[RISCV] Add canonical ISA string as Module metadata in IR. #80760
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
In an LTO build, we don't set the ELF attributes to indicate what extensions were compiled with. The target CPU/Attrs in RISCVTargetMachine do not get set for an LTO build. Each function gets a target-cpu/feature attribute, but this isn't usable to set ELF attributs since we wouldn't know what function to use. We can't just once since it might have been compiler with an attribute likes target_verson. This patch adds the ISA as Module metadata so we can retrieve it in the backend. Individual translation units can still be compiled with different strings so we need to collect the unique set when Modules are merged. The backend will need to combine the unique ISA strings to produce a single value for the ELF attributes. This will be done in a separate patch.
topperc
requested review from
asb,
MaskRay,
jrtc27,
preames,
kito-cheng and
ilovepi
llvmbot
added
clang
|
@llvm/pr-subscribers-clang @llvm/pr-subscribers-clang-codegen Author: Craig Topper (topperc) ChangesIn an LTO build, we don't set the ELF attributes to indicate what extensions were compiled with. The target CPU/Attrs in RISCVTargetMachine do not get set for an LTO build. Each function gets a target-cpu/feature attribute, but this isn't usable to set ELF attributs since we wouldn't know what function to use. We can't just once since it might have been compiler with an attribute likes target_verson. This patch adds the ISA as Module metadata so we can retrieve it in the backend. Individual translation units can still be compiled with different strings so we need to collect the unique set when Modules are merged. The backend will need to combine the unique ISA strings to produce a single value for the ELF attributes. This will be done in a separate patch. Patch is 49.81 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/80760.diff 3 Files Affected:
diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index 36b63d78b06f83..25b8d5cae36d3a 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -67,6 +67,7 @@
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/ConvertUTF.h"
#include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/RISCVISAInfo.h"
#include "llvm/Support/TimeProfiler.h"
#include "llvm/Support/xxhash.h"
#include "llvm/TargetParser/Triple.h"
@@ -1056,6 +1057,16 @@ void CodeGenModule::Release() {
llvm::LLVMContext &Ctx = TheModule.getContext();
getModule().addModuleFlag(llvm::Module::Error, "target-abi",
llvm::MDString::get(Ctx, ABIStr));
+
+ const std::vector<std::string> &Features =
+ getTarget().getTargetOpts().Features;
+ auto ParseResult =
+ llvm::RISCVISAInfo::parseFeatures(T.isRISCV64() ? 64 : 32, Features);
+ if (!errorToBool(ParseResult.takeError()))
+ getModule().addModuleFlag(
+ llvm::Module::AppendUnique, "riscv-arch",
+ llvm::MDNode::get(
+ Ctx, llvm::MDString::get(Ctx, (*ParseResult)->toString())));
}
if (CodeGenOpts.SanitizeCfiCrossDso) {
diff --git a/clang/test/CodeGen/RISCV/ntlh-intrinsics/riscv32-zihintntl.c b/clang/test/CodeGen/RISCV/ntlh-intrinsics/riscv32-zihintntl.c
index 897edbc6450af6..b11c2ca010e7ce 100644
--- a/clang/test/CodeGen/RISCV/ntlh-intrinsics/riscv32-zihintntl.c
+++ b/clang/test/CodeGen/RISCV/ntlh-intrinsics/riscv32-zihintntl.c
@@ -28,190 +28,190 @@ vint8m1_t *scvc1, *scvc2;
// clang-format off
void ntl_all_sizes() { // CHECK-LABEL: ntl_all_sizes
- uc = __riscv_ntl_load(&sc, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !5
- sc = __riscv_ntl_load(&uc, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !5
- us = __riscv_ntl_load(&ss, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !5
- ss = __riscv_ntl_load(&us, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !5
- ui = __riscv_ntl_load(&si, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !5
- si = __riscv_ntl_load(&ui, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !5
- ull = __riscv_ntl_load(&sll, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !5
- sll = __riscv_ntl_load(&ull, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !5
- h1 = __riscv_ntl_load(&h2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load half{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !5
- f1 = __riscv_ntl_load(&f2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load float{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !5
- d1 = __riscv_ntl_load(&d2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load double{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !5
- v4si1 = __riscv_ntl_load(&v4si2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <4 x i32>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !5
- v8ss1 = __riscv_ntl_load(&v8ss2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <8 x i16>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !5
- v16sc1 = __riscv_ntl_load(&v16sc2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <16 x i8>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !5
- *scvi1 = __riscv_ntl_load(scvi2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <vscale x 2 x i32>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !5
- *scvs1 = __riscv_ntl_load(scvs2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <vscale x 4 x i16>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !5
- *scvc1 = __riscv_ntl_load(scvc2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <vscale x 8 x i8>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !5
+ uc = __riscv_ntl_load(&sc, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i8{{.*}}align 1, !nontemporal !6, !riscv-nontemporal-domain !7
+ sc = __riscv_ntl_load(&uc, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i8{{.*}}align 1, !nontemporal !6, !riscv-nontemporal-domain !7
+ us = __riscv_ntl_load(&ss, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i16{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !7
+ ss = __riscv_ntl_load(&us, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i16{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !7
+ ui = __riscv_ntl_load(&si, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i32{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !7
+ si = __riscv_ntl_load(&ui, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i32{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !7
+ ull = __riscv_ntl_load(&sll, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i64{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !7
+ sll = __riscv_ntl_load(&ull, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i64{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !7
+ h1 = __riscv_ntl_load(&h2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load half{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !7
+ f1 = __riscv_ntl_load(&f2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load float{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !7
+ d1 = __riscv_ntl_load(&d2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load double{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !7
+ v4si1 = __riscv_ntl_load(&v4si2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <4 x i32>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !7
+ v8ss1 = __riscv_ntl_load(&v8ss2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <8 x i16>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !7
+ v16sc1 = __riscv_ntl_load(&v16sc2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <16 x i8>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !7
+ *scvi1 = __riscv_ntl_load(scvi2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <vscale x 2 x i32>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !7
+ *scvs1 = __riscv_ntl_load(scvs2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <vscale x 4 x i16>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !7
+ *scvc1 = __riscv_ntl_load(scvc2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <vscale x 8 x i8>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !7
- uc = __riscv_ntl_load(&sc, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !6
- sc = __riscv_ntl_load(&uc, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !6
- us = __riscv_ntl_load(&ss, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !6
- ss = __riscv_ntl_load(&us, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !6
- ui = __riscv_ntl_load(&si, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !6
- si = __riscv_ntl_load(&ui, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !6
- ull = __riscv_ntl_load(&sll, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !6
- sll = __riscv_ntl_load(&ull, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !6
- h1 = __riscv_ntl_load(&h2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load half{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !6
- f1 = __riscv_ntl_load(&f2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load float{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !6
- d1 = __riscv_ntl_load(&d2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load double{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !6
- v4si1 = __riscv_ntl_load(&v4si2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <4 x i32>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !6
- v8ss1 = __riscv_ntl_load(&v8ss2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <8 x i16>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !6
- v16sc1 = __riscv_ntl_load(&v16sc2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <16 x i8>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !6
- *scvi1 = __riscv_ntl_load(scvi2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <vscale x 2 x i32>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !6
- *scvs1 = __riscv_ntl_load(scvs2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <vscale x 4 x i16>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !6
- *scvc1 = __riscv_ntl_load(scvc2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <vscale x 8 x i8>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !6
+ uc = __riscv_ntl_load(&sc, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i8{{.*}}align 1, !nontemporal !6, !riscv-nontemporal-domain !8
+ sc = __riscv_ntl_load(&uc, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i8{{.*}}align 1, !nontemporal !6, !riscv-nontemporal-domain !8
+ us = __riscv_ntl_load(&ss, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i16{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !8
+ ss = __riscv_ntl_load(&us, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i16{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !8
+ ui = __riscv_ntl_load(&si, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i32{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !8
+ si = __riscv_ntl_load(&ui, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i32{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !8
+ ull = __riscv_ntl_load(&sll, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i64{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !8
+ sll = __riscv_ntl_load(&ull, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i64{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !8
+ h1 = __riscv_ntl_load(&h2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load half{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !8
+ f1 = __riscv_ntl_load(&f2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load float{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !8
+ d1 = __riscv_ntl_load(&d2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load double{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !8
+ v4si1 = __riscv_ntl_load(&v4si2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <4 x i32>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !8
+ v8ss1 = __riscv_ntl_load(&v8ss2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <8 x i16>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !8
+ v16sc1 = __riscv_ntl_load(&v16sc2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <16 x i8>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !8
+ *scvi1 = __riscv_ntl_load(scvi2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <vscale x 2 x i32>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !8
+ *scvs1 = __riscv_ntl_load(scvs2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <vscale x 4 x i16>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !8
+ *scvc1 = __riscv_ntl_load(scvc2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <vscale x 8 x i8>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !8
- uc = __riscv_ntl_load(&sc, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !7
- sc = __riscv_ntl_load(&uc, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !7
- us = __riscv_ntl_load(&ss, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !7
- ss = __riscv_ntl_load(&us, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !7
- ui = __riscv_ntl_load(&si, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !7
- si = __riscv_ntl_load(&ui, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !7
- ull = __riscv_ntl_load(&sll, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !7
- sll = __riscv_ntl_load(&ull, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !7
- h1 = __riscv_ntl_load(&h2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load half{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !7
- f1 = __riscv_ntl_load(&f2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load float{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !7
- d1 = __riscv_ntl_load(&d2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load double{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !7
- v4si1 = __riscv_ntl_load(&v4si2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <4 x i32>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !7
- v8ss1 = __riscv_ntl_load(&v8ss2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <8 x i16>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !7
- v16sc1 = __riscv_ntl_load(&v16sc2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <16 x i8>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !7
- *scvi1 = __riscv_ntl_load(scvi2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <vscale x 2 x i32>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !7
- *scvs1 = __riscv_ntl_load(scvs2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <vscale x 4 x i16>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !7
- *scvc1 = __riscv_ntl_load(scvc2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <vscale x 8 x i8>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !7
+ uc = __riscv_ntl_load(&sc, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i8{{.*}}align 1, !nontemporal !6, !riscv-nontemporal-domain !9
+ sc = __riscv_ntl_load(&uc, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i8{{.*}}align 1, !nontemporal !6, !riscv-nontemporal-domain !9
+ us = __riscv_ntl_load(&ss, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i16{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !9
+ ss = __riscv_ntl_load(&us, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i16{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !9
+ ui = __riscv_ntl_load(&si, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i32{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !9
+ si = __riscv_ntl_load(&ui, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i32{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !9
+ ull = __riscv_ntl_load(&sll, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i64{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !9
+ sll = __riscv_ntl_load(&ull, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i64{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !9
+ h1 = __riscv_ntl_load(&h2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load half{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !9
+ f1 = __riscv_ntl_load(&f2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load float{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !9
+ d1 = __riscv_ntl_load(&d2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load double{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !9
+ v4si1 = __riscv_ntl_load(&v4si2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <4 x i32>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !9
+ v8ss1 = __riscv_ntl_load(&v8ss2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <8 x i16>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !9
+ v16sc1 = __riscv_ntl_load(&v16sc2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <16 x i8>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !9
+ *scvi1 = __riscv_ntl_load(scvi2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <vscale x 2 x i32>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !9
+ *scvs1 = __riscv_ntl_load(scvs2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <vscale x 4 x i16>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !9
+ *scvc1 = __riscv_ntl_load(scvc2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <vscale x 8 x i8>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !9
- uc = __riscv_ntl_load(&sc, __RISCV_NTLH_ALL); // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !8
- sc = __riscv_ntl_load(&uc, __RISCV_NTLH_ALL); // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !8
- us = __riscv_ntl_load(&ss, __RISCV_NTLH_ALL); // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !8
- ss = __riscv_ntl_load(&us, __RISCV_NTLH_ALL); // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !8
- ui = __riscv_ntl_load(&si, __RISCV_NTLH_ALL); // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !8
- si = __riscv_ntl_load(&ui, __RISCV_NTLH_ALL); // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !8
- ull = __riscv_ntl_load(&sll, __RISCV_NTLH_ALL); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
- sll = __riscv_ntl_load(&ull, __RISCV_NTLH_ALL); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
- h1 = __riscv_ntl_load(&h2, __RISCV_NTLH_ALL); // CHECK: load half{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !8
- f1 = __riscv_ntl_load(&f2, __RISCV_NTLH_ALL); // CHECK: load float{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !8
- d1 = __riscv_ntl_load(&d2, __RISCV_NTLH_ALL); // CHECK: load double{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
- v4si1 = __riscv_ntl_load(&v4si2, __RISCV_NTLH_ALL); // CHECK: load <4 x i32>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !8
- v8ss1 = __riscv_ntl_load(&v8ss2, __RISCV_NTLH_ALL); // CHECK: load <8 x i16>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !8
- v16sc1 = __riscv_ntl_load(&v16sc2, __RISCV_NTLH_ALL); // CHECK: load <16 x i8>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !8
- *scvi1 = __riscv_ntl_load(scvi2, __RISCV_NTLH_ALL); // CHECK: load <vscale x 2 x i32>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
- *scvs1 = __riscv_ntl_load(scvs2, __RISCV_NTLH_ALL); // CHECK: load <vscale x 4 x i16>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
- *scvc1 = __riscv_ntl_load(scvc2, __RISCV_NTLH_ALL); // CHECK: load <vscale x 8 x i8>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
+ uc = __riscv_ntl_load(&sc, __RISCV_NTLH_ALL); // CHECK: load i8{{.*}}align 1, !nontemporal !6, !riscv-nontemporal-domain !10
+ sc = __riscv_ntl_load(&uc, __RISCV_NTLH_ALL); // CHECK: load i8{{.*}}align 1, !nontemporal !6, !riscv-nontemporal-domain !10
+ us = __riscv_ntl_load(&ss, __RISCV_NTLH_ALL); // CHECK: load i16{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !10
+ ss = __riscv_ntl_load(&us, __RISCV_NTLH_ALL); // CHECK: load i16{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !10
+ ui = __riscv_ntl_load(&si, __RISCV_NTLH_ALL); // CHECK: load i32{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !10
+ si = __riscv_ntl_load(&ui, __RISCV_NTLH_ALL); // CHECK: load i32{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !10
+ ull = __riscv_ntl_load(&sll, __RISCV_NTLH_ALL); // CHECK: load i64{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !10
+ sll = __riscv_ntl_load(&ull, __RISCV_NTLH_ALL); // CHECK: load i64{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !10
+ h1 = __riscv_ntl_load(&h2, __RISCV_NTLH_ALL); // CHECK: load half{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !10
+ f1 = __riscv_ntl_load(&f2, __RISCV_NTLH_ALL); // CHECK: load float{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !10
+ d1 = __riscv_ntl_load(&d2, __RISCV_NTL...
[truncated]
|
|
@llvm/pr-subscribers-backend-risc-v Author: Craig Topper (topperc) ChangesIn an LTO build, we don't set the ELF attributes to indicate what extensions were compiled with. The target CPU/Attrs in RISCVTargetMachine do not get set for an LTO build. Each function gets a target-cpu/feature attribute, but this isn't usable to set ELF attributs since we wouldn't know what function to use. We can't just once since it might have been compiler with an attribute likes target_verson. This patch adds the ISA as Module metadata so we can retrieve it in the backend. Individual translation units can still be compiled with different strings so we need to collect the unique set when Modules are merged. The backend will need to combine the unique ISA strings to produce a single value for the ELF attributes. This will be done in a separate patch. Patch is 49.80 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/80760.diff 3 Files Affected:
diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index 36b63d78b06f8..25b8d5cae36d3 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -67,6 +67,7 @@
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/ConvertUTF.h"
#include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/RISCVISAInfo.h"
#include "llvm/Support/TimeProfiler.h"
#include "llvm/Support/xxhash.h"
#include "llvm/TargetParser/Triple.h"
@@ -1056,6 +1057,16 @@ void CodeGenModule::Release() {
llvm::LLVMContext &Ctx = TheModule.getContext();
getModule().addModuleFlag(llvm::Module::Error, "target-abi",
llvm::MDString::get(Ctx, ABIStr));
+
+ const std::vector<std::string> &Features =
+ getTarget().getTargetOpts().Features;
+ auto ParseResult =
+ llvm::RISCVISAInfo::parseFeatures(T.isRISCV64() ? 64 : 32, Features);
+ if (!errorToBool(ParseResult.takeError()))
+ getModule().addModuleFlag(
+ llvm::Module::AppendUnique, "riscv-arch",
+ llvm::MDNode::get(
+ Ctx, llvm::MDString::get(Ctx, (*ParseResult)->toString())));
}
if (CodeGenOpts.SanitizeCfiCrossDso) {
diff --git a/clang/test/CodeGen/RISCV/ntlh-intrinsics/riscv32-zihintntl.c b/clang/test/CodeGen/RISCV/ntlh-intrinsics/riscv32-zihintntl.c
index 897edbc6450af..b11c2ca010e7c 100644
--- a/clang/test/CodeGen/RISCV/ntlh-intrinsics/riscv32-zihintntl.c
+++ b/clang/test/CodeGen/RISCV/ntlh-intrinsics/riscv32-zihintntl.c
@@ -28,190 +28,190 @@ vint8m1_t *scvc1, *scvc2;
// clang-format off
void ntl_all_sizes() { // CHECK-LABEL: ntl_all_sizes
- uc = __riscv_ntl_load(&sc, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !5
- sc = __riscv_ntl_load(&uc, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !5
- us = __riscv_ntl_load(&ss, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !5
- ss = __riscv_ntl_load(&us, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !5
- ui = __riscv_ntl_load(&si, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !5
- si = __riscv_ntl_load(&ui, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !5
- ull = __riscv_ntl_load(&sll, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !5
- sll = __riscv_ntl_load(&ull, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !5
- h1 = __riscv_ntl_load(&h2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load half{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !5
- f1 = __riscv_ntl_load(&f2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load float{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !5
- d1 = __riscv_ntl_load(&d2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load double{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !5
- v4si1 = __riscv_ntl_load(&v4si2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <4 x i32>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !5
- v8ss1 = __riscv_ntl_load(&v8ss2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <8 x i16>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !5
- v16sc1 = __riscv_ntl_load(&v16sc2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <16 x i8>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !5
- *scvi1 = __riscv_ntl_load(scvi2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <vscale x 2 x i32>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !5
- *scvs1 = __riscv_ntl_load(scvs2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <vscale x 4 x i16>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !5
- *scvc1 = __riscv_ntl_load(scvc2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <vscale x 8 x i8>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !5
+ uc = __riscv_ntl_load(&sc, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i8{{.*}}align 1, !nontemporal !6, !riscv-nontemporal-domain !7
+ sc = __riscv_ntl_load(&uc, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i8{{.*}}align 1, !nontemporal !6, !riscv-nontemporal-domain !7
+ us = __riscv_ntl_load(&ss, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i16{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !7
+ ss = __riscv_ntl_load(&us, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i16{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !7
+ ui = __riscv_ntl_load(&si, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i32{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !7
+ si = __riscv_ntl_load(&ui, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i32{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !7
+ ull = __riscv_ntl_load(&sll, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i64{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !7
+ sll = __riscv_ntl_load(&ull, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load i64{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !7
+ h1 = __riscv_ntl_load(&h2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load half{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !7
+ f1 = __riscv_ntl_load(&f2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load float{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !7
+ d1 = __riscv_ntl_load(&d2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load double{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !7
+ v4si1 = __riscv_ntl_load(&v4si2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <4 x i32>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !7
+ v8ss1 = __riscv_ntl_load(&v8ss2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <8 x i16>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !7
+ v16sc1 = __riscv_ntl_load(&v16sc2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <16 x i8>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !7
+ *scvi1 = __riscv_ntl_load(scvi2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <vscale x 2 x i32>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !7
+ *scvs1 = __riscv_ntl_load(scvs2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <vscale x 4 x i16>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !7
+ *scvc1 = __riscv_ntl_load(scvc2, __RISCV_NTLH_INNERMOST_PRIVATE); // CHECK: load <vscale x 8 x i8>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !7
- uc = __riscv_ntl_load(&sc, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !6
- sc = __riscv_ntl_load(&uc, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !6
- us = __riscv_ntl_load(&ss, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !6
- ss = __riscv_ntl_load(&us, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !6
- ui = __riscv_ntl_load(&si, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !6
- si = __riscv_ntl_load(&ui, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !6
- ull = __riscv_ntl_load(&sll, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !6
- sll = __riscv_ntl_load(&ull, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !6
- h1 = __riscv_ntl_load(&h2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load half{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !6
- f1 = __riscv_ntl_load(&f2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load float{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !6
- d1 = __riscv_ntl_load(&d2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load double{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !6
- v4si1 = __riscv_ntl_load(&v4si2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <4 x i32>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !6
- v8ss1 = __riscv_ntl_load(&v8ss2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <8 x i16>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !6
- v16sc1 = __riscv_ntl_load(&v16sc2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <16 x i8>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !6
- *scvi1 = __riscv_ntl_load(scvi2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <vscale x 2 x i32>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !6
- *scvs1 = __riscv_ntl_load(scvs2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <vscale x 4 x i16>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !6
- *scvc1 = __riscv_ntl_load(scvc2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <vscale x 8 x i8>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !6
+ uc = __riscv_ntl_load(&sc, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i8{{.*}}align 1, !nontemporal !6, !riscv-nontemporal-domain !8
+ sc = __riscv_ntl_load(&uc, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i8{{.*}}align 1, !nontemporal !6, !riscv-nontemporal-domain !8
+ us = __riscv_ntl_load(&ss, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i16{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !8
+ ss = __riscv_ntl_load(&us, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i16{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !8
+ ui = __riscv_ntl_load(&si, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i32{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !8
+ si = __riscv_ntl_load(&ui, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i32{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !8
+ ull = __riscv_ntl_load(&sll, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i64{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !8
+ sll = __riscv_ntl_load(&ull, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load i64{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !8
+ h1 = __riscv_ntl_load(&h2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load half{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !8
+ f1 = __riscv_ntl_load(&f2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load float{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !8
+ d1 = __riscv_ntl_load(&d2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load double{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !8
+ v4si1 = __riscv_ntl_load(&v4si2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <4 x i32>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !8
+ v8ss1 = __riscv_ntl_load(&v8ss2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <8 x i16>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !8
+ v16sc1 = __riscv_ntl_load(&v16sc2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <16 x i8>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !8
+ *scvi1 = __riscv_ntl_load(scvi2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <vscale x 2 x i32>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !8
+ *scvs1 = __riscv_ntl_load(scvs2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <vscale x 4 x i16>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !8
+ *scvc1 = __riscv_ntl_load(scvc2, __RISCV_NTLH_ALL_PRIVATE); // CHECK: load <vscale x 8 x i8>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !8
- uc = __riscv_ntl_load(&sc, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !7
- sc = __riscv_ntl_load(&uc, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !7
- us = __riscv_ntl_load(&ss, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !7
- ss = __riscv_ntl_load(&us, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !7
- ui = __riscv_ntl_load(&si, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !7
- si = __riscv_ntl_load(&ui, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !7
- ull = __riscv_ntl_load(&sll, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !7
- sll = __riscv_ntl_load(&ull, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !7
- h1 = __riscv_ntl_load(&h2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load half{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !7
- f1 = __riscv_ntl_load(&f2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load float{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !7
- d1 = __riscv_ntl_load(&d2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load double{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !7
- v4si1 = __riscv_ntl_load(&v4si2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <4 x i32>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !7
- v8ss1 = __riscv_ntl_load(&v8ss2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <8 x i16>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !7
- v16sc1 = __riscv_ntl_load(&v16sc2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <16 x i8>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !7
- *scvi1 = __riscv_ntl_load(scvi2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <vscale x 2 x i32>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !7
- *scvs1 = __riscv_ntl_load(scvs2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <vscale x 4 x i16>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !7
- *scvc1 = __riscv_ntl_load(scvc2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <vscale x 8 x i8>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !7
+ uc = __riscv_ntl_load(&sc, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i8{{.*}}align 1, !nontemporal !6, !riscv-nontemporal-domain !9
+ sc = __riscv_ntl_load(&uc, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i8{{.*}}align 1, !nontemporal !6, !riscv-nontemporal-domain !9
+ us = __riscv_ntl_load(&ss, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i16{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !9
+ ss = __riscv_ntl_load(&us, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i16{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !9
+ ui = __riscv_ntl_load(&si, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i32{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !9
+ si = __riscv_ntl_load(&ui, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i32{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !9
+ ull = __riscv_ntl_load(&sll, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i64{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !9
+ sll = __riscv_ntl_load(&ull, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load i64{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !9
+ h1 = __riscv_ntl_load(&h2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load half{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !9
+ f1 = __riscv_ntl_load(&f2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load float{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !9
+ d1 = __riscv_ntl_load(&d2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load double{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !9
+ v4si1 = __riscv_ntl_load(&v4si2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <4 x i32>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !9
+ v8ss1 = __riscv_ntl_load(&v8ss2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <8 x i16>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !9
+ v16sc1 = __riscv_ntl_load(&v16sc2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <16 x i8>{{.*}}align 16, !nontemporal !6, !riscv-nontemporal-domain !9
+ *scvi1 = __riscv_ntl_load(scvi2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <vscale x 2 x i32>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !9
+ *scvs1 = __riscv_ntl_load(scvs2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <vscale x 4 x i16>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !9
+ *scvc1 = __riscv_ntl_load(scvc2, __RISCV_NTLH_INNERMOST_SHARED); // CHECK: load <vscale x 8 x i8>{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !9
- uc = __riscv_ntl_load(&sc, __RISCV_NTLH_ALL); // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !8
- sc = __riscv_ntl_load(&uc, __RISCV_NTLH_ALL); // CHECK: load i8{{.*}}align 1, !nontemporal !4, !riscv-nontemporal-domain !8
- us = __riscv_ntl_load(&ss, __RISCV_NTLH_ALL); // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !8
- ss = __riscv_ntl_load(&us, __RISCV_NTLH_ALL); // CHECK: load i16{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !8
- ui = __riscv_ntl_load(&si, __RISCV_NTLH_ALL); // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !8
- si = __riscv_ntl_load(&ui, __RISCV_NTLH_ALL); // CHECK: load i32{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !8
- ull = __riscv_ntl_load(&sll, __RISCV_NTLH_ALL); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
- sll = __riscv_ntl_load(&ull, __RISCV_NTLH_ALL); // CHECK: load i64{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
- h1 = __riscv_ntl_load(&h2, __RISCV_NTLH_ALL); // CHECK: load half{{.*}}align 2, !nontemporal !4, !riscv-nontemporal-domain !8
- f1 = __riscv_ntl_load(&f2, __RISCV_NTLH_ALL); // CHECK: load float{{.*}}align 4, !nontemporal !4, !riscv-nontemporal-domain !8
- d1 = __riscv_ntl_load(&d2, __RISCV_NTLH_ALL); // CHECK: load double{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
- v4si1 = __riscv_ntl_load(&v4si2, __RISCV_NTLH_ALL); // CHECK: load <4 x i32>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !8
- v8ss1 = __riscv_ntl_load(&v8ss2, __RISCV_NTLH_ALL); // CHECK: load <8 x i16>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !8
- v16sc1 = __riscv_ntl_load(&v16sc2, __RISCV_NTLH_ALL); // CHECK: load <16 x i8>{{.*}}align 16, !nontemporal !4, !riscv-nontemporal-domain !8
- *scvi1 = __riscv_ntl_load(scvi2, __RISCV_NTLH_ALL); // CHECK: load <vscale x 2 x i32>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
- *scvs1 = __riscv_ntl_load(scvs2, __RISCV_NTLH_ALL); // CHECK: load <vscale x 4 x i16>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
- *scvc1 = __riscv_ntl_load(scvc2, __RISCV_NTLH_ALL); // CHECK: load <vscale x 8 x i8>{{.*}}align 8, !nontemporal !4, !riscv-nontemporal-domain !8
+ uc = __riscv_ntl_load(&sc, __RISCV_NTLH_ALL); // CHECK: load i8{{.*}}align 1, !nontemporal !6, !riscv-nontemporal-domain !10
+ sc = __riscv_ntl_load(&uc, __RISCV_NTLH_ALL); // CHECK: load i8{{.*}}align 1, !nontemporal !6, !riscv-nontemporal-domain !10
+ us = __riscv_ntl_load(&ss, __RISCV_NTLH_ALL); // CHECK: load i16{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !10
+ ss = __riscv_ntl_load(&us, __RISCV_NTLH_ALL); // CHECK: load i16{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !10
+ ui = __riscv_ntl_load(&si, __RISCV_NTLH_ALL); // CHECK: load i32{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !10
+ si = __riscv_ntl_load(&ui, __RISCV_NTLH_ALL); // CHECK: load i32{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !10
+ ull = __riscv_ntl_load(&sll, __RISCV_NTLH_ALL); // CHECK: load i64{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !10
+ sll = __riscv_ntl_load(&ull, __RISCV_NTLH_ALL); // CHECK: load i64{{.*}}align 8, !nontemporal !6, !riscv-nontemporal-domain !10
+ h1 = __riscv_ntl_load(&h2, __RISCV_NTLH_ALL); // CHECK: load half{{.*}}align 2, !nontemporal !6, !riscv-nontemporal-domain !10
+ f1 = __riscv_ntl_load(&f2, __RISCV_NTLH_ALL); // CHECK: load float{{.*}}align 4, !nontemporal !6, !riscv-nontemporal-domain !10
+ d1 = __riscv_ntl_load(&d2, __RISCV_NTLH_AL...
[truncated]
|
ilovepi
approved these changes
Feb 6, 2024
Maybe other tests are better about using regexes to hide the numbers? |
|
My thoughts on this in the past have been:
RISC-V is the most sensitive to this, but it's just a symptom of not having the general mechanism in place, and I'm not a huge fan of adding ad-hoc stuff specific to RISC-V rather than addressing the underlying problem. Other targets just care less upstream today so get away with this lack of information in practice, but that won't always true (e.g. I believe we need this information on Morello (CHERI+AArch64 prototype) downstream, not just for CHERI-RISC-V). |
|
I agree that finding a generic cross-target solution to this problem is attractive...but I worry it will be hard and slow to proceed. It's possible the previous discussions just lacked someone following up with concrete patches, but I do worry we'd end up stuck in limbo vs starting with a target-specific approach that we later hope to replace with something more generic. With that in mind:
Two simple and early pieces of feedback on the patch:
|
topperc
changed the title
We probably need it to set the I just noticed we only set EF_RISCV_RVC based on C, and not Zca. Is that a bug? |
I think so; binutils sets it for both. |
(R_RISCV_ALIGN can probably fail if you don't, I don't think it's just a missing optimisation) |
|
Does the name of "riscv-arch" give the impression that it might be "march" string? |
|
I agree with @asb's framing above. Assuming this doesn't commit us to something which is hard to forward version for some reason, I support addressing this in a target specific manner for the moment. |
|
That's fair. I just imagine this is going to result in people forgetting about the underlying issue and adding their own patchwork fixes to work around it in their targets, so it would be a nice motivation to fix it properly. |
|
I share a similar concern about not addressing the issue in a target-specific manner. However, I know a lot of confused users by the behavior, and this patch will significantly improve the status quo. I concur with asb's analysis. |
MaskRay
approved these changes
Feb 13, 2024
topperc
added a commit
to topperc/llvm-project
that referenced
this pull request
May 8, 2024
In an LTO build, we don't set the ELF attributes to indicate what extensions were compiled with. The target CPU/Attrs in RISCVTargetMachine do not get set for an LTO build. Each function gets a target-cpu/feature attribute, but this isn't usable to set ELF attributs since we wouldn't know what function to use. We can't just once since it might have been compiler with an attribute likes target_verson. This patch adds the ISA as Module metadata so we can retrieve it in the backend. Individual translation units can still be compiled with different strings so we need to collect the unique set when Modules are merged. The backend will need to combine the unique ISA strings to produce a single value for the ELF attributes. This will be done in a separate patch.
topperc
added a commit
to topperc/llvm-project
that referenced
this pull request
May 8, 2024
In an LTO build, we don't set the ELF attributes to indicate what extensions were compiled with. The target CPU/Attrs in RISCVTargetMachine do not get set for an LTO build. Each function gets a target-cpu/feature attribute, but this isn't usable to set ELF attributs since we wouldn't know what function to use. We can't just once since it might have been compiler with an attribute likes target_verson. This patch adds the ISA as Module metadata so we can retrieve it in the backend. Individual translation units can still be compiled with different strings so we need to collect the unique set when Modules are merged. The backend will need to combine the unique ISA strings to produce a single value for the ELF attributes. This will be done in a separate patch.
tstellar
pushed a commit
to topperc/llvm-project
that referenced
this pull request
May 13, 2024
In an LTO build, we don't set the ELF attributes to indicate what extensions were compiled with. The target CPU/Attrs in RISCVTargetMachine do not get set for an LTO build. Each function gets a target-cpu/feature attribute, but this isn't usable to set ELF attributs since we wouldn't know what function to use. We can't just once since it might have been compiler with an attribute likes target_verson. This patch adds the ISA as Module metadata so we can retrieve it in the backend. Individual translation units can still be compiled with different strings so we need to collect the unique set when Modules are merged. The backend will need to combine the unique ISA strings to produce a single value for the ELF attributes. This will be done in a separate patch.
tstellar
pushed a commit
to topperc/llvm-project
that referenced
this pull request
May 14, 2024
In an LTO build, we don't set the ELF attributes to indicate what extensions were compiled with. The target CPU/Attrs in RISCVTargetMachine do not get set for an LTO build. Each function gets a target-cpu/feature attribute, but this isn't usable to set ELF attributs since we wouldn't know what function to use. We can't just once since it might have been compiler with an attribute likes target_verson. This patch adds the ISA as Module metadata so we can retrieve it in the backend. Individual translation units can still be compiled with different strings so we need to collect the unique set when Modules are merged. The backend will need to combine the unique ISA strings to produce a single value for the ELF attributes. This will be done in a separate patch.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
In an LTO build, we don't set the ELF attributes to indicate what extensions were compiled with. The target CPU/Attrs in RISCVTargetMachine do not get set for an LTO build. Each function gets a target-cpu/feature attribute, but this isn't usable to set ELF attributs since we wouldn't know what function to use. We can't just once since it might have been compiler with an attribute likes target_verson.
This patch adds the ISA as Module metadata so we can retrieve it in the backend. Individual translation units can still be compiled with different strings so we need to collect the unique set when Modules are merged.
The backend will need to combine the unique ISA strings to produce a single value for the ELF attributes. This will be done in a separate patch.