[clang] 8ab7718 - [clang][RISCV] Make the index of riscv_tuple_extract and riscv_tuple_insert be truncated rather than zero extended (#111466)

[via cfe-commits]

Author: Brandon Wu
Date: 2024-10-08T08:28:57-07:00
New Revision: 8ab77184dde2583950fc6e4886ff526e7e598f7e

URL: https://github.com/llvm/llvm-project/commit/8ab77184dde2583950fc6e4886ff526e7e598f7e
DIFF: https://github.com/llvm/llvm-project/commit/8ab77184dde2583950fc6e4886ff526e7e598f7e.diff

LOG: [clang][RISCV] Make the index of riscv_tuple_extract and riscv_tuple_insert be truncated rather than zero extended (#111466)

It's illegal if the index is 64 bits and is zero-extend to 32 bits.

Added: 
    clang/test/CodeGen/RISCV/rvv-intrinsics-handcrafted/rvv-tuple-utils.c

Modified: 
    clang/include/clang/Basic/riscv_vector.td

Removed: 
    


################################################################################
diff  --git a/clang/include/clang/Basic/riscv_vector.td b/clang/include/clang/Basic/riscv_vector.td
index 6e57e51793a712..1387494c4aeb00 100644
--- a/clang/include/clang/Basic/riscv_vector.td
+++ b/clang/include/clang/Basic/riscv_vector.td
@@ -2564,7 +2564,7 @@ let HasMasked = false, HasVL = false, IRName = "" in {
 
         return Builder.CreateIntrinsic(Intrinsic::riscv_tuple_extract,
                                        {ResultType, Ops[0]->getType()},
-                                       {Ops[0], Builder.CreateZExt(Ops[1],
+                                       {Ops[0], Builder.CreateTrunc(Ops[1],
                                         Builder.getInt32Ty())});
       }
       }] in {
@@ -2606,7 +2606,7 @@ let HasMasked = false, HasVL = false, IRName = "" in {
         return Builder.CreateIntrinsic(Intrinsic::riscv_tuple_insert,
                                        {ResultType, Ops[2]->getType()},
                                        {Ops[0], Ops[2],
-                                        Builder.CreateZExt(Ops[1],Builder.getInt32Ty())});
+                                        Builder.CreateTrunc(Ops[1],Builder.getInt32Ty())});
       }
       }] in {
     foreach dst_lmul = ["(LFixedLog2LMUL:1)", "(LFixedLog2LMUL:2)", "(LFixedLog2LMUL:3)"] in {

diff  --git a/clang/test/CodeGen/RISCV/rvv-intrinsics-handcrafted/rvv-tuple-utils.c b/clang/test/CodeGen/RISCV/rvv-intrinsics-handcrafted/rvv-tuple-utils.c
new file mode 100644
index 00000000000000..531cfb4d3e6260
--- /dev/null
+++ b/clang/test/CodeGen/RISCV/rvv-intrinsics-handcrafted/rvv-tuple-utils.c
@@ -0,0 +1,28 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 2
+// REQUIRES: riscv-registered-target
+// RUN: %clang_cc1 -triple riscv64 -target-feature +v -target-feature +zfh \
+// RUN:   -target-feature +zvfhmin -disable-O0-optnone  \
+// RUN:   -emit-llvm %s -o - | opt -S -passes=mem2reg | \
+// RUN:   FileCheck --check-prefix=CHECK-RV64 %s
+
+#include <riscv_vector.h>
+
+// CHECK-RV64-LABEL: define dso_local <vscale x 4 x i64> @test_vget_v_u64m4x2_u64m4
+// CHECK-RV64-SAME: (target("riscv.vector.tuple", <vscale x 32 x i8>, 2) [[SRC:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call <vscale x 4 x i64> @llvm.riscv.tuple.extract.nxv4i64.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) [[SRC]], i32 1)
+// CHECK-RV64-NEXT:    ret <vscale x 4 x i64> [[TMP0]]
+//
+vuint64m4_t test_vget_v_u64m4x2_u64m4(vuint64m4x2_t src) {
+  return __riscv_vget_v_u64m4x2_u64m4(src, 1);
+}
+
+// CHECK-RV64-LABEL: define dso_local target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @test_vset_v_u64m4_u64m4x2
+// CHECK-RV64-SAME: (target("riscv.vector.tuple", <vscale x 32 x i8>, 2) [[DEST:%.*]], <vscale x 4 x i64> [[VAL:%.*]]) #[[ATTR0]] {
+// CHECK-RV64-NEXT:  entry:
+// CHECK-RV64-NEXT:    [[TMP0:%.*]] = call target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @llvm.riscv.tuple.insert.triscv.vector.tuple_nxv32i8_2t.nxv4i64(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) [[DEST]], <vscale x 4 x i64> [[VAL]], i32 1)
+// CHECK-RV64-NEXT:    ret target("riscv.vector.tuple", <vscale x 32 x i8>, 2) [[TMP0]]
+//
+vuint64m4x2_t test_vset_v_u64m4_u64m4x2(vuint64m4x2_t dest, vuint64m4_t val) {
+  return __riscv_vset_v_u64m4_u64m4x2(dest, 1, val);
+}