[PPC] FE support for generating VSX [negated] absolute value instructions

Includes new built-in, conversion of built-in to target-independent intrinsic
and update in the header file. Tests are also updated. There is a second part in
the backend for which I will post a separate code-review. BACKEND PART SHOULD BE
COMMITTED FIRST.

Phabricator: http://reviews.llvm.org/D17816
llvm-svn: 263051
This commit is contained in:
Kit Barton 2016-03-09 19:28:31 +00:00
parent 206ba84413
commit fbab158767
6 changed files with 51 additions and 12 deletions

View File

@ -336,6 +336,9 @@ BUILTIN(__builtin_vsx_xxleqv, "V4UiV4UiV4Ui", "")
BUILTIN(__builtin_vsx_xvcpsgndp, "V2dV2dV2d", "")
BUILTIN(__builtin_vsx_xvcpsgnsp, "V4fV4fV4f", "")
BUILTIN(__builtin_vsx_xvabssp, "V4fV4f", "")
BUILTIN(__builtin_vsx_xvabsdp, "V2dV2d", "")
// HTM builtins
BUILTIN(__builtin_tbegin, "UiUIi", "")
BUILTIN(__builtin_tend, "UiUIi", "")

View File

@ -6992,6 +6992,16 @@ Value *CodeGenFunction::EmitPPCBuiltinExpr(unsigned BuiltinID,
llvm::Function *F = CGM.getIntrinsic(ID, ResultType);
return Builder.CreateCall(F, X);
}
// Absolute value
case PPC::BI__builtin_vsx_xvabsdp:
case PPC::BI__builtin_vsx_xvabssp: {
llvm::Type *ResultType = ConvertType(E->getType());
Value *X = EmitScalarExpr(E->getArg(0));
llvm::Function *F = CGM.getIntrinsic(Intrinsic::fabs, ResultType);
return Builder.CreateCall(F, X);
}
// FMA variations
case PPC::BI__builtin_vsx_xvmaddadp:
case PPC::BI__builtin_vsx_xvmaddasp:

View File

@ -124,16 +124,18 @@ vec_abs(vector signed long long __a) {
#endif
static vector float __ATTRS_o_ai vec_abs(vector float __a) {
#ifdef __VSX__
return __builtin_vsx_xvabssp(__a);
#else
vector unsigned int __res =
(vector unsigned int)__a & (vector unsigned int)(0x7FFFFFFF);
return (vector float)__res;
#endif
}
#if defined(__POWER8_VECTOR__) && defined(__powerpc64__)
static vector double __ATTRS_o_ai vec_abs(vector double __a) {
vector unsigned long long __res = { 0x7FFFFFFFFFFFFFFF, 0x7FFFFFFFFFFFFFFF };
__res &= (vector unsigned int)__a;
return (vector double)__res;
return __builtin_vsx_xvabsdp(__a);
}
#endif

View File

@ -1,7 +1,13 @@
// REQUIRES: powerpc-registered-target
// RUN: %clang_cc1 -faltivec -triple powerpc-unknown-unknown -emit-llvm %s -o - | FileCheck %s
// RUN: %clang_cc1 -faltivec -triple powerpc64-unknown-unknown -emit-llvm %s -o - | FileCheck %s
// RUN: %clang_cc1 -faltivec -triple powerpc64le-unknown-unknown -emit-llvm %s -o - | FileCheck %s -check-prefix=CHECK-LE
// RUN: %clang_cc1 -faltivec -triple powerpc-unknown-unknown -emit-llvm %s \
// RUN: -o - | FileCheck %s
// RUN: %clang_cc1 -faltivec -triple powerpc64-unknown-unknown -emit-llvm %s \
// RUN: -o - | FileCheck %s
// RUN: %clang_cc1 -faltivec -triple powerpc64le-unknown-unknown -emit-llvm %s \
// RUN: -o - | FileCheck %s -check-prefix=CHECK-LE
// RUN: not %clang_cc1 -triple powerpc64le-unknown-unknown -emit-llvm %s \
// RUN: -ferror-limit 0 -o - 2>&1 \
// RUN: | FileCheck %s -check-prefix=CHECK-NOALTIVEC
vector bool char vbc = { 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0 };
vector signed char vsc = { 1, -2, 3, -4, 5, -6, 7, -8, 9, -10, 11, -12, 13, -14, 15, -16 };
@ -27,6 +33,8 @@ vector int res_vi;
vector unsigned int res_vui;
vector float res_vf;
// CHECK-NOALTIVEC: error: unknown type name 'vector'
signed char param_sc;
unsigned char param_uc;
short param_s;
@ -66,8 +74,16 @@ void test1() {
// CHECK-LE: @llvm.ppc.altivec.vmaxsw
vf = vec_abs(vf);
// CHECK: and <4 x i32>
// CHECK-LE: and <4 x i32>
// CHECK: bitcast <4 x float> %{{.*}} to <4 x i32>
// CHECK: and <4 x i32> {{.*}}, <i32 2147483647, i32 2147483647, i32 2147483647, i32 2147483647>
// CHECK: bitcast <4 x i32> %{{.*}} to <4 x float>
// CHECK: store <4 x float> %{{.*}}, <4 x float>* @vf
// CHECK-LE: bitcast <4 x float> %{{.*}} to <4 x i32>
// CHECK-LE: and <4 x i32> {{.*}}, <i32 2147483647, i32 2147483647, i32 2147483647, i32 2147483647>
// CHECK-LE: bitcast <4 x i32> %{{.*}} to <4 x float>
// CHECK-LE: store <4 x float> %{{.*}}, <4 x float>* @vf
// CHECK-NOALTIVEC: error: use of undeclared identifier 'vf'
// CHECK-NOALTIVEC: vf = vec_abs(vf)
/* vec_abs */
vsc = vec_abss(vsc);

View File

@ -73,10 +73,10 @@ void test1() {
// CHECK-PPC: error: call to 'vec_abs' is ambiguous
res_vd = vec_abs(vda);
// CHECK: store <2 x i64> <i64 9223372036854775807, i64 9223372036854775807>, <2 x i64>*
// CHECK: and <2 x i64>
// CHECK-LE: store <2 x i64> <i64 9223372036854775807, i64 9223372036854775807>, <2 x i64>*
// CHECK-LE: and <2 x i64>
// CHECK: call <2 x double> @llvm.fabs.v2f64(<2 x double> %{{.*}})
// CHECK: store <2 x double> %{{.*}}, <2 x double>* @res_vd
// CHECK-LE: call <2 x double> @llvm.fabs.v2f64(<2 x double> %{{.*}})
// CHECK-LE: store <2 x double> %{{.*}}, <2 x double>* @res_vd
// CHECK-PPC: error: call to 'vec_abs' is ambiguous
/* vec_add */

View File

@ -51,6 +51,14 @@ void test1() {
// CHECK-LABEL: define void @test1
// CHECK-LE-LABEL: define void @test1
res_vf = vec_abs(vf);
// CHECK: call <4 x float> @llvm.fabs.v4f32(<4 x float> %{{[0-9]*}})
// CHECK-LE: call <4 x float> @llvm.fabs.v4f32(<4 x float> %{{[0-9]*}})
dummy();
// CHECK: call void @dummy()
// CHECK-LE: call void @dummy()
res_vd = vec_add(vd, vd);
// CHECK: fadd <2 x double>
// CHECK-LE: fadd <2 x double>