%a.gep = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
%b.gep = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
%out.gep = getelementptr i64, i64 addrspace(1)* %out, i32 %tid
- %a = load i64 addrspace(1)* %a.gep, align 8
- %b = load i64 addrspace(1)* %b.gep, align 8
+ %a = load i64, i64 addrspace(1)* %a.gep, align 8
+ %b = load i64, i64 addrspace(1)* %b.gep, align 8
%c = shl i64 %a, %b
%shl = shl i64 %c, 63
%a.gep = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
%b.gep = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
%out.gep = getelementptr i64, i64 addrspace(1)* %out, i32 %tid
- %a = load i64 addrspace(1)* %a.gep, align 8
- %b = load i64 addrspace(1)* %b.gep, align 8
+ %a = load i64, i64 addrspace(1)* %a.gep, align 8
+ %b = load i64, i64 addrspace(1)* %b.gep, align 8
%c = shl i64 %a, %b
%shl = shl i64 %c, 56
%a.gep = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
%b.gep = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
%out.gep = getelementptr i64, i64 addrspace(1)* %out, i32 %tid
- %a = load i64 addrspace(1)* %a.gep, align 8
- %b = load i64 addrspace(1)* %b.gep, align 8
+ %a = load i64, i64 addrspace(1)* %a.gep, align 8
+ %b = load i64, i64 addrspace(1)* %b.gep, align 8
%c = shl i64 %a, %b
%shl = shl i64 %c, 48
%a.gep = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
%b.gep = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
%out.gep = getelementptr i64, i64 addrspace(1)* %out, i32 %tid
- %a = load i64 addrspace(1)* %a.gep, align 8
- %b = load i64 addrspace(1)* %b.gep, align 8
+ %a = load i64, i64 addrspace(1)* %a.gep, align 8
+ %b = load i64, i64 addrspace(1)* %b.gep, align 8
%c = shl i64 %a, %b
%shl = shl i64 %c, 32
}
; FUNC-LABEL: {{^}}sext_in_reg_i1_in_i32_other_amount:
-; SI-NOT: {{[^@]}}bfe
-; SI: s_lshl_b32 [[REG:s[0-9]+]], {{s[0-9]+}}, 6
-; SI: s_ashr_i32 {{s[0-9]+}}, [[REG]], 7
+; SI-NOT: s_lshl
+; SI-NOT: s_ashr
+; SI: s_bfe_i32 {{s[0-9]+}}, {{s[0-9]+}}, 0x190001
; EG: MEM_{{.*}} STORE_{{.*}} [[RES:T[0-9]+\.[XYZW]]], [[ADDR:T[0-9]+.[XYZW]]]
; EG-NOT: BFE
}
; FUNC-LABEL: {{^}}sext_in_reg_v2i1_in_v2i32_other_amount:
-; SI-DAG: s_lshl_b32 [[REG0:s[0-9]+]], {{s[0-9]}}, 6
-; SI-DAG: s_ashr_i32 {{s[0-9]+}}, [[REG0]], 7
-; SI-DAG: s_lshl_b32 [[REG1:s[0-9]+]], {{s[0-9]}}, 6
-; SI-DAG: s_ashr_i32 {{s[0-9]+}}, [[REG1]], 7
+; SI-NOT: s_lshl
+; SI-NOT: s_ashr
+; SI-DAG: s_bfe_i32 {{s[0-9]+}}, {{s[0-9]+}}, 0x190001
+; SI-DAG: s_bfe_i32 {{s[0-9]+}}, {{s[0-9]+}}, 0x190001
; SI: s_endpgm
; EG: MEM_{{.*}} STORE_{{.*}} [[RES:T[0-9]+]]{{\.[XYZW][XYZW]}}, [[ADDR:T[0-9]+.[XYZW]]]
; SI: v_bfe_i32 [[EXTRACT:v[0-9]+]], {{v[0-9]+}}, 0, 8
; SI: v_bfe_i32 [[EXTRACT:v[0-9]+]], {{v[0-9]+}}, 0, 8
define void @vgpr_sext_in_reg_v4i8_to_v4i32(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %a, <4 x i32> addrspace(1)* %b) nounwind {
- %loada = load <4 x i32> addrspace(1)* %a, align 16
- %loadb = load <4 x i32> addrspace(1)* %b, align 16
+ %loada = load <4 x i32>, <4 x i32> addrspace(1)* %a, align 16
+ %loadb = load <4 x i32>, <4 x i32> addrspace(1)* %b, align 16
%c = add <4 x i32> %loada, %loadb ; add to prevent folding into extload
%shl = shl <4 x i32> %c, <i32 24, i32 24, i32 24, i32 24>
%ashr = ashr <4 x i32> %shl, <i32 24, i32 24, i32 24, i32 24>
; SI: v_bfe_i32 [[EXTRACT:v[0-9]+]], {{v[0-9]+}}, 0, 16
; SI: v_bfe_i32 [[EXTRACT:v[0-9]+]], {{v[0-9]+}}, 0, 16
define void @vgpr_sext_in_reg_v4i16_to_v4i32(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %a, <4 x i32> addrspace(1)* %b) nounwind {
- %loada = load <4 x i32> addrspace(1)* %a, align 16
- %loadb = load <4 x i32> addrspace(1)* %b, align 16
+ %loada = load <4 x i32>, <4 x i32> addrspace(1)* %a, align 16
+ %loadb = load <4 x i32>, <4 x i32> addrspace(1)* %b, align 16
%c = add <4 x i32> %loada, %loadb ; add to prevent folding into extload
%shl = shl <4 x i32> %c, <i32 16, i32 16, i32 16, i32 16>
%ashr = ashr <4 x i32> %shl, <i32 16, i32 16, i32 16, i32 16>
; SI: v_bfe_i32
; SI: buffer_store_short
define void @sext_in_reg_to_illegal_type(i16 addrspace(1)* nocapture %out, i8 addrspace(1)* nocapture %src) nounwind {
- %tmp5 = load i8 addrspace(1)* %src, align 1
+ %tmp5 = load i8, i8 addrspace(1)* %src, align 1
%tmp2 = sext i8 %tmp5 to i32
%tmp3 = tail call i32 @llvm.AMDGPU.imax(i32 %tmp2, i32 0) nounwind readnone
%tmp4 = trunc i32 %tmp3 to i8
; SI-NOT: {{[^@]}}bfe
; SI: s_endpgm
define void @bfe_0_width(i32 addrspace(1)* %out, i32 addrspace(1)* %ptr) nounwind {
- %load = load i32 addrspace(1)* %ptr, align 4
+ %load = load i32, i32 addrspace(1)* %ptr, align 4
%bfe = call i32 @llvm.AMDGPU.bfe.i32(i32 %load, i32 8, i32 0) nounwind readnone
store i32 %bfe, i32 addrspace(1)* %out, align 4
ret void
; SI-NOT: {{[^@]}}bfe
; SI: s_endpgm
define void @bfe_8_bfe_8(i32 addrspace(1)* %out, i32 addrspace(1)* %ptr) nounwind {
- %load = load i32 addrspace(1)* %ptr, align 4
+ %load = load i32, i32 addrspace(1)* %ptr, align 4
%bfe0 = call i32 @llvm.AMDGPU.bfe.i32(i32 %load, i32 0, i32 8) nounwind readnone
%bfe1 = call i32 @llvm.AMDGPU.bfe.i32(i32 %bfe0, i32 0, i32 8) nounwind readnone
store i32 %bfe1, i32 addrspace(1)* %out, align 4
; SI: v_bfe_i32 v{{[0-9]+}}, v{{[0-9]+}}, 0, 8
; SI: s_endpgm
define void @bfe_8_bfe_16(i32 addrspace(1)* %out, i32 addrspace(1)* %ptr) nounwind {
- %load = load i32 addrspace(1)* %ptr, align 4
+ %load = load i32, i32 addrspace(1)* %ptr, align 4
%bfe0 = call i32 @llvm.AMDGPU.bfe.i32(i32 %load, i32 0, i32 8) nounwind readnone
%bfe1 = call i32 @llvm.AMDGPU.bfe.i32(i32 %bfe0, i32 0, i32 16) nounwind readnone
store i32 %bfe1, i32 addrspace(1)* %out, align 4
; SI-NOT: {{[^@]}}bfe
; SI: s_endpgm
define void @bfe_16_bfe_8(i32 addrspace(1)* %out, i32 addrspace(1)* %ptr) nounwind {
- %load = load i32 addrspace(1)* %ptr, align 4
+ %load = load i32, i32 addrspace(1)* %ptr, align 4
%bfe0 = call i32 @llvm.AMDGPU.bfe.i32(i32 %load, i32 0, i32 16) nounwind readnone
%bfe1 = call i32 @llvm.AMDGPU.bfe.i32(i32 %bfe0, i32 0, i32 8) nounwind readnone
store i32 %bfe1, i32 addrspace(1)* %out, align 4
; SI-NOT: {{[^@]}}bfe
; SI: s_endpgm
define void @sextload_i8_to_i32_bfe(i32 addrspace(1)* %out, i8 addrspace(1)* %ptr) nounwind {
- %load = load i8 addrspace(1)* %ptr, align 1
+ %load = load i8, i8 addrspace(1)* %ptr, align 1
%sext = sext i8 %load to i32
%bfe = call i32 @llvm.AMDGPU.bfe.i32(i32 %sext, i32 0, i32 8) nounwind readnone
%shl = shl i32 %bfe, 24
ret void
}
-; FUNC-LABEL: {{^}}sextload_i8_to_i32_bfe_0:
; SI: .text
+; FUNC-LABEL: {{^}}sextload_i8_to_i32_bfe_0:{{.*$}}
; SI-NOT: {{[^@]}}bfe
; SI: s_endpgm
define void @sextload_i8_to_i32_bfe_0(i32 addrspace(1)* %out, i8 addrspace(1)* %ptr) nounwind {
- %load = load i8 addrspace(1)* %ptr, align 1
+ %load = load i8, i8 addrspace(1)* %ptr, align 1
%sext = sext i8 %load to i32
%bfe = call i32 @llvm.AMDGPU.bfe.i32(i32 %sext, i32 8, i32 0) nounwind readnone
%shl = shl i32 %bfe, 24
; SI: v_bfe_i32 v{{[0-9]+}}, v{{[0-9]+}}, 0, 1
; SI: s_endpgm
define void @sext_in_reg_i1_bfe_offset_0(i32 addrspace(1)* %out, i32 addrspace(1)* %in) nounwind {
- %x = load i32 addrspace(1)* %in, align 4
+ %x = load i32, i32 addrspace(1)* %in, align 4
%shl = shl i32 %x, 31
%shr = ashr i32 %shl, 31
%bfe = call i32 @llvm.AMDGPU.bfe.i32(i32 %shr, i32 0, i32 1)
; SI: v_bfe_i32 v{{[0-9]+}}, v{{[0-9]+}}, 1, 1
; SI: s_endpgm
define void @sext_in_reg_i1_bfe_offset_1(i32 addrspace(1)* %out, i32 addrspace(1)* %in) nounwind {
- %x = load i32 addrspace(1)* %in, align 4
+ %x = load i32, i32 addrspace(1)* %in, align 4
%shl = shl i32 %x, 30
%shr = ashr i32 %shl, 30
%bfe = call i32 @llvm.AMDGPU.bfe.i32(i32 %shr, i32 1, i32 1)
; FUNC-LABEL: {{^}}sext_in_reg_i2_bfe_offset_1:
; SI: buffer_load_dword
-; SI: v_lshlrev_b32_e32 v{{[0-9]+}}, 30, v{{[0-9]+}}
-; SI: v_ashrrev_i32_e32 v{{[0-9]+}}, 30, v{{[0-9]+}}
+; SI-NOT: v_lshl
+; SI-NOT: v_ashr
+; SI: v_bfe_i32 v{{[0-9]+}}, v{{[0-9]+}}, 0, 2
; SI: v_bfe_i32 v{{[0-9]+}}, v{{[0-9]+}}, 1, 2
; SI: s_endpgm
define void @sext_in_reg_i2_bfe_offset_1(i32 addrspace(1)* %out, i32 addrspace(1)* %in) nounwind {
- %x = load i32 addrspace(1)* %in, align 4
+ %x = load i32, i32 addrspace(1)* %in, align 4
%shl = shl i32 %x, 30
%shr = ashr i32 %shl, 30
%bfe = call i32 @llvm.AMDGPU.bfe.i32(i32 %shr, i32 1, i32 2)