From cf4df61688385af46de58562890dae335f96694b Mon Sep 17 00:00:00 2001
From: Paul Walker <paul.walker@arm.com>
Date: Fri, 10 Feb 2023 12:25:26 +0000
Subject: [PATCH] [SVE] Add intrinsics for floating-point operations that
 explicitly undefine the result for inactive lanes.

This patch is the floating-point equivalent of D141937.

Depends on D143764.

Differential Revision: https://reviews.llvm.org/D143765
---
 llvm/include/llvm/IR/IntrinsicsAArch64.td          |  16 +-
 llvm/lib/Target/AArch64/AArch64ISelLowering.cpp    |  32 +-
 llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td     |  15 +-
 .../AArch64/sve-intrinsics-fp-arith-undef.ll       | 822 +++++++++++++++++++++
 4 files changed, 877 insertions(+), 8 deletions(-)
 create mode 100644 llvm/test/CodeGen/AArch64/sve-intrinsics-fp-arith-undef.ll

diff --git a/llvm/include/llvm/IR/IntrinsicsAArch64.td b/llvm/include/llvm/IR/IntrinsicsAArch64.td
index edf6fdd..b2ae369 100644
--- a/llvm/include/llvm/IR/IntrinsicsAArch64.td
+++ b/llvm/include/llvm/IR/IntrinsicsAArch64.td
@@ -1931,31 +1931,44 @@ def int_aarch64_sve_whilehi : AdvSIMD_SVE_WHILE_Intrinsic;
 //
 
 def int_aarch64_sve_fabd       : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_fabd_u     : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fabs       : AdvSIMD_Merged1VectorArg_Intrinsic;
 def int_aarch64_sve_fadd       : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_fadd_u     : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fcadd      : AdvSIMD_SVE_CADD_Intrinsic;
 def int_aarch64_sve_fcmla      : AdvSIMD_SVE_CMLA_Intrinsic;
 def int_aarch64_sve_fcmla_lane : AdvSIMD_SVE_CMLA_LANE_Intrinsic;
 def int_aarch64_sve_fdiv       : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_fdiv_u     : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fdivr      : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fexpa_x    : AdvSIMD_SVE_EXPA_Intrinsic;
 def int_aarch64_sve_fmad       : AdvSIMD_Pred3VectorArg_Intrinsic;
 def int_aarch64_sve_fmax       : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_fmax_u     : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fmaxnm     : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_fmaxnm_u   : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fmin       : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_fmin_u     : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fminnm     : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_fminnm_u   : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fmla       : AdvSIMD_Pred3VectorArg_Intrinsic;
 def int_aarch64_sve_fmla_lane  : AdvSIMD_3VectorArgIndexed_Intrinsic;
+def int_aarch64_sve_fmla_u     : AdvSIMD_Pred3VectorArg_Intrinsic;
 def int_aarch64_sve_fmls       : AdvSIMD_Pred3VectorArg_Intrinsic;
 def int_aarch64_sve_fmls_lane  : AdvSIMD_3VectorArgIndexed_Intrinsic;
+def int_aarch64_sve_fmls_u     : AdvSIMD_Pred3VectorArg_Intrinsic;
 def int_aarch64_sve_fmsb       : AdvSIMD_Pred3VectorArg_Intrinsic;
 def int_aarch64_sve_fmul       : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_fmul_lane  : AdvSIMD_2VectorArgIndexed_Intrinsic;
+def int_aarch64_sve_fmul_u     : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fmulx      : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_fmulx_u    : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fneg       : AdvSIMD_Merged1VectorArg_Intrinsic;
-def int_aarch64_sve_fmul_lane  : AdvSIMD_2VectorArgIndexed_Intrinsic;
 def int_aarch64_sve_fnmad      : AdvSIMD_Pred3VectorArg_Intrinsic;
 def int_aarch64_sve_fnmla      : AdvSIMD_Pred3VectorArg_Intrinsic;
+def int_aarch64_sve_fnmla_u    : AdvSIMD_Pred3VectorArg_Intrinsic;
 def int_aarch64_sve_fnmls      : AdvSIMD_Pred3VectorArg_Intrinsic;
+def int_aarch64_sve_fnmls_u    : AdvSIMD_Pred3VectorArg_Intrinsic;
 def int_aarch64_sve_fnmsb      : AdvSIMD_Pred3VectorArg_Intrinsic;
 def int_aarch64_sve_frecpe_x   : AdvSIMD_1VectorArg_Intrinsic;
 def int_aarch64_sve_frecps_x   : AdvSIMD_2VectorArg_Intrinsic;
@@ -1972,6 +1985,7 @@ def int_aarch64_sve_frsqrts_x  : AdvSIMD_2VectorArg_Intrinsic;
 def int_aarch64_sve_fscale     : AdvSIMD_SVE_SCALE_Intrinsic;
 def int_aarch64_sve_fsqrt      : AdvSIMD_Merged1VectorArg_Intrinsic;
 def int_aarch64_sve_fsub       : AdvSIMD_Pred2VectorArg_Intrinsic;
+def int_aarch64_sve_fsub_u     : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_fsubr      : AdvSIMD_Pred2VectorArg_Intrinsic;
 def int_aarch64_sve_ftmad_x    : AdvSIMD_2VectorArgIndexed_Intrinsic;
 def int_aarch64_sve_ftsmul_x   : AdvSIMD_SVE_TSMUL_Intrinsic;
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index c69b164..3a9372e 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -18337,10 +18337,38 @@ static SDValue performIntrinsicCombine(SDNode *N,
                        N->getOperand(1), N->getOperand(2), N->getOperand(3));
   case Intrinsic::aarch64_sve_fadd:
     return convertMergedOpToPredOp(N, AArch64ISD::FADD_PRED, DAG);
-  case Intrinsic::aarch64_sve_fsub:
-    return convertMergedOpToPredOp(N, AArch64ISD::FSUB_PRED, DAG);
+  case Intrinsic::aarch64_sve_fadd_u:
+    return DAG.getNode(AArch64ISD::FADD_PRED, SDLoc(N), N->getValueType(0),
+                       N->getOperand(1), N->getOperand(2), N->getOperand(3));
+  case Intrinsic::aarch64_sve_fdiv_u:
+    return DAG.getNode(AArch64ISD::FDIV_PRED, SDLoc(N), N->getValueType(0),
+                       N->getOperand(1), N->getOperand(2), N->getOperand(3));
+  case Intrinsic::aarch64_sve_fmax_u:
+    return DAG.getNode(AArch64ISD::FMAX_PRED, SDLoc(N), N->getValueType(0),
+                       N->getOperand(1), N->getOperand(2), N->getOperand(3));
+  case Intrinsic::aarch64_sve_fmaxnm_u:
+    return DAG.getNode(AArch64ISD::FMAXNM_PRED, SDLoc(N), N->getValueType(0),
+                       N->getOperand(1), N->getOperand(2), N->getOperand(3));
+  case Intrinsic::aarch64_sve_fmla_u:
+    return DAG.getNode(AArch64ISD::FMA_PRED, SDLoc(N), N->getValueType(0),
+                       N->getOperand(1), N->getOperand(3), N->getOperand(4),
+                       N->getOperand(2));
+  case Intrinsic::aarch64_sve_fmin_u:
+    return DAG.getNode(AArch64ISD::FMIN_PRED, SDLoc(N), N->getValueType(0),
+                       N->getOperand(1), N->getOperand(2), N->getOperand(3));
+  case Intrinsic::aarch64_sve_fminnm_u:
+    return DAG.getNode(AArch64ISD::FMINNM_PRED, SDLoc(N), N->getValueType(0),
+                       N->getOperand(1), N->getOperand(2), N->getOperand(3));
   case Intrinsic::aarch64_sve_fmul:
     return convertMergedOpToPredOp(N, AArch64ISD::FMUL_PRED, DAG);
+  case Intrinsic::aarch64_sve_fmul_u:
+    return DAG.getNode(AArch64ISD::FMUL_PRED, SDLoc(N), N->getValueType(0),
+                       N->getOperand(1), N->getOperand(2), N->getOperand(3));
+  case Intrinsic::aarch64_sve_fsub:
+    return convertMergedOpToPredOp(N, AArch64ISD::FSUB_PRED, DAG);
+  case Intrinsic::aarch64_sve_fsub_u:
+    return DAG.getNode(AArch64ISD::FSUB_PRED, SDLoc(N), N->getValueType(0),
+                       N->getOperand(1), N->getOperand(2), N->getOperand(3));
   case Intrinsic::aarch64_sve_add:
     return convertMergedOpToPredOp(N, ISD::ADD, DAG, true);
   case Intrinsic::aarch64_sve_add_u:
diff --git a/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td b/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
index c298736..8ac7559 100644
--- a/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
+++ b/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
@@ -352,24 +352,28 @@ def AArch64fmul_p_oneuse : PatFrag<(ops node:$pred, node:$src1, node:$src2),
 }]>;
 
 
-def AArch64fabd_p : PatFrag<(ops node:$pg, node:$op1, node:$op2),
-                            (AArch64fabs_mt node:$pg, (AArch64fsub_p node:$pg, node:$op1, node:$op2), undef)>;
+def AArch64fabd_p : PatFrags<(ops node:$pg, node:$op1, node:$op2),
+                             [(int_aarch64_sve_fabd_u node:$pg, node:$op1, node:$op2),
+                              (AArch64fabs_mt node:$pg, (AArch64fsub_p node:$pg, node:$op1, node:$op2), undef)]>;
 
 def AArch64fmla_p : PatFrags<(ops node:$pg, node:$za, node:$zn, node:$zm),
                              [(AArch64fma_p node:$pg, node:$zn, node:$zm, node:$za),
                               (vselect node:$pg, (AArch64fma_p (AArch64ptrue 31), node:$zn, node:$zm, node:$za), node:$za)]>;
 
 def AArch64fmls_p : PatFrags<(ops node:$pg, node:$za, node:$zn, node:$zm),
-                             [(AArch64fma_p node:$pg, (AArch64fneg_mt node:$pg, node:$zn, (undef)), node:$zm, node:$za),
+                             [(int_aarch64_sve_fmls_u node:$pg, node:$za, node:$zn, node:$zm),
+                              (AArch64fma_p node:$pg, (AArch64fneg_mt node:$pg, node:$zn, (undef)), node:$zm, node:$za),
                               (AArch64fma_p node:$pg, node:$zm, (AArch64fneg_mt node:$pg, node:$zn, (undef)), node:$za),
                               (vselect node:$pg, (AArch64fma_p (AArch64ptrue 31), (AArch64fneg_mt (AArch64ptrue 31), node:$zn, (undef)), node:$zm, node:$za), node:$za)]>;
 
 def AArch64fnmla_p : PatFrags<(ops node:$pg, node:$za, node:$zn, node:$zm),
-                              [(AArch64fma_p node:$pg, (AArch64fneg_mt node:$pg, node:$zn, (undef)), node:$zm, (AArch64fneg_mt node:$pg, node:$za, (undef))),
+                              [(int_aarch64_sve_fnmla_u node:$pg, node:$za, node:$zn, node:$zm),
+                               (AArch64fma_p node:$pg, (AArch64fneg_mt node:$pg, node:$zn, (undef)), node:$zm, (AArch64fneg_mt node:$pg, node:$za, (undef))),
                                (AArch64fneg_mt_nsz node:$pg, (AArch64fma_p node:$pg, node:$zn, node:$zm, node:$za), (undef))]>;
 
 def AArch64fnmls_p : PatFrags<(ops node:$pg, node:$za, node:$zn, node:$zm),
-                              [(AArch64fma_p node:$pg, node:$zn, node:$zm, (AArch64fneg_mt node:$pg, node:$za, (undef)))]>;
+                              [(int_aarch64_sve_fnmls_u node:$pg, node:$za, node:$zn, node:$zm),
+                               (AArch64fma_p node:$pg, node:$zn, node:$zm, (AArch64fneg_mt node:$pg, node:$za, (undef)))]>;
 
 def AArch64fsubr_p : PatFrag<(ops node:$pg, node:$op1, node:$op2),
                              (AArch64fsub_p node:$pg, node:$op2, node:$op1)>;
@@ -623,6 +627,7 @@ let Predicates = [HasSVEorSME] in {
   defm FMAX_ZPZZ   : sve_fp_bin_pred_hfd<AArch64fmax_p>;
   defm FMIN_ZPZZ   : sve_fp_bin_pred_hfd<AArch64fmin_p>;
   defm FABD_ZPZZ   : sve_fp_bin_pred_hfd<AArch64fabd_p>;
+  defm FMULX_ZPZZ  : sve_fp_bin_pred_hfd<int_aarch64_sve_fmulx_u>;
   defm FDIV_ZPZZ   : sve_fp_bin_pred_hfd<AArch64fdiv_p>;
 } // End HasSVEorSME
 
diff --git a/llvm/test/CodeGen/AArch64/sve-intrinsics-fp-arith-undef.ll b/llvm/test/CodeGen/AArch64/sve-intrinsics-fp-arith-undef.ll
new file mode 100644
index 0000000..a77bcdf
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/sve-intrinsics-fp-arith-undef.ll
@@ -0,0 +1,822 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve < %s | FileCheck %s
+
+;
+; FABD
+;
+
+define <vscale x 8 x half> @fabd_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
+; CHECK-LABEL: fabd_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fabd z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fabd.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                   <vscale x 8 x half> %a,
+                                                                   <vscale x 8 x half> %b)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fabd_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
+; CHECK-LABEL: fabd_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fabd z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fabd.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                    <vscale x 4 x float> %a,
+                                                                    <vscale x 4 x float> %b)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fabd_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
+; CHECK-LABEL: fabd_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fabd z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fabd.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                     <vscale x 2 x double> %a,
+                                                                     <vscale x 2 x double> %b)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FADD
+;
+
+define <vscale x 8 x half> @fadd_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
+; CHECK-LABEL: fadd_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fadd z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fadd.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                   <vscale x 8 x half> %a,
+                                                                   <vscale x 8 x half> %b)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fadd_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
+; CHECK-LABEL: fadd_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fadd z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fadd.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                    <vscale x 4 x float> %a,
+                                                                    <vscale x 4 x float> %b)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fadd_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
+; CHECK-LABEL: fadd_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fadd z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fadd.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                     <vscale x 2 x double> %a,
+                                                                     <vscale x 2 x double> %b)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FDIV
+;
+
+define <vscale x 8 x half> @fdiv_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
+; CHECK-LABEL: fdiv_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fdiv z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fdiv.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                   <vscale x 8 x half> %a,
+                                                                   <vscale x 8 x half> %b)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fdiv_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
+; CHECK-LABEL: fdiv_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fdiv z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fdiv.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                    <vscale x 4 x float> %a,
+                                                                    <vscale x 4 x float> %b)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fdiv_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
+; CHECK-LABEL: fdiv_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fdiv z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fdiv.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                     <vscale x 2 x double> %a,
+                                                                     <vscale x 2 x double> %b)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FDIVR
+;
+
+define <vscale x 8 x half> @fdivr_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
+; CHECK-LABEL: fdivr_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fdivr z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fdiv.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                   <vscale x 8 x half> %b,
+                                                                   <vscale x 8 x half> %a)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fdivr_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
+; CHECK-LABEL: fdivr_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fdivr z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fdiv.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                    <vscale x 4 x float> %b,
+                                                                    <vscale x 4 x float> %a)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fdivr_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
+; CHECK-LABEL: fdivr_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fdivr z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fdiv.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                     <vscale x 2 x double> %b,
+                                                                     <vscale x 2 x double> %a)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FMAD
+;
+
+define <vscale x 8 x half> @fmad_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fmad_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmad z0.h, p0/m, z1.h, z2.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmla.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                   <vscale x 8 x half> %c,
+                                                                   <vscale x 8 x half> %a,
+                                                                   <vscale x 8 x half> %b)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fmad_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fmad_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmad z0.s, p0/m, z1.s, z2.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmla.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                    <vscale x 4 x float> %c,
+                                                                    <vscale x 4 x float> %a,
+                                                                    <vscale x 4 x float> %b)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fmad_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fmad_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmad z0.d, p0/m, z1.d, z2.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmla.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                     <vscale x 2 x double> %c,
+                                                                     <vscale x 2 x double> %a,
+                                                                     <vscale x 2 x double> %b)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FMAX
+;
+
+define <vscale x 8 x half> @fmax_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
+; CHECK-LABEL: fmax_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmax z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmax.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                   <vscale x 8 x half> %a,
+                                                                   <vscale x 8 x half> %b)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fmax_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
+; CHECK-LABEL: fmax_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmax z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmax.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                    <vscale x 4 x float> %a,
+                                                                    <vscale x 4 x float> %b)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fmax_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
+; CHECK-LABEL: fmax_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmax z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmax.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                     <vscale x 2 x double> %a,
+                                                                     <vscale x 2 x double> %b)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FMAXNM
+;
+
+define <vscale x 8 x half> @fmaxnm_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
+; CHECK-LABEL: fmaxnm_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmaxnm z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmaxnm.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                     <vscale x 8 x half> %a,
+                                                                     <vscale x 8 x half> %b)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fmaxnm_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
+; CHECK-LABEL: fmaxnm_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmaxnm z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmaxnm.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                      <vscale x 4 x float> %a,
+                                                                      <vscale x 4 x float> %b)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fmaxnm_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
+; CHECK-LABEL: fmaxnm_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmaxnm z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmaxnm.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                       <vscale x 2 x double> %a,
+                                                                       <vscale x 2 x double> %b)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FMIN
+;
+
+define <vscale x 8 x half> @fmin_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
+; CHECK-LABEL: fmin_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmin z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmin.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                   <vscale x 8 x half> %a,
+                                                                   <vscale x 8 x half> %b)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fmin_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
+; CHECK-LABEL: fmin_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmin z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmin.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                    <vscale x 4 x float> %a,
+                                                                    <vscale x 4 x float> %b)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fmin_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
+; CHECK-LABEL: fmin_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmin z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmin.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                     <vscale x 2 x double> %a,
+                                                                     <vscale x 2 x double> %b)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FMINNM
+;
+
+define <vscale x 8 x half> @fminnm_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
+; CHECK-LABEL: fminnm_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fminnm z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fminnm.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                     <vscale x 8 x half> %a,
+                                                                     <vscale x 8 x half> %b)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fminnm_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
+; CHECK-LABEL: fminnm_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fminnm z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fminnm.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                      <vscale x 4 x float> %a,
+                                                                      <vscale x 4 x float> %b)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fminnm_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
+; CHECK-LABEL: fminnm_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fminnm z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fminnm.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                       <vscale x 2 x double> %a,
+                                                                       <vscale x 2 x double> %b)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FMLA
+;
+
+define <vscale x 8 x half> @fmla_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fmla_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmla z0.h, p0/m, z1.h, z2.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmla.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                   <vscale x 8 x half> %a,
+                                                                   <vscale x 8 x half> %b,
+                                                                   <vscale x 8 x half> %c)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fmla_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fmla_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmla z0.s, p0/m, z1.s, z2.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmla.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                    <vscale x 4 x float> %a,
+                                                                    <vscale x 4 x float> %b,
+                                                                    <vscale x 4 x float> %c)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fmla_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fmla_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmla z0.d, p0/m, z1.d, z2.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmla.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                     <vscale x 2 x double> %a,
+                                                                     <vscale x 2 x double> %b,
+                                                                     <vscale x 2 x double> %c)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FMLS
+;
+
+define <vscale x 8 x half> @fmls_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fmls_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmls z0.h, p0/m, z1.h, z2.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmls.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                   <vscale x 8 x half> %a,
+                                                                   <vscale x 8 x half> %b,
+                                                                   <vscale x 8 x half> %c)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fmls_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fmls_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmls z0.s, p0/m, z1.s, z2.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmls.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                    <vscale x 4 x float> %a,
+                                                                    <vscale x 4 x float> %b,
+                                                                    <vscale x 4 x float> %c)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fmls_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fmls_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmls z0.d, p0/m, z1.d, z2.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmls.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                     <vscale x 2 x double> %a,
+                                                                     <vscale x 2 x double> %b,
+                                                                     <vscale x 2 x double> %c)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FMSB
+;
+
+define <vscale x 8 x half> @fmsb_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fmsb_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmsb z0.h, p0/m, z1.h, z2.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmls.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                   <vscale x 8 x half> %c,
+                                                                   <vscale x 8 x half> %a,
+                                                                   <vscale x 8 x half> %b)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fmsb_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fmsb_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmsb z0.s, p0/m, z1.s, z2.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmls.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                    <vscale x 4 x float> %c,
+                                                                    <vscale x 4 x float> %a,
+                                                                    <vscale x 4 x float> %b)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fmsb_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fmsb_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmsb z0.d, p0/m, z1.d, z2.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmls.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                     <vscale x 2 x double> %c,
+                                                                     <vscale x 2 x double> %a,
+                                                                     <vscale x 2 x double> %b)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FMUL
+;
+
+define <vscale x 8 x half> @fmul_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
+; CHECK-LABEL: fmul_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmul z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmul.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                   <vscale x 8 x half> %a,
+                                                                   <vscale x 8 x half> %b)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fmul_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
+; CHECK-LABEL: fmul_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmul z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmul.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                    <vscale x 4 x float> %a,
+                                                                    <vscale x 4 x float> %b)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fmul_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
+; CHECK-LABEL: fmul_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmul z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmul.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                     <vscale x 2 x double> %a,
+                                                                     <vscale x 2 x double> %b)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FMULX
+;
+
+define <vscale x 8 x half> @fmulx_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
+; CHECK-LABEL: fmulx_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmulx z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmulx.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                    <vscale x 8 x half> %a,
+                                                                    <vscale x 8 x half> %b)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fmulx_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
+; CHECK-LABEL: fmulx_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmulx z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmulx.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                     <vscale x 4 x float> %a,
+                                                                     <vscale x 4 x float> %b)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fmulx_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
+; CHECK-LABEL: fmulx_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmulx z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmulx.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                      <vscale x 2 x double> %a,
+                                                                      <vscale x 2 x double> %b)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FNMAD
+;
+
+define <vscale x 8 x half> @fnmad_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fnmad_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fnmad z0.h, p0/m, z1.h, z2.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fnmla.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                    <vscale x 8 x half> %c,
+                                                                    <vscale x 8 x half> %a,
+                                                                    <vscale x 8 x half> %b)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fnmad_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fnmad_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fnmad z0.s, p0/m, z1.s, z2.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fnmla.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                     <vscale x 4 x float> %c,
+                                                                     <vscale x 4 x float> %a,
+                                                                     <vscale x 4 x float> %b)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fnmad_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fnmad_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fnmad z0.d, p0/m, z1.d, z2.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fnmla.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                      <vscale x 2 x double> %c,
+                                                                      <vscale x 2 x double> %a,
+                                                                      <vscale x 2 x double> %b)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FNMLA
+;
+
+define <vscale x 8 x half> @fnmla_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fnmla_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fnmla z0.h, p0/m, z1.h, z2.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fnmla.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                    <vscale x 8 x half> %a,
+                                                                    <vscale x 8 x half> %b,
+                                                                    <vscale x 8 x half> %c)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fnmla_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fnmla_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fnmla z0.s, p0/m, z1.s, z2.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fnmla.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                     <vscale x 4 x float> %a,
+                                                                     <vscale x 4 x float> %b,
+                                                                     <vscale x 4 x float> %c)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fnmla_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fnmla_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fnmla z0.d, p0/m, z1.d, z2.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fnmla.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                      <vscale x 2 x double> %a,
+                                                                      <vscale x 2 x double> %b,
+                                                                      <vscale x 2 x double> %c)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FNMLS
+;
+
+define <vscale x 8 x half> @fnmls_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fnmls_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fnmls z0.h, p0/m, z1.h, z2.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fnmls.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                    <vscale x 8 x half> %a,
+                                                                    <vscale x 8 x half> %b,
+                                                                    <vscale x 8 x half> %c)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fnmls_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fnmls_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fnmls z0.s, p0/m, z1.s, z2.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fnmls.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                     <vscale x 4 x float> %a,
+                                                                     <vscale x 4 x float> %b,
+                                                                     <vscale x 4 x float> %c)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fnmls_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fnmls_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fnmls z0.d, p0/m, z1.d, z2.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fnmls.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                      <vscale x 2 x double> %a,
+                                                                      <vscale x 2 x double> %b,
+                                                                      <vscale x 2 x double> %c)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FNMSB
+;
+
+define <vscale x 8 x half> @fnmsb_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
+; CHECK-LABEL: fnmsb_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fnmsb z0.h, p0/m, z1.h, z2.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fnmls.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                    <vscale x 8 x half> %c,
+                                                                    <vscale x 8 x half> %a,
+                                                                    <vscale x 8 x half> %b)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fnmsb_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
+; CHECK-LABEL: fnmsb_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fnmsb z0.s, p0/m, z1.s, z2.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fnmls.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                     <vscale x 4 x float> %c,
+                                                                     <vscale x 4 x float> %a,
+                                                                     <vscale x 4 x float> %b)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fnmsb_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
+; CHECK-LABEL: fnmsb_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fnmsb z0.d, p0/m, z1.d, z2.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fnmls.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                      <vscale x 2 x double> %c,
+                                                                      <vscale x 2 x double> %a,
+                                                                      <vscale x 2 x double> %b)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FSUB
+;
+
+define <vscale x 8 x half> @fsub_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
+; CHECK-LABEL: fsub_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fsub z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fsub.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                   <vscale x 8 x half> %a,
+                                                                   <vscale x 8 x half> %b)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fsub_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
+; CHECK-LABEL: fsub_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fsub z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fsub.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                    <vscale x 4 x float> %a,
+                                                                    <vscale x 4 x float> %b)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fsub_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
+; CHECK-LABEL: fsub_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fsub z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fsub.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                     <vscale x 2 x double> %a,
+                                                                     <vscale x 2 x double> %b)
+  ret <vscale x 2 x double> %out
+}
+
+;
+; FSUBR
+;
+
+define <vscale x 8 x half> @fsubr_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
+; CHECK-LABEL: fsubr_h:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fsubr z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    ret
+  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fsub.u.nxv8f16(<vscale x 8 x i1> %pg,
+                                                                   <vscale x 8 x half> %b,
+                                                                   <vscale x 8 x half> %a)
+  ret <vscale x 8 x half> %out
+}
+
+define <vscale x 4 x float> @fsubr_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
+; CHECK-LABEL: fsubr_s:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fsubr z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    ret
+  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fsub.u.nxv4f32(<vscale x 4 x i1> %pg,
+                                                                    <vscale x 4 x float> %b,
+                                                                    <vscale x 4 x float> %a)
+  ret <vscale x 4 x float> %out
+}
+
+define <vscale x 2 x double> @fsubr_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
+; CHECK-LABEL: fsubr_d:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fsubr z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    ret
+  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fsub.u.nxv2f64(<vscale x 2 x i1> %pg,
+                                                                     <vscale x 2 x double> %b,
+                                                                     <vscale x 2 x double> %a)
+  ret <vscale x 2 x double> %out
+}
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fabd.u.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fabd.u.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fabd.u.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fadd.u.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fadd.u.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fadd.u.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fdiv.u.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fdiv.u.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fdiv.u.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fmax.u.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fmax.u.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fmax.u.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fmaxnm.u.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fmaxnm.u.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fmaxnm.u.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fmin.u.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fmin.u.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fmin.u.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fminnm.u.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fminnm.u.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fminnm.u.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fmla.u.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fmla.u.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fmla.u.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fmls.u.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fmls.u.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fmls.u.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fmul.u.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fmul.u.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fmul.u.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fmulx.u.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fmulx.u.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fmulx.u.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fnmla.u.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fnmla.u.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fnmla.u.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fnmls.u.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fnmls.u.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fnmls.u.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fsub.u.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fsub.u.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fsub.u.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
-- 
2.7.4