#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/ADT/Triple.h"
+#include "llvm/Support/CommandLine.h"
using namespace llvm;
+static cl::opt<TargetLibraryInfoImpl::VectorLibrary> ClVectorLibrary(
+ "vector-library", cl::Hidden, cl::desc("Vector functions library"),
+ cl::init(TargetLibraryInfoImpl::NoLibrary),
+ cl::values(clEnumValN(TargetLibraryInfoImpl::NoLibrary, "none",
+ "No vector functions library"),
+ clEnumValN(TargetLibraryInfoImpl::Accelerate, "Accelerate",
+ "Accelerate framework"),
+ clEnumValEnd));
+
const char *const TargetLibraryInfoImpl::StandardNames[LibFunc::NumLibFuncs] = {
#define TLI_DEFINE_STRING
#include "llvm/Analysis/TargetLibraryInfo.def"
TLI.setUnavailable(LibFunc::statvfs64);
TLI.setUnavailable(LibFunc::tmpfile64);
}
+
+ TLI.addVectorizableFunctionsFromVecLib(ClVectorLibrary);
}
TargetLibraryInfoImpl::TargetLibraryInfoImpl() {
std::sort(ScalarDescs.begin(), ScalarDescs.end(), compareByVectorFnName);
}
+void TargetLibraryInfoImpl::addVectorizableFunctionsFromVecLib(
+ enum VectorLibrary VecLib) {
+ switch (VecLib) {
+ case Accelerate: {
+ const VecDesc VecFuncs[] = {
+ {"expf", "vexpf", 4},
+ {"llvm.exp.f32", "vexpf", 4},
+ {"logf", "vlogf", 4},
+ {"llvm.log.f32", "vlogf", 4},
+ {"sqrtf", "vsqrtf", 4},
+ {"llvm.sqrt.f32", "vsqrtf", 4},
+ {"fabsf", "vfabsf", 4},
+ {"llvm.fabs.f32", "vfabsf", 4},
+ };
+ addVectorizableFunctions(VecFuncs);
+ break;
+ }
+ case NoLibrary:
+ break;
+ }
+}
+
bool TargetLibraryInfoImpl::isFunctionVectorizable(StringRef funcName) const {
funcName = sanitizeFunctionName(funcName);
if (funcName.empty())
--- /dev/null
+; RUN: opt < %s -vector-library=Accelerate -loop-vectorize -S | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+;CHECK-LABEL: @sqrt_f32(
+;CHECK: vsqrtf{{.*}}<4 x float>
+;CHECK: ret void
+declare float @sqrtf(float) nounwind readnone
+define void @sqrt_f32(i32 %n, float* noalias %y, float* noalias %x) nounwind uwtable {
+entry:
+ %cmp6 = icmp sgt i32 %n, 0
+ br i1 %cmp6, label %for.body, label %for.end
+
+for.body: ; preds = %entry, %for.body
+ %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
+ %arrayidx = getelementptr inbounds float, float* %y, i64 %indvars.iv
+ %0 = load float, float* %arrayidx, align 4
+ %call = tail call float @sqrtf(float %0) nounwind readnone
+ %arrayidx2 = getelementptr inbounds float, float* %x, i64 %indvars.iv
+ store float %call, float* %arrayidx2, align 4
+ %indvars.iv.next = add i64 %indvars.iv, 1
+ %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+ %exitcond = icmp eq i32 %lftr.wideiv, %n
+ br i1 %exitcond, label %for.end, label %for.body
+
+for.end: ; preds = %for.body, %entry
+ ret void
+}
+
+;CHECK-LABEL: @exp_f32(
+;CHECK: vexpf{{.*}}<4 x float>
+;CHECK: ret void
+declare float @expf(float) nounwind readnone
+define void @exp_f32(i32 %n, float* noalias %y, float* noalias %x) nounwind uwtable {
+entry:
+ %cmp6 = icmp sgt i32 %n, 0
+ br i1 %cmp6, label %for.body, label %for.end
+
+for.body: ; preds = %entry, %for.body
+ %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
+ %arrayidx = getelementptr inbounds float, float* %y, i64 %indvars.iv
+ %0 = load float, float* %arrayidx, align 4
+ %call = tail call float @expf(float %0) nounwind readnone
+ %arrayidx2 = getelementptr inbounds float, float* %x, i64 %indvars.iv
+ store float %call, float* %arrayidx2, align 4
+ %indvars.iv.next = add i64 %indvars.iv, 1
+ %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+ %exitcond = icmp eq i32 %lftr.wideiv, %n
+ br i1 %exitcond, label %for.end, label %for.body
+
+for.end: ; preds = %for.body, %entry
+ ret void
+}
+
+;CHECK-LABEL: @log_f32(
+;CHECK: vlogf{{.*}}<4 x float>
+;CHECK: ret void
+declare float @logf(float) nounwind readnone
+define void @log_f32(i32 %n, float* noalias %y, float* noalias %x) nounwind uwtable {
+entry:
+ %cmp6 = icmp sgt i32 %n, 0
+ br i1 %cmp6, label %for.body, label %for.end
+
+for.body: ; preds = %entry, %for.body
+ %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
+ %arrayidx = getelementptr inbounds float, float* %y, i64 %indvars.iv
+ %0 = load float, float* %arrayidx, align 4
+ %call = tail call float @logf(float %0) nounwind readnone
+ %arrayidx2 = getelementptr inbounds float, float* %x, i64 %indvars.iv
+ store float %call, float* %arrayidx2, align 4
+ %indvars.iv.next = add i64 %indvars.iv, 1
+ %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+ %exitcond = icmp eq i32 %lftr.wideiv, %n
+ br i1 %exitcond, label %for.end, label %for.body
+
+for.end: ; preds = %for.body, %entry
+ ret void
+}
+
+; For abs instruction we'll generate vector intrinsic, as it's cheaper than a lib call.
+;CHECK-LABEL: @fabs_f32(
+;CHECK: fabs{{.*}}<4 x float>
+;CHECK: ret void
+declare float @fabsf(float) nounwind readnone
+define void @fabs_f32(i32 %n, float* noalias %y, float* noalias %x) nounwind uwtable {
+entry:
+ %cmp6 = icmp sgt i32 %n, 0
+ br i1 %cmp6, label %for.body, label %for.end
+
+for.body: ; preds = %entry, %for.body
+ %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
+ %arrayidx = getelementptr inbounds float, float* %y, i64 %indvars.iv
+ %0 = load float, float* %arrayidx, align 4
+ %call = tail call float @fabsf(float %0) nounwind readnone
+ %arrayidx2 = getelementptr inbounds float, float* %x, i64 %indvars.iv
+ store float %call, float* %arrayidx2, align 4
+ %indvars.iv.next = add i64 %indvars.iv, 1
+ %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+ %exitcond = icmp eq i32 %lftr.wideiv, %n
+ br i1 %exitcond, label %for.end, label %for.body
+
+for.end: ; preds = %for.body, %entry
+ ret void
+}
+
+; Test that we can vectorize an intrinsic into a vector call.
+;CHECK-LABEL: @exp_f32_intrin(
+;CHECK: vexpf{{.*}}<4 x float>
+;CHECK: ret void
+declare float @llvm.exp.f32(float) nounwind readnone
+define void @exp_f32_intrin(i32 %n, float* noalias %y, float* noalias %x) nounwind uwtable {
+entry:
+ %cmp6 = icmp sgt i32 %n, 0
+ br i1 %cmp6, label %for.body, label %for.end
+
+for.body: ; preds = %entry, %for.body
+ %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
+ %arrayidx = getelementptr inbounds float, float* %y, i64 %indvars.iv
+ %0 = load float, float* %arrayidx, align 4
+ %call = tail call float @llvm.exp.f32(float %0) nounwind readnone
+ %arrayidx2 = getelementptr inbounds float, float* %x, i64 %indvars.iv
+ store float %call, float* %arrayidx2, align 4
+ %indvars.iv.next = add i64 %indvars.iv, 1
+ %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+ %exitcond = icmp eq i32 %lftr.wideiv, %n
+ br i1 %exitcond, label %for.end, label %for.body
+
+for.end: ; preds = %for.body, %entry
+ ret void
+}
+
+; Test that we don't vectorize arbitrary functions.
+;CHECK-LABEL: @foo_f32(
+;CHECK-NOT: foo{{.*}}<4 x float>
+;CHECK: ret void
+declare float @foo(float) nounwind readnone
+define void @foo_f32(i32 %n, float* noalias %y, float* noalias %x) nounwind uwtable {
+entry:
+ %cmp6 = icmp sgt i32 %n, 0
+ br i1 %cmp6, label %for.body, label %for.end
+
+for.body: ; preds = %entry, %for.body
+ %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
+ %arrayidx = getelementptr inbounds float, float* %y, i64 %indvars.iv
+ %0 = load float, float* %arrayidx, align 4
+ %call = tail call float @foo(float %0) nounwind readnone
+ %arrayidx2 = getelementptr inbounds float, float* %x, i64 %indvars.iv
+ store float %call, float* %arrayidx2, align 4
+ %indvars.iv.next = add i64 %indvars.iv, 1
+ %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+ %exitcond = icmp eq i32 %lftr.wideiv, %n
+ br i1 %exitcond, label %for.end, label %for.body
+
+for.end: ; preds = %for.body, %entry
+ ret void
+}
+
+; Test that we don't vectorize calls with nobuiltin attribute.
+;CHECK-LABEL: @sqrt_f32_nobuiltin(
+;CHECK-NOT: vsqrtf{{.*}}<4 x float>
+;CHECK: ret void
+define void @sqrt_f32_nobuiltin(i32 %n, float* noalias %y, float* noalias %x) nounwind uwtable {
+entry:
+ %cmp6 = icmp sgt i32 %n, 0
+ br i1 %cmp6, label %for.body, label %for.end
+
+for.body: ; preds = %entry, %for.body
+ %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
+ %arrayidx = getelementptr inbounds float, float* %y, i64 %indvars.iv
+ %0 = load float, float* %arrayidx, align 4
+ %call = tail call float @sqrtf(float %0) nounwind readnone nobuiltin
+ %arrayidx2 = getelementptr inbounds float, float* %x, i64 %indvars.iv
+ store float %call, float* %arrayidx2, align 4
+ %indvars.iv.next = add i64 %indvars.iv, 1
+ %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+ %exitcond = icmp eq i32 %lftr.wideiv, %n
+ br i1 %exitcond, label %for.end, label %for.body
+
+for.end: ; preds = %for.body, %entry
+ ret void
+}