wangpc created this revision.
wangpc added a reviewer: craig.topper.
Herald added subscribers: jobnoorman, luke, VincentWu, vkmr, frasercrmck,
luismarques, apazos, sameer.abuasal, s.egerton, Jim, benna, psnobl, jocewei,
PkmX, the_o, brucehoult, MartinMosbeck, rogfer01, edward-jones, zzheng, jrtc27,
shiva0217, kito-cheng, niosHD, sabuasal, simoncook, johnrusso, rbar, asb,
arichardson.
Herald added a project: All.
wangpc requested review of this revision.
Herald added subscribers: cfe-commits, eopXD, MaskRay.
Herald added a project: clang.
We used to convert them to M1 <https://reviews.llvm.org/M1> types in arguments
and return
value, which causes failures in CodeGen since it is not legal
to insert subvectors with LMUL>1 to M1 <https://reviews.llvm.org/M1> vectors.
Fixes 64266
Repository:
rG LLVM Github Monorepo
https://reviews.llvm.org/D156779
Files:
clang/lib/CodeGen/Targets/RISCV.cpp
clang/test/CodeGen/attr-riscv-rvv-vector-bits-codegen.c
Index: clang/test/CodeGen/attr-riscv-rvv-vector-bits-codegen.c
===================================================================
--- clang/test/CodeGen/attr-riscv-rvv-vector-bits-codegen.c
+++ clang/test/CodeGen/attr-riscv-rvv-vector-bits-codegen.c
@@ -42,8 +42,8 @@
// CHECK-NEXT: [[TMP1:%.*]] = load <8 x i32>, ptr [[TMP0]], align 8
// CHECK-NEXT: store <8 x i32> [[TMP1]], ptr [[RETVAL]], align 8
// CHECK-NEXT: [[TMP2:%.*]] = load <8 x i32>, ptr [[RETVAL]], align 8
-// CHECK-NEXT: [[CASTSCALABLESVE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> undef, <8 x i32> [[TMP2]], i64 0)
-// CHECK-NEXT: ret <vscale x 2 x i32> [[CASTSCALABLESVE]]
+// CHECK-NEXT: [[CAST_SCALABLE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> undef, <8 x i32> [[TMP2]], i64 0)
+// CHECK-NEXT: ret <vscale x 2 x i32> [[CAST_SCALABLE]]
//
fixed_int32m1_t test_ptr_to_global() {
fixed_int32m1_t *global_vec_ptr;
@@ -63,8 +63,8 @@
// CHECK-NEXT: [[TMP1:%.*]] = load <8 x i32>, ptr [[ARRAYIDX]], align 8
// CHECK-NEXT: store <8 x i32> [[TMP1]], ptr [[RETVAL]], align 8
// CHECK-NEXT: [[TMP2:%.*]] = load <8 x i32>, ptr [[RETVAL]], align 8
-// CHECK-NEXT: [[CASTSCALABLESVE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> undef, <8 x i32> [[TMP2]], i64 0)
-// CHECK-NEXT: ret <vscale x 2 x i32> [[CASTSCALABLESVE]]
+// CHECK-NEXT: [[CAST_SCALABLE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> undef, <8 x i32> [[TMP2]], i64 0)
+// CHECK-NEXT: ret <vscale x 2 x i32> [[CAST_SCALABLE]]
//
fixed_int32m1_t array_arg(fixed_int32m1_t arr[]) {
return arr[0];
@@ -76,14 +76,14 @@
// CHECK-NEXT: [[VEC_ADDR:%.*]] = alloca <vscale x 2 x i32>, align 4
// CHECK-NEXT: store <vscale x 2 x i32> [[VEC:%.*]], ptr [[VEC_ADDR]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load <8 x i32>, ptr @global_vec, align 8
-// CHECK-NEXT: [[CASTSCALABLESVE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> undef, <8 x i32> [[TMP0]], i64 0)
+// CHECK-NEXT: [[CAST_SCALABLE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> undef, <8 x i32> [[TMP0]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = load <vscale x 2 x i32>, ptr [[VEC_ADDR]], align 4
-// CHECK-NEXT: [[TMP2:%.*]] = call <vscale x 2 x i32> @llvm.riscv.vadd.nxv2i32.nxv2i32.i64(<vscale x 2 x i32> poison, <vscale x 2 x i32> [[CASTSCALABLESVE]], <vscale x 2 x i32> [[TMP1]], i64 8)
-// CHECK-NEXT: [[CASTFIXEDSVE:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[TMP2]], i64 0)
-// CHECK-NEXT: store <8 x i32> [[CASTFIXEDSVE]], ptr [[RETVAL]], align 8
+// CHECK-NEXT: [[TMP2:%.*]] = call <vscale x 2 x i32> @llvm.riscv.vadd.nxv2i32.nxv2i32.i64(<vscale x 2 x i32> poison, <vscale x 2 x i32> [[CAST_SCALABLE]], <vscale x 2 x i32> [[TMP1]], i64 8)
+// CHECK-NEXT: [[CAST_FIXED:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[TMP2]], i64 0)
+// CHECK-NEXT: store <8 x i32> [[CAST_FIXED]], ptr [[RETVAL]], align 8
// CHECK-NEXT: [[TMP3:%.*]] = load <8 x i32>, ptr [[RETVAL]], align 8
-// CHECK-NEXT: [[CASTSCALABLESVE1:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> undef, <8 x i32> [[TMP3]], i64 0)
-// CHECK-NEXT: ret <vscale x 2 x i32> [[CASTSCALABLESVE1]]
+// CHECK-NEXT: [[CAST_SCALABLE1:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> undef, <8 x i32> [[TMP3]], i64 0)
+// CHECK-NEXT: ret <vscale x 2 x i32> [[CAST_SCALABLE1]]
//
fixed_int32m1_t test_cast(vint32m1_t vec) {
return __riscv_vadd(global_vec, vec, __riscv_v_fixed_vlen/32);
@@ -98,8 +98,8 @@
// CHECK-NEXT: [[TMP1:%.*]] = load <16 x i32>, ptr [[TMP0]], align 8
// CHECK-NEXT: store <16 x i32> [[TMP1]], ptr [[RETVAL]], align 8
// CHECK-NEXT: [[TMP2:%.*]] = load <16 x i32>, ptr [[RETVAL]], align 8
-// CHECK-NEXT: [[CASTSCALABLESVE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v16i32(<vscale x 2 x i32> undef, <16 x i32> [[TMP2]], i64 0)
-// CHECK-NEXT: ret <vscale x 2 x i32> [[CASTSCALABLESVE]]
+// CHECK-NEXT: [[CAST_SCALABLE:%.*]] = call <vscale x 4 x i32> @llvm.vector.insert.nxv4i32.v16i32(<vscale x 4 x i32> undef, <16 x i32> [[TMP2]], i64 0)
+// CHECK-NEXT: ret <vscale x 4 x i32> [[CAST_SCALABLE]]
//
fixed_int32m2_t test_ptr_to_global_m2() {
fixed_int32m2_t *global_vec_ptr;
@@ -119,8 +119,8 @@
// CHECK-NEXT: [[TMP1:%.*]] = load <16 x i32>, ptr [[ARRAYIDX]], align 8
// CHECK-NEXT: store <16 x i32> [[TMP1]], ptr [[RETVAL]], align 8
// CHECK-NEXT: [[TMP2:%.*]] = load <16 x i32>, ptr [[RETVAL]], align 8
-// CHECK-NEXT: [[CASTSCALABLESVE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v16i32(<vscale x 2 x i32> undef, <16 x i32> [[TMP2]], i64 0)
-// CHECK-NEXT: ret <vscale x 2 x i32> [[CASTSCALABLESVE]]
+// CHECK-NEXT: [[CAST_SCALABLE:%.*]] = call <vscale x 4 x i32> @llvm.vector.insert.nxv4i32.v16i32(<vscale x 4 x i32> undef, <16 x i32> [[TMP2]], i64 0)
+// CHECK-NEXT: ret <vscale x 4 x i32> [[CAST_SCALABLE]]
//
fixed_int32m2_t array_arg_m2(fixed_int32m2_t arr[]) {
return arr[0];
@@ -132,14 +132,14 @@
// CHECK-NEXT: [[VEC_ADDR:%.*]] = alloca <vscale x 4 x i32>, align 4
// CHECK-NEXT: store <vscale x 4 x i32> [[VEC:%.*]], ptr [[VEC_ADDR]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load <16 x i32>, ptr @global_vec_m2, align 8
-// CHECK-NEXT: [[CASTSCALABLESVE:%.*]] = call <vscale x 4 x i32> @llvm.vector.insert.nxv4i32.v16i32(<vscale x 4 x i32> undef, <16 x i32> [[TMP0]], i64 0)
+// CHECK-NEXT: [[CAST_SCALABLE:%.*]] = call <vscale x 4 x i32> @llvm.vector.insert.nxv4i32.v16i32(<vscale x 4 x i32> undef, <16 x i32> [[TMP0]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = load <vscale x 4 x i32>, ptr [[VEC_ADDR]], align 4
-// CHECK-NEXT: [[TMP2:%.*]] = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32.i64(<vscale x 4 x i32> poison, <vscale x 4 x i32> [[CASTSCALABLESVE]], <vscale x 4 x i32> [[TMP1]], i64 16)
-// CHECK-NEXT: [[CASTFIXEDSVE:%.*]] = call <16 x i32> @llvm.vector.extract.v16i32.nxv4i32(<vscale x 4 x i32> [[TMP2]], i64 0)
-// CHECK-NEXT: store <16 x i32> [[CASTFIXEDSVE]], ptr [[RETVAL]], align 8
+// CHECK-NEXT: [[TMP2:%.*]] = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32.i64(<vscale x 4 x i32> poison, <vscale x 4 x i32> [[CAST_SCALABLE]], <vscale x 4 x i32> [[TMP1]], i64 16)
+// CHECK-NEXT: [[CAST_FIXED:%.*]] = call <16 x i32> @llvm.vector.extract.v16i32.nxv4i32(<vscale x 4 x i32> [[TMP2]], i64 0)
+// CHECK-NEXT: store <16 x i32> [[CAST_FIXED]], ptr [[RETVAL]], align 8
// CHECK-NEXT: [[TMP3:%.*]] = load <16 x i32>, ptr [[RETVAL]], align 8
-// CHECK-NEXT: [[CASTSCALABLESVE1:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v16i32(<vscale x 2 x i32> undef, <16 x i32> [[TMP3]], i64 0)
-// CHECK-NEXT: ret <vscale x 2 x i32> [[CASTSCALABLESVE1]]
+// CHECK-NEXT: [[CAST_SCALABLE1:%.*]] = call <vscale x 4 x i32> @llvm.vector.insert.nxv4i32.v16i32(<vscale x 4 x i32> undef, <16 x i32> [[TMP3]], i64 0)
+// CHECK-NEXT: ret <vscale x 4 x i32> [[CAST_SCALABLE1]]
//
fixed_int32m2_t test_cast_m2(vint32m2_t vec) {
return __riscv_vadd(global_vec_m2, vec, __riscv_v_fixed_vlen/16);
Index: clang/lib/CodeGen/Targets/RISCV.cpp
===================================================================
--- clang/lib/CodeGen/Targets/RISCV.cpp
+++ clang/lib/CodeGen/Targets/RISCV.cpp
@@ -8,7 +8,6 @@
#include "ABIInfoImpl.h"
#include "TargetInfo.h"
-#include "llvm/TargetParser/RISCVTargetParser.h"
using namespace clang;
using namespace clang::CodeGen;
@@ -315,11 +314,15 @@
assert(VT->getElementType()->isBuiltinType() && "expected builtin type!");
- const auto *BT = VT->getElementType()->castAs<BuiltinType>();
- unsigned EltSize = getContext().getTypeSize(BT);
+ auto VScale =
+ getContext().getTargetInfo().getVScaleRange(getContext().getLangOpts());
+ // The MinNumElts is simplified from equation:
+ // NumElts / VScale =
+ // (EltSize * NumElts / (VScale * RVVBitsPerBlock))
+ // * (RVVBitsPerBlock / EltSize)
llvm::ScalableVectorType *ResType =
- llvm::ScalableVectorType::get(CGT.ConvertType(VT->getElementType()),
- llvm::RISCV::RVVBitsPerBlock / EltSize);
+ llvm::ScalableVectorType::get(CGT.ConvertType(VT->getElementType()),
+ VT->getNumElements() / VScale->first);
return ABIArgInfo::getDirect(ResType);
}
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
