[llvm] 6f285d3 - [AArch64] Generate zeroing forms of certain SVE2.2 instructions (2/11) (#116828)
via llvm-commits
llvm-commits at lists.llvm.org
Mon Dec 2 02:38:13 PST 2024
Author: Momchil Velikov
Date: 2024-12-02T10:38:09Z
New Revision: 6f285d31159501050de5563b1a844a3e1ac79a03
URL: https://github.com/llvm/llvm-project/commit/6f285d31159501050de5563b1a844a3e1ac79a03
DIFF: https://github.com/llvm/llvm-project/commit/6f285d31159501050de5563b1a844a3e1ac79a03.diff
LOG: [AArch64] Generate zeroing forms of certain SVE2.2 instructions (2/11) (#116828)
SVE2.2 introduces instructions with predicated forms with zeroing of
the inactive lanes. This allows in some cases to save a `movprfx` or
a `mov` instruction when emitting code for `_x` or `_z` variants of
intrinsics.
This patch adds support for emitting the zeroing forms of certain
`FCVT`, and `BFCVT` instructions.
Added:
llvm/test/CodeGen/AArch64/zeroing-forms-fcvt-bfcvt.ll
Modified:
llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
llvm/lib/Target/AArch64/SVEInstrFormats.td
Removed:
################################################################################
diff --git a/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td b/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
index fb0eb7a80c6d72..6be5eaf221d179 100644
--- a/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
+++ b/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
@@ -4265,7 +4265,7 @@ defm TBLQ_ZZZ : sve2p1_tblq<"tblq", int_aarch64_sve_tblq>;
//===----------------------------------------------------------------------===//
let Predicates = [HasSVE2p2orSME2p2] in {
// SVE Floating-point convert precision, zeroing predicate
- defm FCVT_ZPzZ : sve_fp_z2op_p_zd_b_0<"fcvt">;
+ defm FCVT_ZPzZ : sve_fp_z2op_p_zd_b_0<"fcvt", "int_aarch64_sve_fcvt">;
// SVE2p2 floating-point convert precision down (placing odd), zeroing predicate
defm FCVTNT_ZPzZ : sve_fp_fcvtntz<"fcvtnt">;
@@ -4279,7 +4279,7 @@ let Predicates = [HasSVE2p2orSME2p2] in {
// SVE2p2 floating-point convert single-to-bf (placing odd), zeroing predicate
def BFCVTNT_ZPzZ : sve_fp_fcvt2z<0b1010, "bfcvtnt", ZPR16, ZPR32>;
// Placing corresponding
- def BFCVT_ZPzZ_StoH : sve_fp_z2op_p_zd<0b1001010, "bfcvt", ZPR32, ZPR16>;
+ defm BFCVT_ZPzZ_StoH : sve_fp_z2op_p_zd_bfcvt<"bfcvt", int_aarch64_sve_fcvt_bf16f32_v2>;
// Floating-point convert to integer, zeroing predicate
defm FCVTZS_ZPzZ : sve_fp_z2op_p_zd_d<0b0, "fcvtzs">;
diff --git a/llvm/lib/Target/AArch64/SVEInstrFormats.td b/llvm/lib/Target/AArch64/SVEInstrFormats.td
index faaaca3f28d758..1d32f2ab758528 100644
--- a/llvm/lib/Target/AArch64/SVEInstrFormats.td
+++ b/llvm/lib/Target/AArch64/SVEInstrFormats.td
@@ -576,6 +576,11 @@ multiclass SVE_3_Op_Undef_Pat<ValueType vtd, SDPatternOperator op, ValueType vt1
(inst $Op1, $Op2, $Op3)>;
}
+class SVE_3_Op_UndefZero_Pat<ValueType vtd, SDPatternOperator op, ValueType vt1,
+ ValueType vt2, ValueType vt3, Instruction inst>
+ : Pat<(vtd (op (vt1 (SVEDup0Undef)), vt2:$Op1, vt3:$Op2)),
+ (inst $Op1, $Op2)>;
+
class SVE_4_Op_Pat<ValueType vtd, SDPatternOperator op, ValueType vt1,
ValueType vt2, ValueType vt3, ValueType vt4,
Instruction inst>
@@ -3273,6 +3278,12 @@ multiclass sve_fp_z2op_p_zd_frint<bits<2> opc, string asm> {
def _D : sve_fp_z2op_p_zd<{ 0b0010, opc{1}, 1, opc{0} }, asm, ZPR64, ZPR64>;
}
+multiclass sve_fp_z2op_p_zd_bfcvt<string asm, SDPatternOperator op> {
+ def NAME : sve_fp_z2op_p_zd<0b1001010, asm, ZPR32, ZPR16>;
+
+ def : SVE_3_Op_UndefZero_Pat<nxv8bf16, op, nxv8bf16, nxv4i1, nxv4f32, !cast<Instruction>(NAME)>;
+}
+
multiclass sve_fp_z2op_p_zd_d<bit U, string asm> {
def _HtoH : sve_fp_z2op_p_zd<{ 0b011101, U }, asm, ZPR16, ZPR16>;
def _HtoS : sve_fp_z2op_p_zd<{ 0b011110, U }, asm, ZPR16, ZPR32>;
@@ -3299,13 +3310,20 @@ multiclass sve_fp_z2op_p_zd_d_flogb<string asm> {
def _D : sve_fp_z2op_p_zd<0b0011011, asm, ZPR64, ZPR64>;
}
-multiclass sve_fp_z2op_p_zd_b_0<string asm> {
+multiclass sve_fp_z2op_p_zd_b_0<string asm, string op> {
def _StoH : sve_fp_z2op_p_zd<0b1001000, asm, ZPR32, ZPR16>;
def _HtoS : sve_fp_z2op_p_zd<0b1001001, asm, ZPR16, ZPR32>;
def _DtoH : sve_fp_z2op_p_zd<0b1101000, asm, ZPR64, ZPR16>;
def _HtoD : sve_fp_z2op_p_zd<0b1101001, asm, ZPR16, ZPR64>;
def _DtoS : sve_fp_z2op_p_zd<0b1101010, asm, ZPR64, ZPR32>;
def _StoD : sve_fp_z2op_p_zd<0b1101011, asm, ZPR32, ZPR64>;
+
+ def : SVE_3_Op_UndefZero_Pat<nxv8f16, !cast<SDPatternOperator>(op # _f16f32), nxv8f16, nxv4i1, nxv4f32, !cast<Instruction>(NAME # _StoH)>;
+ def : SVE_3_Op_UndefZero_Pat<nxv8f16, !cast<SDPatternOperator>(op # _f16f64), nxv8f16, nxv2i1, nxv2f64, !cast<Instruction>(NAME # _DtoH)>;
+ def : SVE_3_Op_UndefZero_Pat<nxv4f32, !cast<SDPatternOperator>(op # _f32f64), nxv4f32, nxv2i1, nxv2f64, !cast<Instruction>(NAME # _DtoS)>;
+ def : SVE_3_Op_UndefZero_Pat<nxv4f32, !cast<SDPatternOperator>(op # _f32f16), nxv4f32, nxv4i1, nxv8f16, !cast<Instruction>(NAME # _HtoS)>;
+ def : SVE_3_Op_UndefZero_Pat<nxv2f64, !cast<SDPatternOperator>(op # _f64f16), nxv2f64, nxv2i1, nxv8f16, !cast<Instruction>(NAME # _HtoD)>;
+ def : SVE_3_Op_UndefZero_Pat<nxv2f64, !cast<SDPatternOperator>(op # _f64f32), nxv2f64, nxv2i1, nxv4f32, !cast<Instruction>(NAME # _StoD)>;
}
//===----------------------------------------------------------------------===//
diff --git a/llvm/test/CodeGen/AArch64/zeroing-forms-fcvt-bfcvt.ll b/llvm/test/CodeGen/AArch64/zeroing-forms-fcvt-bfcvt.ll
new file mode 100644
index 00000000000000..cf9ac49ca7b236
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/zeroing-forms-fcvt-bfcvt.ll
@@ -0,0 +1,330 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -mattr=+sve,+bf16 < %s | FileCheck %s
+; RUN: llc -mattr=+sve2p2,+bf16 < %s | FileCheck %s -check-prefix CHECK-2p2
+
+; RUN: llc -mattr=+sme,+bf16 -force-streaming < %s | FileCheck %s
+; RUN: llc -mattr=+sme2p2,+bf16 -force-streaming < %s | FileCheck %s -check-prefix CHECK-2p2
+
+target triple = "aarch64-linux"
+
+define <vscale x 8 x half> @test_svcvt_f16_f32_x_1(<vscale x 4 x i1> %pg, <vscale x 4 x float> %x) {
+; CHECK-LABEL: test_svcvt_f16_f32_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: fcvt z0.h, p0/m, z0.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f16_f32_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.h, p0/z, z0.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.fcvt.f16f32(<vscale x 8 x half> undef, <vscale x 4 x i1> %pg, <vscale x 4 x float> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_svcvt_f16_f32_x_2(<vscale x 4 x i1> %pg, double %z0, <vscale x 4 x float> %x) {
+; CHECK-LABEL: test_svcvt_f16_f32_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: fcvt z0.h, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f16_f32_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.h, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.fcvt.f16f32(<vscale x 8 x half> undef, <vscale x 4 x i1> %pg, <vscale x 4 x float> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_svcvt_f16_f32_z(<vscale x 4 x i1> %pg, double %z0, <vscale x 4 x float> %x) {
+; CHECK-LABEL: test_svcvt_f16_f32_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.h, #0 // =0x0
+; CHECK-NEXT: fcvt z0.h, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f16_f32_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.h, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.fcvt.f16f32(<vscale x 8 x half> zeroinitializer, <vscale x 4 x i1> %pg, <vscale x 4 x float> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x bfloat> @test_svcvt_bf16_f32_x_1(<vscale x 4 x i1> %pg, <vscale x 4 x float> %x) {
+; CHECK-LABEL: test_svcvt_bf16_f32_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: bfcvt z0.h, p0/m, z0.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_bf16_f32_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: bfcvt z0.h, p0/z, z0.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fcvt.bf16f32.v2(<vscale x 8 x bfloat> undef, <vscale x 4 x i1> %pg, <vscale x 4 x float> %x)
+ ret <vscale x 8 x bfloat> %0
+}
+
+define <vscale x 8 x bfloat> @test_svcvt_bf16_f32_x_2(<vscale x 4 x i1> %pg, double %z0, <vscale x 4 x float> %x) {
+; CHECK-LABEL: test_svcvt_bf16_f32_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: bfcvt z0.h, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_bf16_f32_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: bfcvt z0.h, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fcvt.bf16f32.v2(<vscale x 8 x bfloat> undef, <vscale x 4 x i1> %pg, <vscale x 4 x float> %x)
+ ret <vscale x 8 x bfloat> %0
+}
+
+define <vscale x 8 x bfloat> @test_svcvt_bf16_f32_z(<vscale x 4 x i1> %pg, double %z0, <vscale x 4 x float> %x) {
+; CHECK-LABEL: test_svcvt_bf16_f32_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.h, #0 // =0x0
+; CHECK-NEXT: bfcvt z0.h, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_bf16_f32_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: bfcvt z0.h, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fcvt.bf16f32.v2(<vscale x 8 x bfloat> zeroinitializer, <vscale x 4 x i1> %pg, <vscale x 4 x float> %x)
+ ret <vscale x 8 x bfloat> %0
+}
+
+define <vscale x 8 x half> @test_svcvt_f16_f64_x_1(<vscale x 2 x i1> %pg, <vscale x 2 x double> %x) {
+; CHECK-LABEL: test_svcvt_f16_f64_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: fcvt z0.h, p0/m, z0.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f16_f64_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.h, p0/z, z0.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.fcvt.f16f64(<vscale x 8 x half> undef, <vscale x 2 x i1> %pg, <vscale x 2 x double> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_svcvt_f16_f64_x_2(<vscale x 2 x i1> %pg, double %z0, <vscale x 2 x double> %x) {
+; CHECK-LABEL: test_svcvt_f16_f64_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: fcvt z0.h, p0/m, z1.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f16_f64_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.h, p0/z, z1.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.fcvt.f16f64(<vscale x 8 x half> undef, <vscale x 2 x i1> %pg, <vscale x 2 x double> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_svcvt_f16_f64_z(<vscale x 2 x i1> %pg, double %z0, <vscale x 2 x double> %x) {
+; CHECK-LABEL: test_svcvt_f16_f64_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.h, #0 // =0x0
+; CHECK-NEXT: fcvt z0.h, p0/m, z1.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f16_f64_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.h, p0/z, z1.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.fcvt.f16f64(<vscale x 8 x half> zeroinitializer, <vscale x 2 x i1> %pg, <vscale x 2 x double> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 4 x float> @test_svcvt_f32_f64_x_1(<vscale x 2 x i1> %pg, <vscale x 2 x double> %x) {
+; CHECK-LABEL: test_svcvt_f32_f64_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: fcvt z0.s, p0/m, z0.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f32_f64_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.s, p0/z, z0.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.fcvt.f32f64(<vscale x 4 x float> undef, <vscale x 2 x i1> %pg, <vscale x 2 x double> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 4 x float> @test_svcvt_f32_f64_x_2(<vscale x 2 x i1> %pg, double %z0, <vscale x 2 x double> %x) {
+; CHECK-LABEL: test_svcvt_f32_f64_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: fcvt z0.s, p0/m, z1.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f32_f64_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.s, p0/z, z1.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.fcvt.f32f64(<vscale x 4 x float> undef, <vscale x 2 x i1> %pg, <vscale x 2 x double> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 4 x float> @test_svcvt_f32_f64_z(<vscale x 2 x i1> %pg, double %z0, <vscale x 2 x double> %x) {
+; CHECK-LABEL: test_svcvt_f32_f64_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.s, #0 // =0x0
+; CHECK-NEXT: fcvt z0.s, p0/m, z1.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f32_f64_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.s, p0/z, z1.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.fcvt.f32f64(<vscale x 4 x float> zeroinitializer, <vscale x 2 x i1> %pg, <vscale x 2 x double> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 4 x float> @test_svcvt_f32_f16_x_1(<vscale x 4 x i1> %pg, <vscale x 8 x half> %x) {
+; CHECK-LABEL: test_svcvt_f32_f16_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: fcvt z0.s, p0/m, z0.h
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f32_f16_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.s, p0/z, z0.h
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.fcvt.f32f16(<vscale x 4 x float> undef, <vscale x 4 x i1> %pg, <vscale x 8 x half> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 4 x float> @test_svcvt_f32_f16_x_2(<vscale x 4 x i1> %pg, double %z0, <vscale x 8 x half> %x) {
+; CHECK-LABEL: test_svcvt_f32_f16_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: fcvt z0.s, p0/m, z1.h
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f32_f16_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.s, p0/z, z1.h
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.fcvt.f32f16(<vscale x 4 x float> undef, <vscale x 4 x i1> %pg, <vscale x 8 x half> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 4 x float> @test_svcvt_f32_f16_z(<vscale x 4 x i1> %pg, double %z0, <vscale x 8 x half> %x) {
+; CHECK-LABEL: test_svcvt_f32_f16_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.s, #0 // =0x0
+; CHECK-NEXT: fcvt z0.s, p0/m, z1.h
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f32_f16_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.s, p0/z, z1.h
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.fcvt.f32f16(<vscale x 4 x float> zeroinitializer, <vscale x 4 x i1> %pg, <vscale x 8 x half> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 2 x double> @test_svcvt_f64_f16_x_1(<vscale x 2 x i1> %pg, <vscale x 8 x half> %x) {
+; CHECK-LABEL: test_svcvt_f64_f16_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: fcvt z0.d, p0/m, z0.h
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f64_f16_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.d, p0/z, z0.h
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.fcvt.f64f16(<vscale x 2 x double> undef, <vscale x 2 x i1> %pg, <vscale x 8 x half> %x)
+ ret <vscale x 2 x double> %0
+}
+
+define <vscale x 2 x double> @test_svcvt_f64_f16_x_2(<vscale x 2 x i1> %pg, double %z0, <vscale x 8 x half> %x) {
+; CHECK-LABEL: test_svcvt_f64_f16_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: fcvt z0.d, p0/m, z1.h
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f64_f16_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.d, p0/z, z1.h
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.fcvt.f64f16(<vscale x 2 x double> undef, <vscale x 2 x i1> %pg, <vscale x 8 x half> %x)
+ ret <vscale x 2 x double> %0
+}
+
+define <vscale x 2 x double> @test_svcvt_f64_f16_z(<vscale x 2 x i1> %pg, double %z0, <vscale x 8 x half> %x) {
+; CHECK-LABEL: test_svcvt_f64_f16_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.d, #0 // =0x0
+; CHECK-NEXT: fcvt z0.d, p0/m, z1.h
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f64_f16_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.d, p0/z, z1.h
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.fcvt.f64f16(<vscale x 2 x double> zeroinitializer, <vscale x 2 x i1> %pg, <vscale x 8 x half> %x)
+ ret <vscale x 2 x double> %0
+}
+
+define <vscale x 2 x double> @test_svcvt_f64_f32_x_1(<vscale x 2 x i1> %pg, <vscale x 4 x float> %x) {
+; CHECK-LABEL: test_svcvt_f64_f32_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: fcvt z0.d, p0/m, z0.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f64_f32_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.d, p0/z, z0.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.fcvt.f64f32(<vscale x 2 x double> undef, <vscale x 2 x i1> %pg, <vscale x 4 x float> %x)
+ ret <vscale x 2 x double> %0
+}
+
+define <vscale x 2 x double> @test_svcvt_f64_f32_x_2(<vscale x 2 x i1> %pg, double %z0, <vscale x 4 x float> %x) {
+; CHECK-LABEL: test_svcvt_f64_f32_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: fcvt z0.d, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f64_f32_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.d, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.fcvt.f64f32(<vscale x 2 x double> undef, <vscale x 2 x i1> %pg, <vscale x 4 x float> %x)
+ ret <vscale x 2 x double> %0
+}
+
+define <vscale x 2 x double> @test_svcvt_f64_f32_z(<vscale x 2 x i1> %pg, double %z0, <vscale x 4 x float> %x) {
+; CHECK-LABEL: test_svcvt_f64_f32_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.d, #0 // =0x0
+; CHECK-NEXT: fcvt z0.d, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f64_f32_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: fcvt z0.d, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.fcvt.f64f32(<vscale x 2 x double> zeroinitializer, <vscale x 2 x i1> %pg, <vscale x 4 x float> %x)
+ ret <vscale x 2 x double> %0
+}
More information about the llvm-commits
mailing list