From 5612b90694ed2c6b37c89ac9a643b34c873452b0 Mon Sep 17 00:00:00 2001 From: Lei Huang Date: Thu, 5 Jul 2018 15:21:37 +0000 Subject: [PATCH] [Power9] Add lib calls for float128 operations with no equivalent PPC instructions Map the following instructions to the proper float128 lib calls: pow[i], exp[2], log[2|10], sin, cos, fmin, fmax Differential Revision: https://reviews.llvm.org/D48544 llvm-svn: 336361 --- llvm/lib/Target/PowerPC/PPCISelLowering.cpp | 19 +++ llvm/test/CodeGen/PowerPC/f128-arith.ll | 147 +++++++++++++++++++- 2 files changed, 165 insertions(+), 1 deletion(-) diff --git a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp index 18d94ec8a50d..c236bab2deb7 100644 --- a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp +++ b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp @@ -815,6 +815,11 @@ PPCTargetLowering::PPCTargetLowering(const PPCTargetMachine &TM, setTruncStoreAction(MVT::f128, MVT::f64, Expand); setTruncStoreAction(MVT::f128, MVT::f32, Expand); setOperationAction(ISD::BITCAST, MVT::i128, Custom); + // No implementation for these ops for PowerPC. + setOperationAction(ISD::FSIN , MVT::f128, Expand); + setOperationAction(ISD::FCOS , MVT::f128, Expand); + setOperationAction(ISD::FPOW, MVT::f128, Expand); + setOperationAction(ISD::FPOWI, MVT::f128, Expand); } } @@ -1053,6 +1058,20 @@ PPCTargetLowering::PPCTargetLowering(const PPCTargetMachine &TM, setLibcallName(RTLIB::EXP2_PPCF128, "exp2l$LDBL128"); } + if (EnableQuadPrecision) { + setLibcallName(RTLIB::LOG_F128, "logf128"); + setLibcallName(RTLIB::LOG2_F128, "log2f128"); + setLibcallName(RTLIB::LOG10_F128, "log10f128"); + setLibcallName(RTLIB::EXP_F128, "expf128"); + setLibcallName(RTLIB::EXP2_F128, "exp2f128"); + setLibcallName(RTLIB::SIN_F128, "sinf128"); + setLibcallName(RTLIB::COS_F128, "cosf128"); + setLibcallName(RTLIB::POW_F128, "powf128"); + setLibcallName(RTLIB::FMIN_F128, "fminf128"); + setLibcallName(RTLIB::FMAX_F128, "fmaxf128"); + setLibcallName(RTLIB::POWI_F128, "__powikf2"); + } + // With 32 condition bits, we don't need to sink (and duplicate) compares // aggressively in CodeGenPrep. if (Subtarget.useCRBits()) { diff --git a/llvm/test/CodeGen/PowerPC/f128-arith.ll b/llvm/test/CodeGen/PowerPC/f128-arith.ll index ad037e7e1144..534f8f22ebc2 100644 --- a/llvm/test/CodeGen/PowerPC/f128-arith.ll +++ b/llvm/test/CodeGen/PowerPC/f128-arith.ll @@ -1,5 +1,5 @@ ; RUN: llc -mcpu=pwr9 -mtriple=powerpc64le-unknown-unknown \ -; RUN: -enable-ppc-quad-precision < %s | FileCheck %s +; RUN: -enable-ppc-quad-precision -verify-machineinstrs < %s | FileCheck %s ; Function Attrs: norecurse nounwind define void @qpAdd(fp128* nocapture readonly %a, fp128* nocapture %res) { @@ -148,3 +148,148 @@ entry: ; CHECK: stxv ; CHECK: blr } + +define fp128 @qp_sin(fp128* nocapture readonly %a) { +; CHECK-LABEL: qp_sin: +; CHECK: lxv 34, 0(3) +; CHECK: bl sinf128 +; CHECK: blr +entry: + %0 = load fp128, fp128* %a, align 16 + %1 = tail call fp128 @llvm.sin.f128(fp128 %0) + ret fp128 %1 +} +declare fp128 @llvm.sin.f128(fp128 %Val) + +define fp128 @qp_cos(fp128* nocapture readonly %a) { +; CHECK-LABEL: qp_cos: +; CHECK: lxv 34, 0(3) +; CHECK: bl cosf128 +; CHECK: blr +entry: + %0 = load fp128, fp128* %a, align 16 + %1 = tail call fp128 @llvm.cos.f128(fp128 %0) + ret fp128 %1 +} +declare fp128 @llvm.cos.f128(fp128 %Val) + +define fp128 @qp_log(fp128* nocapture readonly %a) { +; CHECK-LABEL: qp_log: +; CHECK: lxv 34, 0(3) +; CHECK: bl logf128 +; CHECK: blr +entry: + %0 = load fp128, fp128* %a, align 16 + %1 = tail call fp128 @llvm.log.f128(fp128 %0) + ret fp128 %1 +} +declare fp128 @llvm.log.f128(fp128 %Val) + +define fp128 @qp_log10(fp128* nocapture readonly %a) { +; CHECK-LABEL: qp_log10: +; CHECK: lxv 34, 0(3) +; CHECK: bl log10f128 +; CHECK: blr +entry: + %0 = load fp128, fp128* %a, align 16 + %1 = tail call fp128 @llvm.log10.f128(fp128 %0) + ret fp128 %1 +} +declare fp128 @llvm.log10.f128(fp128 %Val) + +define fp128 @qp_log2(fp128* nocapture readonly %a) { +; CHECK-LABEL: qp_log2: +; CHECK: lxv 34, 0(3) +; CHECK: bl log2f128 +; CHECK: blr +entry: + %0 = load fp128, fp128* %a, align 16 + %1 = tail call fp128 @llvm.log2.f128(fp128 %0) + ret fp128 %1 +} +declare fp128 @llvm.log2.f128(fp128 %Val) + +define fp128 @qp_minnum(fp128* nocapture readonly %a, + fp128* nocapture readonly %b) { +; CHECK-LABEL: qp_minnum: +; CHECK: lxv 34, 0(3) +; CHECK: lxv 35, 0(4) +; CHECK: bl fminf128 +; CHECK: blr +entry: + %0 = load fp128, fp128* %a, align 16 + %1 = load fp128, fp128* %b, align 16 + %2 = tail call fp128 @llvm.minnum.f128(fp128 %0, fp128 %1) + ret fp128 %2 +} +declare fp128 @llvm.minnum.f128(fp128 %Val0, fp128 %Val1) + +define fp128 @qp_maxnum(fp128* nocapture readonly %a, + fp128* nocapture readonly %b) { +; CHECK-LABEL: qp_maxnum: +; CHECK: lxv 34, 0(3) +; CHECK: lxv 35, 0(4) +; CHECK: bl fmaxf128 +; CHECK: blr +entry: + %0 = load fp128, fp128* %a, align 16 + %1 = load fp128, fp128* %b, align 16 + %2 = tail call fp128 @llvm.maxnum.f128(fp128 %0, fp128 %1) + ret fp128 %2 +} +declare fp128 @llvm.maxnum.f128(fp128 %Val0, fp128 %Val1) + +define fp128 @qp_pow(fp128* nocapture readonly %a, + fp128* nocapture readonly %b) { +; CHECK-LABEL: qp_pow: +; CHECK: lxv 34, 0(3) +; CHECK: lxv 35, 0(4) +; CHECK: bl powf128 +; CHECK: blr +entry: + %0 = load fp128, fp128* %a, align 16 + %1 = load fp128, fp128* %b, align 16 + %2 = tail call fp128 @llvm.pow.f128(fp128 %0, fp128 %1) + ret fp128 %2 +} +declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power) + +define fp128 @qp_exp(fp128* nocapture readonly %a) { +; CHECK-LABEL: qp_exp: +; CHECK: lxv 34, 0(3) +; CHECK: bl expf128 +; CHECK: blr +entry: + %0 = load fp128, fp128* %a, align 16 + %1 = tail call fp128 @llvm.exp.f128(fp128 %0) + ret fp128 %1 +} +declare fp128 @llvm.exp.f128(fp128 %Val) + +define fp128 @qp_exp2(fp128* nocapture readonly %a) { +; CHECK-LABEL: qp_exp2: +; CHECK: lxv 34, 0(3) +; CHECK: bl exp2f128 +; CHECK: blr +entry: + %0 = load fp128, fp128* %a, align 16 + %1 = tail call fp128 @llvm.exp2.f128(fp128 %0) + ret fp128 %1 +} +declare fp128 @llvm.exp2.f128(fp128 %Val) + +define void @qp_powi(fp128* nocapture readonly %a, i32* nocapture readonly %b, + fp128* nocapture %res) { +; CHECK-LABEL: qp_powi: +; CHECK: lxv 34, 0(3) +; CHECK: lwz 3, 0(4) +; CHECK: bl __powikf2 +; CHECK: blr +entry: + %0 = load fp128, fp128* %a, align 16 + %1 = load i32, i32* %b, align 8 + %2 = tail call fp128 @llvm.powi.f128(fp128 %0, i32 %1) + store fp128 %2, fp128* %res, align 16 + ret void +} +declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)