; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GFX1250,GFX1250-SDAG %s
; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GFX1250,GFX1250-GISEL %s

; %D4 should be zero-initialized for gfx1250, which only supports 4 groups of tensor descriptor
declare void @llvm.amdgcn.tensor.load.to.lds(<4 x i32> %D0, <8 x i32> %D1, <4 x i32> %D2, <4 x i32> %D3, <8 x i32> %D4, i32 %cpol)
declare void @llvm.amdgcn.tensor.store.from.lds(<4 x i32> %D0, <8 x i32> %D1, <4 x i32> %D2, <4 x i32> %D3, <8 x i32> %D4, i32 %cpol)

define amdgpu_ps void @tensor_load_to_lds_d4(<4 x i32> inreg %D0, <8 x i32> inreg %D1, <4 x i32> inreg %D2, <4 x i32> inreg %D3) {
; GFX1250-LABEL: tensor_load_to_lds_d4:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT:    v_nop
; GFX1250-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT:    tensor_load_to_lds s[0:3], s[4:11], s[12:15], s[16:19]
; GFX1250-NEXT:    s_endpgm
  call void @llvm.amdgcn.tensor.load.to.lds(<4 x i32> %D0, <8 x i32> %D1, <4 x i32> %D2, <4 x i32> %D3, <8 x i32> zeroinitializer, i32 0)
  ret void
}

define amdgpu_ps void @tensor_load_to_lds_d4_vector(<4 x i32> %D0, <8 x i32> %D1, <4 x i32> %D2, <4 x i32> %D3) {
; GFX1250-SDAG-LABEL: tensor_load_to_lds_d4_vector:
; GFX1250-SDAG:       ; %bb.0:
; GFX1250-SDAG-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-SDAG-NEXT:    v_nop
; GFX1250-SDAG-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s0, v4
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s1, v5
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s2, v6
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s3, v7
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s4, v8
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s8, v0
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s9, v1
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s10, v2
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s11, v3
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s5, v9
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s6, v10
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s7, v11
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s12, v12
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s13, v13
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s14, v14
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s15, v15
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s16, v16
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s17, v17
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s18, v18
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s19, v19
; GFX1250-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-SDAG-NEXT:    tensor_load_to_lds s[8:11], s[0:7], s[12:15], s[16:19]
; GFX1250-SDAG-NEXT:    s_endpgm
;
; GFX1250-GISEL-LABEL: tensor_load_to_lds_d4_vector:
; GFX1250-GISEL:       ; %bb.0:
; GFX1250-GISEL-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-GISEL-NEXT:    v_nop
; GFX1250-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s8, v0
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s9, v1
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s10, v2
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s11, v3
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s0, v4
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s1, v5
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s2, v6
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s3, v7
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s4, v8
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s5, v9
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s6, v10
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s7, v11
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s12, v12
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s13, v13
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s14, v14
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s15, v15
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s16, v16
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s17, v17
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s18, v18
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s19, v19
; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-GISEL-NEXT:    tensor_load_to_lds s[8:11], s[0:7], s[12:15], s[16:19]
; GFX1250-GISEL-NEXT:    s_endpgm
  call void @llvm.amdgcn.tensor.load.to.lds(<4 x i32> %D0, <8 x i32> %D1, <4 x i32> %D2, <4 x i32> %D3, <8 x i32> zeroinitializer, i32 0)
  ret void
}

define amdgpu_ps void @tensor_load_to_lds_d2(<4 x i32> inreg %D0, <8 x i32> inreg %D1) {
; GFX1250-LABEL: tensor_load_to_lds_d2:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT:    v_nop
; GFX1250-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT:    tensor_load_to_lds s[0:3], s[4:11] th:TH_LOAD_BYPASS scope:SCOPE_SYS
; GFX1250-NEXT:    s_endpgm
  call void @llvm.amdgcn.tensor.load.to.lds(<4 x i32> %D0, <8 x i32> %D1, <4 x i32> zeroinitializer, <4 x i32> zeroinitializer, <8 x i32> zeroinitializer, i32 27)
  ret void
}

define amdgpu_ps void @tensor_load_to_lds_d2_vector(<4 x i32> %D0, <8 x i32> %D1) {
; GFX1250-SDAG-LABEL: tensor_load_to_lds_d2_vector:
; GFX1250-SDAG:       ; %bb.0:
; GFX1250-SDAG-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-SDAG-NEXT:    v_nop
; GFX1250-SDAG-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s0, v4
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s8, v0
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s9, v1
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s10, v2
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s11, v3
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s1, v5
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s2, v6
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s3, v7
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s4, v8
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s5, v9
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s6, v10
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s7, v11
; GFX1250-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-SDAG-NEXT:    tensor_load_to_lds s[8:11], s[0:7] th:TH_LOAD_BYPASS scope:SCOPE_SYS
; GFX1250-SDAG-NEXT:    s_endpgm
;
; GFX1250-GISEL-LABEL: tensor_load_to_lds_d2_vector:
; GFX1250-GISEL:       ; %bb.0:
; GFX1250-GISEL-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-GISEL-NEXT:    v_nop
; GFX1250-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s8, v0
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s9, v1
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s10, v2
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s11, v3
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s0, v4
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s1, v5
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s2, v6
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s3, v7
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s4, v8
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s5, v9
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s6, v10
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s7, v11
; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-GISEL-NEXT:    tensor_load_to_lds s[8:11], s[0:7] th:TH_LOAD_BYPASS scope:SCOPE_SYS
; GFX1250-GISEL-NEXT:    s_endpgm
  call void @llvm.amdgcn.tensor.load.to.lds(<4 x i32> %D0, <8 x i32> %D1, <4 x i32> zeroinitializer, <4 x i32> zeroinitializer, <8 x i32> zeroinitializer, i32 27)
  ret void
}

define amdgpu_ps void @tensor_store_from_lds_d4(<4 x i32> inreg %D0, <8 x i32> inreg %D1, <4 x i32> inreg %D2, <4 x i32> inreg %D3) {
; GFX1250-LABEL: tensor_store_from_lds_d4:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT:    v_nop
; GFX1250-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT:    tensor_store_from_lds s[0:3], s[4:11], s[12:15], s[16:19] th:TH_STORE_NT_HT scope:SCOPE_DEV
; GFX1250-NEXT:    s_endpgm
  call void @llvm.amdgcn.tensor.store.from.lds(<4 x i32> %D0, <8 x i32> %D1, <4 x i32> %D2, <4 x i32> %D3, <8 x i32> zeroinitializer, i32 22)
  ret void
}

define amdgpu_ps void @tensor_store_from_lds_d4_vector(<4 x i32> %D0, <8 x i32> %D1, <4 x i32> %D2, <4 x i32> %D3) {
; GFX1250-SDAG-LABEL: tensor_store_from_lds_d4_vector:
; GFX1250-SDAG:       ; %bb.0:
; GFX1250-SDAG-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-SDAG-NEXT:    v_nop
; GFX1250-SDAG-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s0, v4
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s1, v5
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s2, v6
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s3, v7
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s4, v8
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s8, v0
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s9, v1
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s10, v2
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s11, v3
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s5, v9
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s6, v10
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s7, v11
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s12, v12
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s13, v13
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s14, v14
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s15, v15
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s16, v16
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s17, v17
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s18, v18
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s19, v19
; GFX1250-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-SDAG-NEXT:    tensor_store_from_lds s[8:11], s[0:7], s[12:15], s[16:19] th:TH_STORE_NT_HT scope:SCOPE_DEV
; GFX1250-SDAG-NEXT:    s_endpgm
;
; GFX1250-GISEL-LABEL: tensor_store_from_lds_d4_vector:
; GFX1250-GISEL:       ; %bb.0:
; GFX1250-GISEL-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-GISEL-NEXT:    v_nop
; GFX1250-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s8, v0
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s9, v1
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s10, v2
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s11, v3
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s0, v4
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s1, v5
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s2, v6
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s3, v7
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s4, v8
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s5, v9
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s6, v10
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s7, v11
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s12, v12
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s13, v13
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s14, v14
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s15, v15
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s16, v16
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s17, v17
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s18, v18
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s19, v19
; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-GISEL-NEXT:    tensor_store_from_lds s[8:11], s[0:7], s[12:15], s[16:19] th:TH_STORE_NT_HT scope:SCOPE_DEV
; GFX1250-GISEL-NEXT:    s_endpgm
  call void @llvm.amdgcn.tensor.store.from.lds(<4 x i32> %D0, <8 x i32> %D1, <4 x i32> %D2, <4 x i32> %D3, <8 x i32> zeroinitializer, i32 22)
  ret void
}

define amdgpu_ps void @tensor_store_from_lds_d2(<4 x i32> inreg %D0, <8 x i32> inreg %D1) {
; GFX1250-LABEL: tensor_store_from_lds_d2:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT:    v_nop
; GFX1250-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT:    tensor_store_from_lds s[0:3], s[4:11]
; GFX1250-NEXT:    s_endpgm
  call void @llvm.amdgcn.tensor.store.from.lds(<4 x i32> %D0, <8 x i32> %D1, <4 x i32> zeroinitializer, <4 x i32> zeroinitializer, <8 x i32> zeroinitializer, i32 0)
  ret void
}

define amdgpu_ps void @tensor_store_from_lds_d2_vector(<4 x i32> %D0, <8 x i32> %D1) {
; GFX1250-SDAG-LABEL: tensor_store_from_lds_d2_vector:
; GFX1250-SDAG:       ; %bb.0:
; GFX1250-SDAG-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-SDAG-NEXT:    v_nop
; GFX1250-SDAG-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s0, v4
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s8, v0
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s9, v1
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s10, v2
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s11, v3
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s1, v5
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s2, v6
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s3, v7
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s4, v8
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s5, v9
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s6, v10
; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s7, v11
; GFX1250-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-SDAG-NEXT:    tensor_store_from_lds s[8:11], s[0:7]
; GFX1250-SDAG-NEXT:    s_endpgm
;
; GFX1250-GISEL-LABEL: tensor_store_from_lds_d2_vector:
; GFX1250-GISEL:       ; %bb.0:
; GFX1250-GISEL-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-GISEL-NEXT:    v_nop
; GFX1250-GISEL-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s8, v0
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s9, v1
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s10, v2
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s11, v3
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s0, v4
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s1, v5
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s2, v6
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s3, v7
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s4, v8
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s5, v9
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s6, v10
; GFX1250-GISEL-NEXT:    v_readfirstlane_b32 s7, v11
; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-GISEL-NEXT:    tensor_store_from_lds s[8:11], s[0:7]
; GFX1250-GISEL-NEXT:    s_endpgm
  call void @llvm.amdgcn.tensor.store.from.lds(<4 x i32> %D0, <8 x i32> %D1, <4 x i32> zeroinitializer, <4 x i32> zeroinitializer, <8 x i32> zeroinitializer, i32 0)
  ret void
}

;=======================================================================
; It is fine to pass 5 arguments as tensor descriptor, but the fifth one
; will be ignored silently by the CodeGen for gfx1250, which only
; supports D# up to 4 groups.
;========================================================================

define amdgpu_ps void @tensor_load_to_lds_d5(<4 x i32> inreg %D0, <8 x i32> inreg %D1, <4 x i32> inreg %D2, <4 x i32> inreg %D3, <8 x i32> inreg %D4) {
; GFX1250-LABEL: tensor_load_to_lds_d5:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT:    v_nop
; GFX1250-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT:    tensor_load_to_lds s[0:3], s[4:11], s[12:15], s[16:19]
; GFX1250-NEXT:    s_endpgm
  call void @llvm.amdgcn.tensor.load.to.lds(<4 x i32> %D0, <8 x i32> %D1, <4 x i32> %D2, <4 x i32> %D3, <8 x i32> %D4, i32 0)
  ret void
}

define amdgpu_ps void @tensor_store_from_lds_d5(<4 x i32> inreg %D0, <8 x i32> inreg %D1, <4 x i32> inreg %D2, <4 x i32> inreg %D3, <8 x i32> inreg %D4) {
; GFX1250-LABEL: tensor_store_from_lds_d5:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT:    v_nop
; GFX1250-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT:    tensor_store_from_lds s[0:3], s[4:11], s[12:15], s[16:19] th:TH_STORE_NT_HT scope:SCOPE_DEV
; GFX1250-NEXT:    s_endpgm
  call void @llvm.amdgcn.tensor.store.from.lds(<4 x i32> %D0, <8 x i32> %D1, <4 x i32> %D2, <4 x i32> %D3, <8 x i32> %D4, i32 22)
  ret void
}