[TargetLibraryInfo] add mappings from LLVM sin/cos intrinsics to SVML calls

[spatel]

Summary

I'm assuming these were accidentally left out of D19544.
D40044 made it more likely that we'll have LLVM math intrinsics rather than libcalls, so this bug was more easily exposed.
As the tests/code show, we already have the complete mappings for pow/exp/log.

I don't have any experience with SVML, so I don't know if anything else is missing. It's also not clear to me that we should be doing this transform in IR rather than DAG/isel, but that's a separate issue.

[RKSimon]

Makes sense to me, I'm not a SVML expert but we're already doing the same for exp/exp, log/logf and pow/powf.

Any comments @craig.topper et al?

[mmasten]

This change looks good to me. Thanks for the patch.

[spatel]

Marking this as accepted here in phab ahead of commit, so we don't get the warning about unknown state.

lib/Analysis/TargetLibraryInfo.cpp

const VecDesc VecFuncs[] = {
         {"sin", "__svml_sin2", 2},
         {"sin", "__svml_sin4", 4},
         {"sin", "__svml_sin8", 8},
 
         {"sinf", "__svml_sinf4", 4},
         {"sinf", "__svml_sinf8", 8},
         {"sinf", "__svml_sinf16", 16},
 
+        {"llvm.sin.f64", "__svml_sin2", 2},
+        {"llvm.sin.f64", "__svml_sin4", 4},
+        {"llvm.sin.f64", "__svml_sin8", 8},
+
+        {"llvm.sin.f32", "__svml_sinf4", 4},
+        {"llvm.sin.f32", "__svml_sinf8", 8},
+        {"llvm.sin.f32", "__svml_sinf16", 16},
+
         {"cos", "__svml_cos2", 2},
         {"cos", "__svml_cos4", 4},
         {"cos", "__svml_cos8", 8},
 
         {"cosf", "__svml_cosf4", 4},
         {"cosf", "__svml_cosf8", 8},
         {"cosf", "__svml_cosf16", 16},
 
+        {"llvm.cos.f64", "__svml_cos2", 2},
+        {"llvm.cos.f64", "__svml_cos4", 4},
+        {"llvm.cos.f64", "__svml_cos8", 8},
+
+        {"llvm.cos.f32", "__svml_cosf4", 4},
+        {"llvm.cos.f32", "__svml_cosf8", 8},
+        {"llvm.cos.f32", "__svml_cosf16", 16},
+
         {"pow", "__svml_pow2", 2},
         {"pow", "__svml_pow4", 4},
         {"pow", "__svml_pow8", 8},
 
         {"powf", "__svml_powf4", 4},
         {"powf", "__svml_powf8", 8},
         {"powf", "__svml_powf16", 16},
 

test/Transforms/LoopVectorize/X86/svml-calls.ll

for.body:
   br i1 %exitcond, label %for.end, label %for.body
 
 for.end:
   ret void
 }
 
 define void @sin_f64_intrinsic(double* nocapture %varray) {
 ; CHECK-LABEL: @sin_f64_intrinsic(
-; CHECK:    [[TMP5:%.*]] = call <4 x double> @llvm.sin.v4f64(<4 x double> [[TMP4:%.*]])
+; CHECK:    [[TMP5:%.*]] = call <4 x double> @__svml_sin4(<4 x double> [[TMP4:%.*]])
 ; CHECK:    ret void
 ;
 entry:
   br label %for.body
 
 for.body:
   %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
   %tmp = trunc i64 %iv to i32
   %conv = sitofp i32 %tmp to double
   %call = tail call double @llvm.sin.f64(double %conv)
   %arrayidx = getelementptr inbounds double, double* %varray, i64 %iv
   store double %call, double* %arrayidx, align 4
   %iv.next = add nuw nsw i64 %iv, 1
   %exitcond = icmp eq i64 %iv.next, 1000
   br i1 %exitcond, label %for.end, label %for.body
 
 for.end:
   ret void
 }
 
 define void @sin_f32_intrinsic(float* nocapture %varray) {
 ; CHECK-LABEL: @sin_f32_intrinsic(
-; CHECK:    [[TMP5:%.*]] = call <4 x float> @llvm.sin.v4f32(<4 x float> [[TMP4:%.*]])
+; CHECK:    [[TMP5:%.*]] = call <4 x float> @__svml_sinf4(<4 x float> [[TMP4:%.*]])
 ; CHECK:    ret void
 ;
 entry:
   br label %for.body
 
 for.body:
   %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
   %tmp = trunc i64 %iv to i32
for.body:
   br i1 %exitcond, label %for.end, label %for.body
 
 for.end:
   ret void
 }
 
 define void @cos_f64_intrinsic(double* nocapture %varray) {
 ; CHECK-LABEL: @cos_f64_intrinsic(
-; CHECK:    [[TMP5:%.*]] = call <4 x double> @llvm.cos.v4f64(<4 x double> [[TMP4:%.*]])
+; CHECK:    [[TMP5:%.*]] = call <4 x double> @__svml_cos4(<4 x double> [[TMP4:%.*]])
 ; CHECK:    ret void
 ;
 entry:
   br label %for.body
 
 for.body:
   %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
   %tmp = trunc i64 %iv to i32
   %conv = sitofp i32 %tmp to double
   %call = tail call double @llvm.cos.f64(double %conv)
   %arrayidx = getelementptr inbounds double, double* %varray, i64 %iv
   store double %call, double* %arrayidx, align 4
   %iv.next = add nuw nsw i64 %iv, 1
   %exitcond = icmp eq i64 %iv.next, 1000
   br i1 %exitcond, label %for.end, label %for.body
 
 for.end:
   ret void
 }
 
 define void @cos_f32_intrinsic(float* nocapture %varray) {
 ; CHECK-LABEL: @cos_f32_intrinsic(
-; CHECK:    [[TMP5:%.*]] = call <4 x float> @llvm.cos.v4f32(<4 x float> [[TMP4:%.*]])
+; CHECK:    [[TMP5:%.*]] = call <4 x float> @__svml_cosf4(<4 x float> [[TMP4:%.*]])
 ; CHECK:    ret void
 ;
 entry:
   br label %for.body
 
 for.body:
   %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
   %tmp = trunc i64 %iv to i32