97a1f494a6
Previously this combine would undo AMDGPU's new custom legalization of wide vector shuffles into 2 element pieces. The comment also states that this combine is only done before legalization, but the case with a build_vector source was unconditional. We probably don't want to do this if the multiple uses are full scalarization of the vector, but this seems to work well enough. Scalarizing extracts should have folded out pre-legalize.
5435 lines
195 KiB
LLVM
5435 lines
195 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
|
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -check-prefixes=GFX9,GFX900 %s
|
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a < %s | FileCheck -check-prefixes=GFX9,GFX90APLUS,GFX90A %s
|
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx940 < %s | FileCheck -check-prefixes=GFX9,GFX90APLUS,GFX940 %s
|
|
|
|
|
|
define void @v_shuffle_v4f32_v2f32__u_u_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX9-LABEL: v_shuffle_v4f32_v2f32__u_u_u_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> poison
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__0_u_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__0_u_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v2, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__0_u_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v2, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__0_u_u_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: global_store_dwordx4 v2, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 0, i32 poison, i32 poison, i32 poison>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__1_u_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__1_u_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v2, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__1_u_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v2, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__1_u_u_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v2, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 1, i32 poison, i32 poison, i32 poison>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__2_u_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX9-LABEL: v_shuffle_v4f32_v2f32__2_u_u_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 2, i32 poison, i32 poison, i32 poison>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_u_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_u_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v2, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_u_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v2, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_u_u_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v2, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 poison, i32 poison, i32 poison>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_0_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_0_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_0_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_pk_mov_b32 v[0:1], v[2:3], v[0:1] op_sel:[1,0]
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_0_u_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_nop 0
|
|
; GFX940-NEXT: v_pk_mov_b32 v[0:1], v[2:3], v[0:1] op_sel:[1,0]
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 0, i32 poison, i32 poison>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_1_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_1_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_1_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_1_u_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_nop 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 1, i32 poison, i32 poison>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_2_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_2_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v2
|
|
; GFX900-NEXT: global_store_dwordx4 v3, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_2_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_2_u_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_nop 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 2, i32 poison, i32 poison>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v2, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v2, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_u_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v2, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 poison, i32 poison>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_0_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_0_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_0_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_0_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 0, i32 poison>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_1_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_1_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v3, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_1_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_1_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 1, i32 poison>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_2_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_2_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_2_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_2_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_nop 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 2, i32 poison>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_3_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_3_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v3, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_3_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v3, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_3_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v3, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 3, i32 poison>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_3_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_3_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v5, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_3_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[0:1], v[2:3] op_sel:[1,0]
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_3_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_pk_mov_b32 v[2:3], v[0:1], v[2:3] op_sel:[1,0]
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 3, i32 0>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_3_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_3_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_3_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_3_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 3, i32 1>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_3_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_3_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX900-NEXT: global_store_dwordx4 v5, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_3_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[1,0]
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_3_2:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[1,0]
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 3, i32 2>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_3_3:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__u_0_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__u_0_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__u_0_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX90A-NEXT: global_store_dwordx4 v0, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__u_0_0_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX940-NEXT: global_store_dwordx4 v0, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 poison, i32 0, i32 0, i32 0>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__0_0_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__0_0_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__0_0_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__0_0_0_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v1, v0
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v0
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> zeroinitializer
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__1_0_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__1_0_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__1_0_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_pk_mov_b32 v[0:1], v[2:3], v[2:3] op_sel:[1,0]
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__1_0_0_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_pk_mov_b32 v[0:1], v[2:3], v[2:3] op_sel:[1,0]
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 1, i32 0, i32 0, i32 0>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__2_0_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__2_0_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__2_0_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX90A-NEXT: global_store_dwordx4 v0, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__2_0_0_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX940-NEXT: global_store_dwordx4 v0, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 2, i32 0, i32 0, i32 0>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_0_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_0_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_0_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_pk_mov_b32 v[0:1], v[0:1], v[2:3] op_sel:[1,0]
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_0_0_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_pk_mov_b32 v[0:1], v[0:1], v[2:3] op_sel:[1,0]
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 0, i32 0, i32 0>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_u_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_u_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_u_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_u_0_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 poison, i32 0, i32 0>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_1_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_1_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_1_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_1_0_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 1, i32 0, i32 0>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_2_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_2_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v5, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_2_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_pk_mov_b32 v[0:1], v[0:1], v[0:1] op_sel:[1,0]
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_2_0_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_pk_mov_b32 v[0:1], v[0:1], v[0:1] op_sel:[1,0]
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 2, i32 0, i32 0>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_0_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 0, i32 0>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_u_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_u_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_u_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_u_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 poison, i32 0>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_1_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_1_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX900-NEXT: global_store_dwordx4 v5, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_1_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[2:3], v[2:3] op_sel:[1,0]
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_1_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_pk_mov_b32 v[2:3], v[2:3], v[2:3] op_sel:[1,0]
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_2_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_2_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX900-NEXT: global_store_dwordx4 v6, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_2_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX90A-NEXT: global_store_dwordx4 v6, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_2_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[4:5]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX940-NEXT: global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 2, i32 0>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__u_1_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__u_1_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__u_1_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__u_1_1_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_nop 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 poison, i32 1, i32 1, i32 1>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__0_1_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__0_1_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__0_1_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__0_1_1_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_nop 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 1, i32 1>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__1_1_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__1_1_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__1_1_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__1_1_1_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__2_1_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__2_1_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__2_1_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__2_1_1_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_nop 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 2, i32 1, i32 1, i32 1>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_1_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_1_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_1_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_1_1_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 1, i32 1, i32 1>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_u_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_u_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_u_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_u_1_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 poison, i32 1, i32 1>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_0_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_0_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx4 v5, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_0_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_pk_mov_b32 v[0:1], v[0:1], v[2:3] op_sel:[1,0]
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_0_1_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_pk_mov_b32 v[0:1], v[0:1], v[2:3] op_sel:[1,0]
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 0, i32 1, i32 1>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_2_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_2_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_2_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_pk_mov_b32 v[0:1], v[0:1], v[0:1] op_sel:[1,0]
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_2_1_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_pk_mov_b32 v[0:1], v[0:1], v[0:1] op_sel:[1,0]
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 2, i32 1, i32 1>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_1_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_u_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_u_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_u_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_u_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 poison, i32 1>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_0_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_0_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_0_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_0_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 0, i32 1>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_2_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_2_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX900-NEXT: global_store_dwordx4 v5, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_2_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v6, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_2_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[4:5]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX940-NEXT: global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 2, i32 1>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__u_2_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX9-LABEL: v_shuffle_v4f32_v2f32__u_2_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 poison, i32 2, i32 2, i32 2>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__0_2_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__0_2_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v2, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__0_2_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v2, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__0_2_2_2:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: global_store_dwordx4 v2, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 0, i32 2, i32 2, i32 2>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__1_2_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__1_2_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v2, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__1_2_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v2, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__1_2_2_2:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v2, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 1, i32 2, i32 2, i32 2>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__2_2_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX9-LABEL: v_shuffle_v4f32_v2f32__2_2_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_2_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_2_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_2_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_pk_mov_b32 v[0:1], v[2:3], v[2:3] op_sel:[1,0]
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_2_2_2:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_pk_mov_b32 v[0:1], v[2:3], v[2:3] op_sel:[1,0]
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 2, i32 2, i32 2>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_u_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_u_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx4 v1, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_u_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX90A-NEXT: global_store_dwordx4 v1, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_u_2_2:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX940-NEXT: global_store_dwordx4 v1, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 poison, i32 2, i32 2>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_0_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_0_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_0_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_pk_mov_b32 v[0:1], v[2:3], v[0:1] op_sel:[1,0]
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_0_2_2:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_pk_mov_b32 v[0:1], v[2:3], v[0:1] op_sel:[1,0]
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 0, i32 2, i32 2>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_1_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_1_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_1_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_1_2_2:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 1, i32 2, i32 2>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_2_2:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 2, i32 2>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_u_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_u_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_u_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_u_2:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 poison, i32 2>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_0_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_0_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v4
|
|
; GFX900-NEXT: global_store_dwordx4 v5, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_0_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX90A-NEXT: global_store_dwordx4 v6, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_0_2:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[4:5]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_nop 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX940-NEXT: v_mov_b32_e32 v1, v5
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX940-NEXT: global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 0, i32 2>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_1_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_1_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v4
|
|
; GFX900-NEXT: global_store_dwordx4 v5, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_1_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[2:3], v[0:1] op_sel:[1,0]
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_1_2:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_pk_mov_b32 v[2:3], v[2:3], v[0:1] op_sel:[1,0]
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 1, i32 2>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__u_3_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__u_3_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__u_3_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__u_3_3_3:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_nop 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 poison, i32 3, i32 3, i32 3>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__0_3_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__0_3_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__0_3_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__0_3_3_3:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 0, i32 3, i32 3, i32 3>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__1_3_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__1_3_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__1_3_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__1_3_3_3:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 1, i32 3, i32 3, i32 3>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__2_3_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__2_3_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__2_3_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__2_3_3_3:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_nop 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 2, i32 3, i32 3, i32 3>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_u_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_u_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_u_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v1, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_u_3_3:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX940-NEXT: global_store_dwordx4 v1, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 poison, i32 3, i32 3>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_0_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_0_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_0_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_pk_mov_b32 v[0:1], v[2:3], v[0:1] op_sel:[1,0]
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_0_3_3:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_pk_mov_b32 v[0:1], v[2:3], v[0:1] op_sel:[1,0]
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 0, i32 3, i32 3>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_1_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_1_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_1_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_1_3_3:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 1, i32 3, i32 3>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_2_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_2_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_2_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_pk_mov_b32 v[0:1], v[2:3], v[2:3] op_sel:[1,0]
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_2_3_3:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_pk_mov_b32 v[0:1], v[2:3], v[2:3] op_sel:[1,0]
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 2, i32 3, i32 3>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_u_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_u_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v2, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_u_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v2, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_u_3:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v2, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 poison, i32 3>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_0_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_0_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_0_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_0_3:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 0, i32 3>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_1_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_1_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:2]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:1]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_1_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:1]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_1_3:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX940-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 1, i32 3>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v4f32_v2f32__3_3_2_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v4f32_v2f32__3_3_2_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v4f32_v2f32__3_3_2_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: v_shuffle_v4f32_v2f32__3_3_2_3:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def v[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_nop 0
|
|
; GFX940-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX940-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX940-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1] sc0 sc1
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=v"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 2, i32 3>
|
|
store <4 x float> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__u_u_u_u() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__u_u_u_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> poison
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__0_u_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__0_u_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__0_u_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__0_u_u_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[8:9]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_nop 0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 0, i32 poison, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__1_u_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__1_u_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__1_u_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__1_u_u_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s1
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 1, i32 poison, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__2_u_u_u() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__2_u_u_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 2, i32 poison, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_u_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_u_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_u_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_u_u_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s1
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 poison, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_0_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_0_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[6:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_0_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[6:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_0_u_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[2:3]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s3
|
|
; GFX940-NEXT: s_mov_b32 s9, s0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 0, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_1_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_1_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_1_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_1_u_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[8:9]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s1
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 1, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_2_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_2_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_2_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_2_u_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s1
|
|
; GFX940-NEXT: s_mov_b32 s9, s0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 2, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_u_u() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_3_u_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_0_u() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_3_0_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 0, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_1_u() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_3_1_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s9
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_3_1_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s9
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_3_1_u:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[8:9]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s9
|
|
; GFX940-NEXT: s_mov_b32 s10, s1
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 1, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_2_u() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_3_2_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 2, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_3_u() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_3_3_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s9
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 3, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_3_0() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_3_3_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s9
|
|
; GFX900-NEXT: s_mov_b32 s10, s9
|
|
; GFX900-NEXT: s_mov_b32 s11, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_3_3_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s9
|
|
; GFX90A-NEXT: s_mov_b32 s11, s4
|
|
; GFX90A-NEXT: s_mov_b32 s8, s9
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_3_3_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[8:9]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s10, s9
|
|
; GFX940-NEXT: s_mov_b32 s11, s0
|
|
; GFX940-NEXT: s_mov_b32 s8, s9
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 3, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_3_1() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_3_3_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s9
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 3, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_3_2() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_3_3_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s9, s5
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: s_mov_b32 s11, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90APLUS-LABEL: s_shuffle_v4f32_v2f32__3_3_3_2:
|
|
; GFX90APLUS: ; %bb.0:
|
|
; GFX90APLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90APLUS-NEXT: ;;#ASMSTART
|
|
; GFX90APLUS-NEXT: ; def s[8:9]
|
|
; GFX90APLUS-NEXT: ;;#ASMEND
|
|
; GFX90APLUS-NEXT: s_mov_b32 s10, s9
|
|
; GFX90APLUS-NEXT: s_mov_b32 s11, s8
|
|
; GFX90APLUS-NEXT: s_mov_b32 s8, s9
|
|
; GFX90APLUS-NEXT: ;;#ASMSTART
|
|
; GFX90APLUS-NEXT: ; use s[8:11]
|
|
; GFX90APLUS-NEXT: ;;#ASMEND
|
|
; GFX90APLUS-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 3, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_3_3() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_3_3_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s9
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: s_mov_b32 s11, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__u_0_0_0() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__u_0_0_0:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s9, s10
|
|
; GFX9-NEXT: s_mov_b32 s11, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 poison, i32 0, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__0_0_0_0() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__0_0_0_0:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s9, s8
|
|
; GFX9-NEXT: s_mov_b32 s10, s8
|
|
; GFX9-NEXT: s_mov_b32 s11, s8
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> zeroinitializer
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__1_0_0_0() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__1_0_0_0:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: s_mov_b32 s9, s10
|
|
; GFX9-NEXT: s_mov_b32 s11, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 1, i32 0, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__2_0_0_0() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__2_0_0_0:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s9, s10
|
|
; GFX9-NEXT: s_mov_b32 s11, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 2, i32 0, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_0_0_0() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_0_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[10:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s9, s10
|
|
; GFX900-NEXT: s_mov_b32 s11, s10
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_0_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[10:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s9, s10
|
|
; GFX90A-NEXT: s_mov_b32 s11, s10
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_0_0_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[10:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s1
|
|
; GFX940-NEXT: s_mov_b32 s9, s10
|
|
; GFX940-NEXT: s_mov_b32 s11, s10
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 0, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_u_0_0() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_u_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[10:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s11, s10
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_u_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[10:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s11, s10
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_u_0_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[10:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s1
|
|
; GFX940-NEXT: s_mov_b32 s11, s10
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 poison, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_1_0_0() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_1_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[10:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s11, s10
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_1_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[10:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s11, s10
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_1_0_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[10:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s1
|
|
; GFX940-NEXT: s_mov_b32 s9, s11
|
|
; GFX940-NEXT: s_mov_b32 s11, s10
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 1, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_2_0_0() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_2_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[10:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s10
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_2_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[10:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s10
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_2_0_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[10:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s1
|
|
; GFX940-NEXT: s_mov_b32 s9, s0
|
|
; GFX940-NEXT: s_mov_b32 s11, s10
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 2, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_0_0() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_3_0_0:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s9
|
|
; GFX9-NEXT: s_mov_b32 s11, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_u_0() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_3_u_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s9
|
|
; GFX900-NEXT: s_mov_b32 s11, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_3_u_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s9
|
|
; GFX90A-NEXT: s_mov_b32 s11, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_3_u_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[8:9]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s9
|
|
; GFX940-NEXT: s_mov_b32 s11, s0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 poison, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_1_0() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_3_1_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s9
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: s_mov_b32 s11, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_3_1_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: s_mov_b32 s11, s4
|
|
; GFX90A-NEXT: s_mov_b32 s8, s9
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_3_1_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[8:9]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s10, s1
|
|
; GFX940-NEXT: s_mov_b32 s11, s0
|
|
; GFX940-NEXT: s_mov_b32 s8, s9
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_2_0() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_3_2_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[10:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s11, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_3_2_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[10:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s11, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_3_2_0:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[10:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s11
|
|
; GFX940-NEXT: s_mov_b32 s9, s11
|
|
; GFX940-NEXT: s_mov_b32 s11, s0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 2, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__u_1_1_1() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__u_1_1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: s_mov_b32 s11, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 poison, i32 1, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__0_1_1_1() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__0_1_1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: s_mov_b32 s11, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__1_1_1_1() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__1_1_1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s9
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: s_mov_b32 s11, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__2_1_1_1() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__2_1_1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: s_mov_b32 s11, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 2, i32 1, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_1_1_1() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_1_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s10, s9
|
|
; GFX900-NEXT: s_mov_b32 s11, s9
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_1_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s10, s9
|
|
; GFX90A-NEXT: s_mov_b32 s11, s9
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_1_1_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[8:9]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s1
|
|
; GFX940-NEXT: s_mov_b32 s10, s9
|
|
; GFX940-NEXT: s_mov_b32 s11, s9
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 1, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_u_1_1() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_u_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[10:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s10, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_u_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[10:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s10, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_u_1_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[10:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s1
|
|
; GFX940-NEXT: s_mov_b32 s10, s11
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 poison, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_0_1_1() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_0_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[10:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s9, s10
|
|
; GFX900-NEXT: s_mov_b32 s10, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_0_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[10:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s9, s10
|
|
; GFX90A-NEXT: s_mov_b32 s10, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_0_1_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[10:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s1
|
|
; GFX940-NEXT: s_mov_b32 s9, s10
|
|
; GFX940-NEXT: s_mov_b32 s10, s11
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 0, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_2_1_1() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_2_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[10:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_2_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[10:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_2_1_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[10:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s1
|
|
; GFX940-NEXT: s_mov_b32 s9, s0
|
|
; GFX940-NEXT: s_mov_b32 s10, s11
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 2, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_1_1() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_3_1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s9
|
|
; GFX9-NEXT: s_mov_b32 s10, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_u_1() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_3_u_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 poison, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_0_1() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_3_0_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 0, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_2_1() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_3_2_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[10:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_3_2_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[10:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_3_2_1:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[10:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s11
|
|
; GFX940-NEXT: s_mov_b32 s9, s11
|
|
; GFX940-NEXT: s_mov_b32 s11, s1
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 2, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__u_2_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__u_2_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 poison, i32 2, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__0_2_2_2() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__0_2_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__0_2_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__0_2_2_2:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[8:9]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_nop 0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 0, i32 2, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__1_2_2_2() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__1_2_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__1_2_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__1_2_2_2:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s1
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 1, i32 2, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__2_2_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__2_2_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> poison, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_2_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_2_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: s_mov_b32 s9, s10
|
|
; GFX9-NEXT: s_mov_b32 s11, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 2, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_u_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_u_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: s_mov_b32 s11, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 poison, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_0_2_2() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_0_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[10:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s10
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_0_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[10:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s10
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_0_2_2:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[10:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s11
|
|
; GFX940-NEXT: s_mov_b32 s9, s0
|
|
; GFX940-NEXT: s_mov_b32 s11, s10
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 0, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_1_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_1_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: s_mov_b32 s11, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 1, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_3_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: s_mov_b32 s11, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_u_2() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_3_u_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s9, s5
|
|
; GFX900-NEXT: s_mov_b32 s11, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_3_u_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s9, s5
|
|
; GFX90A-NEXT: s_mov_b32 s11, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_3_u_2:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s1
|
|
; GFX940-NEXT: s_mov_b32 s9, s1
|
|
; GFX940-NEXT: s_mov_b32 s11, s0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 poison, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_0_2() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_3_0_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[10:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s9, s5
|
|
; GFX900-NEXT: s_mov_b32 s11, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_3_0_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[10:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s9, s5
|
|
; GFX90A-NEXT: s_mov_b32 s11, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_3_0_2:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[10:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s1
|
|
; GFX940-NEXT: s_mov_b32 s9, s1
|
|
; GFX940-NEXT: s_mov_b32 s11, s0
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 0, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_1_2() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_3_1_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[6:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: s_mov_b32 s11, s6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_3_1_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: s_mov_b32 s11, s8
|
|
; GFX90A-NEXT: s_mov_b32 s8, s9
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_3_1_2:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[8:9]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s10, s1
|
|
; GFX940-NEXT: s_mov_b32 s11, s8
|
|
; GFX940-NEXT: s_mov_b32 s8, s9
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 1, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__u_3_3_3() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__u_3_3_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: s_mov_b32 s11, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 poison, i32 3, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__0_3_3_3() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__0_3_3_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: s_mov_b32 s10, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 0, i32 3, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__1_3_3_3() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__1_3_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s10, s9
|
|
; GFX900-NEXT: s_mov_b32 s11, s9
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__1_3_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s10, s9
|
|
; GFX90A-NEXT: s_mov_b32 s11, s9
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__1_3_3_3:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[8:9]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s1
|
|
; GFX940-NEXT: s_mov_b32 s10, s9
|
|
; GFX940-NEXT: s_mov_b32 s11, s9
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 1, i32 3, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__2_3_3_3() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__2_3_3_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: s_mov_b32 s11, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 2, i32 3, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_u_3_3() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_u_3_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: s_mov_b32 s10, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 poison, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_0_3_3() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_0_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[10:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_0_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[10:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_0_3_3:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[10:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s11
|
|
; GFX940-NEXT: s_mov_b32 s9, s0
|
|
; GFX940-NEXT: s_mov_b32 s10, s11
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 0, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_1_3_3() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_1_3_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: s_mov_b32 s10, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 1, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_2_3_3() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_2_3_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: s_mov_b32 s9, s10
|
|
; GFX9-NEXT: s_mov_b32 s10, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 2, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_u_3() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_3_u_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s9
|
|
; GFX9-NEXT: s_mov_b32 s11, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 poison, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_0_3() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_3_0_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:9]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s9
|
|
; GFX9-NEXT: s_mov_b32 s11, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 0, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_1_3() {
|
|
; GFX900-LABEL: s_shuffle_v4f32_v2f32__3_3_1_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s9
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: s_mov_b32 s11, s9
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v4f32_v2f32__3_3_1_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s9
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: s_mov_b32 s11, s9
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX940-LABEL: s_shuffle_v4f32_v2f32__3_3_1_3:
|
|
; GFX940: ; %bb.0:
|
|
; GFX940-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[8:9]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; def s[0:1]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_mov_b32 s8, s9
|
|
; GFX940-NEXT: s_mov_b32 s10, s1
|
|
; GFX940-NEXT: s_mov_b32 s11, s9
|
|
; GFX940-NEXT: ;;#ASMSTART
|
|
; GFX940-NEXT: ; use s[8:11]
|
|
; GFX940-NEXT: ;;#ASMEND
|
|
; GFX940-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 1, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v4f32_v2f32__3_3_2_3() {
|
|
; GFX9-LABEL: s_shuffle_v4f32_v2f32__3_3_2_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[10:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <2 x float> asm "; def $0", "=s"()
|
|
%vec1 = call <2 x float> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <2 x float> %vec0, <2 x float> %vec1, <4 x i32> <i32 3, i32 3, i32 2, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<4 x float> %shuf)
|
|
ret void
|
|
}
|