[llvm] [AArch64] Generate zeroing forms of certain SVE2.2 instructions (5/11) (PR #116831)
Momchil Velikov via llvm-commits
llvm-commits at lists.llvm.org
Fri Dec 6 10:46:10 PST 2024
https://github.com/momchil-velikov updated https://github.com/llvm/llvm-project/pull/116831
>From 204333323db85282a6cf7ec0d49c144c73d730cb Mon Sep 17 00:00:00 2001
From: Momchil Velikov <momchil.velikov at arm.com>
Date: Tue, 19 Nov 2024 11:53:56 +0000
Subject: [PATCH] [AArch64] Generate zeroing forms of certain SVE2.2
instructions (5/11)
---
.../lib/Target/AArch64/AArch64SVEInstrInfo.td | 4 +-
llvm/lib/Target/AArch64/SVEInstrFormats.td | 11 +-
.../CodeGen/AArch64/zeroing-forms-uscvtf.ll | 658 ++++++++++++++++++
3 files changed, 670 insertions(+), 3 deletions(-)
create mode 100644 llvm/test/CodeGen/AArch64/zeroing-forms-uscvtf.ll
diff --git a/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td b/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
index 574432869471ad..757024e08dcdeb 100644
--- a/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
+++ b/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
@@ -4285,8 +4285,8 @@ let Predicates = [HasSVE2p2orSME2p2] in {
defm FCVTZS_ZPzZ : sve_fp_z2op_p_zd_d<0b0, "fcvtzs", "int_aarch64_sve_fcvtzs", AArch64fcvtzs_mt>;
defm FCVTZU_ZPzZ : sve_fp_z2op_p_zd_d<0b1, "fcvtzu", "int_aarch64_sve_fcvtzu", AArch64fcvtzu_mt>;
// Integer convert to floating-point, zeroing predicate
- defm SCVTF_ZPzZ : sve_fp_z2op_p_zd_c<0b0, "scvtf">;
- defm UCVTF_ZPzZ : sve_fp_z2op_p_zd_c<0b1, "ucvtf">;
+ defm SCVTF_ZPzZ : sve_fp_z2op_p_zd_c<0b0, "scvtf", "int_aarch64_sve_scvtf", AArch64scvtf_mt>;
+ defm UCVTF_ZPzZ : sve_fp_z2op_p_zd_c<0b1, "ucvtf", "int_aarch64_sve_ucvtf", AArch64ucvtf_mt>;
// Signed integer base 2 logarithm of fp value, zeroing predicate
defm FLOGB_ZPzZ : sve_fp_z2op_p_zd_d_flogb<"flogb">;
diff --git a/llvm/lib/Target/AArch64/SVEInstrFormats.td b/llvm/lib/Target/AArch64/SVEInstrFormats.td
index e9595ff89ddb8b..9a9e99da73e513 100644
--- a/llvm/lib/Target/AArch64/SVEInstrFormats.td
+++ b/llvm/lib/Target/AArch64/SVEInstrFormats.td
@@ -3312,7 +3312,7 @@ multiclass sve_fp_z2op_p_zd_d<bit U, string asm, string int_op, SDPatternOperato
def : SVE_1_Op_PassthruUndefZero_Pat<nxv2i64, ir_op, nxv2i1, nxv2f64, !cast<Instruction>(NAME # _DtoD)>;
}
-multiclass sve_fp_z2op_p_zd_c<bit U, string asm> {
+multiclass sve_fp_z2op_p_zd_c<bit U, string asm, string int_op, SDPatternOperator ir_op> {
def _HtoH : sve_fp_z2op_p_zd<{ 0b011001, U }, asm, ZPR16, ZPR16>;
def _StoH : sve_fp_z2op_p_zd<{ 0b011010, U }, asm, ZPR32, ZPR16>;
def _StoS : sve_fp_z2op_p_zd<{ 0b101010, U }, asm, ZPR32, ZPR32>;
@@ -3320,6 +3320,15 @@ multiclass sve_fp_z2op_p_zd_c<bit U, string asm> {
def _DtoS : sve_fp_z2op_p_zd<{ 0b111010, U }, asm, ZPR64, ZPR32>;
def _DtoH : sve_fp_z2op_p_zd<{ 0b011011, U }, asm, ZPR64, ZPR16>;
def _DtoD : sve_fp_z2op_p_zd<{ 0b111011, U }, asm, ZPR64, ZPR64>;
+
+ def : SVE_3_Op_UndefZero_Pat<nxv4f32, !cast<SDPatternOperator>(int_op # _f32i64), nxv4f32, nxv2i1, nxv2i64, !cast<Instruction>(NAME # _DtoS)>;
+ def : SVE_3_Op_UndefZero_Pat<nxv2f64, !cast<SDPatternOperator>(int_op # _f64i32), nxv2f64, nxv2i1, nxv4i32, !cast<Instruction>(NAME # _StoD)>;
+ def : SVE_3_Op_UndefZero_Pat<nxv8f16, !cast<SDPatternOperator>(int_op # _f16i32), nxv8f16, nxv4i1, nxv4i32, !cast<Instruction>(NAME # _StoH)>;
+ def : SVE_3_Op_UndefZero_Pat<nxv8f16, !cast<SDPatternOperator>(int_op # _f16i64), nxv8f16, nxv2i1, nxv2i64, !cast<Instruction>(NAME # _DtoH)>;
+
+ def : SVE_1_Op_PassthruUndefZero_Pat<nxv8f16, ir_op, nxv8i1,nxv8i16, !cast<Instruction>(NAME # _HtoH)>;
+ def : SVE_1_Op_PassthruUndefZero_Pat<nxv4f32, ir_op, nxv4i1,nxv4i32, !cast<Instruction>(NAME # _StoS)>;
+ def : SVE_1_Op_PassthruUndefZero_Pat<nxv2f64, ir_op, nxv2i1,nxv2i64, !cast<Instruction>(NAME # _DtoD)>;
}
multiclass sve_fp_z2op_p_zd_d_flogb<string asm> {
diff --git a/llvm/test/CodeGen/AArch64/zeroing-forms-uscvtf.ll b/llvm/test/CodeGen/AArch64/zeroing-forms-uscvtf.ll
new file mode 100644
index 00000000000000..1ca3b1a6e31cb9
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/zeroing-forms-uscvtf.ll
@@ -0,0 +1,658 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -mattr=+sve < %s | FileCheck %s
+; RUN: llc -mattr=+sve2p2 < %s | FileCheck %s -check-prefix CHECK-2p2
+
+; RUN: llc -mattr=+sme -force-streaming < %s | FileCheck %s
+; RUN: llc -mattr=+sme2p2 -force-streaming < %s | FileCheck %s -check-prefix CHECK-2p2
+
+target triple = "aarch64-linux"
+
+define <vscale x 4 x float> @test_scvtf_f32_s64_x_1(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_scvtf_f32_s64_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: scvtf z0.s, p0/m, z0.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_scvtf_f32_s64_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.s, p0/z, z0.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.scvtf.f32i64(<vscale x 4 x float> undef, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 4 x float> @test_scvtf_f32_s64_x_2(<vscale x 2 x i1> %pg, double %z0, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_scvtf_f32_s64_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: scvtf z0.s, p0/m, z1.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_scvtf_f32_s64_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.s, p0/z, z1.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.scvtf.f32i64(<vscale x 4 x float> undef, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 4 x float> @test_scvtf_f32_s64_z(<vscale x 2 x i1> %pg, double %z0, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_scvtf_f32_s64_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.s, #0 // =0x0
+; CHECK-NEXT: scvtf z0.s, p0/m, z1.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_scvtf_f32_s64_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.s, p0/z, z1.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.scvtf.f32i64(<vscale x 4 x float> zeroinitializer, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 2 x double> @test_scvtf_f64_s32_x_1(<vscale x 2 x i1> %pg, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_scvtf_f64_s32_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: scvtf z0.d, p0/m, z0.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_scvtf_f64_s32_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.d, p0/z, z0.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.scvtf.f64i32(<vscale x 2 x double> undef, <vscale x 2 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 2 x double> %0
+}
+
+define <vscale x 2 x double> @test_scvtf_f64_s32_x_2(<vscale x 2 x i1> %pg, double %z0, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_scvtf_f64_s32_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: scvtf z0.d, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_scvtf_f64_s32_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.d, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.scvtf.f64i32(<vscale x 2 x double> undef, <vscale x 2 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 2 x double> %0
+}
+
+define <vscale x 2 x double> @test_scvtf_f64_s32_z(<vscale x 2 x i1> %pg, double %z0, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_scvtf_f64_s32_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.d, #0 // =0x0
+; CHECK-NEXT: scvtf z0.d, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_scvtf_f64_s32_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.d, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.scvtf.f64i32(<vscale x 2 x double> zeroinitializer, <vscale x 2 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 2 x double> %0
+}
+
+define <vscale x 8 x half> @test_scvtf_f16_s32_x_1(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_scvtf_f16_s32_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: scvtf z0.h, p0/m, z0.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_scvtf_f16_s32_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.h, p0/z, z0.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.scvtf.f16i32(<vscale x 8 x half> undef, <vscale x 4 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_scvtf_f16_s32_x_2(<vscale x 4 x i1> %pg, double %z0, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_scvtf_f16_s32_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: scvtf z0.h, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_scvtf_f16_s32_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.h, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.scvtf.f16i32(<vscale x 8 x half> undef, <vscale x 4 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_scvtf_f16_s32_z(<vscale x 4 x i1> %pg, double %z0, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_scvtf_f16_s32_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.h, #0 // =0x0
+; CHECK-NEXT: scvtf z0.h, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_scvtf_f16_s32_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.h, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.scvtf.f16i32(<vscale x 8 x half> zeroinitializer, <vscale x 4 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_scvtf_f16_s64_x_1(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_scvtf_f16_s64_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: scvtf z0.h, p0/m, z0.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_scvtf_f16_s64_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.h, p0/z, z0.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.scvtf.f16i64(<vscale x 8 x half> undef, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_scvtf_f16_s64_x_2(<vscale x 2 x i1> %pg, double %z0, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_scvtf_f16_s64_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: scvtf z0.h, p0/m, z1.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_scvtf_f16_s64_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.h, p0/z, z1.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.scvtf.f16i64(<vscale x 8 x half> undef, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_scvtf_f16_s64_z(<vscale x 2 x i1> %pg, double %z0, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_scvtf_f16_s64_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.h, #0 // =0x0
+; CHECK-NEXT: scvtf z0.h, p0/m, z1.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_scvtf_f16_s64_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.h, p0/z, z1.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.scvtf.f16i64(<vscale x 8 x half> zeroinitializer, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 4 x float> @test_ucvtf_f32_u64_x_1(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_ucvtf_f32_u64_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: ucvtf z0.s, p0/m, z0.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_ucvtf_f32_u64_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.s, p0/z, z0.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.ucvtf.f32i64(<vscale x 4 x float> undef, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 4 x float> @test_ucvtf_f32_u64_x_2(<vscale x 2 x i1> %pg, double %z0, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_ucvtf_f32_u64_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: ucvtf z0.s, p0/m, z1.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_ucvtf_f32_u64_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.s, p0/z, z1.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.ucvtf.f32i64(<vscale x 4 x float> undef, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 4 x float> @test_ucvtf_f32_u64_z(<vscale x 2 x i1> %pg, double %z0, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_ucvtf_f32_u64_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.s, #0 // =0x0
+; CHECK-NEXT: ucvtf z0.s, p0/m, z1.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_ucvtf_f32_u64_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.s, p0/z, z1.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.ucvtf.f32i64(<vscale x 4 x float> zeroinitializer, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 2 x double> @test_ucvtf_f64_u32_x_1(<vscale x 2 x i1> %pg, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_ucvtf_f64_u32_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: ucvtf z0.d, p0/m, z0.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_ucvtf_f64_u32_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.d, p0/z, z0.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.ucvtf.f64i32(<vscale x 2 x double> undef, <vscale x 2 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 2 x double> %0
+}
+
+define <vscale x 2 x double> @test_ucvtf_f64_u32_x_2(<vscale x 2 x i1> %pg, double %z0, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_ucvtf_f64_u32_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: ucvtf z0.d, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_ucvtf_f64_u32_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.d, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.ucvtf.f64i32(<vscale x 2 x double> undef, <vscale x 2 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 2 x double> %0
+}
+
+define <vscale x 2 x double> @test_ucvtf_f64_u32_z(<vscale x 2 x i1> %pg, double %z0, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_ucvtf_f64_u32_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.d, #0 // =0x0
+; CHECK-NEXT: ucvtf z0.d, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_ucvtf_f64_u32_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.d, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.ucvtf.f64i32(<vscale x 2 x double> zeroinitializer, <vscale x 2 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 2 x double> %0
+}
+
+define <vscale x 8 x half> @test_ucvtf_f16_u32_x_1(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_ucvtf_f16_u32_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: ucvtf z0.h, p0/m, z0.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_ucvtf_f16_u32_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.h, p0/z, z0.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.ucvtf.f16i32(<vscale x 8 x half> undef, <vscale x 4 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_ucvtf_f16_u32_x_2(<vscale x 4 x i1> %pg, double %z0, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_ucvtf_f16_u32_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: ucvtf z0.h, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_ucvtf_f16_u32_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.h, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.ucvtf.f16i32(<vscale x 8 x half> undef, <vscale x 4 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_ucvtf_f16_u32_z(<vscale x 4 x i1> %pg, double %z0, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_ucvtf_f16_u32_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.h, #0 // =0x0
+; CHECK-NEXT: ucvtf z0.h, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_ucvtf_f16_u32_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.h, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.ucvtf.f16i32(<vscale x 8 x half> zeroinitializer, <vscale x 4 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_ucvtf_f16_u64_x_1(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_ucvtf_f16_u64_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: ucvtf z0.h, p0/m, z0.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_ucvtf_f16_u64_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.h, p0/z, z0.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.ucvtf.f16i64(<vscale x 8 x half> undef, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_ucvtf_f16_u64_x_2(<vscale x 2 x i1> %pg, double %z0, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_ucvtf_f16_u64_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: ucvtf z0.h, p0/m, z1.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_ucvtf_f16_u64_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.h, p0/z, z1.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.ucvtf.f16i64(<vscale x 8 x half> undef, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_ucvtf_f16_u64_z(<vscale x 2 x i1> %pg, double %z0, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_ucvtf_f16_u64_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.h, #0 // =0x0
+; CHECK-NEXT: ucvtf z0.h, p0/m, z1.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_ucvtf_f16_u64_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.h, p0/z, z1.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.ucvtf.f16i64(<vscale x 8 x half> zeroinitializer, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_svcvt_f16_s16_x_1(<vscale x 8 x i1> %pg, <vscale x 8 x i16> %x) {
+; CHECK-LABEL: test_svcvt_f16_s16_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: scvtf z0.h, p0/m, z0.h
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f16_s16_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.h, p0/z, z0.h
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.scvtf.nxv8f16.nxv8i16(<vscale x 8 x half> undef, <vscale x 8 x i1> %pg, <vscale x 8 x i16> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_svcvt_f16_s16_x_2(<vscale x 8 x i1> %pg, double %z0, <vscale x 8 x i16> %x) {
+; CHECK-LABEL: test_svcvt_f16_s16_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movprfx z0, z1
+; CHECK-NEXT: scvtf z0.h, p0/m, z1.h
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f16_s16_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.h, p0/z, z1.h
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.scvtf.nxv8f16.nxv8i16(<vscale x 8 x half> undef, <vscale x 8 x i1> %pg, <vscale x 8 x i16> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_svcvt_f16_s16_z(<vscale x 8 x i1> %pg, double %z0, <vscale x 8 x i16> %x) {
+; CHECK-LABEL: test_svcvt_f16_s16_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.h, #0 // =0x0
+; CHECK-NEXT: scvtf z0.h, p0/m, z1.h
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f16_s16_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.h, p0/z, z1.h
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.scvtf.nxv8f16.nxv8i16(<vscale x 8 x half> zeroinitializer, <vscale x 8 x i1> %pg, <vscale x 8 x i16> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_svcvt_f16_u16_x_1(<vscale x 8 x i1> %pg, <vscale x 8 x i16> %x) {
+; CHECK-LABEL: test_svcvt_f16_u16_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: ucvtf z0.h, p0/m, z0.h
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f16_u16_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.h, p0/z, z0.h
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.ucvtf.nxv8f16.nxv8i16(<vscale x 8 x half> undef, <vscale x 8 x i1> %pg, <vscale x 8 x i16> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_svcvt_f16_u16_x_2(<vscale x 8 x i1> %pg, double %z0, <vscale x 8 x i16> %x) {
+; CHECK-LABEL: test_svcvt_f16_u16_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movprfx z0, z1
+; CHECK-NEXT: ucvtf z0.h, p0/m, z1.h
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f16_u16_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.h, p0/z, z1.h
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.ucvtf.nxv8f16.nxv8i16(<vscale x 8 x half> undef, <vscale x 8 x i1> %pg, <vscale x 8 x i16> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 8 x half> @test_svcvt_f16_u16_z(<vscale x 8 x i1> %pg, double %z0, <vscale x 8 x i16> %x) {
+; CHECK-LABEL: test_svcvt_f16_u16_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.h, #0 // =0x0
+; CHECK-NEXT: ucvtf z0.h, p0/m, z1.h
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f16_u16_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.h, p0/z, z1.h
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 8 x half> @llvm.aarch64.sve.ucvtf.nxv8f16.nxv8i16(<vscale x 8 x half> zeroinitializer, <vscale x 8 x i1> %pg, <vscale x 8 x i16> %x)
+ ret <vscale x 8 x half> %0
+}
+
+define <vscale x 4 x float> @test_svcvt_f32_s32_x_1(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_svcvt_f32_s32_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: scvtf z0.s, p0/m, z0.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f32_s32_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.s, p0/z, z0.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.scvtf.nxv4f32.nxv4i32(<vscale x 4 x float> undef, <vscale x 4 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 4 x float> @test_svcvt_f32_s32_x_2(<vscale x 4 x i1> %pg, double %z0, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_svcvt_f32_s32_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movprfx z0, z1
+; CHECK-NEXT: scvtf z0.s, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f32_s32_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.s, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.scvtf.nxv4f32.nxv4i32(<vscale x 4 x float> undef, <vscale x 4 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 4 x float> @test_svcvt_f32_s32_z(<vscale x 4 x i1> %pg, double %z0, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_svcvt_f32_s32_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.s, #0 // =0x0
+; CHECK-NEXT: scvtf z0.s, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f32_s32_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.s, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.scvtf.nxv4f32.nxv4i32(<vscale x 4 x float> zeroinitializer, <vscale x 4 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 4 x float> @test_svcvt_f32_u32_x_1(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_svcvt_f32_u32_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: ucvtf z0.s, p0/m, z0.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f32_u32_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.s, p0/z, z0.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.ucvtf.nxv4f32.nxv4i32(<vscale x 4 x float> undef, <vscale x 4 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 4 x float> @test_svcvt_f32_u32_x_2(<vscale x 4 x i1> %pg, double %z0, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_svcvt_f32_u32_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movprfx z0, z1
+; CHECK-NEXT: ucvtf z0.s, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f32_u32_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.s, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.ucvtf.nxv4f32.nxv4i32(<vscale x 4 x float> undef, <vscale x 4 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 4 x float> @test_svcvt_f32_u32_z(<vscale x 4 x i1> %pg, double %z0, <vscale x 4 x i32> %x) {
+; CHECK-LABEL: test_svcvt_f32_u32_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.s, #0 // =0x0
+; CHECK-NEXT: ucvtf z0.s, p0/m, z1.s
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f32_u32_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.s, p0/z, z1.s
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 4 x float> @llvm.aarch64.sve.ucvtf.nxv4f32.nxv4i32(<vscale x 4 x float> zeroinitializer, <vscale x 4 x i1> %pg, <vscale x 4 x i32> %x)
+ ret <vscale x 4 x float> %0
+}
+
+define <vscale x 2 x double> @test_svcvt_f64_s64_x_1(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_svcvt_f64_s64_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: scvtf z0.d, p0/m, z0.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f64_s64_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.d, p0/z, z0.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.scvtf.nxv2f64.nxv2i64(<vscale x 2 x double> undef, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 2 x double> %0
+}
+
+define <vscale x 2 x double> @test_svcvt_f64_s64_x_2(<vscale x 2 x i1> %pg, double %z0, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_svcvt_f64_s64_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movprfx z0, z1
+; CHECK-NEXT: scvtf z0.d, p0/m, z1.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f64_s64_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.d, p0/z, z1.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.scvtf.nxv2f64.nxv2i64(<vscale x 2 x double> undef, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 2 x double> %0
+}
+
+define <vscale x 2 x double> @test_svcvt_f64_s64_z(<vscale x 2 x i1> %pg, double %z0, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_svcvt_f64_s64_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.d, #0 // =0x0
+; CHECK-NEXT: scvtf z0.d, p0/m, z1.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f64_s64_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: scvtf z0.d, p0/z, z1.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.scvtf.nxv2f64.nxv2i64(<vscale x 2 x double> zeroinitializer, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 2 x double> %0
+}
+
+define <vscale x 2 x double> @test_svcvt_f64_u64_x_1(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_svcvt_f64_u64_x_1:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: ucvtf z0.d, p0/m, z0.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f64_u64_x_1:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.d, p0/z, z0.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.ucvtf.nxv2f64.nxv2i64(<vscale x 2 x double> undef, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 2 x double> %0
+}
+
+define <vscale x 2 x double> @test_svcvt_f64_u64_x_2(<vscale x 2 x i1> %pg, double %z0, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_svcvt_f64_u64_x_2:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movprfx z0, z1
+; CHECK-NEXT: ucvtf z0.d, p0/m, z1.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f64_u64_x_2:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.d, p0/z, z1.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.ucvtf.nxv2f64.nxv2i64(<vscale x 2 x double> undef, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 2 x double> %0
+}
+
+define <vscale x 2 x double> @test_svcvt_f64_u64_z(<vscale x 2 x i1> %pg, double %z0, <vscale x 2 x i64> %x) {
+; CHECK-LABEL: test_svcvt_f64_u64_z:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov z0.d, #0 // =0x0
+; CHECK-NEXT: ucvtf z0.d, p0/m, z1.d
+; CHECK-NEXT: ret
+;
+; CHECK-2p2-LABEL: test_svcvt_f64_u64_z:
+; CHECK-2p2: // %bb.0: // %entry
+; CHECK-2p2-NEXT: ucvtf z0.d, p0/z, z1.d
+; CHECK-2p2-NEXT: ret
+entry:
+ %0 = tail call <vscale x 2 x double> @llvm.aarch64.sve.ucvtf.nxv2f64.nxv2i64(<vscale x 2 x double> zeroinitializer, <vscale x 2 x i1> %pg, <vscale x 2 x i64> %x)
+ ret <vscale x 2 x double> %0
+}
More information about the llvm-commits
mailing list