
; MOVREL-NEXT: v_mov_b32_e32 v7, s6
; MOVREL-NEXT: v_mov_b32_e32 v8, s7
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; MOVREL-NEXT: v_cndmask_b32_e32 v2, v2, v6, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 3, v0
; MOVREL-NEXT: s_mov_b32 s10, s12
; MOVREL-NEXT: s_mov_b32 s11, s13
; MOVREL-NEXT: v_mov_b32_e32 v9, s8
; MOVREL-NEXT: v_mov_b32_e32 v10, s9
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc
; MOVREL-NEXT: v_cndmask_b32_e32 v2, v2, v8, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 4, v0
; MOVREL-NEXT: v_mov_b32_e32 v11, s10
; MOVREL-NEXT: v_mov_b32_e32 v12, s11
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc
; MOVREL-NEXT: v_cndmask_b32_e32 v2, v2, v10, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 5, v0
; MOVREL-NEXT: v_mov_b32_e32 v13, s14
; MOVREL-NEXT: v_mov_b32_e32 v14, s15
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc
; MOVREL-NEXT: v_cndmask_b32_e32 v2, v2, v12, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 6, v0
; MOVREL-NEXT: v_mov_b32_e32 v15, s16
; MOVREL-NEXT: v_mov_b32_e32 v16, s17
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc
; MOVREL-NEXT: v_cndmask_b32_e32 v2, v2, v14, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 7, v0
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v1, v15, vcc
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v2, v16, vcc
; MOVREL-NEXT: flat_store_dwordx2 v[0:1], v[0:1]
; MOVREL-NEXT: s_endpgm
;
; GFX10-LABEL: dyn_extract_v8i64_s_v:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_mov_b32 s0, s2
; GFX10-NEXT: s_mov_b32 s2, s4
; GFX10-NEXT: s_mov_b32 s19, s5
; GFX10-NEXT: v_mov_b32_e32 v1, s2
; GFX10-NEXT: v_mov_b32_e32 v2, s19
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX10-NEXT: s_mov_b32 s1, s3
; GFX10-NEXT: s_mov_b32 s4, s6
; GFX10-NEXT: s_mov_b32 s5, s7
; GFX10-NEXT: s_mov_b32 s6, s8
; GFX10-NEXT: v_cndmask_b32_e32 v1, s0, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v2, s1, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10-NEXT: s_mov_b32 s7, s9
; GFX10-NEXT: s_mov_b32 s8, s10
; GFX10-NEXT: s_mov_b32 s9, s11
; GFX10-NEXT: s_mov_b32 s10, s12
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10-NEXT: s_mov_b32 s11, s13
; GFX10-NEXT: s_mov_b32 s12, s14
; GFX10-NEXT: s_mov_b32 s13, s15
; GFX10-NEXT: s_mov_b32 s14, s16
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX10-NEXT: s_mov_b32 s15, s17
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s13, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, s14, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v1, v2, s15, vcc_lo
; GFX10-NEXT: global_store_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_endpgm
;
; GFX11-LABEL: dyn_extract_v8i64_s_v:
; GFX11: ; %bb.0: ; %entry
; GFX11-NEXT: s_mov_b32 s0, s2
; GFX11-NEXT: s_mov_b32 s2, s4
; GFX11-NEXT: s_mov_b32 s19, s5
; GFX11-NEXT: v_dual_mov_b32 v1, s2 :: v_dual_mov_b32 v2, s19
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX11-NEXT: s_mov_b32 s1, s3
; GFX11-NEXT: s_mov_b32 s4, s6
; GFX11-NEXT: s_mov_b32 s5, s7
; GFX11-NEXT: s_mov_b32 s6, s8
; GFX11-NEXT: v_cndmask_b32_e32 v1, s0, v1, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e32 v2, s1, v2, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX11-NEXT: s_mov_b32 s7, s9
; GFX11-NEXT: s_mov_b32 s8, s10
; GFX11-NEXT: s_mov_b32 s9, s11
; GFX11-NEXT: s_mov_b32 s10, s12
; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s5, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX11-NEXT: s_mov_b32 s11, s13
; GFX11-NEXT: s_mov_b32 s12, s14
; GFX11-NEXT: s_mov_b32 s13, s15
; GFX11-NEXT: s_mov_b32 s14, s16
; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s6, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s7, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX11-NEXT: s_mov_b32 s15, s17
; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s9, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s10, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s11, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0
; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s12, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s13, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v0
; GFX11-NEXT: v_cndmask_b32_e64 v0, v1, s14, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v1, v2, s15, vcc_lo
; GFX11-NEXT: global_store_b64 v[0:1], v[0:1], off
; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX11-NEXT: s_endpgm
entry:
%ext = extractelement <8 x i64> %vec, i32 %sel
store i64 %ext, i64 addrspace(1)* undef
ret void
}
define i64 @dyn_extract_v8i64_v_v(<8 x i64> %vec, i32 %sel) {
; GCN-LABEL: dyn_extract_v8i64_v_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v16
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v16
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v16
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v16
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v16
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 6, v16
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 7, v16
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: dyn_extract_v8i64_v_v:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: dyn_extract_v8i64_v_v:
; GFX11: ; %bb.0: ; %entry
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v16
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v2 :: v_dual_cndmask_b32 v1, v1, v3
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v16
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v4 :: v_dual_cndmask_b32 v1, v1, v5
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v16
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v6 :: v_dual_cndmask_b32 v1, v1, v7
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v16
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v8 :: v_dual_cndmask_b32 v1, v1, v9
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v16
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v10 :: v_dual_cndmask_b32 v1, v1, v11
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v16
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v12 :: v_dual_cndmask_b32 v1, v1, v13
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v16
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v14 :: v_dual_cndmask_b32 v1, v1, v15
; GFX11-NEXT: s_setpc_b64 s[30:31]
entry:
%ext = extractelement <8 x i64> %vec, i32 %sel
ret i64 %ext
}
define amdgpu_ps void @dyn_extract_v8i64_v_s(<8 x i64> %vec, i32 inreg %sel) {
; GPRIDX-LABEL: dyn_extract_v8i64_v_s:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_lshl_b32 s0, s2, 1
; GPRIDX-NEXT: s_set_gpr_idx_on s0, gpr_idx(SRC0)
; GPRIDX-NEXT: v_mov_b32_e32 v16, v0
; GPRIDX-NEXT: v_mov_b32_e32 v17, v1
; GPRIDX-NEXT: s_set_gpr_idx_off
; GPRIDX-NEXT: global_store_dwordx2 v[0:1], v[16:17], off
; GPRIDX-NEXT: s_endpgm
;
; MOVREL-LABEL: dyn_extract_v8i64_v_s:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_lshl_b32 m0, s2, 1
; MOVREL-NEXT: v_movrels_b32_e32 v16, v0
; MOVREL-NEXT: v_movrels_b32_e32 v17, v1
; MOVREL-NEXT: flat_store_dwordx2 v[0:1], v[16:17]
; MOVREL-NEXT: s_endpgm
;
; GFX10-LABEL: dyn_extract_v8i64_v_s:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_lshl_b32 m0, s2, 1
; GFX10-NEXT: v_movrels_b32_e32 v16, v0
; GFX10-NEXT: v_movrels_b32_e32 v17, v1
; GFX10-NEXT: global_store_dwordx2 v[0:1], v[16:17], off
; GFX10-NEXT: s_endpgm
;
; GFX11-LABEL: dyn_extract_v8i64_v_s:
; GFX11: ; %bb.0: ; %entry
; GFX11-NEXT: s_lshl_b32 m0, s2, 1
; GFX11-NEXT: v_movrels_b32_e32 v16, v0
; GFX11-NEXT: v_movrels_b32_e32 v17, v1
; GFX11-NEXT: global_store_b64 v[0:1], v[16:17], off
; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX11-NEXT: s_endpgm
entry:
%ext = extractelement <8 x i64> %vec, i32 %sel
store i64 %ext, i64 addrspace(1)* undef
ret void
}
define amdgpu_ps void @dyn_extract_v8i64_s_s(<8 x i64> inreg %vec, i32 inreg %sel) {
; GPRIDX-LABEL: dyn_extract_v8i64_s_s:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_mov_b32 s0, s2
; GPRIDX-NEXT: s_mov_b32 s1, s3
; GPRIDX-NEXT: s_mov_b32 m0, s18
; GPRIDX-NEXT: s_mov_b32 s2, s4
; GPRIDX-NEXT: s_mov_b32 s3, s5
; GPRIDX-NEXT: s_mov_b32 s4, s6
; GPRIDX-NEXT: s_mov_b32 s5, s7
; GPRIDX-NEXT: s_mov_b32 s6, s8
; GPRIDX-NEXT: s_mov_b32 s7, s9
; GPRIDX-NEXT: s_mov_b32 s8, s10
; GPRIDX-NEXT: s_mov_b32 s9, s11
; GPRIDX-NEXT: s_mov_b32 s10, s12
; GPRIDX-NEXT: s_mov_b32 s11, s13
; GPRIDX-NEXT: s_mov_b32 s12, s14
; GPRIDX-NEXT: s_mov_b32 s13, s15
; GPRIDX-NEXT: s_mov_b32 s14, s16
; GPRIDX-NEXT: s_mov_b32 s15, s17
; GPRIDX-NEXT: s_movrels_b64 s[0:1], s[0:1]
; GPRIDX-NEXT: v_mov_b32_e32 v0, s0
; GPRIDX-NEXT: v_mov_b32_e32 v1, s1
; GPRIDX-NEXT: global_store_dwordx2 v[0:1], v[0:1], off
; GPRIDX-NEXT: s_endpgm
;
; MOVREL-LABEL: dyn_extract_v8i64_s_s:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_mov_b32 s0, s2
; MOVREL-NEXT: s_mov_b32 s1, s3
; MOVREL-NEXT: s_mov_b32 m0, s18
; MOVREL-NEXT: s_mov_b32 s2, s4
; MOVREL-NEXT: s_mov_b32 s3, s5
; MOVREL-NEXT: s_mov_b32 s4, s6
; MOVREL-NEXT: s_mov_b32 s5, s7
; MOVREL-NEXT: s_mov_b32 s6, s8
; MOVREL-NEXT: s_mov_b32 s7, s9
; MOVREL-NEXT: s_mov_b32 s8, s10
; MOVREL-NEXT: s_mov_b32 s9, s11
; MOVREL-NEXT: s_mov_b32 s10, s12
; MOVREL-NEXT: s_mov_b32 s11, s13
; MOVREL-NEXT: s_mov_b32 s12, s14
; MOVREL-NEXT: s_mov_b32 s13, s15
; MOVREL-NEXT: s_mov_b32 s14, s16
; MOVREL-NEXT: s_mov_b32 s15, s17
; MOVREL-NEXT: s_movrels_b64 s[0:1], s[0:1]
; MOVREL-NEXT: v_mov_b32_e32 v0, s0
; MOVREL-NEXT: v_mov_b32_e32 v1, s1
; MOVREL-NEXT: flat_store_dwordx2 v[0:1], v[0:1]
; MOVREL-NEXT: s_endpgm
;
; GFX10-LABEL: dyn_extract_v8i64_s_s:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_mov_b32 s0, s2
; GFX10-NEXT: s_mov_b32 s1, s3
; GFX10-NEXT: s_mov_b32 m0, s18
; GFX10-NEXT: s_mov_b32 s2, s4
; GFX10-NEXT: s_mov_b32 s3, s5
; GFX10-NEXT: s_mov_b32 s4, s6
; GFX10-NEXT: s_mov_b32 s5, s7
; GFX10-NEXT: s_mov_b32 s6, s8
; GFX10-NEXT: s_mov_b32 s7, s9
; GFX10-NEXT: s_mov_b32 s8, s10
; GFX10-NEXT: s_mov_b32 s9, s11
; GFX10-NEXT: s_mov_b32 s10, s12
; GFX10-NEXT: s_mov_b32 s11, s13
; GFX10-NEXT: s_mov_b32 s12, s14
; GFX10-NEXT: s_mov_b32 s13, s15
; GFX10-NEXT: s_mov_b32 s14, s16
; GFX10-NEXT: s_mov_b32 s15, s17
; GFX10-NEXT: s_movrels_b64 s[0:1], s[0:1]
; GFX10-NEXT: v_mov_b32_e32 v0, s0
; GFX10-NEXT: v_mov_b32_e32 v1, s1
; GFX10-NEXT: global_store_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_endpgm
;
; GFX11-LABEL: dyn_extract_v8i64_s_s:
; GFX11: ; %bb.0: ; %entry
; GFX11-NEXT: s_mov_b32 s0, s2
; GFX11-NEXT: s_mov_b32 s1, s3
; GFX11-NEXT: s_mov_b32 m0, s18
; GFX11-NEXT: s_mov_b32 s2, s4
; GFX11-NEXT: s_mov_b32 s3, s5
; GFX11-NEXT: s_mov_b32 s4, s6
; GFX11-NEXT: s_mov_b32 s5, s7
; GFX11-NEXT: s_mov_b32 s6, s8
; GFX11-NEXT: s_mov_b32 s7, s9
; GFX11-NEXT: s_mov_b32 s8, s10
; GFX11-NEXT: s_mov_b32 s9, s11
; GFX11-NEXT: s_mov_b32 s10, s12
; GFX11-NEXT: s_mov_b32 s11, s13
; GFX11-NEXT: s_mov_b32 s12, s14
; GFX11-NEXT: s_mov_b32 s13, s15
; GFX11-NEXT: s_mov_b32 s14, s16
; GFX11-NEXT: s_mov_b32 s15, s17
; GFX11-NEXT: s_movrels_b64 s[0:1], s[0:1]
; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
; GFX11-NEXT: global_store_b64 v[0:1], v[0:1], off
; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX11-NEXT: s_endpgm
entry:
%ext = extractelement <8 x i64> %vec, i32 %sel
store i64 %ext, i64 addrspace(1)* undef
ret void
}
define amdgpu_ps float @dyn_extract_v8f32_s_s_offset3(<8 x float> inreg %vec, i32 inreg %sel) {
; GPRIDX-LABEL: dyn_extract_v8f32_s_s_offset3:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_add_i32 s10, s10, 3
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 1
; GPRIDX-NEXT: s_cselect_b32 s0, s3, s2
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 2
; GPRIDX-NEXT: s_cselect_b32 s0, s4, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 3
; GPRIDX-NEXT: s_cselect_b32 s0, s5, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 4
; GPRIDX-NEXT: s_cselect_b32 s0, s6, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 5
; GPRIDX-NEXT: s_cselect_b32 s0, s7, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 6
; GPRIDX-NEXT: s_cselect_b32 s0, s8, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 7
; GPRIDX-NEXT: s_cselect_b32 s0, s9, s0
; GPRIDX-NEXT: v_mov_b32_e32 v0, s0
; GPRIDX-NEXT: ; return to shader part epilog
;
; MOVREL-LABEL: dyn_extract_v8f32_s_s_offset3:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_mov_b32 s0, s2
; MOVREL-NEXT: s_mov_b32 s1, s3
; MOVREL-NEXT: s_mov_b32 s3, s5
; MOVREL-NEXT: s_mov_b32 m0, s10
; MOVREL-NEXT: s_mov_b32 s2, s4
; MOVREL-NEXT: s_mov_b32 s4, s6
; MOVREL-NEXT: s_mov_b32 s5, s7
; MOVREL-NEXT: s_mov_b32 s6, s8
; MOVREL-NEXT: s_mov_b32 s7, s9
; MOVREL-NEXT: s_movrels_b32 s0, s3
; MOVREL-NEXT: v_mov_b32_e32 v0, s0
; MOVREL-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v8f32_s_s_offset3:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 m0, s10
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: s_mov_b32 s6, s8
; GFX10PLUS-NEXT: s_mov_b32 s7, s9
; GFX10PLUS-NEXT: s_movrels_b32 s0, s3
; GFX10PLUS-NEXT: v_mov_b32_e32 v0, s0
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%add = add i32 %sel, 3
%ext = extractelement <8 x float> %vec, i32 %add
ret float %ext
}
define float @dyn_extract_v8f32_v_v_offset3(<8 x float> %vec, i32 %sel) {
; GPRIDX-LABEL: dyn_extract_v8f32_v_v_offset3:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GPRIDX-NEXT: v_add_u32_e32 v8, 3, v8
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 1, v8
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 2, v8
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 3, v8
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 4, v8
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 5, v8
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 6, v8
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 7, v8
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc
; GPRIDX-NEXT: s_setpc_b64 s[30:31]
;
; MOVREL-LABEL: dyn_extract_v8f32_v_v_offset3:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; MOVREL-NEXT: v_add_u32_e32 v8, vcc, 3, v8
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 1, v8
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 2, v8
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 3, v8
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 4, v8
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 5, v8
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 6, v8
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 7, v8
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc
; MOVREL-NEXT: s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: dyn_extract_v8f32_v_v_offset3:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10PLUS-NEXT: v_add_nc_u32_e32 v8, 3, v8
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc_lo
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
entry:
%add = add i32 %sel, 3
%ext = extractelement <8 x float> %vec, i32 %add
ret float %ext
}
define amdgpu_ps double @dyn_extract_v8f64_s_s_offset1(<8 x double> inreg %vec, i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v8f64_s_s_offset1:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 s1, s3
; GCN-NEXT: s_mov_b32 s2, s4
; GCN-NEXT: s_mov_b32 s3, s5
; GCN-NEXT: s_mov_b32 m0, s18
; GCN-NEXT: s_mov_b32 s4, s6
; GCN-NEXT: s_mov_b32 s5, s7
; GCN-NEXT: s_mov_b32 s6, s8
; GCN-NEXT: s_mov_b32 s7, s9
; GCN-NEXT: s_mov_b32 s8, s10
; GCN-NEXT: s_mov_b32 s9, s11
; GCN-NEXT: s_mov_b32 s10, s12
; GCN-NEXT: s_mov_b32 s11, s13
; GCN-NEXT: s_mov_b32 s12, s14
; GCN-NEXT: s_mov_b32 s13, s15
; GCN-NEXT: s_mov_b32 s14, s16
; GCN-NEXT: s_mov_b32 s15, s17
; GCN-NEXT: s_movrels_b64 s[0:1], s[2:3]
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v8f64_s_s_offset1:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 m0, s18
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: s_mov_b32 s6, s8
; GFX10PLUS-NEXT: s_mov_b32 s7, s9
; GFX10PLUS-NEXT: s_mov_b32 s8, s10
; GFX10PLUS-NEXT: s_mov_b32 s9, s11
; GFX10PLUS-NEXT: s_mov_b32 s10, s12
; GFX10PLUS-NEXT: s_mov_b32 s11, s13
; GFX10PLUS-NEXT: s_mov_b32 s12, s14
; GFX10PLUS-NEXT: s_mov_b32 s13, s15
; GFX10PLUS-NEXT: s_mov_b32 s14, s16
; GFX10PLUS-NEXT: s_mov_b32 s15, s17
; GFX10PLUS-NEXT: s_movrels_b64 s[0:1], s[2:3]
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%add = add i32 %sel, 1
%ext = extractelement <8 x double> %vec, i32 %add
ret double %ext
}
define amdgpu_ps double @dyn_extract_v8f64_s_s_offset2(<8 x double> inreg %vec, i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v8f64_s_s_offset2:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 s1, s3
; GCN-NEXT: s_mov_b32 s2, s4
; GCN-NEXT: s_mov_b32 s3, s5
; GCN-NEXT: s_mov_b32 s4, s6
; GCN-NEXT: s_mov_b32 s5, s7
; GCN-NEXT: s_mov_b32 m0, s18
; GCN-NEXT: s_mov_b32 s6, s8
; GCN-NEXT: s_mov_b32 s7, s9
; GCN-NEXT: s_mov_b32 s8, s10
; GCN-NEXT: s_mov_b32 s9, s11
; GCN-NEXT: s_mov_b32 s10, s12
; GCN-NEXT: s_mov_b32 s11, s13
; GCN-NEXT: s_mov_b32 s12, s14
; GCN-NEXT: s_mov_b32 s13, s15
; GCN-NEXT: s_mov_b32 s14, s16
; GCN-NEXT: s_mov_b32 s15, s17
; GCN-NEXT: s_movrels_b64 s[0:1], s[4:5]
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v8f64_s_s_offset2:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: s_mov_b32 m0, s18
; GFX10PLUS-NEXT: s_mov_b32 s6, s8
; GFX10PLUS-NEXT: s_mov_b32 s7, s9
; GFX10PLUS-NEXT: s_mov_b32 s8, s10
; GFX10PLUS-NEXT: s_mov_b32 s9, s11
; GFX10PLUS-NEXT: s_mov_b32 s10, s12
; GFX10PLUS-NEXT: s_mov_b32 s11, s13
; GFX10PLUS-NEXT: s_mov_b32 s12, s14
; GFX10PLUS-NEXT: s_mov_b32 s13, s15
; GFX10PLUS-NEXT: s_mov_b32 s14, s16
; GFX10PLUS-NEXT: s_mov_b32 s15, s17
; GFX10PLUS-NEXT: s_movrels_b64 s[0:1], s[4:5]
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%add = add i32 %sel, 2
%ext = extractelement <8 x double> %vec, i32 %add
ret double %ext
}
define amdgpu_ps double @dyn_extract_v8f64_s_s_offset3(<8 x double> inreg %vec, i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v8f64_s_s_offset3:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 s1, s3
; GCN-NEXT: s_mov_b32 s2, s4
; GCN-NEXT: s_mov_b32 s3, s5
; GCN-NEXT: s_mov_b32 s4, s6
; GCN-NEXT: s_mov_b32 s5, s7
; GCN-NEXT: s_mov_b32 s6, s8
; GCN-NEXT: s_mov_b32 s7, s9
; GCN-NEXT: s_mov_b32 m0, s18
; GCN-NEXT: s_mov_b32 s8, s10
; GCN-NEXT: s_mov_b32 s9, s11
; GCN-NEXT: s_mov_b32 s10, s12
; GCN-NEXT: s_mov_b32 s11, s13
; GCN-NEXT: s_mov_b32 s12, s14
; GCN-NEXT: s_mov_b32 s13, s15
; GCN-NEXT: s_mov_b32 s14, s16
; GCN-NEXT: s_mov_b32 s15, s17
; GCN-NEXT: s_movrels_b64 s[0:1], s[6:7]
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v8f64_s_s_offset3:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: s_mov_b32 s6, s8
; GFX10PLUS-NEXT: s_mov_b32 s7, s9
; GFX10PLUS-NEXT: s_mov_b32 m0, s18
; GFX10PLUS-NEXT: s_mov_b32 s8, s10
; GFX10PLUS-NEXT: s_mov_b32 s9, s11
; GFX10PLUS-NEXT: s_mov_b32 s10, s12
; GFX10PLUS-NEXT: s_mov_b32 s11, s13
; GFX10PLUS-NEXT: s_mov_b32 s12, s14
; GFX10PLUS-NEXT: s_mov_b32 s13, s15
; GFX10PLUS-NEXT: s_mov_b32 s14, s16
; GFX10PLUS-NEXT: s_mov_b32 s15, s17
; GFX10PLUS-NEXT: s_movrels_b64 s[0:1], s[6:7]
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%add = add i32 %sel, 3
%ext = extractelement <8 x double> %vec, i32 %add
ret double %ext
}
define amdgpu_ps double @dyn_extract_v8f64_s_s_offset4(<8 x double> inreg %vec, i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v8f64_s_s_offset4:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 s1, s3
; GCN-NEXT: s_mov_b32 s2, s4
; GCN-NEXT: s_mov_b32 s3, s5
; GCN-NEXT: s_mov_b32 s4, s6
; GCN-NEXT: s_mov_b32 s5, s7
; GCN-NEXT: s_mov_b32 s6, s8
; GCN-NEXT: s_mov_b32 s7, s9
; GCN-NEXT: s_mov_b32 s8, s10
; GCN-NEXT: s_mov_b32 s9, s11
; GCN-NEXT: s_mov_b32 m0, s18
; GCN-NEXT: s_mov_b32 s10, s12
; GCN-NEXT: s_mov_b32 s11, s13
; GCN-NEXT: s_mov_b32 s12, s14
; GCN-NEXT: s_mov_b32 s13, s15
; GCN-NEXT: s_mov_b32 s14, s16
; GCN-NEXT: s_mov_b32 s15, s17
; GCN-NEXT: s_movrels_b64 s[0:1], s[8:9]
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v8f64_s_s_offset4:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: s_mov_b32 s6, s8
; GFX10PLUS-NEXT: s_mov_b32 s7, s9
; GFX10PLUS-NEXT: s_mov_b32 s8, s10
; GFX10PLUS-NEXT: s_mov_b32 s9, s11
; GFX10PLUS-NEXT: s_mov_b32 m0, s18
; GFX10PLUS-NEXT: s_mov_b32 s10, s12
; GFX10PLUS-NEXT: s_mov_b32 s11, s13
; GFX10PLUS-NEXT: s_mov_b32 s12, s14
; GFX10PLUS-NEXT: s_mov_b32 s13, s15
; GFX10PLUS-NEXT: s_mov_b32 s14, s16
; GFX10PLUS-NEXT: s_mov_b32 s15, s17
; GFX10PLUS-NEXT: s_movrels_b64 s[0:1], s[8:9]
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%add = add i32 %sel, 4
%ext = extractelement <8 x double> %vec, i32 %add
ret double %ext
}
define amdgpu_ps double @dyn_extract_v8f64_s_s_offset5(<8 x double> inreg %vec, i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v8f64_s_s_offset5:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 s1, s3
; GCN-NEXT: s_mov_b32 s2, s4
; GCN-NEXT: s_mov_b32 s3, s5
; GCN-NEXT: s_mov_b32 s4, s6
; GCN-NEXT: s_mov_b32 s5, s7
; GCN-NEXT: s_mov_b32 s6, s8
; GCN-NEXT: s_mov_b32 s7, s9
; GCN-NEXT: s_mov_b32 s8, s10
; GCN-NEXT: s_mov_b32 s9, s11
; GCN-NEXT: s_mov_b32 s10, s12
; GCN-NEXT: s_mov_b32 s11, s13
; GCN-NEXT: s_mov_b32 m0, s18
; GCN-NEXT: s_mov_b32 s12, s14
; GCN-NEXT: s_mov_b32 s13, s15
; GCN-NEXT: s_mov_b32 s14, s16
; GCN-NEXT: s_mov_b32 s15, s17
; GCN-NEXT: s_movrels_b64 s[0:1], s[10:11]
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v8f64_s_s_offset5:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: s_mov_b32 s6, s8
; GFX10PLUS-NEXT: s_mov_b32 s7, s9
; GFX10PLUS-NEXT: s_mov_b32 s8, s10
; GFX10PLUS-NEXT: s_mov_b32 s9, s11
; GFX10PLUS-NEXT: s_mov_b32 s10, s12
; GFX10PLUS-NEXT: s_mov_b32 s11, s13
; GFX10PLUS-NEXT: s_mov_b32 m0, s18
; GFX10PLUS-NEXT: s_mov_b32 s12, s14
; GFX10PLUS-NEXT: s_mov_b32 s13, s15
; GFX10PLUS-NEXT: s_mov_b32 s14, s16
; GFX10PLUS-NEXT: s_mov_b32 s15, s17
; GFX10PLUS-NEXT: s_movrels_b64 s[0:1], s[10:11]
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%add = add i32 %sel, 5
%ext = extractelement <8 x double> %vec, i32 %add
ret double %ext
}
define amdgpu_ps double @dyn_extract_v8f64_s_s_offset6(<8 x double> inreg %vec, i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v8f64_s_s_offset6:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 s1, s3
; GCN-NEXT: s_mov_b32 s2, s4
; GCN-NEXT: s_mov_b32 s3, s5
; GCN-NEXT: s_mov_b32 s4, s6
; GCN-NEXT: s_mov_b32 s5, s7
; GCN-NEXT: s_mov_b32 s6, s8
; GCN-NEXT: s_mov_b32 s7, s9
; GCN-NEXT: s_mov_b32 s8, s10
; GCN-NEXT: s_mov_b32 s9, s11
; GCN-NEXT: s_mov_b32 s10, s12
; GCN-NEXT: s_mov_b32 s11, s13
; GCN-NEXT: s_mov_b32 s12, s14
; GCN-NEXT: s_mov_b32 s13, s15
; GCN-NEXT: s_mov_b32 m0, s18
; GCN-NEXT: s_mov_b32 s14, s16
; GCN-NEXT: s_mov_b32 s15, s17
; GCN-NEXT: s_movrels_b64 s[0:1], s[12:13]
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v8f64_s_s_offset6:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: s_mov_b32 s6, s8
; GFX10PLUS-NEXT: s_mov_b32 s7, s9
; GFX10PLUS-NEXT: s_mov_b32 s8, s10
; GFX10PLUS-NEXT: s_mov_b32 s9, s11
; GFX10PLUS-NEXT: s_mov_b32 s10, s12
; GFX10PLUS-NEXT: s_mov_b32 s11, s13
; GFX10PLUS-NEXT: s_mov_b32 s12, s14
; GFX10PLUS-NEXT: s_mov_b32 s13, s15
; GFX10PLUS-NEXT: s_mov_b32 m0, s18
; GFX10PLUS-NEXT: s_mov_b32 s14, s16
; GFX10PLUS-NEXT: s_mov_b32 s15, s17
; GFX10PLUS-NEXT: s_movrels_b64 s[0:1], s[12:13]
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%add = add i32 %sel, 6
%ext = extractelement <8 x double> %vec, i32 %add
ret double %ext
}
define amdgpu_ps double @dyn_extract_v8f64_s_s_offset7(<8 x double> inreg %vec, i32 inreg %sel) {
; GPRIDX-LABEL: dyn_extract_v8f64_s_s_offset7:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_mov_b32 s0, s2
; GPRIDX-NEXT: s_mov_b32 s1, s3
; GPRIDX-NEXT: s_mov_b32 s2, s4
; GPRIDX-NEXT: s_mov_b32 s3, s5
; GPRIDX-NEXT: s_mov_b32 s4, s6
; GPRIDX-NEXT: s_mov_b32 s5, s7
; GPRIDX-NEXT: s_mov_b32 s6, s8
; GPRIDX-NEXT: s_mov_b32 s7, s9
; GPRIDX-NEXT: s_mov_b32 s8, s10
; GPRIDX-NEXT: s_mov_b32 s9, s11
; GPRIDX-NEXT: s_mov_b32 s10, s12
; GPRIDX-NEXT: s_mov_b32 s11, s13
; GPRIDX-NEXT: s_mov_b32 s12, s14
; GPRIDX-NEXT: s_mov_b32 s13, s15
; GPRIDX-NEXT: s_mov_b32 s14, s16
; GPRIDX-NEXT: s_mov_b32 s15, s17
; GPRIDX-NEXT: s_mov_b32 m0, s18
; GPRIDX-NEXT: s_nop 0
; GPRIDX-NEXT: s_movrels_b64 s[0:1], s[14:15]
; GPRIDX-NEXT: ; return to shader part epilog
;
; MOVREL-LABEL: dyn_extract_v8f64_s_s_offset7:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_mov_b32 s0, s2
; MOVREL-NEXT: s_mov_b32 s1, s3
; MOVREL-NEXT: s_mov_b32 s2, s4
; MOVREL-NEXT: s_mov_b32 s3, s5
; MOVREL-NEXT: s_mov_b32 s4, s6
; MOVREL-NEXT: s_mov_b32 s5, s7
; MOVREL-NEXT: s_mov_b32 s6, s8
; MOVREL-NEXT: s_mov_b32 s7, s9
; MOVREL-NEXT: s_mov_b32 s8, s10
; MOVREL-NEXT: s_mov_b32 s9, s11
; MOVREL-NEXT: s_mov_b32 s10, s12
; MOVREL-NEXT: s_mov_b32 s11, s13
; MOVREL-NEXT: s_mov_b32 s12, s14
; MOVREL-NEXT: s_mov_b32 s13, s15
; MOVREL-NEXT: s_mov_b32 s14, s16
; MOVREL-NEXT: s_mov_b32 s15, s17
; MOVREL-NEXT: s_mov_b32 m0, s18
; MOVREL-NEXT: s_movrels_b64 s[0:1], s[14:15]
; MOVREL-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v8f64_s_s_offset7:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: s_mov_b32 s6, s8
; GFX10PLUS-NEXT: s_mov_b32 s7, s9
; GFX10PLUS-NEXT: s_mov_b32 s8, s10
; GFX10PLUS-NEXT: s_mov_b32 s9, s11
; GFX10PLUS-NEXT: s_mov_b32 s10, s12
; GFX10PLUS-NEXT: s_mov_b32 s11, s13
; GFX10PLUS-NEXT: s_mov_b32 s12, s14
; GFX10PLUS-NEXT: s_mov_b32 s13, s15
; GFX10PLUS-NEXT: s_mov_b32 s14, s16
; GFX10PLUS-NEXT: s_mov_b32 s15, s17
; GFX10PLUS-NEXT: s_mov_b32 m0, s18
; GFX10PLUS-NEXT: s_movrels_b64 s[0:1], s[14:15]
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%add = add i32 %sel, 7
%ext = extractelement <8 x double> %vec, i32 %add
ret double %ext
}
define amdgpu_ps double @dyn_extract_v8f64_s_s_offsetm1(<8 x double> inreg %vec, i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v8f64_s_s_offsetm1:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 s1, s3
; GCN-NEXT: s_add_i32 m0, s18, -1
; GCN-NEXT: s_mov_b32 s2, s4
; GCN-NEXT: s_mov_b32 s3, s5
; GCN-NEXT: s_mov_b32 s4, s6
; GCN-NEXT: s_mov_b32 s5, s7
; GCN-NEXT: s_mov_b32 s6, s8
; GCN-NEXT: s_mov_b32 s7, s9
; GCN-NEXT: s_mov_b32 s8, s10
; GCN-NEXT: s_mov_b32 s9, s11
; GCN-NEXT: s_mov_b32 s10, s12
; GCN-NEXT: s_mov_b32 s11, s13
; GCN-NEXT: s_mov_b32 s12, s14
; GCN-NEXT: s_mov_b32 s13, s15
; GCN-NEXT: s_mov_b32 s14, s16
; GCN-NEXT: s_mov_b32 s15, s17
; GCN-NEXT: s_movrels_b64 s[0:1], s[0:1]
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v8f64_s_s_offsetm1:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: s_add_i32 m0, s18, -1
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: s_mov_b32 s6, s8
; GFX10PLUS-NEXT: s_mov_b32 s7, s9
; GFX10PLUS-NEXT: s_mov_b32 s8, s10
; GFX10PLUS-NEXT: s_mov_b32 s9, s11
; GFX10PLUS-NEXT: s_mov_b32 s10, s12
; GFX10PLUS-NEXT: s_mov_b32 s11, s13
; GFX10PLUS-NEXT: s_mov_b32 s12, s14
; GFX10PLUS-NEXT: s_mov_b32 s13, s15
; GFX10PLUS-NEXT: s_mov_b32 s14, s16
; GFX10PLUS-NEXT: s_mov_b32 s15, s17
; GFX10PLUS-NEXT: s_movrels_b64 s[0:1], s[0:1]
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%add = add i32 %sel, -1
%ext = extractelement <8 x double> %vec, i32 %add
ret double %ext
}
define double @dyn_extract_v8f64_v_v_offset3(<8 x double> %vec, i32 %sel) {
; GPRIDX-LABEL: dyn_extract_v8f64_v_v_offset3:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GPRIDX-NEXT: v_add_u32_e32 v16, 3, v16
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 1, v16
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GPRIDX-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 2, v16
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GPRIDX-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 3, v16
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; GPRIDX-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 4, v16
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc
; GPRIDX-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 5, v16
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc
; GPRIDX-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 6, v16
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc
; GPRIDX-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 7, v16
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc
; GPRIDX-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc
; GPRIDX-NEXT: s_setpc_b64 s[30:31]
;
; MOVREL-LABEL: dyn_extract_v8f64_v_v_offset3:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; MOVREL-NEXT: v_add_u32_e32 v16, vcc, 3, v16
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 1, v16
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 2, v16
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 3, v16
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 4, v16
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 5, v16
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 6, v16
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 7, v16
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc
; MOVREL-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc
; MOVREL-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: dyn_extract_v8f64_v_v_offset3:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_add_nc_u32_e32 v16, 3, v16
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: dyn_extract_v8f64_v_v_offset3:
; GFX11: ; %bb.0: ; %entry
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-NEXT: v_add_nc_u32_e32 v16, 3, v16
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v16
; GFX11-NEXT: v_dual_cndmask_b32 v1, v1, v3 :: v_dual_cndmask_b32 v0, v0, v2
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v16
; GFX11-NEXT: v_dual_cndmask_b32 v1, v1, v5 :: v_dual_cndmask_b32 v0, v0, v4
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v16
; GFX11-NEXT: v_dual_cndmask_b32 v1, v1, v7 :: v_dual_cndmask_b32 v0, v0, v6
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v16
; GFX11-NEXT: v_dual_cndmask_b32 v1, v1, v9 :: v_dual_cndmask_b32 v0, v0, v8
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v16
; GFX11-NEXT: v_dual_cndmask_b32 v1, v1, v11 :: v_dual_cndmask_b32 v0, v0, v10
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v16
; GFX11-NEXT: v_dual_cndmask_b32 v1, v1, v13 :: v_dual_cndmask_b32 v0, v0, v12
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v16
; GFX11-NEXT: v_dual_cndmask_b32 v1, v1, v15 :: v_dual_cndmask_b32 v0, v0, v14
; GFX11-NEXT: s_setpc_b64 s[30:31]
entry:
%add = add i32 %sel, 3
%ext = extractelement <8 x double> %vec, i32 %add
ret double %ext
}
define i8 addrspace(3)* @dyn_extract_v8p3_v_v(<8 x i8 addrspace(3)*> %vec, i32 %idx) {
; GCN-LABEL: dyn_extract_v8p3_v_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v8
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v8
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v8
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v8
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v8
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 6, v8
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 7, v8
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: dyn_extract_v8p3_v_v:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v8
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc_lo
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
entry:
%ext = extractelement <8 x i8 addrspace(3)*> %vec, i32 %idx
ret i8 addrspace(3)* %ext
}
define amdgpu_ps void @dyn_extract_v8p3_s_s(<8 x i8 addrspace(3)*> inreg %vec, i32 inreg %idx) {
; GPRIDX-LABEL: dyn_extract_v8p3_s_s:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 1
; GPRIDX-NEXT: s_cselect_b32 s0, s3, s2
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 2
; GPRIDX-NEXT: s_cselect_b32 s0, s4, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 3
; GPRIDX-NEXT: s_cselect_b32 s0, s5, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 4
; GPRIDX-NEXT: s_cselect_b32 s0, s6, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 5
; GPRIDX-NEXT: s_cselect_b32 s0, s7, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 6
; GPRIDX-NEXT: s_cselect_b32 s0, s8, s0
; GPRIDX-NEXT: s_cmp_eq_u32 s10, 7
; GPRIDX-NEXT: s_cselect_b32 s0, s9, s0
; GPRIDX-NEXT: v_mov_b32_e32 v0, s0
; GPRIDX-NEXT: ds_write_b32 v0, v0
; GPRIDX-NEXT: s_endpgm
;
; MOVREL-LABEL: dyn_extract_v8p3_s_s:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_mov_b32 s0, s2
; MOVREL-NEXT: s_mov_b32 m0, s10
; MOVREL-NEXT: s_mov_b32 s1, s3
; MOVREL-NEXT: s_mov_b32 s2, s4
; MOVREL-NEXT: s_mov_b32 s3, s5
; MOVREL-NEXT: s_mov_b32 s4, s6
; MOVREL-NEXT: s_mov_b32 s5, s7
; MOVREL-NEXT: s_mov_b32 s6, s8
; MOVREL-NEXT: s_mov_b32 s7, s9
; MOVREL-NEXT: s_movrels_b32 s0, s0
; MOVREL-NEXT: v_mov_b32_e32 v0, s0
; MOVREL-NEXT: s_mov_b32 m0, -1
; MOVREL-NEXT: ds_write_b32 v0, v0
; MOVREL-NEXT: s_endpgm
;
; GFX10-LABEL: dyn_extract_v8p3_s_s:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_mov_b32 s0, s2
; GFX10-NEXT: s_mov_b32 m0, s10
; GFX10-NEXT: s_mov_b32 s1, s3
; GFX10-NEXT: s_mov_b32 s2, s4
; GFX10-NEXT: s_mov_b32 s3, s5
; GFX10-NEXT: s_mov_b32 s4, s6
; GFX10-NEXT: s_mov_b32 s5, s7
; GFX10-NEXT: s_mov_b32 s6, s8
; GFX10-NEXT: s_mov_b32 s7, s9
; GFX10-NEXT: s_movrels_b32 s0, s0
; GFX10-NEXT: v_mov_b32_e32 v0, s0
; GFX10-NEXT: ds_write_b32 v0, v0
; GFX10-NEXT: s_endpgm
;
; GFX11-LABEL: dyn_extract_v8p3_s_s:
; GFX11: ; %bb.0: ; %entry
; GFX11-NEXT: s_mov_b32 s0, s2
; GFX11-NEXT: s_mov_b32 m0, s10
; GFX11-NEXT: s_mov_b32 s1, s3
; GFX11-NEXT: s_mov_b32 s2, s4
; GFX11-NEXT: s_mov_b32 s3, s5
; GFX11-NEXT: s_mov_b32 s4, s6
; GFX11-NEXT: s_mov_b32 s5, s7
; GFX11-NEXT: s_mov_b32 s6, s8
; GFX11-NEXT: s_mov_b32 s7, s9
; GFX11-NEXT: s_movrels_b32 s0, s0
; GFX11-NEXT: v_mov_b32_e32 v0, s0
; GFX11-NEXT: ds_store_b32 v0, v0
; GFX11-NEXT: s_endpgm
entry:
%ext = extractelement <8 x i8 addrspace(3)*> %vec, i32 %idx
store i8 addrspace(3)* %ext, i8 addrspace(3)* addrspace(3)* undef
ret void
}
define i8 addrspace(1)* @dyn_extract_v8p1_v_v(<8 x i8 addrspace(1)*> %vec, i32 %idx) {
; GCN-LABEL: dyn_extract_v8p1_v_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v16
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v16
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v16
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v16
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v16
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 6, v16
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 7, v16
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: dyn_extract_v8p1_v_v:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v16
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v15, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: dyn_extract_v8p1_v_v:
; GFX11: ; %bb.0: ; %entry
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v16
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v2 :: v_dual_cndmask_b32 v1, v1, v3
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v16
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v4 :: v_dual_cndmask_b32 v1, v1, v5
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v16
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v6 :: v_dual_cndmask_b32 v1, v1, v7
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v16
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v8 :: v_dual_cndmask_b32 v1, v1, v9
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v16
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v10 :: v_dual_cndmask_b32 v1, v1, v11
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v16
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v12 :: v_dual_cndmask_b32 v1, v1, v13
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v16
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v14 :: v_dual_cndmask_b32 v1, v1, v15
; GFX11-NEXT: s_setpc_b64 s[30:31]
entry:
%ext = extractelement <8 x i8 addrspace(1)*> %vec, i32 %idx
ret i8 addrspace(1)* %ext
}
define amdgpu_ps void @dyn_extract_v8p1_s_s(<8 x i8 addrspace(1)*> inreg %vec, i32 inreg %idx) {
; GPRIDX-LABEL: dyn_extract_v8p1_s_s:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_mov_b32 s0, s2
; GPRIDX-NEXT: s_mov_b32 s1, s3
; GPRIDX-NEXT: s_mov_b32 m0, s18
; GPRIDX-NEXT: s_mov_b32 s2, s4
; GPRIDX-NEXT: s_mov_b32 s3, s5
; GPRIDX-NEXT: s_mov_b32 s4, s6
; GPRIDX-NEXT: s_mov_b32 s5, s7
; GPRIDX-NEXT: s_mov_b32 s6, s8
; GPRIDX-NEXT: s_mov_b32 s7, s9
; GPRIDX-NEXT: s_mov_b32 s8, s10
; GPRIDX-NEXT: s_mov_b32 s9, s11
; GPRIDX-NEXT: s_mov_b32 s10, s12
; GPRIDX-NEXT: s_mov_b32 s11, s13
; GPRIDX-NEXT: s_mov_b32 s12, s14
; GPRIDX-NEXT: s_mov_b32 s13, s15
; GPRIDX-NEXT: s_mov_b32 s14, s16
; GPRIDX-NEXT: s_mov_b32 s15, s17
; GPRIDX-NEXT: s_movrels_b64 s[0:1], s[0:1]
; GPRIDX-NEXT: v_mov_b32_e32 v0, s0
; GPRIDX-NEXT: v_mov_b32_e32 v1, s1
; GPRIDX-NEXT: global_store_dwordx2 v[0:1], v[0:1], off
; GPRIDX-NEXT: s_endpgm
;
; MOVREL-LABEL: dyn_extract_v8p1_s_s:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_mov_b32 s0, s2
; MOVREL-NEXT: s_mov_b32 s1, s3
; MOVREL-NEXT: s_mov_b32 m0, s18
; MOVREL-NEXT: s_mov_b32 s2, s4
; MOVREL-NEXT: s_mov_b32 s3, s5
; MOVREL-NEXT: s_mov_b32 s4, s6
; MOVREL-NEXT: s_mov_b32 s5, s7
; MOVREL-NEXT: s_mov_b32 s6, s8
; MOVREL-NEXT: s_mov_b32 s7, s9
; MOVREL-NEXT: s_mov_b32 s8, s10
; MOVREL-NEXT: s_mov_b32 s9, s11
; MOVREL-NEXT: s_mov_b32 s10, s12
; MOVREL-NEXT: s_mov_b32 s11, s13
; MOVREL-NEXT: s_mov_b32 s12, s14
; MOVREL-NEXT: s_mov_b32 s13, s15
; MOVREL-NEXT: s_mov_b32 s14, s16
; MOVREL-NEXT: s_mov_b32 s15, s17
; MOVREL-NEXT: s_movrels_b64 s[0:1], s[0:1]
; MOVREL-NEXT: v_mov_b32_e32 v0, s0
; MOVREL-NEXT: v_mov_b32_e32 v1, s1
; MOVREL-NEXT: flat_store_dwordx2 v[0:1], v[0:1]
; MOVREL-NEXT: s_endpgm
;
; GFX10-LABEL: dyn_extract_v8p1_s_s:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_mov_b32 s0, s2
; GFX10-NEXT: s_mov_b32 s1, s3
; GFX10-NEXT: s_mov_b32 m0, s18
; GFX10-NEXT: s_mov_b32 s2, s4
; GFX10-NEXT: s_mov_b32 s3, s5
; GFX10-NEXT: s_mov_b32 s4, s6
; GFX10-NEXT: s_mov_b32 s5, s7
; GFX10-NEXT: s_mov_b32 s6, s8
; GFX10-NEXT: s_mov_b32 s7, s9
; GFX10-NEXT: s_mov_b32 s8, s10
; GFX10-NEXT: s_mov_b32 s9, s11
; GFX10-NEXT: s_mov_b32 s10, s12
; GFX10-NEXT: s_mov_b32 s11, s13
; GFX10-NEXT: s_mov_b32 s12, s14
; GFX10-NEXT: s_mov_b32 s13, s15
; GFX10-NEXT: s_mov_b32 s14, s16
; GFX10-NEXT: s_mov_b32 s15, s17
; GFX10-NEXT: s_movrels_b64 s[0:1], s[0:1]
; GFX10-NEXT: v_mov_b32_e32 v0, s0
; GFX10-NEXT: v_mov_b32_e32 v1, s1
; GFX10-NEXT: global_store_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_endpgm
;
; GFX11-LABEL: dyn_extract_v8p1_s_s:
; GFX11: ; %bb.0: ; %entry
; GFX11-NEXT: s_mov_b32 s0, s2
; GFX11-NEXT: s_mov_b32 s1, s3
; GFX11-NEXT: s_mov_b32 m0, s18
; GFX11-NEXT: s_mov_b32 s2, s4
; GFX11-NEXT: s_mov_b32 s3, s5
; GFX11-NEXT: s_mov_b32 s4, s6
; GFX11-NEXT: s_mov_b32 s5, s7
; GFX11-NEXT: s_mov_b32 s6, s8
; GFX11-NEXT: s_mov_b32 s7, s9
; GFX11-NEXT: s_mov_b32 s8, s10
; GFX11-NEXT: s_mov_b32 s9, s11
; GFX11-NEXT: s_mov_b32 s10, s12
; GFX11-NEXT: s_mov_b32 s11, s13
; GFX11-NEXT: s_mov_b32 s12, s14
; GFX11-NEXT: s_mov_b32 s13, s15
; GFX11-NEXT: s_mov_b32 s14, s16
; GFX11-NEXT: s_mov_b32 s15, s17
; GFX11-NEXT: s_movrels_b64 s[0:1], s[0:1]
; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
; GFX11-NEXT: global_store_b64 v[0:1], v[0:1], off
; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX11-NEXT: s_endpgm
entry:
%ext = extractelement <8 x i8 addrspace(1)*> %vec, i32 %idx
store i8 addrspace(1)* %ext, i8 addrspace(1)* addrspace(1)* undef
ret void
}
define amdgpu_ps float @dyn_extract_v16f32_v_s(<16 x float> %vec, i32 inreg %sel) {
; GPRIDX-LABEL: dyn_extract_v16f32_v_s:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_set_gpr_idx_on s2, gpr_idx(SRC0)
; GPRIDX-NEXT: v_mov_b32_e32 v0, v0
; GPRIDX-NEXT: s_set_gpr_idx_off
; GPRIDX-NEXT: ; return to shader part epilog
;
; MOVREL-LABEL: dyn_extract_v16f32_v_s:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_mov_b32 m0, s2
; MOVREL-NEXT: v_movrels_b32_e32 v0, v0
; MOVREL-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v16f32_v_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 m0, s2
; GFX10PLUS-NEXT: v_movrels_b32_e32 v0, v0
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <16 x float> %vec, i32 %sel
ret float %ext
}
define amdgpu_ps float @dyn_extract_v32f32_v_s(<32 x float> %vec, i32 inreg %sel) {
; GPRIDX-LABEL: dyn_extract_v32f32_v_s:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_set_gpr_idx_on s2, gpr_idx(SRC0)
; GPRIDX-NEXT: v_mov_b32_e32 v0, v0
; GPRIDX-NEXT: s_set_gpr_idx_off
; GPRIDX-NEXT: ; return to shader part epilog
;
; MOVREL-LABEL: dyn_extract_v32f32_v_s:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_mov_b32 m0, s2
; MOVREL-NEXT: v_movrels_b32_e32 v0, v0
; MOVREL-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v32f32_v_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 m0, s2
; GFX10PLUS-NEXT: v_movrels_b32_e32 v0, v0
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <32 x float> %vec, i32 %sel
ret float %ext
}
define amdgpu_ps double @dyn_extract_v16f64_v_s(<16 x double> %vec, i32 inreg %sel) {
; GPRIDX-LABEL: dyn_extract_v16f64_v_s:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_lshl_b32 s0, s2, 1
; GPRIDX-NEXT: s_set_gpr_idx_on s0, gpr_idx(SRC0)
; GPRIDX-NEXT: v_mov_b32_e32 v32, v0
; GPRIDX-NEXT: v_mov_b32_e32 v0, v1
; GPRIDX-NEXT: s_set_gpr_idx_off
; GPRIDX-NEXT: v_readfirstlane_b32 s0, v32
; GPRIDX-NEXT: v_readfirstlane_b32 s1, v0
; GPRIDX-NEXT: ; return to shader part epilog
;
; MOVREL-LABEL: dyn_extract_v16f64_v_s:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_lshl_b32 m0, s2, 1
; MOVREL-NEXT: v_movrels_b32_e32 v32, v0
; MOVREL-NEXT: v_movrels_b32_e32 v0, v1
; MOVREL-NEXT: v_readfirstlane_b32 s0, v32
; MOVREL-NEXT: v_readfirstlane_b32 s1, v0
; MOVREL-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v16f64_v_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_lshl_b32 m0, s2, 1
; GFX10PLUS-NEXT: v_movrels_b32_e32 v32, v0
; GFX10PLUS-NEXT: v_movrels_b32_e32 v0, v1
; GFX10PLUS-NEXT: v_readfirstlane_b32 s0, v32
; GFX10PLUS-NEXT: v_readfirstlane_b32 s1, v0
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <16 x double> %vec, i32 %sel
ret double %ext
}
define amdgpu_ps float @dyn_extract_v16f32_s_s(i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v16f32_s_s:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s4, 1.0
; GCN-NEXT: s_mov_b32 m0, s2
; GCN-NEXT: s_mov_b32 s19, 0x41800000
; GCN-NEXT: s_mov_b32 s18, 0x41700000
; GCN-NEXT: s_mov_b32 s17, 0x41600000
; GCN-NEXT: s_mov_b32 s16, 0x41500000
; GCN-NEXT: s_mov_b32 s15, 0x41400000
; GCN-NEXT: s_mov_b32 s14, 0x41300000
; GCN-NEXT: s_mov_b32 s13, 0x41200000
; GCN-NEXT: s_mov_b32 s12, 0x41100000
; GCN-NEXT: s_mov_b32 s11, 0x41000000
; GCN-NEXT: s_mov_b32 s10, 0x40e00000
; GCN-NEXT: s_mov_b32 s9, 0x40c00000
; GCN-NEXT: s_mov_b32 s8, 0x40a00000
; GCN-NEXT: s_mov_b32 s7, 4.0
; GCN-NEXT: s_mov_b32 s6, 0x40400000
; GCN-NEXT: s_mov_b32 s5, 2.0
; GCN-NEXT: s_movrels_b32 s0, s4
; GCN-NEXT: v_mov_b32_e32 v0, s0
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v16f32_s_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s4, 1.0
; GFX10PLUS-NEXT: s_mov_b32 m0, s2
; GFX10PLUS-NEXT: s_mov_b32 s19, 0x41800000
; GFX10PLUS-NEXT: s_mov_b32 s18, 0x41700000
; GFX10PLUS-NEXT: s_mov_b32 s17, 0x41600000
; GFX10PLUS-NEXT: s_mov_b32 s16, 0x41500000
; GFX10PLUS-NEXT: s_mov_b32 s15, 0x41400000
; GFX10PLUS-NEXT: s_mov_b32 s14, 0x41300000
; GFX10PLUS-NEXT: s_mov_b32 s13, 0x41200000
; GFX10PLUS-NEXT: s_mov_b32 s12, 0x41100000
; GFX10PLUS-NEXT: s_mov_b32 s11, 0x41000000
; GFX10PLUS-NEXT: s_mov_b32 s10, 0x40e00000
; GFX10PLUS-NEXT: s_mov_b32 s9, 0x40c00000
; GFX10PLUS-NEXT: s_mov_b32 s8, 0x40a00000
; GFX10PLUS-NEXT: s_mov_b32 s7, 4.0
; GFX10PLUS-NEXT: s_mov_b32 s6, 0x40400000
; GFX10PLUS-NEXT: s_mov_b32 s5, 2.0
; GFX10PLUS-NEXT: s_movrels_b32 s0, s4
; GFX10PLUS-NEXT: v_mov_b32_e32 v0, s0
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <16 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0, float 9.0, float 10.0, float 11.0, float 12.0, float 13.0, float 14.0, float 15.0, float 16.0>, i32 %sel
ret float %ext
}
define amdgpu_ps float @dyn_extract_v32f32_s_s(i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v32f32_s_s:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s36, 1.0
; GCN-NEXT: s_mov_b32 m0, s2
; GCN-NEXT: s_mov_b32 s67, 0x42000000
; GCN-NEXT: s_mov_b32 s66, 0x41f80000
; GCN-NEXT: s_mov_b32 s65, 0x41f00000
; GCN-NEXT: s_mov_b32 s64, 0x41e80000
; GCN-NEXT: s_mov_b32 s63, 0x41e00000
; GCN-NEXT: s_mov_b32 s62, 0x41d80000
; GCN-NEXT: s_mov_b32 s61, 0x41d00000
; GCN-NEXT: s_mov_b32 s60, 0x41c80000
; GCN-NEXT: s_mov_b32 s59, 0x41c00000
; GCN-NEXT: s_mov_b32 s58, 0x41b80000
; GCN-NEXT: s_mov_b32 s57, 0x41b00000
; GCN-NEXT: s_mov_b32 s56, 0x41a80000
; GCN-NEXT: s_mov_b32 s55, 0x41a00000
; GCN-NEXT: s_mov_b32 s54, 0x41980000
; GCN-NEXT: s_mov_b32 s53, 0x41900000
; GCN-NEXT: s_mov_b32 s52, 0x41880000
; GCN-NEXT: s_mov_b32 s51, 0x41800000
; GCN-NEXT: s_mov_b32 s50, 0x41700000
; GCN-NEXT: s_mov_b32 s49, 0x41600000
; GCN-NEXT: s_mov_b32 s48, 0x41500000
; GCN-NEXT: s_mov_b32 s47, 0x41400000
; GCN-NEXT: s_mov_b32 s46, 0x41300000
; GCN-NEXT: s_mov_b32 s45, 0x41200000
; GCN-NEXT: s_mov_b32 s44, 0x41100000
; GCN-NEXT: s_mov_b32 s43, 0x41000000
; GCN-NEXT: s_mov_b32 s42, 0x40e00000
; GCN-NEXT: s_mov_b32 s41, 0x40c00000
; GCN-NEXT: s_mov_b32 s40, 0x40a00000
; GCN-NEXT: s_mov_b32 s39, 4.0
; GCN-NEXT: s_mov_b32 s38, 0x40400000
; GCN-NEXT: s_mov_b32 s37, 2.0
; GCN-NEXT: s_movrels_b32 s0, s36
; GCN-NEXT: v_mov_b32_e32 v0, s0
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v32f32_s_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s36, 1.0
; GFX10PLUS-NEXT: s_mov_b32 m0, s2
; GFX10PLUS-NEXT: s_mov_b32 s67, 0x42000000
; GFX10PLUS-NEXT: s_mov_b32 s66, 0x41f80000
; GFX10PLUS-NEXT: s_mov_b32 s65, 0x41f00000
; GFX10PLUS-NEXT: s_mov_b32 s64, 0x41e80000
; GFX10PLUS-NEXT: s_mov_b32 s63, 0x41e00000
; GFX10PLUS-NEXT: s_mov_b32 s62, 0x41d80000
; GFX10PLUS-NEXT: s_mov_b32 s61, 0x41d00000
; GFX10PLUS-NEXT: s_mov_b32 s60, 0x41c80000
; GFX10PLUS-NEXT: s_mov_b32 s59, 0x41c00000
; GFX10PLUS-NEXT: s_mov_b32 s58, 0x41b80000
; GFX10PLUS-NEXT: s_mov_b32 s57, 0x41b00000
; GFX10PLUS-NEXT: s_mov_b32 s56, 0x41a80000
; GFX10PLUS-NEXT: s_mov_b32 s55, 0x41a00000
; GFX10PLUS-NEXT: s_mov_b32 s54, 0x41980000
; GFX10PLUS-NEXT: s_mov_b32 s53, 0x41900000
; GFX10PLUS-NEXT: s_mov_b32 s52, 0x41880000
; GFX10PLUS-NEXT: s_mov_b32 s51, 0x41800000
; GFX10PLUS-NEXT: s_mov_b32 s50, 0x41700000
; GFX10PLUS-NEXT: s_mov_b32 s49, 0x41600000
; GFX10PLUS-NEXT: s_mov_b32 s48, 0x41500000
; GFX10PLUS-NEXT: s_mov_b32 s47, 0x41400000
; GFX10PLUS-NEXT: s_mov_b32 s46, 0x41300000
; GFX10PLUS-NEXT: s_mov_b32 s45, 0x41200000
; GFX10PLUS-NEXT: s_mov_b32 s44, 0x41100000
; GFX10PLUS-NEXT: s_mov_b32 s43, 0x41000000
; GFX10PLUS-NEXT: s_mov_b32 s42, 0x40e00000
; GFX10PLUS-NEXT: s_mov_b32 s41, 0x40c00000
; GFX10PLUS-NEXT: s_mov_b32 s40, 0x40a00000
; GFX10PLUS-NEXT: s_mov_b32 s39, 4.0
; GFX10PLUS-NEXT: s_mov_b32 s38, 0x40400000
; GFX10PLUS-NEXT: s_mov_b32 s37, 2.0
; GFX10PLUS-NEXT: s_movrels_b32 s0, s36
; GFX10PLUS-NEXT: v_mov_b32_e32 v0, s0
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <32 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0, float 9.0, float 10.0, float 11.0, float 12.0, float 13.0, float 14.0, float 15.0, float 16.0, float 17.0, float 18.0, float 19.0, float 20.0, float 21.0, float 22.0, float 23.0, float 24.0, float 25.0, float 26.0, float 27.0, float 28.0, float 29.0, float 30.0, float 31.0, float 32.0>, i32 %sel
ret float %ext
}
define amdgpu_ps double @dyn_extract_v16f64_s_s(i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v16f64_s_s:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s66, 0
; GCN-NEXT: s_mov_b64 s[36:37], 1.0
; GCN-NEXT: s_mov_b32 m0, s2
; GCN-NEXT: s_mov_b32 s67, 0x40300000
; GCN-NEXT: s_mov_b32 s65, 0x402e0000
; GCN-NEXT: s_mov_b32 s64, s66
; GCN-NEXT: s_mov_b32 s63, 0x402c0000
; GCN-NEXT: s_mov_b32 s62, s66
; GCN-NEXT: s_mov_b32 s61, 0x402a0000
; GCN-NEXT: s_mov_b32 s60, s66
; GCN-NEXT: s_mov_b32 s59, 0x40280000
; GCN-NEXT: s_mov_b32 s58, s66
; GCN-NEXT: s_mov_b32 s57, 0x40260000
; GCN-NEXT: s_mov_b32 s56, s66
; GCN-NEXT: s_mov_b32 s55, 0x40240000
; GCN-NEXT: s_mov_b32 s54, s66
; GCN-NEXT: s_mov_b32 s53, 0x40220000
; GCN-NEXT: s_mov_b32 s52, s66
; GCN-NEXT: s_mov_b32 s51, 0x40200000
; GCN-NEXT: s_mov_b32 s50, s66
; GCN-NEXT: s_mov_b32 s49, 0x401c0000
; GCN-NEXT: s_mov_b32 s48, s66
; GCN-NEXT: s_mov_b32 s47, 0x40180000
; GCN-NEXT: s_mov_b32 s46, s66
; GCN-NEXT: s_mov_b32 s45, 0x40140000
; GCN-NEXT: s_mov_b32 s44, s66
; GCN-NEXT: s_mov_b64 s[42:43], 4.0
; GCN-NEXT: s_mov_b32 s41, 0x40080000
; GCN-NEXT: s_mov_b32 s40, s66
; GCN-NEXT: s_mov_b64 s[38:39], 2.0
; GCN-NEXT: s_movrels_b64 s[0:1], s[36:37]
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v16f64_s_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s66, 0
; GFX10PLUS-NEXT: s_mov_b64 s[36:37], 1.0
; GFX10PLUS-NEXT: s_mov_b32 m0, s2
; GFX10PLUS-NEXT: s_mov_b32 s67, 0x40300000
; GFX10PLUS-NEXT: s_mov_b32 s65, 0x402e0000
; GFX10PLUS-NEXT: s_mov_b32 s64, s66
; GFX10PLUS-NEXT: s_mov_b32 s63, 0x402c0000
; GFX10PLUS-NEXT: s_mov_b32 s62, s66
; GFX10PLUS-NEXT: s_mov_b32 s61, 0x402a0000
; GFX10PLUS-NEXT: s_mov_b32 s60, s66
; GFX10PLUS-NEXT: s_mov_b32 s59, 0x40280000
; GFX10PLUS-NEXT: s_mov_b32 s58, s66
; GFX10PLUS-NEXT: s_mov_b32 s57, 0x40260000
; GFX10PLUS-NEXT: s_mov_b32 s56, s66
; GFX10PLUS-NEXT: s_mov_b32 s55, 0x40240000
; GFX10PLUS-NEXT: s_mov_b32 s54, s66
; GFX10PLUS-NEXT: s_mov_b32 s53, 0x40220000
; GFX10PLUS-NEXT: s_mov_b32 s52, s66
; GFX10PLUS-NEXT: s_mov_b32 s51, 0x40200000
; GFX10PLUS-NEXT: s_mov_b32 s50, s66
; GFX10PLUS-NEXT: s_mov_b32 s49, 0x401c0000
; GFX10PLUS-NEXT: s_mov_b32 s48, s66
; GFX10PLUS-NEXT: s_mov_b32 s47, 0x40180000
; GFX10PLUS-NEXT: s_mov_b32 s46, s66
; GFX10PLUS-NEXT: s_mov_b32 s45, 0x40140000
; GFX10PLUS-NEXT: s_mov_b32 s44, s66
; GFX10PLUS-NEXT: s_mov_b64 s[42:43], 4.0
; GFX10PLUS-NEXT: s_mov_b32 s41, 0x40080000
; GFX10PLUS-NEXT: s_mov_b32 s40, s66
; GFX10PLUS-NEXT: s_mov_b64 s[38:39], 2.0
; GFX10PLUS-NEXT: s_movrels_b64 s[0:1], s[36:37]
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <16 x double> <double 1.0, double 2.0, double 3.0, double 4.0, double 5.0, double 6.0, double 7.0, double 8.0, double 9.0, double 10.0, double 11.0, double 12.0, double 13.0, double 14.0, double 15.0, double 16.0>, i32 %sel
ret double %ext
}
define amdgpu_ps float @dyn_extract_v6f32_s_v(<6 x float> inreg %vec, i32 %sel) {
; GCN-LABEL: dyn_extract_v6f32_s_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 s1, s3
; GCN-NEXT: v_mov_b32_e32 v1, s0
; GCN-NEXT: v_mov_b32_e32 v2, s1
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; GCN-NEXT: v_mov_b32_e32 v3, s4
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v0
; GCN-NEXT: v_mov_b32_e32 v4, s5
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v0
; GCN-NEXT: v_mov_b32_e32 v5, s6
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v4, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v0
; GCN-NEXT: v_mov_b32_e32 v6, s7
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v0
; GCN-NEXT: v_cndmask_b32_e32 v0, v1, v6, vcc
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v6f32_s_v:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX10PLUS-NEXT: v_mov_b32_e32 v1, s1
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v1, s0, v1, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s3, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v0, v1, s5, vcc_lo
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <6 x float> %vec, i32 %sel
ret float %ext
}
define float @dyn_extract_v6f32_v_v(<6 x float> %vec, i32 %sel) {
; GCN-LABEL: dyn_extract_v6f32_v_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v6
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v6
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v6
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v6
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v6
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: dyn_extract_v6f32_v_v:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v6
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v6
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v6
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v6
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v6
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
entry:
%ext = extractelement <6 x float> %vec, i32 %sel
ret float %ext
}
define amdgpu_ps float @dyn_extract_v6f32_v_s(<6 x float> %vec, i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v6f32_v_s:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: v_cmp_eq_u32_e64 vcc, s2, 1
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GCN-NEXT: v_cmp_eq_u32_e64 vcc, s2, 2
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GCN-NEXT: v_cmp_eq_u32_e64 vcc, s2, 3
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e64 vcc, s2, 4
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GCN-NEXT: v_cmp_eq_u32_e64 vcc, s2, 5
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v6f32_v_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 1
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 2
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 3
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 4
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 5
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <6 x float> %vec, i32 %sel
ret float %ext
}
define amdgpu_ps float @dyn_extract_v6f32_s_s(<6 x float> inreg %vec, i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v6f32_s_s:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_cmp_eq_u32 s8, 1
; GCN-NEXT: s_cselect_b32 s0, s3, s2
; GCN-NEXT: s_cmp_eq_u32 s8, 2
; GCN-NEXT: s_cselect_b32 s0, s4, s0
; GCN-NEXT: s_cmp_eq_u32 s8, 3
; GCN-NEXT: s_cselect_b32 s0, s5, s0
; GCN-NEXT: s_cmp_eq_u32 s8, 4
; GCN-NEXT: s_cselect_b32 s0, s6, s0
; GCN-NEXT: s_cmp_eq_u32 s8, 5
; GCN-NEXT: s_cselect_b32 s0, s7, s0
; GCN-NEXT: v_mov_b32_e32 v0, s0
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v6f32_s_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_cmp_eq_u32 s8, 1
; GFX10PLUS-NEXT: s_cselect_b32 s0, s3, s2
; GFX10PLUS-NEXT: s_cmp_eq_u32 s8, 2
; GFX10PLUS-NEXT: s_cselect_b32 s0, s4, s0
; GFX10PLUS-NEXT: s_cmp_eq_u32 s8, 3
; GFX10PLUS-NEXT: s_cselect_b32 s0, s5, s0
; GFX10PLUS-NEXT: s_cmp_eq_u32 s8, 4
; GFX10PLUS-NEXT: s_cselect_b32 s0, s6, s0
; GFX10PLUS-NEXT: s_cmp_eq_u32 s8, 5
; GFX10PLUS-NEXT: s_cselect_b32 s0, s7, s0
; GFX10PLUS-NEXT: v_mov_b32_e32 v0, s0
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <6 x float> %vec, i32 %sel
ret float %ext
}
define amdgpu_ps float @dyn_extract_v7f32_s_v(<7 x float> inreg %vec, i32 %sel) {
; GCN-LABEL: dyn_extract_v7f32_s_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 s1, s3
; GCN-NEXT: s_mov_b32 s2, s4
; GCN-NEXT: v_mov_b32_e32 v1, s0
; GCN-NEXT: v_mov_b32_e32 v2, s1
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; GCN-NEXT: v_mov_b32_e32 v3, s2
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v0
; GCN-NEXT: v_mov_b32_e32 v4, s5
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v0
; GCN-NEXT: v_mov_b32_e32 v5, s6
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v4, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v0
; GCN-NEXT: v_mov_b32_e32 v6, s7
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v0
; GCN-NEXT: v_mov_b32_e32 v7, s8
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v6, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 6, v0
; GCN-NEXT: v_cndmask_b32_e32 v0, v1, v7, vcc
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v7f32_s_v:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX10PLUS-NEXT: v_mov_b32_e32 v1, s1
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v1, s0, v1, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: s_mov_b32 s6, s8
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s3, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s5, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v0, v1, s6, vcc_lo
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <7 x float> %vec, i32 %sel
ret float %ext
}
define float @dyn_extract_v7f32_v_v(<7 x float> %vec, i32 %sel) {
; GCN-LABEL: dyn_extract_v7f32_v_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v7
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v7
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v7
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v7
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v7
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 6, v7
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: dyn_extract_v7f32_v_v:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v7
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v7
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v7
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v7
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v7
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v7
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
entry:
%ext = extractelement <7 x float> %vec, i32 %sel
ret float %ext
}
define amdgpu_ps float @dyn_extract_v7f32_v_s(<7 x float> %vec, i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v7f32_v_s:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: v_cmp_eq_u32_e64 vcc, s2, 1
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GCN-NEXT: v_cmp_eq_u32_e64 vcc, s2, 2
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GCN-NEXT: v_cmp_eq_u32_e64 vcc, s2, 3
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e64 vcc, s2, 4
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GCN-NEXT: v_cmp_eq_u32_e64 vcc, s2, 5
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc
; GCN-NEXT: v_cmp_eq_u32_e64 vcc, s2, 6
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v7f32_v_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 1
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 2
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 3
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 4
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 5
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 6
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <7 x float> %vec, i32 %sel
ret float %ext
}
define amdgpu_ps float @dyn_extract_v7f32_s_s(<7 x float> inreg %vec, i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v7f32_s_s:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_cmp_eq_u32 s9, 1
; GCN-NEXT: s_cselect_b32 s0, s3, s2
; GCN-NEXT: s_cmp_eq_u32 s9, 2
; GCN-NEXT: s_cselect_b32 s0, s4, s0
; GCN-NEXT: s_cmp_eq_u32 s9, 3
; GCN-NEXT: s_cselect_b32 s0, s5, s0
; GCN-NEXT: s_cmp_eq_u32 s9, 4
; GCN-NEXT: s_cselect_b32 s0, s6, s0
; GCN-NEXT: s_cmp_eq_u32 s9, 5
; GCN-NEXT: s_cselect_b32 s0, s7, s0
; GCN-NEXT: s_cmp_eq_u32 s9, 6
; GCN-NEXT: s_cselect_b32 s0, s8, s0
; GCN-NEXT: v_mov_b32_e32 v0, s0
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v7f32_s_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_cmp_eq_u32 s9, 1
; GFX10PLUS-NEXT: s_cselect_b32 s0, s3, s2
; GFX10PLUS-NEXT: s_cmp_eq_u32 s9, 2
; GFX10PLUS-NEXT: s_cselect_b32 s0, s4, s0
; GFX10PLUS-NEXT: s_cmp_eq_u32 s9, 3
; GFX10PLUS-NEXT: s_cselect_b32 s0, s5, s0
; GFX10PLUS-NEXT: s_cmp_eq_u32 s9, 4
; GFX10PLUS-NEXT: s_cselect_b32 s0, s6, s0
; GFX10PLUS-NEXT: s_cmp_eq_u32 s9, 5
; GFX10PLUS-NEXT: s_cselect_b32 s0, s7, s0
; GFX10PLUS-NEXT: s_cmp_eq_u32 s9, 6
; GFX10PLUS-NEXT: s_cselect_b32 s0, s8, s0
; GFX10PLUS-NEXT: v_mov_b32_e32 v0, s0
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <7 x float> %vec, i32 %sel
ret float %ext
}
define amdgpu_ps double @dyn_extract_v6f64_s_v(<6 x double> inreg %vec, i32 %sel) {
; GCN-LABEL: dyn_extract_v6f64_s_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 s1, s3
; GCN-NEXT: s_mov_b32 s2, s4
; GCN-NEXT: s_mov_b32 s3, s5
; GCN-NEXT: s_mov_b32 s4, s6
; GCN-NEXT: s_mov_b32 s5, s7
; GCN-NEXT: v_mov_b32_e32 v1, s0
; GCN-NEXT: v_mov_b32_e32 v2, s1
; GCN-NEXT: v_mov_b32_e32 v3, s2
; GCN-NEXT: v_mov_b32_e32 v4, s3
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; GCN-NEXT: s_mov_b32 s6, s8
; GCN-NEXT: s_mov_b32 s7, s9
; GCN-NEXT: v_mov_b32_e32 v5, s4
; GCN-NEXT: v_mov_b32_e32 v6, s5
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GCN-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v0
; GCN-NEXT: v_mov_b32_e32 v7, s6
; GCN-NEXT: v_mov_b32_e32 v8, s7
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; GCN-NEXT: v_cndmask_b32_e32 v2, v2, v6, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v0
; GCN-NEXT: v_mov_b32_e32 v9, s10
; GCN-NEXT: v_mov_b32_e32 v10, s11
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc
; GCN-NEXT: v_cndmask_b32_e32 v2, v2, v8, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v0
; GCN-NEXT: v_mov_b32_e32 v11, s12
; GCN-NEXT: v_mov_b32_e32 v12, s13
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc
; GCN-NEXT: v_cndmask_b32_e32 v2, v2, v10, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v0
; GCN-NEXT: v_cndmask_b32_e32 v0, v1, v11, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v2, v12, vcc
; GCN-NEXT: v_readfirstlane_b32 s0, v0
; GCN-NEXT: v_readfirstlane_b32 s1, v1
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: dyn_extract_v6f64_s_v:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_mov_b32 s0, s2
; GFX10-NEXT: s_mov_b32 s2, s4
; GFX10-NEXT: s_mov_b32 s15, s5
; GFX10-NEXT: v_mov_b32_e32 v1, s2
; GFX10-NEXT: v_mov_b32_e32 v2, s15
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX10-NEXT: s_mov_b32 s1, s3
; GFX10-NEXT: s_mov_b32 s4, s6
; GFX10-NEXT: s_mov_b32 s5, s7
; GFX10-NEXT: s_mov_b32 s6, s8
; GFX10-NEXT: v_cndmask_b32_e32 v1, s0, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v2, s1, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10-NEXT: s_mov_b32 s7, s9
; GFX10-NEXT: s_mov_b32 s8, s10
; GFX10-NEXT: s_mov_b32 s9, s11
; GFX10-NEXT: s_mov_b32 s10, s12
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10-NEXT: s_mov_b32 s11, s13
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, s10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v1, v2, s11, vcc_lo
; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: v_readfirstlane_b32 s1, v1
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: dyn_extract_v6f64_s_v:
; GFX11: ; %bb.0: ; %entry
; GFX11-NEXT: s_mov_b32 s0, s2
; GFX11-NEXT: s_mov_b32 s2, s4
; GFX11-NEXT: s_mov_b32 s15, s5
; GFX11-NEXT: v_dual_mov_b32 v1, s2 :: v_dual_mov_b32 v2, s15
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX11-NEXT: s_mov_b32 s1, s3
; GFX11-NEXT: s_mov_b32 s4, s6
; GFX11-NEXT: s_mov_b32 s5, s7
; GFX11-NEXT: s_mov_b32 s6, s8
; GFX11-NEXT: v_cndmask_b32_e32 v1, s0, v1, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e32 v2, s1, v2, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX11-NEXT: s_mov_b32 s7, s9
; GFX11-NEXT: s_mov_b32 s8, s10
; GFX11-NEXT: s_mov_b32 s9, s11
; GFX11-NEXT: s_mov_b32 s10, s12
; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s5, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX11-NEXT: s_mov_b32 s11, s13
; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s6, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s7, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s9, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX11-NEXT: v_cndmask_b32_e64 v0, v1, s10, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v1, v2, s11, vcc_lo
; GFX11-NEXT: v_readfirstlane_b32 s0, v0
; GFX11-NEXT: v_readfirstlane_b32 s1, v1
; GFX11-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <6 x double> %vec, i32 %sel
ret double %ext
}
define double @dyn_extract_v6f64_v_v(<6 x double> %vec, i32 %sel) {
; GCN-LABEL: dyn_extract_v6f64_v_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v12
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v12
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v12
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v12
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v12
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: dyn_extract_v6f64_v_v:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v12
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v12
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v12
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v12
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v12
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: dyn_extract_v6f64_v_v:
; GFX11: ; %bb.0: ; %entry
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v12
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v2 :: v_dual_cndmask_b32 v1, v1, v3
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v12
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v4 :: v_dual_cndmask_b32 v1, v1, v5
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v12
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v6 :: v_dual_cndmask_b32 v1, v1, v7
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v12
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v8 :: v_dual_cndmask_b32 v1, v1, v9
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v12
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v10 :: v_dual_cndmask_b32 v1, v1, v11
; GFX11-NEXT: s_setpc_b64 s[30:31]
entry:
%ext = extractelement <6 x double> %vec, i32 %sel
ret double %ext
}
define amdgpu_ps double @dyn_extract_v6f64_v_s(<6 x double> %vec, i32 inreg %sel) {
; GPRIDX-LABEL: dyn_extract_v6f64_v_s:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_lshl_b32 s0, s2, 1
; GPRIDX-NEXT: s_set_gpr_idx_on s0, gpr_idx(SRC0)
; GPRIDX-NEXT: v_mov_b32_e32 v12, v0
; GPRIDX-NEXT: v_mov_b32_e32 v0, v1
; GPRIDX-NEXT: s_set_gpr_idx_off
; GPRIDX-NEXT: v_readfirstlane_b32 s0, v12
; GPRIDX-NEXT: v_readfirstlane_b32 s1, v0
; GPRIDX-NEXT: ; return to shader part epilog
;
; MOVREL-LABEL: dyn_extract_v6f64_v_s:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_lshl_b32 m0, s2, 1
; MOVREL-NEXT: v_movrels_b32_e32 v12, v0
; MOVREL-NEXT: v_movrels_b32_e32 v0, v1
; MOVREL-NEXT: v_readfirstlane_b32 s0, v12
; MOVREL-NEXT: v_readfirstlane_b32 s1, v0
; MOVREL-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v6f64_v_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_lshl_b32 m0, s2, 1
; GFX10PLUS-NEXT: v_movrels_b32_e32 v12, v0
; GFX10PLUS-NEXT: v_movrels_b32_e32 v0, v1
; GFX10PLUS-NEXT: v_readfirstlane_b32 s0, v12
; GFX10PLUS-NEXT: v_readfirstlane_b32 s1, v0
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <6 x double> %vec, i32 %sel
ret double %ext
}
define amdgpu_ps double @dyn_extract_v6f64_s_s(<6 x double> inreg %vec, i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v6f64_s_s:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 s1, s3
; GCN-NEXT: s_mov_b32 m0, s14
; GCN-NEXT: s_mov_b32 s2, s4
; GCN-NEXT: s_mov_b32 s3, s5
; GCN-NEXT: s_mov_b32 s4, s6
; GCN-NEXT: s_mov_b32 s5, s7
; GCN-NEXT: s_mov_b32 s6, s8
; GCN-NEXT: s_mov_b32 s7, s9
; GCN-NEXT: s_mov_b32 s8, s10
; GCN-NEXT: s_mov_b32 s9, s11
; GCN-NEXT: s_mov_b32 s10, s12
; GCN-NEXT: s_mov_b32 s11, s13
; GCN-NEXT: s_movrels_b64 s[0:1], s[0:1]
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v6f64_s_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: s_mov_b32 m0, s14
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: s_mov_b32 s6, s8
; GFX10PLUS-NEXT: s_mov_b32 s7, s9
; GFX10PLUS-NEXT: s_mov_b32 s8, s10
; GFX10PLUS-NEXT: s_mov_b32 s9, s11
; GFX10PLUS-NEXT: s_mov_b32 s10, s12
; GFX10PLUS-NEXT: s_mov_b32 s11, s13
; GFX10PLUS-NEXT: s_movrels_b64 s[0:1], s[0:1]
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <6 x double> %vec, i32 %sel
ret double %ext
}
define amdgpu_ps double @dyn_extract_v7f64_s_v(<7 x double> inreg %vec, i32 %sel) {
; GCN-LABEL: dyn_extract_v7f64_s_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 s1, s3
; GCN-NEXT: s_mov_b32 s2, s4
; GCN-NEXT: s_mov_b32 s3, s5
; GCN-NEXT: s_mov_b32 s4, s6
; GCN-NEXT: s_mov_b32 s5, s7
; GCN-NEXT: v_mov_b32_e32 v1, s0
; GCN-NEXT: v_mov_b32_e32 v2, s1
; GCN-NEXT: v_mov_b32_e32 v3, s2
; GCN-NEXT: v_mov_b32_e32 v4, s3
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; GCN-NEXT: s_mov_b32 s6, s8
; GCN-NEXT: s_mov_b32 s7, s9
; GCN-NEXT: v_mov_b32_e32 v5, s4
; GCN-NEXT: v_mov_b32_e32 v6, s5
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GCN-NEXT: v_cndmask_b32_e32 v2, v2, v4, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v0
; GCN-NEXT: s_mov_b32 s8, s10
; GCN-NEXT: s_mov_b32 s9, s11
; GCN-NEXT: v_mov_b32_e32 v7, s6
; GCN-NEXT: v_mov_b32_e32 v8, s7
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; GCN-NEXT: v_cndmask_b32_e32 v2, v2, v6, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v0
; GCN-NEXT: v_mov_b32_e32 v9, s8
; GCN-NEXT: v_mov_b32_e32 v10, s9
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc
; GCN-NEXT: v_cndmask_b32_e32 v2, v2, v8, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v0
; GCN-NEXT: v_mov_b32_e32 v11, s12
; GCN-NEXT: v_mov_b32_e32 v12, s13
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc
; GCN-NEXT: v_cndmask_b32_e32 v2, v2, v10, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v0
; GCN-NEXT: v_mov_b32_e32 v13, s14
; GCN-NEXT: v_mov_b32_e32 v14, s15
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc
; GCN-NEXT: v_cndmask_b32_e32 v2, v2, v12, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 6, v0
; GCN-NEXT: v_cndmask_b32_e32 v0, v1, v13, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v2, v14, vcc
; GCN-NEXT: v_readfirstlane_b32 s0, v0
; GCN-NEXT: v_readfirstlane_b32 s1, v1
; GCN-NEXT: ; return to shader part epilog
;
; GFX10-LABEL: dyn_extract_v7f64_s_v:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_mov_b32 s0, s2
; GFX10-NEXT: s_mov_b32 s2, s4
; GFX10-NEXT: s_mov_b32 s19, s5
; GFX10-NEXT: v_mov_b32_e32 v1, s2
; GFX10-NEXT: v_mov_b32_e32 v2, s19
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX10-NEXT: s_mov_b32 s1, s3
; GFX10-NEXT: s_mov_b32 s4, s6
; GFX10-NEXT: s_mov_b32 s5, s7
; GFX10-NEXT: s_mov_b32 s6, s8
; GFX10-NEXT: v_cndmask_b32_e32 v1, s0, v1, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v2, s1, v2, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10-NEXT: s_mov_b32 s7, s9
; GFX10-NEXT: s_mov_b32 s8, s10
; GFX10-NEXT: s_mov_b32 s9, s11
; GFX10-NEXT: s_mov_b32 s10, s12
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10-NEXT: s_mov_b32 s11, s13
; GFX10-NEXT: s_mov_b32 s12, s14
; GFX10-NEXT: s_mov_b32 s13, s15
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0
; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, s12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e64 v1, v2, s13, vcc_lo
; GFX10-NEXT: v_readfirstlane_b32 s0, v0
; GFX10-NEXT: v_readfirstlane_b32 s1, v1
; GFX10-NEXT: ; return to shader part epilog
;
; GFX11-LABEL: dyn_extract_v7f64_s_v:
; GFX11: ; %bb.0: ; %entry
; GFX11-NEXT: s_mov_b32 s0, s2
; GFX11-NEXT: s_mov_b32 s2, s4
; GFX11-NEXT: s_mov_b32 s19, s5
; GFX11-NEXT: v_dual_mov_b32 v1, s2 :: v_dual_mov_b32 v2, s19
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX11-NEXT: s_mov_b32 s1, s3
; GFX11-NEXT: s_mov_b32 s4, s6
; GFX11-NEXT: s_mov_b32 s5, s7
; GFX11-NEXT: s_mov_b32 s6, s8
; GFX11-NEXT: v_cndmask_b32_e32 v1, s0, v1, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e32 v2, s1, v2, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX11-NEXT: s_mov_b32 s7, s9
; GFX11-NEXT: s_mov_b32 s8, s10
; GFX11-NEXT: s_mov_b32 s9, s11
; GFX11-NEXT: s_mov_b32 s10, s12
; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s5, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX11-NEXT: s_mov_b32 s11, s13
; GFX11-NEXT: s_mov_b32 s12, s14
; GFX11-NEXT: s_mov_b32 s13, s15
; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s6, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s7, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s9, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s10, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s11, vcc_lo
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0
; GFX11-NEXT: v_cndmask_b32_e64 v0, v1, s12, vcc_lo
; GFX11-NEXT: v_cndmask_b32_e64 v1, v2, s13, vcc_lo
; GFX11-NEXT: v_readfirstlane_b32 s0, v0
; GFX11-NEXT: v_readfirstlane_b32 s1, v1
; GFX11-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <7 x double> %vec, i32 %sel
ret double %ext
}
define double @dyn_extract_v7f64_v_v(<7 x double> %vec, i32 %sel) {
; GCN-LABEL: dyn_extract_v7f64_v_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v14
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v14
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v14
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v14
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v14
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 6, v14
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: dyn_extract_v7f64_v_v:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v14
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v14
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v14
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v14
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v14
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc_lo
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v14
; GFX10-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc_lo
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc_lo
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: dyn_extract_v7f64_v_v:
; GFX11: ; %bb.0: ; %entry
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v14
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v2 :: v_dual_cndmask_b32 v1, v1, v3
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v14
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v4 :: v_dual_cndmask_b32 v1, v1, v5
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v14
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v6 :: v_dual_cndmask_b32 v1, v1, v7
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v14
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v8 :: v_dual_cndmask_b32 v1, v1, v9
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v14
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v10 :: v_dual_cndmask_b32 v1, v1, v11
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v14
; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v12 :: v_dual_cndmask_b32 v1, v1, v13
; GFX11-NEXT: s_setpc_b64 s[30:31]
entry:
%ext = extractelement <7 x double> %vec, i32 %sel
ret double %ext
}
define amdgpu_ps double @dyn_extract_v7f64_v_s(<7 x double> %vec, i32 inreg %sel) {
; GPRIDX-LABEL: dyn_extract_v7f64_v_s:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_lshl_b32 s0, s2, 1
; GPRIDX-NEXT: s_set_gpr_idx_on s0, gpr_idx(SRC0)
; GPRIDX-NEXT: v_mov_b32_e32 v14, v0
; GPRIDX-NEXT: v_mov_b32_e32 v0, v1
; GPRIDX-NEXT: s_set_gpr_idx_off
; GPRIDX-NEXT: v_readfirstlane_b32 s0, v14
; GPRIDX-NEXT: v_readfirstlane_b32 s1, v0
; GPRIDX-NEXT: ; return to shader part epilog
;
; MOVREL-LABEL: dyn_extract_v7f64_v_s:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_lshl_b32 m0, s2, 1
; MOVREL-NEXT: v_movrels_b32_e32 v14, v0
; MOVREL-NEXT: v_movrels_b32_e32 v0, v1
; MOVREL-NEXT: v_readfirstlane_b32 s0, v14
; MOVREL-NEXT: v_readfirstlane_b32 s1, v0
; MOVREL-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v7f64_v_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_lshl_b32 m0, s2, 1
; GFX10PLUS-NEXT: v_movrels_b32_e32 v14, v0
; GFX10PLUS-NEXT: v_movrels_b32_e32 v0, v1
; GFX10PLUS-NEXT: v_readfirstlane_b32 s0, v14
; GFX10PLUS-NEXT: v_readfirstlane_b32 s1, v0
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <7 x double> %vec, i32 %sel
ret double %ext
}
define amdgpu_ps double @dyn_extract_v7f64_s_s(<7 x double> inreg %vec, i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v7f64_s_s:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 s1, s3
; GCN-NEXT: s_mov_b32 m0, s16
; GCN-NEXT: s_mov_b32 s2, s4
; GCN-NEXT: s_mov_b32 s3, s5
; GCN-NEXT: s_mov_b32 s4, s6
; GCN-NEXT: s_mov_b32 s5, s7
; GCN-NEXT: s_mov_b32 s6, s8
; GCN-NEXT: s_mov_b32 s7, s9
; GCN-NEXT: s_mov_b32 s8, s10
; GCN-NEXT: s_mov_b32 s9, s11
; GCN-NEXT: s_mov_b32 s10, s12
; GCN-NEXT: s_mov_b32 s11, s13
; GCN-NEXT: s_mov_b32 s12, s14
; GCN-NEXT: s_mov_b32 s13, s15
; GCN-NEXT: s_movrels_b64 s[0:1], s[0:1]
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v7f64_s_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: s_mov_b32 m0, s16
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: s_mov_b32 s6, s8
; GFX10PLUS-NEXT: s_mov_b32 s7, s9
; GFX10PLUS-NEXT: s_mov_b32 s8, s10
; GFX10PLUS-NEXT: s_mov_b32 s9, s11
; GFX10PLUS-NEXT: s_mov_b32 s10, s12
; GFX10PLUS-NEXT: s_mov_b32 s11, s13
; GFX10PLUS-NEXT: s_mov_b32 s12, s14
; GFX10PLUS-NEXT: s_mov_b32 s13, s15
; GFX10PLUS-NEXT: s_movrels_b64 s[0:1], s[0:1]
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <7 x double> %vec, i32 %sel
ret double %ext
}
define amdgpu_kernel void @dyn_extract_v5f64_s_s(double addrspace(1)* %out, i32 %sel) {
; GPRIDX-LABEL: dyn_extract_v5f64_s_s:
; GPRIDX: .amd_kernel_code_t
; GPRIDX-NEXT: amd_code_version_major = 1
; GPRIDX-NEXT: amd_code_version_minor = 2
; GPRIDX-NEXT: amd_machine_kind = 1
; GPRIDX-NEXT: amd_machine_version_major = 9
; GPRIDX-NEXT: amd_machine_version_minor = 0
; GPRIDX-NEXT: amd_machine_version_stepping = 0
; GPRIDX-NEXT: kernel_code_entry_byte_offset = 256
; GPRIDX-NEXT: kernel_code_prefetch_byte_size = 0
; GPRIDX-NEXT: granulated_workitem_vgpr_count = 0
; GPRIDX-NEXT: granulated_wavefront_sgpr_count = 1
; GPRIDX-NEXT: priority = 0
; GPRIDX-NEXT: float_mode = 240
; GPRIDX-NEXT: priv = 0
; GPRIDX-NEXT: enable_dx10_clamp = 1
; GPRIDX-NEXT: debug_mode = 0
; GPRIDX-NEXT: enable_ieee_mode = 1
; GPRIDX-NEXT: enable_wgp_mode = 0
; GPRIDX-NEXT: enable_mem_ordered = 0
; GPRIDX-NEXT: enable_fwd_progress = 0
; GPRIDX-NEXT: enable_sgpr_private_segment_wave_byte_offset = 0
; GPRIDX-NEXT: user_sgpr_count = 6
; GPRIDX-NEXT: enable_trap_handler = 0
; GPRIDX-NEXT: enable_sgpr_workgroup_id_x = 1
; GPRIDX-NEXT: enable_sgpr_workgroup_id_y = 0
; GPRIDX-NEXT: enable_sgpr_workgroup_id_z = 0
; GPRIDX-NEXT: enable_sgpr_workgroup_info = 0
; GPRIDX-NEXT: enable_vgpr_workitem_id = 0
; GPRIDX-NEXT: enable_exception_msb = 0
; GPRIDX-NEXT: granulated_lds_size = 0
; GPRIDX-NEXT: enable_exception = 0
; GPRIDX-NEXT: enable_sgpr_private_segment_buffer = 1
; GPRIDX-NEXT: enable_sgpr_dispatch_ptr = 0
; GPRIDX-NEXT: enable_sgpr_queue_ptr = 0
; GPRIDX-NEXT: enable_sgpr_kernarg_segment_ptr = 1
; GPRIDX-NEXT: enable_sgpr_dispatch_id = 0
; GPRIDX-NEXT: enable_sgpr_flat_scratch_init = 0
; GPRIDX-NEXT: enable_sgpr_private_segment_size = 0
; GPRIDX-NEXT: enable_sgpr_grid_workgroup_count_x = 0
; GPRIDX-NEXT: enable_sgpr_grid_workgroup_count_y = 0
; GPRIDX-NEXT: enable_sgpr_grid_workgroup_count_z = 0
; GPRIDX-NEXT: enable_wavefront_size32 = 0
; GPRIDX-NEXT: enable_ordered_append_gds = 0
; GPRIDX-NEXT: private_element_size = 1
; GPRIDX-NEXT: is_ptr64 = 1
; GPRIDX-NEXT: is_dynamic_callstack = 0
; GPRIDX-NEXT: is_debug_enabled = 0
; GPRIDX-NEXT: is_xnack_enabled = 1
; GPRIDX-NEXT: workitem_private_segment_byte_size = 0
; GPRIDX-NEXT: workgroup_group_segment_byte_size = 0
; GPRIDX-NEXT: gds_segment_byte_size = 0
; GPRIDX-NEXT: kernarg_segment_byte_size = 12
; GPRIDX-NEXT: workgroup_fbarrier_count = 0
; GPRIDX-NEXT: wavefront_sgpr_count = 9
; GPRIDX-NEXT: workitem_vgpr_count = 3
; GPRIDX-NEXT: reserved_vgpr_first = 0
; GPRIDX-NEXT: reserved_vgpr_count = 0
; GPRIDX-NEXT: reserved_sgpr_first = 0
; GPRIDX-NEXT: reserved_sgpr_count = 0
; GPRIDX-NEXT: debug_wavefront_private_segment_offset_sgpr = 0
; GPRIDX-NEXT: debug_private_segment_buffer_sgpr = 0
; GPRIDX-NEXT: kernarg_segment_alignment = 4
; GPRIDX-NEXT: group_segment_alignment = 4
; GPRIDX-NEXT: private_segment_alignment = 4
; GPRIDX-NEXT: wavefront_size = 6
; GPRIDX-NEXT: call_convention = -1
; GPRIDX-NEXT: runtime_loader_kernel_symbol = 0
; GPRIDX-NEXT: .end_amd_kernel_code_t
; GPRIDX-NEXT: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
; GPRIDX-NEXT: s_load_dword s8, s[4:5], 0x8
; GPRIDX-NEXT: s_mov_b32 s2, 0
; GPRIDX-NEXT: s_mov_b32 s3, 0x40140000
; GPRIDX-NEXT: s_mov_b32 s5, 0x40080000
; GPRIDX-NEXT: s_mov_b32 s4, s2
; GPRIDX-NEXT: s_waitcnt lgkmcnt(0)
; GPRIDX-NEXT: s_cmp_eq_u32 s8, 1
; GPRIDX-NEXT: s_cselect_b64 s[6:7], 2.0, 1.0
; GPRIDX-NEXT: s_cmp_eq_u32 s8, 2
; GPRIDX-NEXT: s_cselect_b64 s[4:5], s[4:5], s[6:7]
; GPRIDX-NEXT: s_cmp_eq_u32 s8, 3
; GPRIDX-NEXT: s_cselect_b64 s[4:5], 4.0, s[4:5]
; GPRIDX-NEXT: s_cmp_eq_u32 s8, 4
; GPRIDX-NEXT: s_cselect_b64 s[2:3], s[2:3], s[4:5]
; GPRIDX-NEXT: v_mov_b32_e32 v0, s2
; GPRIDX-NEXT: v_mov_b32_e32 v1, s3
; GPRIDX-NEXT: v_mov_b32_e32 v2, 0
; GPRIDX-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1]
; GPRIDX-NEXT: s_endpgm
;
; MOVREL-LABEL: dyn_extract_v5f64_s_s:
; MOVREL: .amd_kernel_code_t
; MOVREL-NEXT: amd_code_version_major = 1
; MOVREL-NEXT: amd_code_version_minor = 2
; MOVREL-NEXT: amd_machine_kind = 1
; MOVREL-NEXT: amd_machine_version_major = 8
; MOVREL-NEXT: amd_machine_version_minor = 0
; MOVREL-NEXT: amd_machine_version_stepping = 3
; MOVREL-NEXT: kernel_code_entry_byte_offset = 256
; MOVREL-NEXT: kernel_code_prefetch_byte_size = 0
; MOVREL-NEXT: granulated_workitem_vgpr_count = 0
; MOVREL-NEXT: granulated_wavefront_sgpr_count = 1
; MOVREL-NEXT: priority = 0
; MOVREL-NEXT: float_mode = 240
; MOVREL-NEXT: priv = 0
; MOVREL-NEXT: enable_dx10_clamp = 1
; MOVREL-NEXT: debug_mode = 0
; MOVREL-NEXT: enable_ieee_mode = 1
; MOVREL-NEXT: enable_wgp_mode = 0
; MOVREL-NEXT: enable_mem_ordered = 0
; MOVREL-NEXT: enable_fwd_progress = 0
; MOVREL-NEXT: enable_sgpr_private_segment_wave_byte_offset = 0
; MOVREL-NEXT: user_sgpr_count = 6
; MOVREL-NEXT: enable_trap_handler = 0
; MOVREL-NEXT: enable_sgpr_workgroup_id_x = 1
; MOVREL-NEXT: enable_sgpr_workgroup_id_y = 0
; MOVREL-NEXT: enable_sgpr_workgroup_id_z = 0
; MOVREL-NEXT: enable_sgpr_workgroup_info = 0
; MOVREL-NEXT: enable_vgpr_workitem_id = 0
; MOVREL-NEXT: enable_exception_msb = 0
; MOVREL-NEXT: granulated_lds_size = 0
; MOVREL-NEXT: enable_exception = 0
; MOVREL-NEXT: enable_sgpr_private_segment_buffer = 1
; MOVREL-NEXT: enable_sgpr_dispatch_ptr = 0
; MOVREL-NEXT: enable_sgpr_queue_ptr = 0
; MOVREL-NEXT: enable_sgpr_kernarg_segment_ptr = 1
; MOVREL-NEXT: enable_sgpr_dispatch_id = 0
; MOVREL-NEXT: enable_sgpr_flat_scratch_init = 0
; MOVREL-NEXT: enable_sgpr_private_segment_size = 0
; MOVREL-NEXT: enable_sgpr_grid_workgroup_count_x = 0
; MOVREL-NEXT: enable_sgpr_grid_workgroup_count_y = 0
; MOVREL-NEXT: enable_sgpr_grid_workgroup_count_z = 0
; MOVREL-NEXT: enable_wavefront_size32 = 0
; MOVREL-NEXT: enable_ordered_append_gds = 0
; MOVREL-NEXT: private_element_size = 1
; MOVREL-NEXT: is_ptr64 = 1
; MOVREL-NEXT: is_dynamic_callstack = 0
; MOVREL-NEXT: is_debug_enabled = 0
; MOVREL-NEXT: is_xnack_enabled = 0
; MOVREL-NEXT: workitem_private_segment_byte_size = 0
; MOVREL-NEXT: workgroup_group_segment_byte_size = 0
; MOVREL-NEXT: gds_segment_byte_size = 0
; MOVREL-NEXT: kernarg_segment_byte_size = 12
; MOVREL-NEXT: workgroup_fbarrier_count = 0
; MOVREL-NEXT: wavefront_sgpr_count = 9
; MOVREL-NEXT: workitem_vgpr_count = 4
; MOVREL-NEXT: reserved_vgpr_first = 0
; MOVREL-NEXT: reserved_vgpr_count = 0
; MOVREL-NEXT: reserved_sgpr_first = 0
; MOVREL-NEXT: reserved_sgpr_count = 0
; MOVREL-NEXT: debug_wavefront_private_segment_offset_sgpr = 0
; MOVREL-NEXT: debug_private_segment_buffer_sgpr = 0
; MOVREL-NEXT: kernarg_segment_alignment = 4
; MOVREL-NEXT: group_segment_alignment = 4
; MOVREL-NEXT: private_segment_alignment = 4
; MOVREL-NEXT: wavefront_size = 6
; MOVREL-NEXT: call_convention = -1
; MOVREL-NEXT: runtime_loader_kernel_symbol = 0
; MOVREL-NEXT: .end_amd_kernel_code_t
; MOVREL-NEXT: ; %bb.0: ; %entry
; MOVREL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
; MOVREL-NEXT: s_load_dword s8, s[4:5], 0x8
; MOVREL-NEXT: s_mov_b32 s2, 0
; MOVREL-NEXT: s_mov_b32 s3, 0x40140000
; MOVREL-NEXT: s_mov_b32 s5, 0x40080000
; MOVREL-NEXT: s_mov_b32 s4, s2
; MOVREL-NEXT: s_waitcnt lgkmcnt(0)
; MOVREL-NEXT: s_cmp_eq_u32 s8, 1
; MOVREL-NEXT: s_cselect_b64 s[6:7], 2.0, 1.0
; MOVREL-NEXT: s_cmp_eq_u32 s8, 2
; MOVREL-NEXT: s_cselect_b64 s[4:5], s[4:5], s[6:7]
; MOVREL-NEXT: s_cmp_eq_u32 s8, 3
; MOVREL-NEXT: s_cselect_b64 s[4:5], 4.0, s[4:5]
; MOVREL-NEXT: s_cmp_eq_u32 s8, 4
; MOVREL-NEXT: s_cselect_b64 s[2:3], s[2:3], s[4:5]
; MOVREL-NEXT: v_mov_b32_e32 v0, s2
; MOVREL-NEXT: v_mov_b32_e32 v3, s1
; MOVREL-NEXT: v_mov_b32_e32 v1, s3
; MOVREL-NEXT: v_mov_b32_e32 v2, s0
; MOVREL-NEXT: flat_store_dwordx2 v[2:3], v[0:1]
; MOVREL-NEXT: s_endpgm
;
; GFX10-LABEL: dyn_extract_v5f64_s_s:
; GFX10: .amd_kernel_code_t
; GFX10-NEXT: amd_code_version_major = 1
; GFX10-NEXT: amd_code_version_minor = 2
; GFX10-NEXT: amd_machine_kind = 1
; GFX10-NEXT: amd_machine_version_major = 10
; GFX10-NEXT: amd_machine_version_minor = 1
; GFX10-NEXT: amd_machine_version_stepping = 0
; GFX10-NEXT: kernel_code_entry_byte_offset = 256
; GFX10-NEXT: kernel_code_prefetch_byte_size = 0
; GFX10-NEXT: granulated_workitem_vgpr_count = 0
; GFX10-NEXT: granulated_wavefront_sgpr_count = 1
; GFX10-NEXT: priority = 0
; GFX10-NEXT: float_mode = 240
; GFX10-NEXT: priv = 0
; GFX10-NEXT: enable_dx10_clamp = 1
; GFX10-NEXT: debug_mode = 0
; GFX10-NEXT: enable_ieee_mode = 1
; GFX10-NEXT: enable_wgp_mode = 1
; GFX10-NEXT: enable_mem_ordered = 1
; GFX10-NEXT: enable_fwd_progress = 0
; GFX10-NEXT: enable_sgpr_private_segment_wave_byte_offset = 0
; GFX10-NEXT: user_sgpr_count = 6
; GFX10-NEXT: enable_trap_handler = 0
; GFX10-NEXT: enable_sgpr_workgroup_id_x = 1
; GFX10-NEXT: enable_sgpr_workgroup_id_y = 0
; GFX10-NEXT: enable_sgpr_workgroup_id_z = 0
; GFX10-NEXT: enable_sgpr_workgroup_info = 0
; GFX10-NEXT: enable_vgpr_workitem_id = 0
; GFX10-NEXT: enable_exception_msb = 0
; GFX10-NEXT: granulated_lds_size = 0
; GFX10-NEXT: enable_exception = 0
; GFX10-NEXT: enable_sgpr_private_segment_buffer = 1
; GFX10-NEXT: enable_sgpr_dispatch_ptr = 0
; GFX10-NEXT: enable_sgpr_queue_ptr = 0
; GFX10-NEXT: enable_sgpr_kernarg_segment_ptr = 1
; GFX10-NEXT: enable_sgpr_dispatch_id = 0
; GFX10-NEXT: enable_sgpr_flat_scratch_init = 0
; GFX10-NEXT: enable_sgpr_private_segment_size = 0
; GFX10-NEXT: enable_sgpr_grid_workgroup_count_x = 0
; GFX10-NEXT: enable_sgpr_grid_workgroup_count_y = 0
; GFX10-NEXT: enable_sgpr_grid_workgroup_count_z = 0
; GFX10-NEXT: enable_wavefront_size32 = 1
; GFX10-NEXT: enable_ordered_append_gds = 0
; GFX10-NEXT: private_element_size = 1
; GFX10-NEXT: is_ptr64 = 1
; GFX10-NEXT: is_dynamic_callstack = 0
; GFX10-NEXT: is_debug_enabled = 0
; GFX10-NEXT: is_xnack_enabled = 1
; GFX10-NEXT: workitem_private_segment_byte_size = 0
; GFX10-NEXT: workgroup_group_segment_byte_size = 0
; GFX10-NEXT: gds_segment_byte_size = 0
; GFX10-NEXT: kernarg_segment_byte_size = 12
; GFX10-NEXT: workgroup_fbarrier_count = 0
; GFX10-NEXT: wavefront_sgpr_count = 9
; GFX10-NEXT: workitem_vgpr_count = 3
; GFX10-NEXT: reserved_vgpr_first = 0
; GFX10-NEXT: reserved_vgpr_count = 0
; GFX10-NEXT: reserved_sgpr_first = 0
; GFX10-NEXT: reserved_sgpr_count = 0
; GFX10-NEXT: debug_wavefront_private_segment_offset_sgpr = 0
; GFX10-NEXT: debug_private_segment_buffer_sgpr = 0
; GFX10-NEXT: kernarg_segment_alignment = 4
; GFX10-NEXT: group_segment_alignment = 4
; GFX10-NEXT: private_segment_alignment = 4
; GFX10-NEXT: wavefront_size = 5
; GFX10-NEXT: call_convention = -1
; GFX10-NEXT: runtime_loader_kernel_symbol = 0
; GFX10-NEXT: .end_amd_kernel_code_t
; GFX10-NEXT: ; %bb.0: ; %entry
; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: s_load_dword s8, s[4:5], 0x8
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
; GFX10-NEXT: s_mov_b32 s2, 0
; GFX10-NEXT: s_mov_b32 s3, 0x40140000
; GFX10-NEXT: s_mov_b32 s5, 0x40080000
; GFX10-NEXT: s_mov_b32 s4, s2
; GFX10-NEXT: v_mov_b32_e32 v2, 0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_cmp_eq_u32 s8, 1
; GFX10-NEXT: s_cselect_b64 s[6:7], 2.0, 1.0
; GFX10-NEXT: s_cmp_eq_u32 s8, 2
; GFX10-NEXT: s_cselect_b64 s[4:5], s[4:5], s[6:7]
; GFX10-NEXT: s_cmp_eq_u32 s8, 3
; GFX10-NEXT: s_cselect_b64 s[4:5], 4.0, s[4:5]
; GFX10-NEXT: s_cmp_eq_u32 s8, 4
; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[4:5]
; GFX10-NEXT: v_mov_b32_e32 v0, s2
; GFX10-NEXT: v_mov_b32_e32 v1, s3
; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1]
; GFX10-NEXT: s_endpgm
;
; GFX11-LABEL: dyn_extract_v5f64_s_s:
; GFX11: .amd_kernel_code_t
; GFX11-NEXT: amd_code_version_major = 1
; GFX11-NEXT: amd_code_version_minor = 2
; GFX11-NEXT: amd_machine_kind = 1
; GFX11-NEXT: amd_machine_version_major = 11
; GFX11-NEXT: amd_machine_version_minor = 0
; GFX11-NEXT: amd_machine_version_stepping = 0
; GFX11-NEXT: kernel_code_entry_byte_offset = 256
; GFX11-NEXT: kernel_code_prefetch_byte_size = 0
; GFX11-NEXT: granulated_workitem_vgpr_count = 0
; GFX11-NEXT: granulated_wavefront_sgpr_count = 1
; GFX11-NEXT: priority = 0
; GFX11-NEXT: float_mode = 240
; GFX11-NEXT: priv = 0
; GFX11-NEXT: enable_dx10_clamp = 1
; GFX11-NEXT: debug_mode = 0
; GFX11-NEXT: enable_ieee_mode = 1
; GFX11-NEXT: enable_wgp_mode = 1
; GFX11-NEXT: enable_mem_ordered = 1
; GFX11-NEXT: enable_fwd_progress = 0
; GFX11-NEXT: enable_sgpr_private_segment_wave_byte_offset = 0
; GFX11-NEXT: user_sgpr_count = 15
; GFX11-NEXT: enable_trap_handler = 0
; GFX11-NEXT: enable_sgpr_workgroup_id_x = 1
; GFX11-NEXT: enable_sgpr_workgroup_id_y = 0
; GFX11-NEXT: enable_sgpr_workgroup_id_z = 0
; GFX11-NEXT: enable_sgpr_workgroup_info = 0
; GFX11-NEXT: enable_vgpr_workitem_id = 0
; GFX11-NEXT: enable_exception_msb = 0
; GFX11-NEXT: granulated_lds_size = 0
; GFX11-NEXT: enable_exception = 0
; GFX11-NEXT: enable_sgpr_private_segment_buffer = 0
; GFX11-NEXT: enable_sgpr_dispatch_ptr = 0
; GFX11-NEXT: enable_sgpr_queue_ptr = 0
; GFX11-NEXT: enable_sgpr_kernarg_segment_ptr = 1
; GFX11-NEXT: enable_sgpr_dispatch_id = 0
; GFX11-NEXT: enable_sgpr_flat_scratch_init = 0
; GFX11-NEXT: enable_sgpr_private_segment_size = 0
; GFX11-NEXT: enable_sgpr_grid_workgroup_count_x = 0
; GFX11-NEXT: enable_sgpr_grid_workgroup_count_y = 0
; GFX11-NEXT: enable_sgpr_grid_workgroup_count_z = 0
; GFX11-NEXT: enable_wavefront_size32 = 1
; GFX11-NEXT: enable_ordered_append_gds = 0
; GFX11-NEXT: private_element_size = 1
; GFX11-NEXT: is_ptr64 = 1
; GFX11-NEXT: is_dynamic_callstack = 0
; GFX11-NEXT: is_debug_enabled = 0
; GFX11-NEXT: is_xnack_enabled = 0
; GFX11-NEXT: workitem_private_segment_byte_size = 0
; GFX11-NEXT: workgroup_group_segment_byte_size = 0
; GFX11-NEXT: gds_segment_byte_size = 0
; GFX11-NEXT: kernarg_segment_byte_size = 12
; GFX11-NEXT: workgroup_fbarrier_count = 0
; GFX11-NEXT: wavefront_sgpr_count = 9
; GFX11-NEXT: workitem_vgpr_count = 3
; GFX11-NEXT: reserved_vgpr_first = 0
; GFX11-NEXT: reserved_vgpr_count = 0
; GFX11-NEXT: reserved_sgpr_first = 0
; GFX11-NEXT: reserved_sgpr_count = 0
; GFX11-NEXT: debug_wavefront_private_segment_offset_sgpr = 0
; GFX11-NEXT: debug_private_segment_buffer_sgpr = 0
; GFX11-NEXT: kernarg_segment_alignment = 4
; GFX11-NEXT: group_segment_alignment = 4
; GFX11-NEXT: private_segment_alignment = 4
; GFX11-NEXT: wavefront_size = 5
; GFX11-NEXT: call_convention = -1
; GFX11-NEXT: runtime_loader_kernel_symbol = 0
; GFX11-NEXT: .end_amd_kernel_code_t
; GFX11-NEXT: ; %bb.0: ; %entry
; GFX11-NEXT: s_clause 0x1
; GFX11-NEXT: s_load_b32 s8, s[0:1], 0x8
; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x0
; GFX11-NEXT: s_mov_b32 s2, 0
; GFX11-NEXT: s_mov_b32 s3, 0x40140000
; GFX11-NEXT: s_mov_b32 s5, 0x40080000
; GFX11-NEXT: s_mov_b32 s4, s2
; GFX11-NEXT: v_mov_b32_e32 v2, 0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_cmp_eq_u32 s8, 1
; GFX11-NEXT: s_cselect_b64 s[6:7], 2.0, 1.0
; GFX11-NEXT: s_cmp_eq_u32 s8, 2
; GFX11-NEXT: s_cselect_b64 s[4:5], s[4:5], s[6:7]
; GFX11-NEXT: s_cmp_eq_u32 s8, 3
; GFX11-NEXT: s_cselect_b64 s[4:5], 4.0, s[4:5]
; GFX11-NEXT: s_cmp_eq_u32 s8, 4
; GFX11-NEXT: s_cselect_b64 s[2:3], s[2:3], s[4:5]
; GFX11-NEXT: v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX11-NEXT: s_endpgm
entry:
%ext = extractelement <5 x double> <double 1.0, double 2.0, double 3.0, double 4.0, double 5.0>, i32 %sel
store double %ext, double addrspace(1)* %out
ret void
}
define float @dyn_extract_v15f32_const_s_v(i32 %sel) {
; GCN-LABEL: dyn_extract_v15f32_const_s_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; GCN-NEXT: v_mov_b32_e32 v1, 0x40400000
; GCN-NEXT: v_cndmask_b32_e64 v13, 1.0, 2.0, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v0
; GCN-NEXT: v_cndmask_b32_e32 v1, v13, v1, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v0
; GCN-NEXT: v_mov_b32_e32 v2, 0x40a00000
; GCN-NEXT: v_cndmask_b32_e64 v1, v1, 4.0, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v0
; GCN-NEXT: v_mov_b32_e32 v3, 0x40c00000
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v0
; GCN-NEXT: v_mov_b32_e32 v4, 0x40e00000
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 6, v0
; GCN-NEXT: v_mov_b32_e32 v5, 0x41000000
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v4, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 7, v0
; GCN-NEXT: v_mov_b32_e32 v6, 0x41100000
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 8, v0
; GCN-NEXT: v_mov_b32_e32 v7, 0x41200000
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v6, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 9, v0
; GCN-NEXT: v_mov_b32_e32 v8, 0x41300000
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 10, v0
; GCN-NEXT: v_mov_b32_e32 v9, 0x41400000
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v8, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 11, v0
; GCN-NEXT: v_mov_b32_e32 v10, 0x41500000
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 12, v0
; GCN-NEXT: v_mov_b32_e32 v11, 0x41600000
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v10, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 13, v0
; GCN-NEXT: v_mov_b32_e32 v12, 0x41700000
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 14, v0
; GCN-NEXT: v_cndmask_b32_e32 v0, v1, v12, vcc
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: dyn_extract_v15f32_const_s_v:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, 1.0, 2.0, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 0x40400000, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 4.0, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 0x40a00000, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 0x40c00000, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 0x40e00000, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 0x41000000, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 8, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 0x41100000, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 9, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 0x41200000, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 10, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 0x41300000, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 11, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 0x41400000, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 12, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 0x41500000, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 13, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 0x41600000, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 14, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v0, v1, 0x41700000, vcc_lo
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
entry:
%ext = extractelement <15 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0, float 9.0, float 10.0, float 11.0, float 12.0, float 13.0, float 14.0, float 15.0>, i32 %sel
ret float %ext
}
define amdgpu_ps float @dyn_extract_v15f32_const_s_s(i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v15f32_const_s_s:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s4, 1.0
; GCN-NEXT: s_mov_b32 m0, s2
; GCN-NEXT: s_mov_b32 s18, 0x41700000
; GCN-NEXT: s_mov_b32 s17, 0x41600000
; GCN-NEXT: s_mov_b32 s16, 0x41500000
; GCN-NEXT: s_mov_b32 s15, 0x41400000
; GCN-NEXT: s_mov_b32 s14, 0x41300000
; GCN-NEXT: s_mov_b32 s13, 0x41200000
; GCN-NEXT: s_mov_b32 s12, 0x41100000
; GCN-NEXT: s_mov_b32 s11, 0x41000000
; GCN-NEXT: s_mov_b32 s10, 0x40e00000
; GCN-NEXT: s_mov_b32 s9, 0x40c00000
; GCN-NEXT: s_mov_b32 s8, 0x40a00000
; GCN-NEXT: s_mov_b32 s7, 4.0
; GCN-NEXT: s_mov_b32 s6, 0x40400000
; GCN-NEXT: s_mov_b32 s5, 2.0
; GCN-NEXT: s_movrels_b32 s0, s4
; GCN-NEXT: v_mov_b32_e32 v0, s0
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v15f32_const_s_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s4, 1.0
; GFX10PLUS-NEXT: s_mov_b32 m0, s2
; GFX10PLUS-NEXT: s_mov_b32 s18, 0x41700000
; GFX10PLUS-NEXT: s_mov_b32 s17, 0x41600000
; GFX10PLUS-NEXT: s_mov_b32 s16, 0x41500000
; GFX10PLUS-NEXT: s_mov_b32 s15, 0x41400000
; GFX10PLUS-NEXT: s_mov_b32 s14, 0x41300000
; GFX10PLUS-NEXT: s_mov_b32 s13, 0x41200000
; GFX10PLUS-NEXT: s_mov_b32 s12, 0x41100000
; GFX10PLUS-NEXT: s_mov_b32 s11, 0x41000000
; GFX10PLUS-NEXT: s_mov_b32 s10, 0x40e00000
; GFX10PLUS-NEXT: s_mov_b32 s9, 0x40c00000
; GFX10PLUS-NEXT: s_mov_b32 s8, 0x40a00000
; GFX10PLUS-NEXT: s_mov_b32 s7, 4.0
; GFX10PLUS-NEXT: s_mov_b32 s6, 0x40400000
; GFX10PLUS-NEXT: s_mov_b32 s5, 2.0
; GFX10PLUS-NEXT: s_movrels_b32 s0, s4
; GFX10PLUS-NEXT: v_mov_b32_e32 v0, s0
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <15 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0, float 9.0, float 10.0, float 11.0, float 12.0, float 13.0, float 14.0, float 15.0>, i32 %sel
ret float %ext
}
define amdgpu_ps float @dyn_extract_v15f32_s_v(<15 x float> inreg %vec, i32 %sel) {
; GCN-LABEL: dyn_extract_v15f32_s_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 s1, s3
; GCN-NEXT: s_mov_b32 s2, s4
; GCN-NEXT: v_mov_b32_e32 v1, s0
; GCN-NEXT: v_mov_b32_e32 v2, s1
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; GCN-NEXT: s_mov_b32 s3, s5
; GCN-NEXT: v_mov_b32_e32 v3, s2
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v0
; GCN-NEXT: s_mov_b32 s4, s6
; GCN-NEXT: v_mov_b32_e32 v4, s3
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v0
; GCN-NEXT: s_mov_b32 s5, s7
; GCN-NEXT: v_mov_b32_e32 v5, s4
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v4, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v0
; GCN-NEXT: s_mov_b32 s6, s8
; GCN-NEXT: v_mov_b32_e32 v6, s5
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v0
; GCN-NEXT: s_mov_b32 s7, s9
; GCN-NEXT: v_mov_b32_e32 v7, s6
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v6, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 6, v0
; GCN-NEXT: s_mov_b32 s8, s10
; GCN-NEXT: v_mov_b32_e32 v8, s7
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 7, v0
; GCN-NEXT: s_mov_b32 s9, s11
; GCN-NEXT: v_mov_b32_e32 v9, s8
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v8, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 8, v0
; GCN-NEXT: s_mov_b32 s10, s12
; GCN-NEXT: v_mov_b32_e32 v10, s9
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 9, v0
; GCN-NEXT: v_mov_b32_e32 v11, s10
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v10, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 10, v0
; GCN-NEXT: v_mov_b32_e32 v12, s13
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v11, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 11, v0
; GCN-NEXT: v_mov_b32_e32 v13, s14
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v12, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 12, v0
; GCN-NEXT: v_mov_b32_e32 v14, s15
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v13, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 13, v0
; GCN-NEXT: v_mov_b32_e32 v15, s16
; GCN-NEXT: v_cndmask_b32_e32 v1, v1, v14, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 14, v0
; GCN-NEXT: v_cndmask_b32_e32 v0, v1, v15, vcc
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v15f32_s_v:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX10PLUS-NEXT: v_mov_b32_e32 v1, s1
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v1, s0, v1, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v0
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: s_mov_b32 s6, s8
; GFX10PLUS-NEXT: s_mov_b32 s7, s9
; GFX10PLUS-NEXT: s_mov_b32 s8, s10
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v0
; GFX10PLUS-NEXT: s_mov_b32 s9, s11
; GFX10PLUS-NEXT: s_mov_b32 s10, s12
; GFX10PLUS-NEXT: s_mov_b32 s11, s13
; GFX10PLUS-NEXT: s_mov_b32 s12, s14
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s3, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v0
; GFX10PLUS-NEXT: s_mov_b32 s13, s15
; GFX10PLUS-NEXT: s_mov_b32 s14, s16
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s5, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s6, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s7, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 8, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s8, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 9, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s9, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 10, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s10, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 11, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s11, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 12, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s12, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 13, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, s13, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 14, v0
; GFX10PLUS-NEXT: v_cndmask_b32_e64 v0, v1, s14, vcc_lo
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <15 x float> %vec, i32 %sel
ret float %ext
}
define float @dyn_extract_v15f32_v_v(<15 x float> %vec, i32 %sel) {
; GCN-LABEL: dyn_extract_v15f32_v_v:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v15
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 2, v15
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 3, v15
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 4, v15
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 5, v15
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 6, v15
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 7, v15
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 8, v15
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 9, v15
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v9, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 10, v15
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 11, v15
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v11, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 12, v15
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 13, v15
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v13, vcc
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 14, v15
; GCN-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: dyn_extract_v15f32_v_v:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 8, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 9, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v9, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 10, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 11, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v11, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 12, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 13, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v13, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 14, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
entry:
%ext = extractelement <15 x float> %vec, i32 %sel
ret float %ext
}
define amdgpu_ps float @dyn_extract_v15f32_v_s(<15 x float> %vec, i32 inreg %sel) {
; GPRIDX-LABEL: dyn_extract_v15f32_v_s:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_set_gpr_idx_on s2, gpr_idx(SRC0)
; GPRIDX-NEXT: v_mov_b32_e32 v0, v0
; GPRIDX-NEXT: s_set_gpr_idx_off
; GPRIDX-NEXT: ; return to shader part epilog
;
; MOVREL-LABEL: dyn_extract_v15f32_v_s:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_mov_b32 m0, s2
; MOVREL-NEXT: v_movrels_b32_e32 v0, v0
; MOVREL-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v15f32_v_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 m0, s2
; GFX10PLUS-NEXT: v_movrels_b32_e32 v0, v0
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <15 x float> %vec, i32 %sel
ret float %ext
}
define amdgpu_ps float @dyn_extract_v15f32_s_s(<15 x float> inreg %vec, i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v15f32_s_s:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 m0, s17
; GCN-NEXT: s_mov_b32 s1, s3
; GCN-NEXT: s_mov_b32 s2, s4
; GCN-NEXT: s_mov_b32 s3, s5
; GCN-NEXT: s_mov_b32 s4, s6
; GCN-NEXT: s_mov_b32 s5, s7
; GCN-NEXT: s_mov_b32 s6, s8
; GCN-NEXT: s_mov_b32 s7, s9
; GCN-NEXT: s_mov_b32 s8, s10
; GCN-NEXT: s_mov_b32 s9, s11
; GCN-NEXT: s_mov_b32 s10, s12
; GCN-NEXT: s_mov_b32 s11, s13
; GCN-NEXT: s_mov_b32 s12, s14
; GCN-NEXT: s_mov_b32 s13, s15
; GCN-NEXT: s_mov_b32 s14, s16
; GCN-NEXT: s_movrels_b32 s0, s0
; GCN-NEXT: v_mov_b32_e32 v0, s0
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v15f32_s_s:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 m0, s17
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: s_mov_b32 s6, s8
; GFX10PLUS-NEXT: s_mov_b32 s7, s9
; GFX10PLUS-NEXT: s_mov_b32 s8, s10
; GFX10PLUS-NEXT: s_mov_b32 s9, s11
; GFX10PLUS-NEXT: s_mov_b32 s10, s12
; GFX10PLUS-NEXT: s_mov_b32 s11, s13
; GFX10PLUS-NEXT: s_mov_b32 s12, s14
; GFX10PLUS-NEXT: s_mov_b32 s13, s15
; GFX10PLUS-NEXT: s_mov_b32 s14, s16
; GFX10PLUS-NEXT: s_movrels_b32 s0, s0
; GFX10PLUS-NEXT: v_mov_b32_e32 v0, s0
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%ext = extractelement <15 x float> %vec, i32 %sel
ret float %ext
}
define amdgpu_ps float @dyn_extract_v15f32_s_s_offset3(<15 x float> inreg %vec, i32 inreg %sel) {
; GCN-LABEL: dyn_extract_v15f32_s_s_offset3:
; GCN: ; %bb.0: ; %entry
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 s1, s3
; GCN-NEXT: s_mov_b32 s3, s5
; GCN-NEXT: s_mov_b32 m0, s17
; GCN-NEXT: s_mov_b32 s2, s4
; GCN-NEXT: s_mov_b32 s4, s6
; GCN-NEXT: s_mov_b32 s5, s7
; GCN-NEXT: s_mov_b32 s6, s8
; GCN-NEXT: s_mov_b32 s7, s9
; GCN-NEXT: s_mov_b32 s8, s10
; GCN-NEXT: s_mov_b32 s9, s11
; GCN-NEXT: s_mov_b32 s10, s12
; GCN-NEXT: s_mov_b32 s11, s13
; GCN-NEXT: s_mov_b32 s12, s14
; GCN-NEXT: s_mov_b32 s13, s15
; GCN-NEXT: s_mov_b32 s14, s16
; GCN-NEXT: s_movrels_b32 s0, s3
; GCN-NEXT: v_mov_b32_e32 v0, s0
; GCN-NEXT: ; return to shader part epilog
;
; GFX10PLUS-LABEL: dyn_extract_v15f32_s_s_offset3:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_mov_b32 s1, s3
; GFX10PLUS-NEXT: s_mov_b32 s3, s5
; GFX10PLUS-NEXT: s_mov_b32 m0, s17
; GFX10PLUS-NEXT: s_mov_b32 s0, s2
; GFX10PLUS-NEXT: s_mov_b32 s2, s4
; GFX10PLUS-NEXT: s_mov_b32 s4, s6
; GFX10PLUS-NEXT: s_mov_b32 s5, s7
; GFX10PLUS-NEXT: s_mov_b32 s6, s8
; GFX10PLUS-NEXT: s_mov_b32 s7, s9
; GFX10PLUS-NEXT: s_mov_b32 s8, s10
; GFX10PLUS-NEXT: s_mov_b32 s9, s11
; GFX10PLUS-NEXT: s_mov_b32 s10, s12
; GFX10PLUS-NEXT: s_mov_b32 s11, s13
; GFX10PLUS-NEXT: s_mov_b32 s12, s14
; GFX10PLUS-NEXT: s_mov_b32 s13, s15
; GFX10PLUS-NEXT: s_mov_b32 s14, s16
; GFX10PLUS-NEXT: s_movrels_b32 s0, s3
; GFX10PLUS-NEXT: v_mov_b32_e32 v0, s0
; GFX10PLUS-NEXT: ; return to shader part epilog
entry:
%add = add i32 %sel, 3
%ext = extractelement <15 x float> %vec, i32 %add
ret float %ext
}
define float @dyn_extract_v15f32_v_v_offset3(<15 x float> %vec, i32 %sel) {
; GPRIDX-LABEL: dyn_extract_v15f32_v_v_offset3:
; GPRIDX: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GPRIDX-NEXT: v_add_u32_e32 v15, 3, v15
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 1, v15
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 2, v15
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 3, v15
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 4, v15
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 5, v15
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 6, v15
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 7, v15
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 8, v15
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 9, v15
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v9, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 10, v15
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 11, v15
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v11, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 12, v15
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 13, v15
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v13, vcc
; GPRIDX-NEXT: v_cmp_eq_u32_e32 vcc, 14, v15
; GPRIDX-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc
; GPRIDX-NEXT: s_setpc_b64 s[30:31]
;
; MOVREL-LABEL: dyn_extract_v15f32_v_v_offset3:
; MOVREL: ; %bb.0: ; %entry
; MOVREL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; MOVREL-NEXT: v_add_u32_e32 v15, vcc, 3, v15
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 1, v15
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 2, v15
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 3, v15
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 4, v15
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 5, v15
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 6, v15
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 7, v15
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 8, v15
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 9, v15
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v9, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 10, v15
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 11, v15
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v11, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 12, v15
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 13, v15
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v13, vcc
; MOVREL-NEXT: v_cmp_eq_u32_e32 vcc, 14, v15
; MOVREL-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc
; MOVREL-NEXT: s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: dyn_extract_v15f32_v_v_offset3:
; GFX10PLUS: ; %bb.0: ; %entry
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10PLUS-NEXT: v_add_nc_u32_e32 v15, 3, v15
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 2, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 3, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v3, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 4, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v4, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 5, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v5, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 6, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v6, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 7, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v7, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 8, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v8, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 9, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v9, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 10, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v10, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 11, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v11, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 12, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v12, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 13, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v13, vcc_lo
; GFX10PLUS-NEXT: v_cmp_eq_u32_e32 vcc_lo, 14, v15
; GFX10PLUS-NEXT: v_cndmask_b32_e32 v0, v0, v14, vcc_lo
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
entry:
%add = add i32 %sel, 3
%ext = extractelement <15 x float> %vec, i32 %add
ret float %ext
}
define amdgpu_kernel void @dyn_extract_v4f32_s_s_s(float addrspace(1)* %out, i32 %sel) {
; GPRIDX-LABEL: dyn_extract_v4f32_s_s_s:
; GPRIDX: .amd_kernel_code_t
; GPRIDX-NEXT: amd_code_version_major = 1
; GPRIDX-NEXT: amd_code_version_minor = 2
; GPRIDX-NEXT: amd_machine_kind = 1
; GPRIDX-NEXT: amd_machine_version_major = 9
; GPRIDX-NEXT: amd_machine_version_minor = 0
; GPRIDX-NEXT: amd_machine_version_stepping = 0
; GPRIDX-NEXT: kernel_code_entry_byte_offset = 256
; GPRIDX-NEXT: kernel_code_prefetch_byte_size = 0
; GPRIDX-NEXT: granulated_workitem_vgpr_count = 0
; GPRIDX-NEXT: granulated_wavefront_sgpr_count = 0
; GPRIDX-NEXT: priority = 0
; GPRIDX-NEXT: float_mode = 240
; GPRIDX-NEXT: priv = 0
; GPRIDX-NEXT: enable_dx10_clamp = 1
; GPRIDX-NEXT: debug_mode = 0
; GPRIDX-NEXT: enable_ieee_mode = 1
; GPRIDX-NEXT: enable_wgp_mode = 0
; GPRIDX-NEXT: enable_mem_ordered = 0
; GPRIDX-NEXT: enable_fwd_progress = 0
; GPRIDX-NEXT: enable_sgpr_private_segment_wave_byte_offset = 0
; GPRIDX-NEXT: user_sgpr_count = 6
; GPRIDX-NEXT: enable_trap_handler = 0
; GPRIDX-NEXT: enable_sgpr_workgroup_id_x = 1
; GPRIDX-NEXT: enable_sgpr_workgroup_id_y = 0
; GPRIDX-NEXT: enable_sgpr_workgroup_id_z = 0
; GPRIDX-NEXT: enable_sgpr_workgroup_info = 0
; GPRIDX-NEXT: enable_vgpr_workitem_id = 0
; GPRIDX-NEXT: enable_exception_msb = 0
; GPRIDX-NEXT: granulated_lds_size = 0
; GPRIDX-NEXT: enable_exception = 0
; GPRIDX-NEXT: enable_sgpr_private_segment_buffer = 1
; GPRIDX-NEXT: enable_sgpr_dispatch_ptr = 0
; GPRIDX-NEXT: enable_sgpr_queue_ptr = 0
; GPRIDX-NEXT: enable_sgpr_kernarg_segment_ptr = 1
; GPRIDX-NEXT: enable_sgpr_dispatch_id = 0
; GPRIDX-NEXT: enable_sgpr_flat_scratch_init = 0
; GPRIDX-NEXT: enable_sgpr_private_segment_size = 0
; GPRIDX-NEXT: enable_sgpr_grid_workgroup_count_x = 0
; GPRIDX-NEXT: enable_sgpr_grid_workgroup_count_y = 0
; GPRIDX-NEXT: enable_sgpr_grid_workgroup_count_z = 0
; GPRIDX-NEXT: enable_wavefront_size32 = 0
; GPRIDX-NEXT: enable_ordered_append_gds = 0
; GPRIDX-NEXT: private_element_size = 1
; GPRIDX-NEXT: is_ptr64 = 1
; GPRIDX-NEXT: is_dynamic_callstack = 0
; GPRIDX-NEXT: is_debug_enabled = 0
; GPRIDX-NEXT: is_xnack_enabled = 1
; GPRIDX-NEXT: workitem_private_segment_byte_size = 0
; GPRIDX-NEXT: workgroup_group_segment_byte_size = 0
; GPRIDX-NEXT: gds_segment_byte_size = 0
; GPRIDX-NEXT: kernarg_segment_byte_size = 12
; GPRIDX-NEXT: workgroup_fbarrier_count = 0
; GPRIDX-NEXT: wavefront_sgpr_count = 6
; GPRIDX-NEXT: workitem_vgpr_count = 2
; GPRIDX-NEXT: reserved_vgpr_first = 0
; GPRIDX-NEXT: reserved_vgpr_count = 0
; GPRIDX-NEXT: reserved_sgpr_first = 0
; GPRIDX-NEXT: reserved_sgpr_count = 0
; GPRIDX-NEXT: debug_wavefront_private_segment_offset_sgpr = 0
; GPRIDX-NEXT: debug_private_segment_buffer_sgpr = 0
; GPRIDX-NEXT: kernarg_segment_alignment = 4
; GPRIDX-NEXT: group_segment_alignment = 4
; GPRIDX-NEXT: private_segment_alignment = 4
; GPRIDX-NEXT: wavefront_size = 6
; GPRIDX-NEXT: call_convention = -1
; GPRIDX-NEXT: runtime_loader_kernel_symbol = 0
; GPRIDX-NEXT: .end_amd_kernel_code_t
; GPRIDX-NEXT: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_load_dword s2, s[4:5], 0x8
; GPRIDX-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
; GPRIDX-NEXT: v_mov_b32_e32 v1, 0
; GPRIDX-NEXT: s_waitcnt lgkmcnt(0)
; GPRIDX-NEXT: s_cmp_eq_u32 s2, 1
; GPRIDX-NEXT: s_cselect_b32 s3, 2.0, 1.0
; GPRIDX-NEXT: s_cmp_eq_u32 s2, 2
; GPRIDX-NEXT: s_cselect_b32 s3, 0x40400000, s3
; GPRIDX-NEXT: s_cmp_eq_u32 s2, 3
; GPRIDX-NEXT: s_cselect_b32 s2, 4.0, s3
; GPRIDX-NEXT: v_mov_b32_e32 v0, s2
; GPRIDX-NEXT: global_store_dword v1, v0, s[0:1]
; GPRIDX-NEXT: s_endpgm
;
; MOVREL-LABEL: dyn_extract_v4f32_s_s_s:
; MOVREL: .amd_kernel_code_t
; MOVREL-NEXT: amd_code_version_major = 1
; MOVREL-NEXT: amd_code_version_minor = 2
; MOVREL-NEXT: amd_machine_kind = 1
; MOVREL-NEXT: amd_machine_version_major = 8
; MOVREL-NEXT: amd_machine_version_minor = 0
; MOVREL-NEXT: amd_machine_version_stepping = 3
; MOVREL-NEXT: kernel_code_entry_byte_offset = 256
; MOVREL-NEXT: kernel_code_prefetch_byte_size = 0
; MOVREL-NEXT: granulated_workitem_vgpr_count = 0
; MOVREL-NEXT: granulated_wavefront_sgpr_count = 0
; MOVREL-NEXT: priority = 0
; MOVREL-NEXT: float_mode = 240
; MOVREL-NEXT: priv = 0
; MOVREL-NEXT: enable_dx10_clamp = 1
; MOVREL-NEXT: debug_mode = 0
; MOVREL-NEXT: enable_ieee_mode = 1
; MOVREL-NEXT: enable_wgp_mode = 0
; MOVREL-NEXT: enable_mem_ordered = 0
; MOVREL-NEXT: enable_fwd_progress = 0
; MOVREL-NEXT: enable_sgpr_private_segment_wave_byte_offset = 0
; MOVREL-NEXT: user_sgpr_count = 6
; MOVREL-NEXT: enable_trap_handler = 0
; MOVREL-NEXT: enable_sgpr_workgroup_id_x = 1
; MOVREL-NEXT: enable_sgpr_workgroup_id_y = 0
; MOVREL-NEXT: enable_sgpr_workgroup_id_z = 0
; MOVREL-NEXT: enable_sgpr_workgroup_info = 0
; MOVREL-NEXT: enable_vgpr_workitem_id = 0
; MOVREL-NEXT: enable_exception_msb = 0
; MOVREL-NEXT: granulated_lds_size = 0
; MOVREL-NEXT: enable_exception = 0
; MOVREL-NEXT: enable_sgpr_private_segment_buffer = 1
; MOVREL-NEXT: enable_sgpr_dispatch_ptr = 0
; MOVREL-NEXT: enable_sgpr_queue_ptr = 0
; MOVREL-NEXT: enable_sgpr_kernarg_segment_ptr = 1
; MOVREL-NEXT: enable_sgpr_dispatch_id = 0
; MOVREL-NEXT: enable_sgpr_flat_scratch_init = 0
; MOVREL-NEXT: enable_sgpr_private_segment_size = 0
; MOVREL-NEXT: enable_sgpr_grid_workgroup_count_x = 0
; MOVREL-NEXT: enable_sgpr_grid_workgroup_count_y = 0
; MOVREL-NEXT: enable_sgpr_grid_workgroup_count_z = 0
; MOVREL-NEXT: enable_wavefront_size32 = 0
; MOVREL-NEXT: enable_ordered_append_gds = 0
; MOVREL-NEXT: private_element_size = 1
; MOVREL-NEXT: is_ptr64 = 1
; MOVREL-NEXT: is_dynamic_callstack = 0
; MOVREL-NEXT: is_debug_enabled = 0
; MOVREL-NEXT: is_xnack_enabled = 0
; MOVREL-NEXT: workitem_private_segment_byte_size = 0
; MOVREL-NEXT: workgroup_group_segment_byte_size = 0
; MOVREL-NEXT: gds_segment_byte_size = 0
; MOVREL-NEXT: kernarg_segment_byte_size = 12
; MOVREL-NEXT: workgroup_fbarrier_count = 0
; MOVREL-NEXT: wavefront_sgpr_count = 6
; MOVREL-NEXT: workitem_vgpr_count = 3
; MOVREL-NEXT: reserved_vgpr_first = 0
; MOVREL-NEXT: reserved_vgpr_count = 0
; MOVREL-NEXT: reserved_sgpr_first = 0
; MOVREL-NEXT: reserved_sgpr_count = 0
; MOVREL-NEXT: debug_wavefront_private_segment_offset_sgpr = 0
; MOVREL-NEXT: debug_private_segment_buffer_sgpr = 0
; MOVREL-NEXT: kernarg_segment_alignment = 4
; MOVREL-NEXT: group_segment_alignment = 4
; MOVREL-NEXT: private_segment_alignment = 4
; MOVREL-NEXT: wavefront_size = 6
; MOVREL-NEXT: call_convention = -1
; MOVREL-NEXT: runtime_loader_kernel_symbol = 0
; MOVREL-NEXT: .end_amd_kernel_code_t
; MOVREL-NEXT: ; %bb.0: ; %entry
; MOVREL-NEXT: s_load_dword s2, s[4:5], 0x8
; MOVREL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
; MOVREL-NEXT: s_waitcnt lgkmcnt(0)
; MOVREL-NEXT: s_cmp_eq_u32 s2, 1
; MOVREL-NEXT: s_cselect_b32 s3, 2.0, 1.0
; MOVREL-NEXT: s_cmp_eq_u32 s2, 2
; MOVREL-NEXT: s_cselect_b32 s3, 0x40400000, s3
; MOVREL-NEXT: s_cmp_eq_u32 s2, 3
; MOVREL-NEXT: s_cselect_b32 s2, 4.0, s3
; MOVREL-NEXT: v_mov_b32_e32 v0, s0
; MOVREL-NEXT: v_mov_b32_e32 v2, s2
; MOVREL-NEXT: v_mov_b32_e32 v1, s1
; MOVREL-NEXT: flat_store_dword v[0:1], v2
; MOVREL-NEXT: s_endpgm
;
; GFX10-LABEL: dyn_extract_v4f32_s_s_s:
; GFX10: .amd_kernel_code_t
; GFX10-NEXT: amd_code_version_major = 1
; GFX10-NEXT: amd_code_version_minor = 2
; GFX10-NEXT: amd_machine_kind = 1
; GFX10-NEXT: amd_machine_version_major = 10
; GFX10-NEXT: amd_machine_version_minor = 1
; GFX10-NEXT: amd_machine_version_stepping = 0
; GFX10-NEXT: kernel_code_entry_byte_offset = 256
; GFX10-NEXT: kernel_code_prefetch_byte_size = 0
; GFX10-NEXT: granulated_workitem_vgpr_count = 0
; GFX10-NEXT: granulated_wavefront_sgpr_count = 0
; GFX10-NEXT: priority = 0
; GFX10-NEXT: float_mode = 240
; GFX10-NEXT: priv = 0
; GFX10-NEXT: enable_dx10_clamp = 1
; GFX10-NEXT: debug_mode = 0
; GFX10-NEXT: enable_ieee_mode = 1
; GFX10-NEXT: enable_wgp_mode = 1
; GFX10-NEXT: enable_mem_ordered = 1
; GFX10-NEXT: enable_fwd_progress = 0
; GFX10-NEXT: enable_sgpr_private_segment_wave_byte_offset = 0
; GFX10-NEXT: user_sgpr_count = 6
; GFX10-NEXT: enable_trap_handler = 0
; GFX10-NEXT: enable_sgpr_workgroup_id_x = 1
; GFX10-NEXT: enable_sgpr_workgroup_id_y = 0
; GFX10-NEXT: enable_sgpr_workgroup_id_z = 0
; GFX10-NEXT: enable_sgpr_workgroup_info = 0
; GFX10-NEXT: enable_vgpr_workitem_id = 0
; GFX10-NEXT: enable_exception_msb = 0
; GFX10-NEXT: granulated_lds_size = 0
; GFX10-NEXT: enable_exception = 0
; GFX10-NEXT: enable_sgpr_private_segment_buffer = 1
; GFX10-NEXT: enable_sgpr_dispatch_ptr = 0
; GFX10-NEXT: enable_sgpr_queue_ptr = 0
; GFX10-NEXT: enable_sgpr_kernarg_segment_ptr = 1
; GFX10-NEXT: enable_sgpr_dispatch_id = 0
; GFX10-NEXT: enable_sgpr_flat_scratch_init = 0
; GFX10-NEXT: enable_sgpr_private_segment_size = 0
; GFX10-NEXT: enable_sgpr_grid_workgroup_count_x = 0
; GFX10-NEXT: enable_sgpr_grid_workgroup_count_y = 0
; GFX10-NEXT: enable_sgpr_grid_workgroup_count_z = 0
; GFX10-NEXT: enable_wavefront_size32 = 1
; GFX10-NEXT: enable_ordered_append_gds = 0
; GFX10-NEXT: private_element_size = 1
; GFX10-NEXT: is_ptr64 = 1
; GFX10-NEXT: is_dynamic_callstack = 0
; GFX10-NEXT: is_debug_enabled = 0
; GFX10-NEXT: is_xnack_enabled = 1
; GFX10-NEXT: workitem_private_segment_byte_size = 0
; GFX10-NEXT: workgroup_group_segment_byte_size = 0
; GFX10-NEXT: gds_segment_byte_size = 0
; GFX10-NEXT: kernarg_segment_byte_size = 12
; GFX10-NEXT: workgroup_fbarrier_count = 0
; GFX10-NEXT: wavefront_sgpr_count = 6
; GFX10-NEXT: workitem_vgpr_count = 2
; GFX10-NEXT: reserved_vgpr_first = 0
; GFX10-NEXT: reserved_vgpr_count = 0
; GFX10-NEXT: reserved_sgpr_first = 0
; GFX10-NEXT: reserved_sgpr_count = 0
; GFX10-NEXT: debug_wavefront_private_segment_offset_sgpr = 0
; GFX10-NEXT: debug_private_segment_buffer_sgpr = 0
; GFX10-NEXT: kernarg_segment_alignment = 4
; GFX10-NEXT: group_segment_alignment = 4
; GFX10-NEXT: private_segment_alignment = 4
; GFX10-NEXT: wavefront_size = 5
; GFX10-NEXT: call_convention = -1
; GFX10-NEXT: runtime_loader_kernel_symbol = 0
; GFX10-NEXT: .end_amd_kernel_code_t
; GFX10-NEXT: ; %bb.0: ; %entry
; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: s_load_dword s2, s[4:5], 0x8
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
; GFX10-NEXT: v_mov_b32_e32 v1, 0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_cmp_eq_u32 s2, 1
; GFX10-NEXT: s_cselect_b32 s3, 2.0, 1.0
; GFX10-NEXT: s_cmp_eq_u32 s2, 2
; GFX10-NEXT: s_cselect_b32 s3, 0x40400000, s3
; GFX10-NEXT: s_cmp_eq_u32 s2, 3
; GFX10-NEXT: s_cselect_b32 s2, 4.0, s3
; GFX10-NEXT: v_mov_b32_e32 v0, s2
; GFX10-NEXT: global_store_dword v1, v0, s[0:1]
; GFX10-NEXT: s_endpgm
;
; GFX11-LABEL: dyn_extract_v4f32_s_s_s:
; GFX11: .amd_kernel_code_t
; GFX11-NEXT: amd_code_version_major = 1
; GFX11-NEXT: amd_code_version_minor = 2
; GFX11-NEXT: amd_machine_kind = 1
; GFX11-NEXT: amd_machine_version_major = 11
; GFX11-NEXT: amd_machine_version_minor = 0
; GFX11-NEXT: amd_machine_version_stepping = 0
; GFX11-NEXT: kernel_code_entry_byte_offset = 256
; GFX11-NEXT: kernel_code_prefetch_byte_size = 0
; GFX11-NEXT: granulated_workitem_vgpr_count = 0
; GFX11-NEXT: granulated_wavefront_sgpr_count = 0
; GFX11-NEXT: priority = 0
; GFX11-NEXT: float_mode = 240
; GFX11-NEXT: priv = 0
; GFX11-NEXT: enable_dx10_clamp = 1
; GFX11-NEXT: debug_mode = 0
; GFX11-NEXT: enable_ieee_mode = 1
; GFX11-NEXT: enable_wgp_mode = 1
; GFX11-NEXT: enable_mem_ordered = 1
; GFX11-NEXT: enable_fwd_progress = 0
; GFX11-NEXT: enable_sgpr_private_segment_wave_byte_offset = 0
; GFX11-NEXT: user_sgpr_count = 15
; GFX11-NEXT: enable_trap_handler = 0
; GFX11-NEXT: enable_sgpr_workgroup_id_x = 1
; GFX11-NEXT: enable_sgpr_workgroup_id_y = 0
; GFX11-NEXT: enable_sgpr_workgroup_id_z = 0
; GFX11-NEXT: enable_sgpr_workgroup_info = 0
; GFX11-NEXT: enable_vgpr_workitem_id = 0
; GFX11-NEXT: enable_exception_msb = 0
; GFX11-NEXT: granulated_lds_size = 0
; GFX11-NEXT: enable_exception = 0
; GFX11-NEXT: enable_sgpr_private_segment_buffer = 0
; GFX11-NEXT: enable_sgpr_dispatch_ptr = 0
; GFX11-NEXT: enable_sgpr_queue_ptr = 0
; GFX11-NEXT: enable_sgpr_kernarg_segment_ptr = 1
; GFX11-NEXT: enable_sgpr_dispatch_id = 0
; GFX11-NEXT: enable_sgpr_flat_scratch_init = 0
; GFX11-NEXT: enable_sgpr_private_segment_size = 0
; GFX11-NEXT: enable_sgpr_grid_workgroup_count_x = 0
; GFX11-NEXT: enable_sgpr_grid_workgroup_count_y = 0
; GFX11-NEXT: enable_sgpr_grid_workgroup_count_z = 0
; GFX11-NEXT: enable_wavefront_size32 = 1
; GFX11-NEXT: enable_ordered_append_gds = 0
; GFX11-NEXT: private_element_size = 1
; GFX11-NEXT: is_ptr64 = 1
; GFX11-NEXT: is_dynamic_callstack = 0
; GFX11-NEXT: is_debug_enabled = 0
; GFX11-NEXT: is_xnack_enabled = 0
; GFX11-NEXT: workitem_private_segment_byte_size = 0
; GFX11-NEXT: workgroup_group_segment_byte_size = 0
; GFX11-NEXT: gds_segment_byte_size = 0
; GFX11-NEXT: kernarg_segment_byte_size = 12
; GFX11-NEXT: workgroup_fbarrier_count = 0
; GFX11-NEXT: wavefront_sgpr_count = 4
; GFX11-NEXT: workitem_vgpr_count = 2
; GFX11-NEXT: reserved_vgpr_first = 0
; GFX11-NEXT: reserved_vgpr_count = 0
; GFX11-NEXT: reserved_sgpr_first = 0
; GFX11-NEXT: reserved_sgpr_count = 0
; GFX11-NEXT: debug_wavefront_private_segment_offset_sgpr = 0
; GFX11-NEXT: debug_private_segment_buffer_sgpr = 0
; GFX11-NEXT: kernarg_segment_alignment = 4
; GFX11-NEXT: group_segment_alignment = 4
; GFX11-NEXT: private_segment_alignment = 4
; GFX11-NEXT: wavefront_size = 5
; GFX11-NEXT: call_convention = -1
; GFX11-NEXT: runtime_loader_kernel_symbol = 0
; GFX11-NEXT: .end_amd_kernel_code_t
; GFX11-NEXT: ; %bb.0: ; %entry
; GFX11-NEXT: s_clause 0x1
; GFX11-NEXT: s_load_b32 s2, s[0:1], 0x8
; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x0
; GFX11-NEXT: v_mov_b32_e32 v1, 0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_cmp_eq_u32 s2, 1
; GFX11-NEXT: s_cselect_b32 s3, 2.0, 1.0
; GFX11-NEXT: s_cmp_eq_u32 s2, 2
; GFX11-NEXT: s_cselect_b32 s3, 0x40400000, s3
; GFX11-NEXT: s_cmp_eq_u32 s2, 3
; GFX11-NEXT: s_cselect_b32 s2, 4.0, s3
; GFX11-NEXT: v_mov_b32_e32 v0, s2
; GFX11-NEXT: global_store_b32 v1, v0, s[0:1]
; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX11-NEXT: s_endpgm
entry:
%ext = extractelement <4 x float> <float 1.0, float 2.0, float 3.0, float 4.0>, i32 %sel
store float %ext, float addrspace(1)* %out
ret void
}
define amdgpu_kernel void @dyn_extract_v4f64_s_s_s(double addrspace(1)* %out, i32 %sel) {
; GPRIDX-LABEL: dyn_extract_v4f64_s_s_s:
; GPRIDX: .amd_kernel_code_t
; GPRIDX-NEXT: amd_code_version_major = 1
; GPRIDX-NEXT: amd_code_version_minor = 2
; GPRIDX-NEXT: amd_machine_kind = 1
; GPRIDX-NEXT: amd_machine_version_major = 9
; GPRIDX-NEXT: amd_machine_version_minor = 0
; GPRIDX-NEXT: amd_machine_version_stepping = 0
; GPRIDX-NEXT: kernel_code_entry_byte_offset = 256
; GPRIDX-NEXT: kernel_code_prefetch_byte_size = 0
; GPRIDX-NEXT: granulated_workitem_vgpr_count = 0
; GPRIDX-NEXT: granulated_wavefront_sgpr_count = 0
; GPRIDX-NEXT: priority = 0
; GPRIDX-NEXT: float_mode = 240
; GPRIDX-NEXT: priv = 0
; GPRIDX-NEXT: enable_dx10_clamp = 1
; GPRIDX-NEXT: debug_mode = 0
; GPRIDX-NEXT: enable_ieee_mode = 1
; GPRIDX-NEXT: enable_wgp_mode = 0
; GPRIDX-NEXT: enable_mem_ordered = 0
; GPRIDX-NEXT: enable_fwd_progress = 0
; GPRIDX-NEXT: enable_sgpr_private_segment_wave_byte_offset = 0
; GPRIDX-NEXT: user_sgpr_count = 6
; GPRIDX-NEXT: enable_trap_handler = 0
; GPRIDX-NEXT: enable_sgpr_workgroup_id_x = 1
; GPRIDX-NEXT: enable_sgpr_workgroup_id_y = 0
; GPRIDX-NEXT: enable_sgpr_workgroup_id_z = 0
; GPRIDX-NEXT: enable_sgpr_workgroup_info = 0
; GPRIDX-NEXT: enable_vgpr_workitem_id = 0
; GPRIDX-NEXT: enable_exception_msb = 0
; GPRIDX-NEXT: granulated_lds_size = 0
; GPRIDX-NEXT: enable_exception = 0
; GPRIDX-NEXT: enable_sgpr_private_segment_buffer = 1
; GPRIDX-NEXT: enable_sgpr_dispatch_ptr = 0
; GPRIDX-NEXT: enable_sgpr_queue_ptr = 0
; GPRIDX-NEXT: enable_sgpr_kernarg_segment_ptr = 1
; GPRIDX-NEXT: enable_sgpr_dispatch_id = 0
; GPRIDX-NEXT: enable_sgpr_flat_scratch_init = 0
; GPRIDX-NEXT: enable_sgpr_private_segment_size = 0
; GPRIDX-NEXT: enable_sgpr_grid_workgroup_count_x = 0
; GPRIDX-NEXT: enable_sgpr_grid_workgroup_count_y = 0
; GPRIDX-NEXT: enable_sgpr_grid_workgroup_count_z = 0
; GPRIDX-NEXT: enable_wavefront_size32 = 0
; GPRIDX-NEXT: enable_ordered_append_gds = 0
; GPRIDX-NEXT: private_element_size = 1
; GPRIDX-NEXT: is_ptr64 = 1
; GPRIDX-NEXT: is_dynamic_callstack = 0
; GPRIDX-NEXT: is_debug_enabled = 0
; GPRIDX-NEXT: is_xnack_enabled = 1
; GPRIDX-NEXT: workitem_private_segment_byte_size = 0
; GPRIDX-NEXT: workgroup_group_segment_byte_size = 0
; GPRIDX-NEXT: gds_segment_byte_size = 0
; GPRIDX-NEXT: kernarg_segment_byte_size = 12
; GPRIDX-NEXT: workgroup_fbarrier_count = 0
; GPRIDX-NEXT: wavefront_sgpr_count = 7
; GPRIDX-NEXT: workitem_vgpr_count = 3
; GPRIDX-NEXT: reserved_vgpr_first = 0
; GPRIDX-NEXT: reserved_vgpr_count = 0
; GPRIDX-NEXT: reserved_sgpr_first = 0
; GPRIDX-NEXT: reserved_sgpr_count = 0
; GPRIDX-NEXT: debug_wavefront_private_segment_offset_sgpr = 0
; GPRIDX-NEXT: debug_private_segment_buffer_sgpr = 0
; GPRIDX-NEXT: kernarg_segment_alignment = 4
; GPRIDX-NEXT: group_segment_alignment = 4
; GPRIDX-NEXT: private_segment_alignment = 4
; GPRIDX-NEXT: wavefront_size = 6
; GPRIDX-NEXT: call_convention = -1
; GPRIDX-NEXT: runtime_loader_kernel_symbol = 0
; GPRIDX-NEXT: .end_amd_kernel_code_t
; GPRIDX-NEXT: ; %bb.0: ; %entry
; GPRIDX-NEXT: s_load_dword s6, s[4:5], 0x8
; GPRIDX-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
; GPRIDX-NEXT: s_mov_b32 s2, 0
; GPRIDX-NEXT: s_mov_b32 s3, 0x40080000
; GPRIDX-NEXT: v_mov_b32_e32 v2, 0
; GPRIDX-NEXT: s_waitcnt lgkmcnt(0)
; GPRIDX-NEXT: s_cmp_eq_u32 s6, 1
; GPRIDX-NEXT: s_cselect_b64 s[4:5], 2.0, 1.0
; GPRIDX-NEXT: s_cmp_eq_u32 s6, 2
; GPRIDX-NEXT: s_cselect_b64 s[2:3], s[2:3], s[4:5]
; GPRIDX-NEXT: s_cmp_eq_u32 s6, 3
; GPRIDX-NEXT: s_cselect_b64 s[2:3], 4.0, s[2:3]
; GPRIDX-NEXT: v_mov_b32_e32 v0, s2
; GPRIDX-NEXT: v_mov_b32_e32 v1, s3
; GPRIDX-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1]
; GPRIDX-NEXT: s_endpgm
;
; MOVREL-LABEL: dyn_extract_v4f64_s_s_s:
; MOVREL: .amd_kernel_code_t
; MOVREL-NEXT: amd_code_version_major = 1
; MOVREL-NEXT: amd_code_version_minor = 2
; MOVREL-NEXT: amd_machine_kind = 1
; MOVREL-NEXT: amd_machine_version_major = 8
; MOVREL-NEXT: amd_machine_version_minor = 0
; MOVREL-NEXT: amd_machine_version_stepping = 3
; MOVREL-NEXT: kernel_code_entry_byte_offset = 256
; MOVREL-NEXT: kernel_code_prefetch_byte_size = 0
; MOVREL-NEXT: granulated_workitem_vgpr_count = 0
; MOVREL-NEXT: granulated_wavefront_sgpr_count = 0
; MOVREL-NEXT: priority = 0
; MOVREL-NEXT: float_mode = 240
; MOVREL-NEXT: priv = 0
; MOVREL-NEXT: enable_dx10_clamp = 1
; MOVREL-NEXT: debug_mode = 0
; MOVREL-NEXT: enable_ieee_mode = 1
; MOVREL-NEXT: enable_wgp_mode = 0
; MOVREL-NEXT: enable_mem_ordered = 0
; MOVREL-NEXT: enable_fwd_progress = 0
; MOVREL-NEXT: enable_sgpr_private_segment_wave_byte_offset = 0
; MOVREL-NEXT: user_sgpr_count = 6
; MOVREL-NEXT: enable_trap_handler = 0
; MOVREL-NEXT: enable_sgpr_workgroup_id_x = 1
; MOVREL-NEXT: enable_sgpr_workgroup_id_y = 0
; MOVREL-NEXT: enable_sgpr_workgroup_id_z = 0
; MOVREL-NEXT: enable_sgpr_workgroup_info = 0
; MOVREL-NEXT: enable_vgpr_workitem_id = 0
; MOVREL-NEXT: enable_exception_msb = 0
; MOVREL-NEXT: granulated_lds_size = 0
; MOVREL-NEXT: enable_exception = 0
; MOVREL-NEXT: enable_sgpr_private_segment_buffer = 1
; MOVREL-NEXT: enable_sgpr_dispatch_ptr = 0
; MOVREL-NEXT: enable_sgpr_queue_ptr = 0
; MOVREL-NEXT: enable_sgpr_kernarg_segment_ptr = 1
; MOVREL-NEXT: enable_sgpr_dispatch_id = 0
; MOVREL-NEXT: enable_sgpr_flat_scratch_init = 0
; MOVREL-NEXT: enable_sgpr_private_segment_size = 0
; MOVREL-NEXT: enable_sgpr_grid_workgroup_count_x = 0
; MOVREL-NEXT: enable_sgpr_grid_workgroup_count_y = 0
; MOVREL-NEXT: enable_sgpr_grid_workgroup_count_z = 0
; MOVREL-NEXT: enable_wavefront_size32 = 0
; MOVREL-NEXT: enable_ordered_append_gds = 0
; MOVREL-NEXT: private_element_size = 1
; MOVREL-NEXT: is_ptr64 = 1
; MOVREL-NEXT: is_dynamic_callstack = 0
; MOVREL-NEXT: is_debug_enabled = 0
; MOVREL-NEXT: is_xnack_enabled = 0
; MOVREL-NEXT: workitem_private_segment_byte_size = 0
; MOVREL-NEXT: workgroup_group_segment_byte_size = 0
; MOVREL-NEXT: gds_segment_byte_size = 0
; MOVREL-NEXT: kernarg_segment_byte_size = 12
; MOVREL-NEXT: workgroup_fbarrier_count = 0
; MOVREL-NEXT: wavefront_sgpr_count = 7
; MOVREL-NEXT: workitem_vgpr_count = 4
; MOVREL-NEXT: reserved_vgpr_first = 0
; MOVREL-NEXT: reserved_vgpr_count = 0
; MOVREL-NEXT: reserved_sgpr_first = 0
; MOVREL-NEXT: reserved_sgpr_count = 0
; MOVREL-NEXT: debug_wavefront_private_segment_offset_sgpr = 0
; MOVREL-NEXT: debug_private_segment_buffer_sgpr = 0
; MOVREL-NEXT: kernarg_segment_alignment = 4
; MOVREL-NEXT: group_segment_alignment = 4
; MOVREL-NEXT: private_segment_alignment = 4
; MOVREL-NEXT: wavefront_size = 6
; MOVREL-NEXT: call_convention = -1
; MOVREL-NEXT: runtime_loader_kernel_symbol = 0
; MOVREL-NEXT: .end_amd_kernel_code_t
; MOVREL-NEXT: ; %bb.0: ; %entry
; MOVREL-NEXT: s_load_dword s6, s[4:5], 0x8
; MOVREL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
; MOVREL-NEXT: s_mov_b32 s2, 0
; MOVREL-NEXT: s_mov_b32 s3, 0x40080000
; MOVREL-NEXT: s_waitcnt lgkmcnt(0)
; MOVREL-NEXT: s_cmp_eq_u32 s6, 1
; MOVREL-NEXT: s_cselect_b64 s[4:5], 2.0, 1.0
; MOVREL-NEXT: s_cmp_eq_u32 s6, 2
; MOVREL-NEXT: s_cselect_b64 s[2:3], s[2:3], s[4:5]
; MOVREL-NEXT: s_cmp_eq_u32 s6, 3
; MOVREL-NEXT: s_cselect_b64 s[2:3], 4.0, s[2:3]
; MOVREL-NEXT: v_mov_b32_e32 v0, s2
; MOVREL-NEXT: v_mov_b32_e32 v3, s1
; MOVREL-NEXT: v_mov_b32_e32 v1, s3
; MOVREL-NEXT: v_mov_b32_e32 v2, s0
; MOVREL-NEXT: flat_store_dwordx2 v[2:3], v[0:1]
; MOVREL-NEXT: s_endpgm
;
; GFX10-LABEL: dyn_extract_v4f64_s_s_s:
; GFX10: .amd_kernel_code_t
; GFX10-NEXT: amd_code_version_major = 1
; GFX10-NEXT: amd_code_version_minor = 2
; GFX10-NEXT: amd_machine_kind = 1
; GFX10-NEXT: amd_machine_version_major = 10
; GFX10-NEXT: amd_machine_version_minor = 1
; GFX10-NEXT: amd_machine_version_stepping = 0
; GFX10-NEXT: kernel_code_entry_byte_offset = 256
; GFX10-NEXT: kernel_code_prefetch_byte_size = 0
; GFX10-NEXT: granulated_workitem_vgpr_count = 0
; GFX10-NEXT: granulated_wavefront_sgpr_count = 0
; GFX10-NEXT: priority = 0
; GFX10-NEXT: float_mode = 240
; GFX10-NEXT: priv = 0
; GFX10-NEXT: enable_dx10_clamp = 1
; GFX10-NEXT: debug_mode = 0
; GFX10-NEXT: enable_ieee_mode = 1
; GFX10-NEXT: enable_wgp_mode = 1
; GFX10-NEXT: enable_mem_ordered = 1
; GFX10-NEXT: enable_fwd_progress = 0
; GFX10-NEXT: enable_sgpr_private_segment_wave_byte_offset = 0
; GFX10-NEXT: user_sgpr_count = 6
; GFX10-NEXT: enable_trap_handler = 0
; GFX10-NEXT: enable_sgpr_workgroup_id_x = 1
; GFX10-NEXT: enable_sgpr_workgroup_id_y = 0
; GFX10-NEXT: enable_sgpr_workgroup_id_z = 0
; GFX10-NEXT: enable_sgpr_workgroup_info = 0
; GFX10-NEXT: enable_vgpr_workitem_id = 0
; GFX10-NEXT: enable_exception_msb = 0
; GFX10-NEXT: granulated_lds_size = 0
; GFX10-NEXT: enable_exception = 0
; GFX10-NEXT: enable_sgpr_private_segment_buffer = 1
; GFX10-NEXT: enable_sgpr_dispatch_ptr = 0
; GFX10-NEXT: enable_sgpr_queue_ptr = 0
; GFX10-NEXT: enable_sgpr_kernarg_segment_ptr = 1
; GFX10-NEXT: enable_sgpr_dispatch_id = 0
; GFX10-NEXT: enable_sgpr_flat_scratch_init = 0
; GFX10-NEXT: enable_sgpr_private_segment_size = 0
; GFX10-NEXT: enable_sgpr_grid_workgroup_count_x = 0
; GFX10-NEXT: enable_sgpr_grid_workgroup_count_y = 0
; GFX10-NEXT: enable_sgpr_grid_workgroup_count_z = 0
; GFX10-NEXT: enable_wavefront_size32 = 1
; GFX10-NEXT: enable_ordered_append_gds = 0
; GFX10-NEXT: private_element_size = 1
; GFX10-NEXT: is_ptr64 = 1
; GFX10-NEXT: is_dynamic_callstack = 0
; GFX10-NEXT: is_debug_enabled = 0
; GFX10-NEXT: is_xnack_enabled = 1
; GFX10-NEXT: workitem_private_segment_byte_size = 0
; GFX10-NEXT: workgroup_group_segment_byte_size = 0
; GFX10-NEXT: gds_segment_byte_size = 0
; GFX10-NEXT: kernarg_segment_byte_size = 12
; GFX10-NEXT: workgroup_fbarrier_count = 0
; GFX10-NEXT: wavefront_sgpr_count = 7
; GFX10-NEXT: workitem_vgpr_count = 3
; GFX10-NEXT: reserved_vgpr_first = 0
; GFX10-NEXT: reserved_vgpr_count = 0
; GFX10-NEXT: reserved_sgpr_first = 0
; GFX10-NEXT: reserved_sgpr_count = 0
; GFX10-NEXT: debug_wavefront_private_segment_offset_sgpr = 0
; GFX10-NEXT: debug_private_segment_buffer_sgpr = 0
; GFX10-NEXT: kernarg_segment_alignment = 4
; GFX10-NEXT: group_segment_alignment = 4
; GFX10-NEXT: private_segment_alignment = 4
; GFX10-NEXT: wavefront_size = 5
; GFX10-NEXT: call_convention = -1
; GFX10-NEXT: runtime_loader_kernel_symbol = 0
; GFX10-NEXT: .end_amd_kernel_code_t
; GFX10-NEXT: ; %bb.0: ; %entry
; GFX10-NEXT: s_clause 0x1
; GFX10-NEXT: s_load_dword s6, s[4:5], 0x8
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x0
; GFX10-NEXT: s_mov_b32 s2, 0
; GFX10-NEXT: s_mov_b32 s3, 0x40080000
; GFX10-NEXT: v_mov_b32_e32 v2, 0
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: s_cmp_eq_u32 s6, 1
; GFX10-NEXT: s_cselect_b64 s[4:5], 2.0, 1.0
; GFX10-NEXT: s_cmp_eq_u32 s6, 2
; GFX10-NEXT: s_cselect_b64 s[2:3], s[2:3], s[4:5]
; GFX10-NEXT: s_cmp_eq_u32 s6, 3
; GFX10-NEXT: s_cselect_b64 s[2:3], 4.0, s[2:3]
; GFX10-NEXT: v_mov_b32_e32 v0, s2
; GFX10-NEXT: v_mov_b32_e32 v1, s3
; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1]
; GFX10-NEXT: s_endpgm
;
; GFX11-LABEL: dyn_extract_v4f64_s_s_s:
; GFX11: .amd_kernel_code_t
; GFX11-NEXT: amd_code_version_major = 1
; GFX11-NEXT: amd_code_version_minor = 2
; GFX11-NEXT: amd_machine_kind = 1
; GFX11-NEXT: amd_machine_version_major = 11
; GFX11-NEXT: amd_machine_version_minor = 0
; GFX11-NEXT: amd_machine_version_stepping = 0
; GFX11-NEXT: kernel_code_entry_byte_offset = 256
; GFX11-NEXT: kernel_code_prefetch_byte_size = 0
; GFX11-NEXT: granulated_workitem_vgpr_count = 0
; GFX11-NEXT: granulated_wavefront_sgpr_count = 0
; GFX11-NEXT: priority = 0
; GFX11-NEXT: float_mode = 240
; GFX11-NEXT: priv = 0
; GFX11-NEXT: enable_dx10_clamp = 1
; GFX11-NEXT: debug_mode = 0
; GFX11-NEXT: enable_ieee_mode = 1
; GFX11-NEXT: enable_wgp_mode = 1
; GFX11-NEXT: enable_mem_ordered = 1
; GFX11-NEXT: enable_fwd_progress = 0
; GFX11-NEXT: enable_sgpr_private_segment_wave_byte_offset = 0
; GFX11-NEXT: user_sgpr_count = 15
; GFX11-NEXT: enable_trap_handler = 0
; GFX11-NEXT: enable_sgpr_workgroup_id_x = 1
; GFX11-NEXT: enable_sgpr_workgroup_id_y = 0
; GFX11-NEXT: enable_sgpr_workgroup_id_z = 0
; GFX11-NEXT: enable_sgpr_workgroup_info = 0
; GFX11-NEXT: enable_vgpr_workitem_id = 0
; GFX11-NEXT: enable_exception_msb = 0
; GFX11-NEXT: granulated_lds_size = 0
; GFX11-NEXT: enable_exception = 0
; GFX11-NEXT: enable_sgpr_private_segment_buffer = 0
; GFX11-NEXT: enable_sgpr_dispatch_ptr = 0
; GFX11-NEXT: enable_sgpr_queue_ptr = 0
; GFX11-NEXT: enable_sgpr_kernarg_segment_ptr = 1
; GFX11-NEXT: enable_sgpr_dispatch_id = 0
; GFX11-NEXT: enable_sgpr_flat_scratch_init = 0
; GFX11-NEXT: enable_sgpr_private_segment_size = 0
; GFX11-NEXT: enable_sgpr_grid_workgroup_count_x = 0
; GFX11-NEXT: enable_sgpr_grid_workgroup_count_y = 0
; GFX11-NEXT: enable_sgpr_grid_workgroup_count_z = 0
; GFX11-NEXT: enable_wavefront_size32 = 1
; GFX11-NEXT: enable_ordered_append_gds = 0
; GFX11-NEXT: private_element_size = 1
; GFX11-NEXT: is_ptr64 = 1
; GFX11-NEXT: is_dynamic_callstack = 0
; GFX11-NEXT: is_debug_enabled = 0
; GFX11-NEXT: is_xnack_enabled = 0
; GFX11-NEXT: workitem_private_segment_byte_size = 0
; GFX11-NEXT: workgroup_group_segment_byte_size = 0
; GFX11-NEXT: gds_segment_byte_size = 0
; GFX11-NEXT: kernarg_segment_byte_size = 12
; GFX11-NEXT: workgroup_fbarrier_count = 0
; GFX11-NEXT: wavefront_sgpr_count = 7
; GFX11-NEXT: workitem_vgpr_count = 3
; GFX11-NEXT: reserved_vgpr_first = 0
; GFX11-NEXT: reserved_vgpr_count = 0
; GFX11-NEXT: reserved_sgpr_first = 0
; GFX11-NEXT: reserved_sgpr_count = 0
; GFX11-NEXT: debug_wavefront_private_segment_offset_sgpr = 0
; GFX11-NEXT: debug_private_segment_buffer_sgpr = 0
; GFX11-NEXT: kernarg_segment_alignment = 4
; GFX11-NEXT: group_segment_alignment = 4
; GFX11-NEXT: private_segment_alignment = 4
; GFX11-NEXT: wavefront_size = 5
; GFX11-NEXT: call_convention = -1
; GFX11-NEXT: runtime_loader_kernel_symbol = 0
; GFX11-NEXT: .end_amd_kernel_code_t
; GFX11-NEXT: ; %bb.0: ; %entry
; GFX11-NEXT: s_clause 0x1
; GFX11-NEXT: s_load_b32 s6, s[0:1], 0x8
; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x0
; GFX11-NEXT: s_mov_b32 s2, 0
; GFX11-NEXT: s_mov_b32 s3, 0x40080000
; GFX11-NEXT: v_mov_b32_e32 v2, 0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_cmp_eq_u32 s6, 1
; GFX11-NEXT: s_cselect_b64 s[4:5], 2.0, 1.0
; GFX11-NEXT: s_cmp_eq_u32 s6, 2
; GFX11-NEXT: s_cselect_b64 s[2:3], s[2:3], s[4:5]
; GFX11-NEXT: s_cmp_eq_u32 s6, 3
; GFX11-NEXT: s_cselect_b64 s[2:3], 4.0, s[2:3]
; GFX11-NEXT: v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
; GFX11-NEXT: s_endpgm
entry:
%ext = extractelement <4 x double> <double 1.0, double 2.0, double 3.0, double 4.0>, i32 %sel
store double %ext, double addrspace(1)* %out
ret void
}
define i32 @v_extract_v64i32_7(<64 x i32> addrspace(1)* %ptr) {
; GPRIDX-LABEL: v_extract_v64i32_7:
; GPRIDX: ; %bb.0:
; GPRIDX-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GPRIDX-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:16
; GPRIDX-NEXT: s_waitcnt vmcnt(0)
; GPRIDX-NEXT: v_mov_b32_e32 v0, v7
; GPRIDX-NEXT: s_setpc_b64 s[30:31]
;
; MOVREL-LABEL: v_extract_v64i32_7:
; MOVREL: ; %bb.0:
; MOVREL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; MOVREL-NEXT: v_add_u32_e32 v0, vcc, 16, v0
; MOVREL-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
; MOVREL-NEXT: flat_load_dwordx4 v[4:7], v[0:1]
; MOVREL-NEXT: s_waitcnt vmcnt(0)
; MOVREL-NEXT: v_mov_b32_e32 v0, v7
; MOVREL-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_extract_v64i32_7:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:16
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v0, v7
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: v_extract_v64i32_7:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-NEXT: global_load_b128 v[4:7], v[0:1], off offset:16
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_mov_b32_e32 v0, v7
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vec = load <64 x i32>, <64 x i32> addrspace(1)* %ptr
%elt = extractelement <64 x i32> %vec, i32 7
ret i32 %elt
}
define i32 @v_extract_v64i32_32(<64 x i32> addrspace(1)* %ptr) {
; GPRIDX-LABEL: v_extract_v64i32_32:
; GPRIDX: ; %bb.0:
; GPRIDX-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GPRIDX-NEXT: global_load_dwordx4 v[0:3], v[0:1], off offset:128
; GPRIDX-NEXT: s_waitcnt vmcnt(0)
; GPRIDX-NEXT: s_setpc_b64 s[30:31]
;
; MOVREL-LABEL: v_extract_v64i32_32:
; MOVREL: ; %bb.0:
; MOVREL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; MOVREL-NEXT: s_mov_b64 s[4:5], 0x80
; MOVREL-NEXT: v_mov_b32_e32 v2, s4
; MOVREL-NEXT: v_mov_b32_e32 v3, s5
; MOVREL-NEXT: v_add_u32_e32 v0, vcc, v0, v2
; MOVREL-NEXT: v_addc_u32_e32 v1, vcc, v1, v3, vcc
; MOVREL-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; MOVREL-NEXT: s_waitcnt vmcnt(0)
; MOVREL-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_extract_v64i32_32:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off offset:128
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: v_extract_v64i32_32:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off offset:128
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vec = load <64 x i32>, <64 x i32> addrspace(1)* %ptr
%elt = extractelement <64 x i32> %vec, i32 32
ret i32 %elt
}
define i32 @v_extract_v64i32_33(<64 x i32> addrspace(1)* %ptr) {
; GPRIDX-LABEL: v_extract_v64i32_33:
; GPRIDX: ; %bb.0:
; GPRIDX-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GPRIDX-NEXT: global_load_dwordx4 v[0:3], v[0:1], off offset:128
; GPRIDX-NEXT: s_waitcnt vmcnt(0)
; GPRIDX-NEXT: v_mov_b32_e32 v0, v1
; GPRIDX-NEXT: s_setpc_b64 s[30:31]
;
; MOVREL-LABEL: v_extract_v64i32_33:
; MOVREL: ; %bb.0:
; MOVREL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; MOVREL-NEXT: s_mov_b64 s[4:5], 0x80
; MOVREL-NEXT: v_mov_b32_e32 v2, s4
; MOVREL-NEXT: v_mov_b32_e32 v3, s5
; MOVREL-NEXT: v_add_u32_e32 v0, vcc, v0, v2
; MOVREL-NEXT: v_addc_u32_e32 v1, vcc, v1, v3, vcc
; MOVREL-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
; MOVREL-NEXT: s_waitcnt vmcnt(0)
; MOVREL-NEXT: v_mov_b32_e32 v0, v1
; MOVREL-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_extract_v64i32_33:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off offset:128
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v0, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: v_extract_v64i32_33:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off offset:128
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_mov_b32_e32 v0, v1
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vec = load <64 x i32>, <64 x i32> addrspace(1)* %ptr
%elt = extractelement <64 x i32> %vec, i32 33
ret i32 %elt
}
define i32 @v_extract_v64i32_37(<64 x i32> addrspace(1)* %ptr) {
; GPRIDX-LABEL: v_extract_v64i32_37:
; GPRIDX: ; %bb.0:
; GPRIDX-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GPRIDX-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:144
; GPRIDX-NEXT: s_waitcnt vmcnt(0)
; GPRIDX-NEXT: v_mov_b32_e32 v0, v5
; GPRIDX-NEXT: s_setpc_b64 s[30:31]
;
; MOVREL-LABEL: v_extract_v64i32_37:
; MOVREL: ; %bb.0:
; MOVREL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; MOVREL-NEXT: v_add_u32_e32 v0, vcc, 0x90, v0
; MOVREL-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
; MOVREL-NEXT: flat_load_dwordx4 v[4:7], v[0:1]
; MOVREL-NEXT: s_waitcnt vmcnt(0)
; MOVREL-NEXT: v_mov_b32_e32 v0, v5
; MOVREL-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: v_extract_v64i32_37:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
; GFX10-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:144
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v0, v5
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-LABEL: v_extract_v64i32_37:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-NEXT: global_load_b128 v[4:7], v[0:1], off offset:144
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: v_mov_b32_e32 v0, v5
; GFX11-NEXT: s_setpc_b64 s[30:31]
%vec = load <64 x i32>, <64 x i32> addrspace(1)* %ptr
%elt = extractelement <64 x i32> %vec, i32 37
ret i32 %elt
}