Skip to content

Commit 582df8e

Browse files
MDevereauMDevereau
committed
Rebase and rename ImmToTile->ImmToReg
1 parent 6b419bb commit 582df8e

File tree

5 files changed

+38
-40
lines changed

5 files changed

+38
-40
lines changed

‎clang/test/Sema/aarch64-sme2-intrinsics/acle_sme2_imm.cpp

Lines changed: 0 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -24,8 +24,6 @@ void test_ldr_str_zt(const void *const_base, void *base) __arm_streaming_compati
2424
svldr_zt(1, const_base); // expected-error {{argument value 1 is outside the valid range [0, 0]}}
2525
svstr_zt(1, base); // expected-error {{argument value 1 is outside the valid range [0, 0]}}
2626
}
27-
// REQUIRES: aarch64-registered-target
28-
#include <arm_sme_draft_spec_subject_to_change.h>
2927

3028
void test_svluti2_lane_zt_x4(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
3129
// Test Reg Offset

‎llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -327,7 +327,7 @@ class AArch64DAGToDAGISel : public SelectionDAGISel {
327327
}
328328

329329
template <unsigned BaseReg, unsigned Max>
330-
bool ImmToTile(SDValue N, SDValue &Imm) {
330+
bool ImmToReg(SDValue N, SDValue &Imm) {
331331
if (auto *CI = dyn_cast<ConstantSDNode>(N)) {
332332
uint64_t C = CI->getZExtValue();
333333

@@ -1875,7 +1875,7 @@ void AArch64DAGToDAGISel::SelectMultiVectorLuti(SDNode *Node,
18751875
return;
18761876

18771877
SDValue ZtValue;
1878-
if (!ImmToTile<AArch64::ZT0, 0>(Node->getOperand(2), ZtValue))
1878+
if (!ImmToReg<AArch64::ZT0, 0>(Node->getOperand(2), ZtValue))
18791879
return;
18801880
SDValue Ops[] = {ZtValue, Node->getOperand(3), Node->getOperand(4)};
18811881
SDLoc DL(Node);

‎llvm/lib/Target/AArch64/SMEInstrFormats.td

Lines changed: 6 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -10,12 +10,12 @@
1010
//
1111
//===----------------------------------------------------------------------===//
1212

13-
def imm_to_tile8 : ComplexPattern<i32, 1, "ImmToTile<AArch64::ZAB0, 0>", []>;
14-
def imm_to_tile16 : ComplexPattern<i32, 1, "ImmToTile<AArch64::ZAH0, 1>", []>;
15-
def imm_to_tile32 : ComplexPattern<i32, 1, "ImmToTile<AArch64::ZAS0, 3>", []>;
16-
def imm_to_tile64 : ComplexPattern<i32, 1, "ImmToTile<AArch64::ZAD0, 7>", []>;
17-
def imm_to_tile128 : ComplexPattern<i32, 1, "ImmToTile<AArch64::ZAQ0, 15>", []>;
18-
def imm_to_zt : ComplexPattern<i32, 1, "ImmToTile<AArch64::ZT0, 0>", []>;
13+
def imm_to_tile8 : ComplexPattern<i32, 1, "ImmToReg<AArch64::ZAB0, 0>", []>;
14+
def imm_to_tile16 : ComplexPattern<i32, 1, "ImmToReg<AArch64::ZAH0, 1>", []>;
15+
def imm_to_tile32 : ComplexPattern<i32, 1, "ImmToReg<AArch64::ZAS0, 3>", []>;
16+
def imm_to_tile64 : ComplexPattern<i32, 1, "ImmToReg<AArch64::ZAD0, 7>", []>;
17+
def imm_to_tile128 : ComplexPattern<i32, 1, "ImmToReg<AArch64::ZAQ0, 15>", []>;
18+
def imm_to_zt : ComplexPattern<i32, 1, "ImmToReg<AArch64::ZT0, 0>", []>;
1919

2020
def tileslice8 : ComplexPattern<i32 , 2, "SelectSMETileSlice<15, 1>", []>;
2121
def tileslice16 : ComplexPattern<i32 , 2, "SelectSMETileSlice<7, 1>", []>;

‎llvm/test/CodeGen/AArch64/sme2-intrinsics-luti2-lane-x4.ll

Lines changed: 15 additions & 15 deletions
Original file line numberDiff line numberDiff line change
@@ -12,55 +12,55 @@ define {<vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16
1212
ret {<vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>} %res
1313
}
1414

15-
define {<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>} @luti2_i16(<vscale x 8 x i16> %x) {
15+
define {<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>} @luti2_i16(<vscale x 16 x i8> %x) {
1616
; CHECK-LABEL: luti2_i16:
1717
; CHECK: // %bb.0:
1818
; CHECK-NEXT: luti2 { z0.h - z3.h }, zt0, z0[3]
1919
; CHECK-NEXT: ret
20-
%res = call {<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8i16(i32 0, <vscale x 8 x i16> %x, i32 3)
20+
%res = call {<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8i16(i32 0, <vscale x 16 x i8> %x, i32 3)
2121
ret {<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>} %res
2222
}
2323

24-
define {<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>} @luti2_i32(<vscale x 4 x i32> %x) {
24+
define {<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>} @luti2_i32(<vscale x 16 x i8> %x) {
2525
; CHECK-LABEL: luti2_i32:
2626
; CHECK: // %bb.0:
2727
; CHECK-NEXT: luti2 { z0.s - z3.s }, zt0, z0[3]
2828
; CHECK-NEXT: ret
29-
%res = call {<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv4i32(i32 0, <vscale x 4 x i32> %x, i32 3)
29+
%res = call {<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv4i32(i32 0, <vscale x 16 x i8> %x, i32 3)
3030
ret {<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>}%res
3131
}
3232

33-
define {<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>} @luti2_f16(<vscale x 8 x half> %x) {
33+
define {<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>} @luti2_f16(<vscale x 16 x i8> %x) {
3434
; CHECK-LABEL: luti2_f16:
3535
; CHECK: // %bb.0:
3636
; CHECK-NEXT: luti2 { z0.h - z3.h }, zt0, z0[3]
3737
; CHECK-NEXT: ret
38-
%res = call {<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8f16(i32 0, <vscale x 8 x half> %x, i32 3)
38+
%res = call {<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8f16(i32 0, <vscale x 16 x i8> %x, i32 3)
3939
ret {<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>} %res
4040
}
4141

42-
define {<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>} @luti2_bf16(<vscale x 8 x bfloat> %x) {
42+
define {<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>} @luti2_bf16(<vscale x 16 x i8> %x) {
4343
; CHECK-LABEL: luti2_bf16:
4444
; CHECK: // %bb.0:
4545
; CHECK-NEXT: luti2 { z0.h - z3.h }, zt0, z0[3]
4646
; CHECK-NEXT: ret
47-
%res = call {<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8bf16(i32 0, <vscale x 8 x bfloat> %x, i32 3)
47+
%res = call {<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8bf16(i32 0, <vscale x 16 x i8> %x, i32 3)
4848
ret {<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>} %res
4949
}
5050

51-
define {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} @luti2_f32(<vscale x 4 x float> %x) {
51+
define {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} @luti2_f32(<vscale x 16 x i8> %x) {
5252
; CHECK-LABEL: luti2_f32:
5353
; CHECK: // %bb.0:
5454
; CHECK-NEXT: luti2 { z0.s - z3.s }, zt0, z0[3]
5555
; CHECK-NEXT: ret
56-
%res = call {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv4f32(i32 0, <vscale x 4 x float> %x, i32 3)
56+
%res = call {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv4f32(i32 0, <vscale x 16 x i8> %x, i32 3)
5757
ret {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>}%res
5858
}
5959

6060

6161
declare {<vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv16i8(i32, <vscale x 16 x i8>, i32)
62-
declare {<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8i16(i32, <vscale x 8 x i16>, i32)
63-
declare {<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv4i32(i32, <vscale x 4 x i32>, i32)
64-
declare {<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8f16(i32, <vscale x 8 x half>, i32)
65-
declare {<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8bf16(i32, <vscale x 8 x bfloat>, i32)
66-
declare {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv4f32(i32, <vscale x 4 x float>, i32)
62+
declare {<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8i16(i32, <vscale x 16 x i8>, i32)
63+
declare {<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv4i32(i32, <vscale x 16 x i8>, i32)
64+
declare {<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8f16(i32, <vscale x 16 x i8>, i32)
65+
declare {<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8bf16(i32, <vscale x 16 x i8>, i32)
66+
declare {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} @llvm.aarch64.sme.luti2.lane.zt.x4.nxv4f32(i32, <vscale x 16 x i8>, i32)

‎llvm/test/CodeGen/AArch64/sme2-intrinsics-luti4-lane-x4.ll

Lines changed: 15 additions & 15 deletions
Original file line numberDiff line numberDiff line change
@@ -3,53 +3,53 @@
33

44
; lookup table expand one register
55

6-
define {<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>} @luti4_i16(<vscale x 8 x i16> %x) {
6+
define {<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>} @luti4_i16(<vscale x 16 x i8> %x) {
77
; CHECK-LABEL: luti4_i16:
88
; CHECK: // %bb.0:
99
; CHECK-NEXT: luti4 { z0.h - z3.h }, zt0, z0[0]
1010
; CHECK-NEXT: ret
11-
%res = call {<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8i16(i32 0, <vscale x 8 x i16> %x, i32 0)
11+
%res = call {<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8i16(i32 0, <vscale x 16 x i8> %x, i32 0)
1212
ret {<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>} %res
1313
}
1414

15-
define {<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>} @luti4_i32(<vscale x 4 x i32> %x) {
15+
define {<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>} @luti4_i32(<vscale x 16 x i8> %x) {
1616
; CHECK-LABEL: luti4_i32:
1717
; CHECK: // %bb.0:
1818
; CHECK-NEXT: luti4 { z0.s - z3.s }, zt0, z0[1]
1919
; CHECK-NEXT: ret
20-
%res = call {<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv4i32(i32 0, <vscale x 4 x i32> %x, i32 1)
20+
%res = call {<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv4i32(i32 0, <vscale x 16 x i8> %x, i32 1)
2121
ret {<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>} %res
2222
}
2323

24-
define {<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>} @luti4_bf16(<vscale x 8 x bfloat> %x) {
24+
define {<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>} @luti4_bf16(<vscale x 16 x i8> %x) {
2525
; CHECK-LABEL: luti4_bf16:
2626
; CHECK: // %bb.0:
2727
; CHECK-NEXT: luti4 { z0.h - z3.h }, zt0, z0[1]
2828
; CHECK-NEXT: ret
29-
%res = call {<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8bf16(i32 0, <vscale x 8 x bfloat> %x, i32 1)
29+
%res = call {<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8bf16(i32 0, <vscale x 16 x i8> %x, i32 1)
3030
ret {<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>} %res
3131
}
3232

33-
define {<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>} @luti4_f16(<vscale x 8 x half> %x) {
33+
define {<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>} @luti4_f16(<vscale x 16 x i8> %x) {
3434
; CHECK-LABEL: luti4_f16:
3535
; CHECK: // %bb.0:
3636
; CHECK-NEXT: luti4 { z0.h - z3.h }, zt0, z0[1]
3737
; CHECK-NEXT: ret
38-
%res = call {<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8f16(i32 0, <vscale x 8 x half> %x, i32 1)
38+
%res = call {<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8f16(i32 0, <vscale x 16 x i8> %x, i32 1)
3939
ret {<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>} %res
4040
}
4141

42-
define {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} @luti4_f32(<vscale x 4 x float> %x) {
42+
define {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} @luti4_f32(<vscale x 16 x i8> %x) {
4343
; CHECK-LABEL: luti4_f32:
4444
; CHECK: // %bb.0:
4545
; CHECK-NEXT: luti4 { z0.s - z3.s }, zt0, z0[1]
4646
; CHECK-NEXT: ret
47-
%res = call {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv4f32(i32 0, <vscale x 4 x float> %x, i32 1)
47+
%res = call {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv4f32(i32 0, <vscale x 16 x i8> %x, i32 1)
4848
ret {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} %res
4949
}
5050

51-
declare {<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8i16(i32, <vscale x 8 x i16>, i32)
52-
declare {<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv4i32(i32, <vscale x 4 x i32>, i32)
53-
declare {<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8bf16(i32, <vscale x 8 x bfloat>, i32)
54-
declare {<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8f16(i32, <vscale x 8 x half>, i32)
55-
declare {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv4f32(i32, <vscale x 4 x float>, i32)
51+
declare {<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8i16(i32, <vscale x 16 x i8>, i32)
52+
declare {<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv4i32(i32, <vscale x 16 x i8>, i32)
53+
declare {<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8bf16(i32, <vscale x 16 x i8>, i32)
54+
declare {<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8f16(i32, <vscale x 16 x i8>, i32)
55+
declare {<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>} @llvm.aarch64.sme.luti4.lane.zt.x4.nxv4f32(i32, <vscale x 16 x i8>, i32)

0 commit comments

Comments
 (0)