[clang] [llvm] [RISCV][FMV] Support target_clones (PR #85786)
Piyou Chen via cfe-commits
cfe-commits at lists.llvm.org
Wed Jul 17 23:07:08 PDT 2024
https://github.com/BeMg updated https://github.com/llvm/llvm-project/pull/85786
>From 38e5bef5d20d7e81170720eb56354c0392f5c902 Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Wed, 5 Jun 2024 01:17:03 -0700
Subject: [PATCH 01/19] [RISCV] Add groupid/bitmask for RISC-V extension
Base on https://github.com/riscv-non-isa/riscv-c-api-doc/pull/74.
This patch defines the groupid/bitmask in RISCVFeatures.td and generates the corresponding table in RISCVTargetParserDef.inc.
The groupid/bitmask of extensions provides an abstraction layer between the compiler and runtime functions.
---
.../llvm/TargetParser/RISCVTargetParser.h | 8 ++
llvm/lib/Target/RISCV/RISCVFeatures.td | 131 ++++++++++++------
llvm/lib/TargetParser/RISCVTargetParser.cpp | 32 +++++
llvm/test/TableGen/riscv-target-def.td | 18 ++-
llvm/utils/TableGen/RISCVTargetDefEmitter.cpp | 37 +++++
5 files changed, 183 insertions(+), 43 deletions(-)
diff --git a/llvm/include/llvm/TargetParser/RISCVTargetParser.h b/llvm/include/llvm/TargetParser/RISCVTargetParser.h
index 5b1494efe7bdc..e998bc4ca59ee 100644
--- a/llvm/include/llvm/TargetParser/RISCVTargetParser.h
+++ b/llvm/include/llvm/TargetParser/RISCVTargetParser.h
@@ -24,6 +24,14 @@ class Triple;
namespace RISCV {
+namespace RISCVExtensionBitmaskTable {
+struct RISCVExtensionBitmask {
+ const char *Name;
+ unsigned GroupID;
+ unsigned BitPosition;
+};
+} // namespace RISCVExtensionBitmaskTable
+
// We use 64 bits as the known part in the scalable vector types.
static constexpr unsigned RVVBitsPerBlock = 64;
diff --git a/llvm/lib/Target/RISCV/RISCVFeatures.td b/llvm/lib/Target/RISCV/RISCVFeatures.td
index d3cb2aeab41cb..125e9e2f53953 100644
--- a/llvm/lib/Target/RISCV/RISCVFeatures.td
+++ b/llvm/lib/Target/RISCV/RISCVFeatures.td
@@ -37,6 +37,15 @@ class RISCVExtension<string name, int major, int minor, string desc,
bit Experimental = false;
}
+// The groupID/bitmask of RISCVExtension is used to retrieve a specific bit value
+// from __riscv_feature_bit based on the groupID and bitmask.
+// groupID - groupID of extension
+// bitmaskShift - bitmask shift of extension
+class RISCVExtensionBitmask<bits<3> groupID, int bitmaskShift> {
+ int GroupID = groupID;
+ int BitPos = bitmaskShift;
+}
+
// Version of RISCVExtension to be used for Experimental extensions. This
// sets the Experimental flag and prepends experimental- to the -mattr name.
class RISCVExperimentalExtension<string name, int major, int minor, string desc,
@@ -52,7 +61,8 @@ class RISCVExperimentalExtension<string name, int major, int minor, string desc,
def FeatureStdExtI
: RISCVExtension<"i", 2, 1,
- "'I' (Base Integer Instruction Set)">;
+ "'I' (Base Integer Instruction Set)">,
+ RISCVExtensionBitmask<0, 8>;
def FeatureStdExtE
: RISCVExtension<"e", 2, 0,
@@ -78,7 +88,8 @@ def HasStdExtZicbop : Predicate<"Subtarget->hasStdExtZicbop()">,
def FeatureStdExtZicboz
: RISCVExtension<"zicboz", 1, 0,
- "'Zicboz' (Cache-Block Zero Instructions)">;
+ "'Zicboz' (Cache-Block Zero Instructions)">,
+ RISCVExtensionBitmask<0, 37>;
def HasStdExtZicboz : Predicate<"Subtarget->hasStdExtZicboz()">,
AssemblerPredicate<(all_of FeatureStdExtZicboz),
"'Zicboz' (Cache-Block Zero Instructions)">;
@@ -113,7 +124,8 @@ def FeatureStdExtZicntr
def FeatureStdExtZicond
: RISCVExtension<"zicond", 1, 0,
- "'Zicond' (Integer Conditional Operations)">;
+ "'Zicond' (Integer Conditional Operations)">,
+ RISCVExtensionBitmask<0, 38>;
def HasStdExtZicond : Predicate<"Subtarget->hasStdExtZicond()">,
AssemblerPredicate<(all_of FeatureStdExtZicond),
"'Zicond' (Integer Conditional Operations)">;
@@ -134,7 +146,8 @@ def HasStdExtZihintpause : Predicate<"Subtarget->hasStdExtZihintpause()">,
def FeatureStdExtZihintntl
: RISCVExtension<"zihintntl", 1, 0,
- "'Zihintntl' (Non-Temporal Locality Hints)">;
+ "'Zihintntl' (Non-Temporal Locality Hints)">,
+ RISCVExtensionBitmask<0, 39>;
def HasStdExtZihintntl : Predicate<"Subtarget->hasStdExtZihintntl()">,
AssemblerPredicate<(all_of FeatureStdExtZihintntl),
"'Zihintntl' (Non-Temporal Locality Hints)">;
@@ -181,7 +194,8 @@ def HasStdExtZmmul : Predicate<"Subtarget->hasStdExtZmmul()">,
def FeatureStdExtM
: RISCVExtension<"m", 2, 0,
"'M' (Integer Multiplication and Division)",
- [FeatureStdExtZmmul]>;
+ [FeatureStdExtZmmul]>,
+ RISCVExtensionBitmask<0, 12>;
def HasStdExtM : Predicate<"Subtarget->hasStdExtM()">,
AssemblerPredicate<(all_of FeatureStdExtM),
"'M' (Integer Multiplication and Division)">;
@@ -190,14 +204,16 @@ def HasStdExtM : Predicate<"Subtarget->hasStdExtM()">,
def FeatureStdExtA
: RISCVExtension<"a", 2, 1,
- "'A' (Atomic Instructions)">;
+ "'A' (Atomic Instructions)">,
+ RISCVExtensionBitmask<0, 0>;
def HasStdExtA : Predicate<"Subtarget->hasStdExtA()">,
AssemblerPredicate<(all_of FeatureStdExtA),
"'A' (Atomic Instructions)">;
def FeatureStdExtZtso
: RISCVExtension<"ztso", 1, 0,
- "'Ztso' (Memory Model - Total Store Order)">;
+ "'Ztso' (Memory Model - Total Store Order)">,
+ RISCVExtensionBitmask<0, 47>;
def HasStdExtZtso : Predicate<"Subtarget->hasStdExtZtso()">,
AssemblerPredicate<(all_of FeatureStdExtZtso),
"'Ztso' (Memory Model - Total Store Order)">;
@@ -227,7 +243,8 @@ def HasStdExtZabha : Predicate<"Subtarget->hasStdExtZabha()">,
def FeatureStdExtZacas
: RISCVExtension<"zacas", 1, 0,
- "'Zacas' (Atomic Compare-And-Swap Instructions)">;
+ "'Zacas' (Atomic Compare-And-Swap Instructions)">,
+ RISCVExtensionBitmask<0, 26>;
def HasStdExtZacas : Predicate<"Subtarget->hasStdExtZacas()">,
AssemblerPredicate<(all_of FeatureStdExtZacas),
"'Zacas' (Atomic Compare-And-Swap Instructions)">;
@@ -264,7 +281,8 @@ def HasStdExtZawrs : Predicate<"Subtarget->hasStdExtZawrs()">,
def FeatureStdExtF
: RISCVExtension<"f", 2, 2,
"'F' (Single-Precision Floating-Point)",
- [FeatureStdExtZicsr]>;
+ [FeatureStdExtZicsr]>,
+ RISCVExtensionBitmask<0, 5>;
def HasStdExtF : Predicate<"Subtarget->hasStdExtF()">,
AssemblerPredicate<(all_of FeatureStdExtF),
"'F' (Single-Precision Floating-Point)">;
@@ -272,7 +290,8 @@ def HasStdExtF : Predicate<"Subtarget->hasStdExtF()">,
def FeatureStdExtD
: RISCVExtension<"d", 2, 2,
"'D' (Double-Precision Floating-Point)",
- [FeatureStdExtF]>;
+ [FeatureStdExtF]>,
+ RISCVExtensionBitmask<0, 3>;
def HasStdExtD : Predicate<"Subtarget->hasStdExtD()">,
AssemblerPredicate<(all_of FeatureStdExtD),
"'D' (Double-Precision Floating-Point)">;
@@ -280,7 +299,8 @@ def HasStdExtD : Predicate<"Subtarget->hasStdExtD()">,
def FeatureStdExtZfhmin
: RISCVExtension<"zfhmin", 1, 0,
"'Zfhmin' (Half-Precision Floating-Point Minimal)",
- [FeatureStdExtF]>;
+ [FeatureStdExtF]>,
+ RISCVExtensionBitmask<0, 36>;
def HasStdExtZfhmin : Predicate<"Subtarget->hasStdExtZfhmin()">,
AssemblerPredicate<(all_of FeatureStdExtZfhmin),
"'Zfh' (Half-Precision Floating-Point) or "
@@ -289,7 +309,8 @@ def HasStdExtZfhmin : Predicate<"Subtarget->hasStdExtZfhmin()">,
def FeatureStdExtZfh
: RISCVExtension<"zfh", 1, 0,
"'Zfh' (Half-Precision Floating-Point)",
- [FeatureStdExtZfhmin]>;
+ [FeatureStdExtZfhmin]>,
+ RISCVExtensionBitmask<0, 35>;
def HasStdExtZfh : Predicate<"Subtarget->hasStdExtZfh()">,
AssemblerPredicate<(all_of FeatureStdExtZfh),
"'Zfh' (Half-Precision Floating-Point)">;
@@ -313,7 +334,8 @@ def HasHalfFPLoadStoreMove
def FeatureStdExtZfa
: RISCVExtension<"zfa", 1, 0,
"'Zfa' (Additional Floating-Point)",
- [FeatureStdExtF]>;
+ [FeatureStdExtF]>,
+ RISCVExtensionBitmask<0, 34>;
def HasStdExtZfa : Predicate<"Subtarget->hasStdExtZfa()">,
AssemblerPredicate<(all_of FeatureStdExtZfa),
"'Zfa' (Additional Floating-Point)">;
@@ -356,7 +378,8 @@ def NoStdExtZhinx : Predicate<"!Subtarget->hasStdExtZhinx()">;
def FeatureStdExtC
: RISCVExtension<"c", 2, 0,
- "'C' (Compressed Instructions)">;
+ "'C' (Compressed Instructions)">,
+ RISCVExtensionBitmask<0, 2>;
def HasStdExtC : Predicate<"Subtarget->hasStdExtC()">,
AssemblerPredicate<(all_of FeatureStdExtC),
"'C' (Compressed Instructions)">;
@@ -445,7 +468,8 @@ def HasStdExtZcmop : Predicate<"Subtarget->hasStdExtZcmop()">,
def FeatureStdExtZba
: RISCVExtension<"zba", 1, 0,
- "'Zba' (Address Generation Instructions)">;
+ "'Zba' (Address Generation Instructions)">,
+ RISCVExtensionBitmask<0, 27>;
def HasStdExtZba : Predicate<"Subtarget->hasStdExtZba()">,
AssemblerPredicate<(all_of FeatureStdExtZba),
"'Zba' (Address Generation Instructions)">;
@@ -453,7 +477,8 @@ def NotHasStdExtZba : Predicate<"!Subtarget->hasStdExtZba()">;
def FeatureStdExtZbb
: RISCVExtension<"zbb", 1, 0,
- "'Zbb' (Basic Bit-Manipulation)">;
+ "'Zbb' (Basic Bit-Manipulation)">,
+ RISCVExtensionBitmask<0, 28>;
def HasStdExtZbb : Predicate<"Subtarget->hasStdExtZbb()">,
AssemblerPredicate<(all_of FeatureStdExtZbb),
"'Zbb' (Basic Bit-Manipulation)">;
@@ -462,14 +487,16 @@ def NoStdExtZbb : Predicate<"!Subtarget->hasStdExtZbb()">,
def FeatureStdExtZbc
: RISCVExtension<"zbc", 1, 0,
- "'Zbc' (Carry-Less Multiplication)">;
+ "'Zbc' (Carry-Less Multiplication)">,
+ RISCVExtensionBitmask<0, 29>;
def HasStdExtZbc : Predicate<"Subtarget->hasStdExtZbc()">,
AssemblerPredicate<(all_of FeatureStdExtZbc),
"'Zbc' (Carry-Less Multiplication)">;
def FeatureStdExtZbs
: RISCVExtension<"zbs", 1, 0,
- "'Zbs' (Single-Bit Instructions)">;
+ "'Zbs' (Single-Bit Instructions)">,
+ RISCVExtensionBitmask<0, 33>;
def HasStdExtZbs : Predicate<"Subtarget->hasStdExtZbs()">,
AssemblerPredicate<(all_of FeatureStdExtZbs),
"'Zbs' (Single-Bit Instructions)">;
@@ -486,14 +513,16 @@ def HasStdExtB : Predicate<"Subtarget->hasStdExtB()">,
def FeatureStdExtZbkb
: RISCVExtension<"zbkb", 1, 0,
- "'Zbkb' (Bitmanip instructions for Cryptography)">;
+ "'Zbkb' (Bitmanip instructions for Cryptography)">,
+ RISCVExtensionBitmask<0, 30>;
def HasStdExtZbkb : Predicate<"Subtarget->hasStdExtZbkb()">,
AssemblerPredicate<(all_of FeatureStdExtZbkb),
"'Zbkb' (Bitmanip instructions for Cryptography)">;
def FeatureStdExtZbkx
: RISCVExtension<"zbkx", 1, 0,
- "'Zbkx' (Crossbar permutation instructions)">;
+ "'Zbkx' (Crossbar permutation instructions)">,
+ RISCVExtensionBitmask<0, 32>;
def HasStdExtZbkx : Predicate<"Subtarget->hasStdExtZbkx()">,
AssemblerPredicate<(all_of FeatureStdExtZbkx),
"'Zbkx' (Crossbar permutation instructions)">;
@@ -510,7 +539,8 @@ def HasStdExtZbbOrZbkb
def FeatureStdExtZbkc
: RISCVExtension<"zbkc", 1, 0,
"'Zbkc' (Carry-less multiply instructions for "
- "Cryptography)">;
+ "Cryptography)">,
+ RISCVExtensionBitmask<0, 31>;
def HasStdExtZbkc
: Predicate<"Subtarget->hasStdExtZbkc()">,
AssemblerPredicate<(all_of FeatureStdExtZbkc),
@@ -527,14 +557,16 @@ def HasStdExtZbcOrZbkc
def FeatureStdExtZknd
: RISCVExtension<"zknd", 1, 0,
- "'Zknd' (NIST Suite: AES Decryption)">;
+ "'Zknd' (NIST Suite: AES Decryption)">,
+ RISCVExtensionBitmask<0, 41>;
def HasStdExtZknd : Predicate<"Subtarget->hasStdExtZknd()">,
AssemblerPredicate<(all_of FeatureStdExtZknd),
"'Zknd' (NIST Suite: AES Decryption)">;
def FeatureStdExtZkne
: RISCVExtension<"zkne", 1, 0,
- "'Zkne' (NIST Suite: AES Encryption)">;
+ "'Zkne' (NIST Suite: AES Encryption)">,
+ RISCVExtensionBitmask<0, 42>;
def HasStdExtZkne : Predicate<"Subtarget->hasStdExtZkne()">,
AssemblerPredicate<(all_of FeatureStdExtZkne),
"'Zkne' (NIST Suite: AES Encryption)">;
@@ -549,21 +581,24 @@ def HasStdExtZkndOrZkne
def FeatureStdExtZknh
: RISCVExtension<"zknh", 1, 0,
- "'Zknh' (NIST Suite: Hash Function Instructions)">;
+ "'Zknh' (NIST Suite: Hash Function Instructions)">,
+ RISCVExtensionBitmask<0, 43>;
def HasStdExtZknh : Predicate<"Subtarget->hasStdExtZknh()">,
AssemblerPredicate<(all_of FeatureStdExtZknh),
"'Zknh' (NIST Suite: Hash Function Instructions)">;
def FeatureStdExtZksed
: RISCVExtension<"zksed", 1, 0,
- "'Zksed' (ShangMi Suite: SM4 Block Cipher Instructions)">;
+ "'Zksed' (ShangMi Suite: SM4 Block Cipher Instructions)">,
+ RISCVExtensionBitmask<0, 44>;
def HasStdExtZksed : Predicate<"Subtarget->hasStdExtZksed()">,
AssemblerPredicate<(all_of FeatureStdExtZksed),
"'Zksed' (ShangMi Suite: SM4 Block Cipher Instructions)">;
def FeatureStdExtZksh
: RISCVExtension<"zksh", 1, 0,
- "'Zksh' (ShangMi Suite: SM3 Hash Function Instructions)">;
+ "'Zksh' (ShangMi Suite: SM3 Hash Function Instructions)">,
+ RISCVExtensionBitmask<0, 45>;
def HasStdExtZksh : Predicate<"Subtarget->hasStdExtZksh()">,
AssemblerPredicate<(all_of FeatureStdExtZksh),
"'Zksh' (ShangMi Suite: SM3 Hash Function Instructions)">;
@@ -596,7 +631,8 @@ def FeatureStdExtZks
def FeatureStdExtZkt
: RISCVExtension<"zkt", 1, 0,
- "'Zkt' (Data Independent Execution Latency)">;
+ "'Zkt' (Data Independent Execution Latency)">,
+ RISCVExtensionBitmask<0, 46>;
def FeatureStdExtZk
: RISCVExtension<"zk", 1, 0,
@@ -626,6 +662,7 @@ def FeatureStdExtZve32x
"with maximal 32 EEW)",
[FeatureStdExtZicsr, FeatureStdExtZvl32b]>;
+
def FeatureStdExtZve32f
: RISCVExtension<"zve32f", 1, 0,
"'Zve32f' (Vector Extensions for Embedded Processors "
@@ -653,7 +690,8 @@ def FeatureStdExtZve64d
def FeatureStdExtV
: RISCVExtension<"v", 1, 0,
"'V' (Vector Extension for Application Processors)",
- [FeatureStdExtZvl128b, FeatureStdExtZve64d]>;
+ [FeatureStdExtZvl128b, FeatureStdExtZve64d]>,
+ RISCVExtensionBitmask<0, 21>;
def FeatureStdExtZvfbfmin
: RISCVExtension<"zvfbfmin", 1, 0, "'Zvbfmin' (Vector BF16 Converts)",
@@ -673,12 +711,14 @@ def HasStdExtZvfbfwma : Predicate<"Subtarget->hasStdExtZvfbfwma()">,
def FeatureStdExtZvfhmin
: RISCVExtension<"zvfhmin", 1, 0,
"'Zvfhmin' (Vector Half-Precision Floating-Point Minimal)",
- [FeatureStdExtZve32f]>;
+ [FeatureStdExtZve32f]>,
+ RISCVExtensionBitmask<0, 51>;
def FeatureStdExtZvfh
: RISCVExtension<"zvfh", 1, 0,
"'Zvfh' (Vector Half-Precision Floating-Point)",
- [FeatureStdExtZvfhmin, FeatureStdExtZfhmin]>;
+ [FeatureStdExtZvfhmin, FeatureStdExtZfhmin]>,
+ RISCVExtensionBitmask<0, 50>;
def HasStdExtZfhOrZvfh
: Predicate<"Subtarget->hasStdExtZfh() || Subtarget->hasStdExtZvfh()">,
@@ -690,7 +730,8 @@ def HasStdExtZfhOrZvfh
def FeatureStdExtZvkb
: RISCVExtension<"zvkb", 1, 0,
- "'Zvkb' (Vector Bit-manipulation used in Cryptography)">;
+ "'Zvkb' (Vector Bit-manipulation used in Cryptography)">,
+ RISCVExtensionBitmask<0, 52>;
def HasStdExtZvkb : Predicate<"Subtarget->hasStdExtZvkb()">,
AssemblerPredicate<(all_of FeatureStdExtZvkb),
"'Zvkb' (Vector Bit-manipulation used in Cryptography)">;
@@ -698,35 +739,40 @@ def HasStdExtZvkb : Predicate<"Subtarget->hasStdExtZvkb()">,
def FeatureStdExtZvbb
: RISCVExtension<"zvbb", 1, 0,
"'Zvbb' (Vector basic bit-manipulation instructions)",
- [FeatureStdExtZvkb]>;
+ [FeatureStdExtZvkb]>,
+ RISCVExtensionBitmask<0, 48>;
def HasStdExtZvbb : Predicate<"Subtarget->hasStdExtZvbb()">,
AssemblerPredicate<(all_of FeatureStdExtZvbb),
"'Zvbb' (Vector basic bit-manipulation instructions)">;
def FeatureStdExtZvbc
: RISCVExtension<"zvbc", 1, 0,
- "'Zvbc' (Vector Carryless Multiplication)">;
+ "'Zvbc' (Vector Carryless Multiplication)">,
+ RISCVExtensionBitmask<0, 49>;
def HasStdExtZvbc : Predicate<"Subtarget->hasStdExtZvbc()">,
AssemblerPredicate<(all_of FeatureStdExtZvbc),
"'Zvbc' (Vector Carryless Multiplication)">;
def FeatureStdExtZvkg
: RISCVExtension<"zvkg", 1, 0,
- "'Zvkg' (Vector GCM instructions for Cryptography)">;
+ "'Zvkg' (Vector GCM instructions for Cryptography)">,
+ RISCVExtensionBitmask<0, 53>;
def HasStdExtZvkg : Predicate<"Subtarget->hasStdExtZvkg()">,
AssemblerPredicate<(all_of FeatureStdExtZvkg),
"'Zvkg' (Vector GCM instructions for Cryptography)">;
def FeatureStdExtZvkned
: RISCVExtension<"zvkned", 1, 0,
- "'Zvkned' (Vector AES Encryption & Decryption (Single Round))">;
+ "'Zvkned' (Vector AES Encryption & Decryption (Single Round))">,
+ RISCVExtensionBitmask<0, 54>;
def HasStdExtZvkned : Predicate<"Subtarget->hasStdExtZvkned()">,
AssemblerPredicate<(all_of FeatureStdExtZvkned),
"'Zvkned' (Vector AES Encryption & Decryption (Single Round))">;
def FeatureStdExtZvknha
: RISCVExtension<"zvknha", 1, 0,
- "'Zvknha' (Vector SHA-2 (SHA-256 only))">;
+ "'Zvknha' (Vector SHA-2 (SHA-256 only))">,
+ RISCVExtensionBitmask<0, 55>;
def HasStdExtZvknha : Predicate<"Subtarget->hasStdExtZvknha()">,
AssemblerPredicate<(all_of FeatureStdExtZvknha),
"'Zvknha' (Vector SHA-2 (SHA-256 only))">;
@@ -734,7 +780,8 @@ def HasStdExtZvknha : Predicate<"Subtarget->hasStdExtZvknha()">,
def FeatureStdExtZvknhb
: RISCVExtension<"zvknhb", 1, 0,
"'Zvknhb' (Vector SHA-2 (SHA-256 and SHA-512))",
- [FeatureStdExtZve64x]>;
+ [FeatureStdExtZve64x]>,
+ RISCVExtensionBitmask<0, 56>;
def HasStdExtZvknhb : Predicate<"Subtarget->hasStdExtZvknhb()">,
AssemblerPredicate<(all_of FeatureStdExtZvknhb),
"'Zvknhb' (Vector SHA-2 (SHA-256 and SHA-512))">;
@@ -745,21 +792,24 @@ def HasStdExtZvknhaOrZvknhb : Predicate<"Subtarget->hasStdExtZvknha() || Subtarg
def FeatureStdExtZvksed
: RISCVExtension<"zvksed", 1, 0,
- "'Zvksed' (SM4 Block Cipher Instructions)">;
+ "'Zvksed' (SM4 Block Cipher Instructions)">,
+ RISCVExtensionBitmask<0, 57>;
def HasStdExtZvksed : Predicate<"Subtarget->hasStdExtZvksed()">,
AssemblerPredicate<(all_of FeatureStdExtZvksed),
"'Zvksed' (SM4 Block Cipher Instructions)">;
def FeatureStdExtZvksh
: RISCVExtension<"zvksh", 1, 0,
- "'Zvksh' (SM3 Hash Function Instructions)">;
+ "'Zvksh' (SM3 Hash Function Instructions)">,
+ RISCVExtensionBitmask<0, 58>;
def HasStdExtZvksh : Predicate<"Subtarget->hasStdExtZvksh()">,
AssemblerPredicate<(all_of FeatureStdExtZvksh),
"'Zvksh' (SM3 Hash Function Instructions)">;
def FeatureStdExtZvkt
: RISCVExtension<"zvkt", 1, 0,
- "'Zvkt' (Vector Data-Independent Execution Latency)">;
+ "'Zvkt' (Vector Data-Independent Execution Latency)">,
+ RISCVExtensionBitmask<0, 59>;
// Zvk short-hand extensions
@@ -796,7 +846,6 @@ def FeatureStdExtZvksg
: RISCVExtension<"zvksg", 1, 0,
"'Zvksg' (shorthand for 'Zvks' and 'Zvkg')",
[FeatureStdExtZvks, FeatureStdExtZvkg]>;
-
// Vector instruction predicates
def HasVInstructions : Predicate<"Subtarget->hasVInstructions()">,
diff --git a/llvm/lib/TargetParser/RISCVTargetParser.cpp b/llvm/lib/TargetParser/RISCVTargetParser.cpp
index 9003f9beffa7e..adb594777be80 100644
--- a/llvm/lib/TargetParser/RISCVTargetParser.cpp
+++ b/llvm/lib/TargetParser/RISCVTargetParser.cpp
@@ -119,6 +119,38 @@ void getFeaturesForCPU(StringRef CPU,
else
EnabledFeatures.push_back(F.substr(1));
}
+
+namespace RISCVExtensionBitmaskTable {
+#define GET_RISCVExtensionBitmaskTable_IMPL
+#include "llvm/TargetParser/RISCVTargetParserDef.inc"
+
+} // namespace RISCVExtensionBitmaskTable
+
+namespace {
+struct LessExtName {
+ bool operator()(const RISCVExtensionBitmaskTable::RISCVExtensionBitmask &LHS,
+ StringRef RHS) {
+ return StringRef(LHS.Name) < RHS;
+ }
+ bool
+ operator()(StringRef LHS,
+ const RISCVExtensionBitmaskTable::RISCVExtensionBitmask &RHS) {
+ return LHS < StringRef(RHS.Name);
+ }
+};
+} // namespace
+
+static RISCVExtensionBitmaskTable::RISCVExtensionBitmask
+getExtensionBitmask(StringRef ExtName) {
+ ArrayRef<RISCVExtensionBitmaskTable::RISCVExtensionBitmask> ExtBitmasks =
+ RISCVExtensionBitmaskTable::ExtensionBitmask;
+ auto *I = llvm::lower_bound(ExtBitmasks, ExtName, LessExtName());
+
+ if (I != ExtBitmasks.end())
+ return *I;
+
+ return RISCVExtensionBitmaskTable::RISCVExtensionBitmask();
+}
} // namespace RISCV
namespace RISCVVType {
diff --git a/llvm/test/TableGen/riscv-target-def.td b/llvm/test/TableGen/riscv-target-def.td
index fb58448d7ce88..9e027d18b76fb 100644
--- a/llvm/test/TableGen/riscv-target-def.td
+++ b/llvm/test/TableGen/riscv-target-def.td
@@ -12,6 +12,11 @@ class RISCVExtension<string name, int major, int minor, string desc,
bit Experimental = false;
}
+class RISCVExtensionBitmask<bits<3> groupID, int bitmaskShift> {
+ int GroupID = groupID;
+ int BitPos = bitmaskShift;
+}
+
class RISCVExperimentalExtension<string name, int major, int minor, string desc,
list<RISCVExtension> implies = [],
string fieldname = !subst("Feature", "Has", NAME),
@@ -23,7 +28,8 @@ class RISCVExperimentalExtension<string name, int major, int minor, string desc,
def FeatureStdExtI
: RISCVExtension<"i", 2, 1,
- "'I' (Base Integer Instruction Set)">;
+ "'I' (Base Integer Instruction Set)">,
+ RISCVExtensionBitmask<0, 8>;
def FeatureStdExtZicsr
: RISCVExtension<"zicsr", 2, 0,
@@ -36,7 +42,8 @@ def FeatureStdExtZifencei
def FeatureStdExtF
: RISCVExtension<"f", 2, 2,
"'F' (Single-Precision Floating-Point)",
- [FeatureStdExtZicsr]>;
+ [FeatureStdExtZicsr]>,
+ RISCVExtensionBitmask<0, 5>;
def FeatureStdExtZidummy
: RISCVExperimentalExtension<"zidummy", 0, 1,
@@ -171,3 +178,10 @@ def ROCKET : RISCVTuneProcessorModel<"rocket",
// CHECK-NEXT: TUNE_PROC(ROCKET, "rocket")
// CHECK: #undef TUNE_PROC
+
+// CHECK: #ifdef GET_RISCVExtensionBitmaskTable_IMPL
+// CHECK-NEXT: static const RISCVExtensionBitmask ExtensionBitmask[]={
+// CHECK-NEXT: {"f", 0, 5ULL},
+// CHECK-NEXT: {"i", 0, 8ULL},
+// CHECK-NEXT: };
+// CHECK-NEXT: #endif
diff --git a/llvm/utils/TableGen/RISCVTargetDefEmitter.cpp b/llvm/utils/TableGen/RISCVTargetDefEmitter.cpp
index b76ba05954aa5..edb1a4d5d8531 100644
--- a/llvm/utils/TableGen/RISCVTargetDefEmitter.cpp
+++ b/llvm/utils/TableGen/RISCVTargetDefEmitter.cpp
@@ -11,6 +11,7 @@
//
//===----------------------------------------------------------------------===//
+#include "llvm/ADT/DenseSet.h"
#include "llvm/Support/RISCVISAUtils.h"
#include "llvm/TableGen/Record.h"
#include "llvm/TableGen/TableGenBackend.h"
@@ -210,10 +211,46 @@ static void emitRISCVProcs(RecordKeeper &RK, raw_ostream &OS) {
OS << "\n#undef TUNE_PROC\n";
}
+static void emitRISCVExtensionBitmask(RecordKeeper &RK, raw_ostream &OS) {
+
+ std::vector<Record *> Extensions =
+ RK.getAllDerivedDefinitionsIfDefined("RISCVExtensionBitmask");
+ llvm::sort(Extensions, [](const Record *Rec1, const Record *Rec2) {
+ return getExtensionName(Rec1) < getExtensionName(Rec2);
+ });
+
+#ifndef NDEBUG
+ llvm::DenseSet<std::pair<uint64_t, uint64_t>> Seen;
+#endif
+
+ OS << "#ifdef GET_RISCVExtensionBitmaskTable_IMPL\n";
+ OS << "static const RISCVExtensionBitmask ExtensionBitmask[]={\n";
+ for (const Record *Rec : Extensions) {
+ unsigned GroupIDVal = Rec->getValueAsInt("GroupID");
+ unsigned BitPosVal = Rec->getValueAsInt("BitPos");
+
+ StringRef ExtName = Rec->getValueAsString("Name");
+ ExtName.consume_front("experimental-");
+
+#ifndef NDEBUG
+ assert(Seen.insert(std::make_pair(GroupIDVal, BitPosVal)).second &&
+ "duplicated bitmask");
+#endif
+
+ OS << " {"
+ << "\"" << ExtName << "\""
+ << ", " << GroupIDVal << ", " << BitPosVal << "ULL"
+ << "},\n";
+ }
+ OS << "};\n";
+ OS << "#endif\n";
+}
+
static void EmitRISCVTargetDef(RecordKeeper &RK, raw_ostream &OS) {
emitRISCVExtensions(RK, OS);
emitRISCVProfiles(RK, OS);
emitRISCVProcs(RK, OS);
+ emitRISCVExtensionBitmask(RK, OS);
}
static TableGen::Emitter::Opt X("gen-riscv-target-def", EmitRISCVTargetDef,
>From 205f27e3261a3ba30b608e399f93ab6e8c03b3dd Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Thu, 11 Jul 2024 20:48:02 -0700
Subject: [PATCH 02/19] Add zihintpause
---
llvm/lib/Target/RISCV/RISCVFeatures.td | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/llvm/lib/Target/RISCV/RISCVFeatures.td b/llvm/lib/Target/RISCV/RISCVFeatures.td
index 125e9e2f53953..ff6ec42a2e72f 100644
--- a/llvm/lib/Target/RISCV/RISCVFeatures.td
+++ b/llvm/lib/Target/RISCV/RISCVFeatures.td
@@ -139,7 +139,8 @@ def HasStdExtZifencei : Predicate<"Subtarget->hasStdExtZifencei()">,
def FeatureStdExtZihintpause
: RISCVExtension<"zihintpause", 2, 0,
- "'Zihintpause' (Pause Hint)">;
+ "'Zihintpause' (Pause Hint)">,
+ RISCVExtensionBitmask<0, 40>;
def HasStdExtZihintpause : Predicate<"Subtarget->hasStdExtZihintpause()">,
AssemblerPredicate<(all_of FeatureStdExtZihintpause),
"'Zihintpause' (Pause Hint)">;
>From 0a873d0897fbbf94d46e6fbf129aee3090b4b9c1 Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Mon, 10 Jun 2024 22:22:42 -0700
Subject: [PATCH 03/19] [RISCV][FMV] Support target_clones
---
.../clang/Basic/DiagnosticFrontendKinds.td | 4 +
clang/include/clang/Basic/TargetInfo.h | 3 +-
clang/lib/AST/ASTContext.cpp | 12 +
clang/lib/CodeGen/CGBuiltin.cpp | 78 +++++++
clang/lib/CodeGen/CodeGenFunction.cpp | 91 +++++++-
clang/lib/CodeGen/CodeGenFunction.h | 7 +
clang/lib/CodeGen/CodeGenModule.cpp | 5 +-
clang/lib/CodeGen/Targets/RISCV.cpp | 23 ++
clang/lib/Sema/SemaDeclAttr.cpp | 26 +++
.../attr-target-clones-riscv-invaild.c | 8 +
clang/test/CodeGen/attr-target-clones-riscv.c | 205 ++++++++++++++++++
.../CodeGenCXX/attr-target-clones-riscv.cpp | 204 +++++++++++++++++
.../test/SemaCXX/attr-target-clones-riscv.cpp | 32 +++
.../llvm/TargetParser/RISCVTargetParser.h | 4 +
llvm/lib/TargetParser/RISCVTargetParser.cpp | 18 +-
15 files changed, 715 insertions(+), 5 deletions(-)
create mode 100644 clang/test/CodeGen/attr-target-clones-riscv-invaild.c
create mode 100644 clang/test/CodeGen/attr-target-clones-riscv.c
create mode 100644 clang/test/CodeGenCXX/attr-target-clones-riscv.cpp
create mode 100644 clang/test/SemaCXX/attr-target-clones-riscv.cpp
diff --git a/clang/include/clang/Basic/DiagnosticFrontendKinds.td b/clang/include/clang/Basic/DiagnosticFrontendKinds.td
index 12a4617c64d87..b2b63674ecc07 100644
--- a/clang/include/clang/Basic/DiagnosticFrontendKinds.td
+++ b/clang/include/clang/Basic/DiagnosticFrontendKinds.td
@@ -375,4 +375,8 @@ def warn_missing_symbol_graph_dir : Warning<
def err_ast_action_on_llvm_ir : Error<
"cannot apply AST actions to LLVM IR file '%0'">,
DefaultFatal;
+
+def err_os_unsupport_riscv_target_clones : Error<
+ "target_clones is currently only supported on Linux">;
+
}
diff --git a/clang/include/clang/Basic/TargetInfo.h b/clang/include/clang/Basic/TargetInfo.h
index cf7628553647c..40fbc3a1116cb 100644
--- a/clang/include/clang/Basic/TargetInfo.h
+++ b/clang/include/clang/Basic/TargetInfo.h
@@ -1519,7 +1519,8 @@ class TargetInfo : public TransferrableTargetInfo,
/// Identify whether this target supports multiversioning of functions,
/// which requires support for cpu_supports and cpu_is functionality.
bool supportsMultiVersioning() const {
- return getTriple().isX86() || getTriple().isAArch64();
+ return getTriple().isX86() || getTriple().isAArch64() ||
+ getTriple().isRISCV();
}
/// Identify whether this target supports IFuncs.
diff --git a/clang/lib/AST/ASTContext.cpp b/clang/lib/AST/ASTContext.cpp
index 8e35e71d4a23e..29dce55f349ec 100644
--- a/clang/lib/AST/ASTContext.cpp
+++ b/clang/lib/AST/ASTContext.cpp
@@ -13839,6 +13839,18 @@ void ASTContext::getFunctionFeatureMap(llvm::StringMap<bool> &FeatureMap,
Target->getTargetOpts().FeaturesAsWritten.begin(),
Target->getTargetOpts().FeaturesAsWritten.end());
Target->initFeatureMap(FeatureMap, getDiagnostics(), TargetCPU, Features);
+ } else if (Target->getTriple().isRISCV()) {
+ StringRef VersionStr = TC->getFeatureStr(GD.getMultiVersionIndex());
+ std::vector<std::string> Features;
+ if (VersionStr != "default") {
+ ParsedTargetAttr ParsedAttr = Target->parseTargetAttr(VersionStr);
+ Features.insert(Features.begin(), ParsedAttr.Features.begin(),
+ ParsedAttr.Features.end());
+ }
+ Features.insert(Features.begin(),
+ Target->getTargetOpts().FeaturesAsWritten.begin(),
+ Target->getTargetOpts().FeaturesAsWritten.end());
+ Target->initFeatureMap(FeatureMap, getDiagnostics(), TargetCPU, Features);
} else {
std::vector<std::string> Features;
StringRef VersionStr = TC->getFeatureStr(GD.getMultiVersionIndex());
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 6cc0d9485720c..42647224a88a4 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -62,6 +62,7 @@
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/ScopedPrinter.h"
#include "llvm/TargetParser/AArch64TargetParser.h"
+#include "llvm/TargetParser/RISCVTargetParser.h"
#include "llvm/TargetParser/X86TargetParser.h"
#include <optional>
#include <sstream>
@@ -14214,6 +14215,16 @@ Value *CodeGenFunction::EmitAArch64CpuInit() {
return Builder.CreateCall(Func);
}
+Value *CodeGenFunction::EmitRISCVCpuInit() {
+ llvm::FunctionType *FTy = llvm::FunctionType::get(VoidTy, false);
+ llvm::FunctionCallee Func =
+ CGM.CreateRuntimeFunction(FTy, "__init_riscv_features_bit");
+ cast<llvm::GlobalValue>(Func.getCallee())->setDSOLocal(true);
+ cast<llvm::GlobalValue>(Func.getCallee())
+ ->setDLLStorageClass(llvm::GlobalValue::DefaultStorageClass);
+ return Builder.CreateCall(Func);
+}
+
Value *CodeGenFunction::EmitX86CpuInit() {
llvm::FunctionType *FTy = llvm::FunctionType::get(VoidTy,
/*Variadic*/ false);
@@ -14266,6 +14277,73 @@ CodeGenFunction::EmitAArch64CpuSupports(ArrayRef<StringRef> FeaturesStrs) {
return Result;
}
+Value *CodeGenFunction::EmitRISCVCpuSupports(ArrayRef<StringRef> FeaturesStrs,
+ unsigned &MaxGroupIDUsed) {
+
+ const unsigned FeatureBitSize = llvm::RISCV::RISCVFeatureBitSize;
+ llvm::ArrayType *ArrayOfInt64Ty =
+ llvm::ArrayType::get(Int64Ty, FeatureBitSize);
+ llvm::Type *StructTy = llvm::StructType::get(Int32Ty, ArrayOfInt64Ty);
+ llvm::Constant *RISCVFeaturesBits =
+ CGM.CreateRuntimeVariable(StructTy, "__riscv_feature_bits");
+ cast<llvm::GlobalValue>(RISCVFeaturesBits)->setDSOLocal(true);
+
+ auto LoadFeatureBit = [&](unsigned Index) {
+ // Create GEP then load.
+ Value *IndexVal = llvm::ConstantInt::get(Int32Ty, Index);
+ std::vector<llvm::Value *> GEPIndices = {llvm::ConstantInt::get(Int32Ty, 0),
+ llvm::ConstantInt::get(Int32Ty, 1),
+ IndexVal};
+ Value *Ptr =
+ Builder.CreateInBoundsGEP(StructTy, RISCVFeaturesBits, GEPIndices);
+ Value *FeaturesBit =
+ Builder.CreateAlignedLoad(Int64Ty, Ptr, CharUnits::fromQuantity(8));
+ return FeaturesBit;
+ };
+
+ SmallVector<unsigned long long> RequireFeatureBits =
+ llvm::RISCV::getRequireFeatureBitMask(FeaturesStrs);
+ Value *Result = Builder.getTrue();
+ for (unsigned i = 0; i < RequireFeatureBits.size(); i++) {
+ if (!RequireFeatureBits[i])
+ continue;
+ MaxGroupIDUsed = i;
+ Value *Mask = Builder.getInt64(RequireFeatureBits[i]);
+ Value *Bitset = Builder.CreateAnd(LoadFeatureBit(i), Mask);
+ Value *Cmp = Builder.CreateICmpEQ(Bitset, Mask);
+ Result = Builder.CreateAnd(Result, Cmp);
+ }
+
+ return Result;
+}
+
+Value *CodeGenFunction::EmitRISCVFeatureBitsLength(unsigned MaxGroupIDUsed) {
+
+ const unsigned FeatureBitSize = llvm::RISCV::RISCVFeatureBitSize;
+ llvm::ArrayType *ArrayOfInt64Ty =
+ llvm::ArrayType::get(Int64Ty, FeatureBitSize);
+ llvm::Type *StructTy = llvm::StructType::get(Int32Ty, ArrayOfInt64Ty);
+ llvm::Constant *RISCVFeaturesBits =
+ CGM.CreateRuntimeVariable(StructTy, "__riscv_feature_bits");
+ cast<llvm::GlobalValue>(RISCVFeaturesBits)->setDSOLocal(true);
+
+ auto LoadMaxGroupID = [&]() {
+ std::vector<llvm::Value *> GEPIndices = {
+ llvm::ConstantInt::get(Int32Ty, 0), llvm::ConstantInt::get(Int32Ty, 0)};
+ llvm::Value *Ptr =
+ Builder.CreateInBoundsGEP(StructTy, RISCVFeaturesBits, GEPIndices);
+ Value *Length =
+ Builder.CreateAlignedLoad(Int64Ty, Ptr, CharUnits::fromQuantity(8));
+ return Length;
+ };
+
+ Value *UsedMaxGroupID = Builder.getInt64(MaxGroupIDUsed);
+ Value *GroupIDResult =
+ Builder.CreateICmpULT(UsedMaxGroupID, LoadMaxGroupID());
+
+ return GroupIDResult;
+}
+
Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
const CallExpr *E) {
if (BuiltinID == Builtin::BI__builtin_cpu_is)
diff --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp
index 26deeca95d326..27305e0f4b7b4 100644
--- a/clang/lib/CodeGen/CodeGenFunction.cpp
+++ b/clang/lib/CodeGen/CodeGenFunction.cpp
@@ -2854,10 +2854,99 @@ void CodeGenFunction::EmitMultiVersionResolver(
case llvm::Triple::aarch64:
EmitAArch64MultiVersionResolver(Resolver, Options);
return;
+ case llvm::Triple::riscv32:
+ case llvm::Triple::riscv64:
+ EmitRISCVMultiVersionResolver(Resolver, Options);
+ return;
default:
- assert(false && "Only implemented for x86 and AArch64 targets");
+ assert(false && "Only implemented for x86, AArch64 and RISC-V targets");
+ }
+}
+
+void CodeGenFunction::EmitRISCVMultiVersionResolver(
+ llvm::Function *Resolver, ArrayRef<MultiVersionResolverOption> Options) {
+
+ if (getContext().getTargetInfo().getTriple().getOS() !=
+ llvm::Triple::OSType::Linux) {
+ CGM.getDiags().Report(diag::err_os_unsupport_riscv_target_clones);
+ return;
+ }
+
+ llvm::BasicBlock *CurBlock = createBasicBlock("resolver_entry", Resolver);
+ Builder.SetInsertPoint(CurBlock);
+ EmitRISCVCpuInit();
+
+ bool SupportsIFunc = getContext().getTargetInfo().supportsIFunc();
+ bool HasDefault = false;
+ int DefaultIndex = 0;
+ // Check the each candidate function.
+ for (unsigned Index = 0; Index < Options.size(); Index++) {
+
+ if (Options[Index].Conditions.Features[0].starts_with("default")) {
+ HasDefault = true;
+ DefaultIndex = Index;
+ continue;
+ }
+
+ Builder.SetInsertPoint(CurBlock);
+
+ std::vector<std::string> TargetAttrFeats =
+ getContext()
+ .getTargetInfo()
+ .parseTargetAttr(Options[Index].Conditions.Features[0])
+ .Features;
+
+ if (!TargetAttrFeats.empty()) {
+
+ llvm::BasicBlock *SecondCond =
+ createBasicBlock("resovler_cond", Resolver);
+
+ Builder.SetInsertPoint(SecondCond);
+ unsigned MaxGroupIDUsed = 0;
+ llvm::SmallVector<StringRef, 8> CurrTargetAttrFeats;
+
+ for (auto Feat: TargetAttrFeats)
+ CurrTargetAttrFeats.push_back(StringRef(Feat).substr(1));
+
+ llvm::Value *SecondCondition =
+ EmitRISCVCpuSupports(CurrTargetAttrFeats, MaxGroupIDUsed);
+
+ Builder.SetInsertPoint(CurBlock);
+ llvm::Value *FirstCondition = EmitRISCVFeatureBitsLength(MaxGroupIDUsed);
+
+ llvm::BasicBlock *RetBlock =
+ createBasicBlock("resolver_return", Resolver);
+ CGBuilderTy RetBuilder(*this, RetBlock);
+ CreateMultiVersionResolverReturn(CGM, Resolver, RetBuilder,
+ Options[Index].Function, SupportsIFunc);
+ llvm::BasicBlock *ElseBlock = createBasicBlock("resolver_else", Resolver);
+
+ Builder.SetInsertPoint(CurBlock);
+ Builder.CreateCondBr(FirstCondition, SecondCond, ElseBlock);
+
+ Builder.SetInsertPoint(SecondCond);
+ Builder.CreateCondBr(SecondCondition, RetBlock, ElseBlock);
+
+ CurBlock = ElseBlock;
+ }
}
+
+ // Finally, emit the default one.
+ if (HasDefault) {
+ Builder.SetInsertPoint(CurBlock);
+ CreateMultiVersionResolverReturn(
+ CGM, Resolver, Builder, Options[DefaultIndex].Function, SupportsIFunc);
+ return;
+ }
+
+ // If no generic/default, emit an unreachable.
+ Builder.SetInsertPoint(CurBlock);
+ llvm::CallInst *TrapCall = EmitTrapCall(llvm::Intrinsic::trap);
+ TrapCall->setDoesNotReturn();
+ TrapCall->setDoesNotThrow();
+ Builder.CreateUnreachable();
+ Builder.ClearInsertionPoint();
}
void CodeGenFunction::EmitAArch64MultiVersionResolver(
diff --git a/clang/lib/CodeGen/CodeGenFunction.h b/clang/lib/CodeGen/CodeGenFunction.h
index 13f12b5d878a6..107701d3f1faa 100644
--- a/clang/lib/CodeGen/CodeGenFunction.h
+++ b/clang/lib/CodeGen/CodeGenFunction.h
@@ -5291,6 +5291,9 @@ class CodeGenFunction : public CodeGenTypeCache {
void
EmitAArch64MultiVersionResolver(llvm::Function *Resolver,
ArrayRef<MultiVersionResolverOption> Options);
+ void
+ EmitRISCVMultiVersionResolver(llvm::Function *Resolver,
+ ArrayRef<MultiVersionResolverOption> Options);
private:
QualType getVarArgType(const Expr *Arg);
@@ -5315,6 +5318,10 @@ class CodeGenFunction : public CodeGenTypeCache {
FormAArch64ResolverCondition(const MultiVersionResolverOption &RO);
llvm::Value *EmitAArch64CpuSupports(const CallExpr *E);
llvm::Value *EmitAArch64CpuSupports(ArrayRef<StringRef> FeatureStrs);
+ llvm::Value *EmitRISCVCpuInit();
+ llvm::Value *EmitRISCVCpuSupports(ArrayRef<StringRef> FeatureStrs,
+ unsigned &MaxGroupIDUsed);
+ llvm::Value *EmitRISCVFeatureBitsLength(unsigned MaxGroupIDUsed);
};
inline DominatingLLVMValue::saved_type
diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index 5c810cd332185..d876dff3d5755 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -4257,7 +4257,10 @@ void CodeGenModule::emitMultiVersionFunctions() {
Feats.clear();
if (getTarget().getTriple().isAArch64())
TC->getFeatures(Feats, I);
- else {
+ else if (getTarget().getTriple().isRISCV()) {
+ StringRef Version = TC->getFeatureStr(I);
+ Feats.push_back(Version);
+ } else {
StringRef Version = TC->getFeatureStr(I);
if (Version.starts_with("arch="))
Architecture = Version.drop_front(sizeof("arch=") - 1);
diff --git a/clang/lib/CodeGen/Targets/RISCV.cpp b/clang/lib/CodeGen/Targets/RISCV.cpp
index f2add9351c03c..ba81bf7d1dd0a 100644
--- a/clang/lib/CodeGen/Targets/RISCV.cpp
+++ b/clang/lib/CodeGen/Targets/RISCV.cpp
@@ -63,9 +63,32 @@ class RISCVABIInfo : public DefaultABIInfo {
CharUnits Field2Off) const;
ABIArgInfo coerceVLSVector(QualType Ty) const;
+
+ using ABIInfo::appendAttributeMangling;
+ void appendAttributeMangling(TargetClonesAttr *Attr, unsigned Index,
+ raw_ostream &Out) const override;
+ void appendAttributeMangling(StringRef AttrStr,
+ raw_ostream &Out) const override;
};
} // end anonymous namespace
+void RISCVABIInfo::appendAttributeMangling(TargetClonesAttr *Attr,
+ unsigned Index,
+ raw_ostream &Out) const {
+ appendAttributeMangling(Attr->getFeatureStr(Index), Out);
+}
+
+void RISCVABIInfo::appendAttributeMangling(StringRef AttrStr,
+ raw_ostream &Out) const {
+ if (AttrStr == "default") {
+ Out << ".default";
+ return;
+ }
+
+ Out << '.';
+ Out << AttrStr;
+}
+
void RISCVABIInfo::computeInfo(CGFunctionInfo &FI) const {
QualType RetTy = FI.getReturnType();
if (!getCXXABI().classifyReturnType(FI))
diff --git a/clang/lib/Sema/SemaDeclAttr.cpp b/clang/lib/Sema/SemaDeclAttr.cpp
index f2cd46d1e7c93..9b420f90b222f 100644
--- a/clang/lib/Sema/SemaDeclAttr.cpp
+++ b/clang/lib/Sema/SemaDeclAttr.cpp
@@ -3127,6 +3127,32 @@ bool Sema::checkTargetClonesAttrString(
/*IncludeLocallyStreaming=*/false))
return Diag(LiteralLoc,
diag::err_sme_streaming_cannot_be_multiversioned);
+ } else if (TInfo.getTriple().isRISCV()) {
+ // Suppress warn_target_clone_mixed_values
+ HasCommas = false;
+
+ // Only support arch=+ext,... syntax.
+ if (Str.starts_with("arch=+")) {
+ // parseTargetAttr will parse full version string,
+ // the following split Cur string is no longer interesting.
+ if ((!Cur.starts_with("arch=")))
+ continue;
+
+ ParsedTargetAttr TargetAttr =
+ Context.getTargetInfo().parseTargetAttr(Str);
+ if (TargetAttr.Features.empty())
+ return Diag(CurLoc, diag::warn_unsupported_target_attribute)
+ << Unsupported << None << Str << TargetClones;
+ } else if (Str == "default") {
+ DefaultIsDupe = HasDefault;
+ HasDefault = true;
+ } else {
+ return Diag(CurLoc, diag::warn_unsupported_target_attribute)
+ << Unsupported << None << Str << TargetClones;
+ }
+ if (llvm::is_contained(StringsBuffer, Str) || DefaultIsDupe)
+ Diag(CurLoc, diag::warn_target_clone_duplicate_options);
+ StringsBuffer.push_back(Str);
} else {
// Other targets ( currently X86 )
if (Cur.starts_with("arch=")) {
diff --git a/clang/test/CodeGen/attr-target-clones-riscv-invaild.c b/clang/test/CodeGen/attr-target-clones-riscv-invaild.c
new file mode 100644
index 0000000000000..f5e8f40d7a8f3
--- /dev/null
+++ b/clang/test/CodeGen/attr-target-clones-riscv-invaild.c
@@ -0,0 +1,8 @@
+// RUN: not %clang_cc1 -triple riscv64 -target-feature +i -emit-llvm -o - %s 2>&1 | FileCheck %s --check-prefix=CHECK-UNSUPPORT-OS
+
+// CHECK-UNSUPPORT-OS: error: target_clones is currently only supported on Linux
+__attribute__((target_clones("default", "arch=+c"))) int foo2(void) {
+ return 2;
+}
+
+int bar() { return foo1()+foo2(); }
diff --git a/clang/test/CodeGen/attr-target-clones-riscv.c b/clang/test/CodeGen/attr-target-clones-riscv.c
new file mode 100644
index 0000000000000..1088dd56a4d83
--- /dev/null
+++ b/clang/test/CodeGen/attr-target-clones-riscv.c
@@ -0,0 +1,205 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --check-globals all --include-generated-funcs --version 4
+// RUN: %clang_cc1 -triple riscv64-linux-gnu -target-feature +i -emit-llvm -o - %s | FileCheck %s
+
+__attribute__((target_clones("default", "arch=+m"))) int foo1(void) {
+ return 1;
+}
+__attribute__((target_clones("default", "arch=+zbb", "arch=+m"))) int foo2(void) { return 2; }
+__attribute__((target_clones("default", "arch=+zbb,+c"))) int foo3(void) { return 3; }
+__attribute__((target_clones("default", "arch=+zbb,+v"))) int
+foo4(void) {
+ return 4;
+}
+__attribute__((target_clones("default"))) int foo5(void) { return 5; }
+__attribute__((target_clones("default", "arch=+zvkt"))) int foo6(void) { return 2; }
+
+int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
+
+//.
+// CHECK: @__riscv_feature_bits = external dso_local global { i32, [1 x i64] }
+// CHECK: @foo1.ifunc = weak_odr alias i32 (), ptr @foo1
+// CHECK: @foo2.ifunc = weak_odr alias i32 (), ptr @foo2
+// CHECK: @foo3.ifunc = weak_odr alias i32 (), ptr @foo3
+// CHECK: @foo4.ifunc = weak_odr alias i32 (), ptr @foo4
+// CHECK: @foo5.ifunc = weak_odr alias i32 (), ptr @foo5
+// CHECK: @foo6.ifunc = weak_odr alias i32 (), ptr @foo6
+// CHECK: @foo1 = weak_odr ifunc i32 (), ptr @foo1.resolver
+// CHECK: @foo2 = weak_odr ifunc i32 (), ptr @foo2.resolver
+// CHECK: @foo3 = weak_odr ifunc i32 (), ptr @foo3.resolver
+// CHECK: @foo4 = weak_odr ifunc i32 (), ptr @foo4.resolver
+// CHECK: @foo5 = weak_odr ifunc i32 (), ptr @foo5.resolver
+// CHECK: @foo6 = weak_odr ifunc i32 (), ptr @foo6.resolver
+//.
+// CHECK-LABEL: define dso_local signext i32 @foo1.default(
+// CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: ret i32 1
+//
+//
+// CHECK-LABEL: define weak_odr ptr @foo1.resolver() comdat {
+// CHECK-NEXT: resolver_entry:
+// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
+// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 4096
+// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 4096
+// CHECK-NEXT: [[TMP5:%.*]] = and i1 true, [[TMP4]]
+// CHECK-NEXT: br i1 [[TMP5]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE]]
+// CHECK: resolver_return:
+// CHECK-NEXT: ret ptr @"foo1.arch=+m"
+// CHECK: resolver_else:
+// CHECK-NEXT: ret ptr @foo1.default
+//
+//
+// CHECK-LABEL: define dso_local signext i32 @foo2.default(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: ret i32 2
+//
+//
+// CHECK-LABEL: define weak_odr ptr @foo2.resolver() comdat {
+// CHECK-NEXT: resolver_entry:
+// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
+// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 268435456
+// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 268435456
+// CHECK-NEXT: [[TMP5:%.*]] = and i1 true, [[TMP4]]
+// CHECK-NEXT: br i1 [[TMP5]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE]]
+// CHECK: resolver_return:
+// CHECK-NEXT: ret ptr @"foo2.arch=+zbb"
+// CHECK: resolver_else:
+// CHECK-NEXT: [[TMP6:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
+// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i64 0, [[TMP6]]
+// CHECK-NEXT: br i1 [[TMP7]], label [[RESOVLER_COND1:%.*]], label [[RESOLVER_ELSE3:%.*]]
+// CHECK: resovler_cond1:
+// CHECK-NEXT: [[TMP8:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
+// CHECK-NEXT: [[TMP9:%.*]] = and i64 [[TMP8]], 4096
+// CHECK-NEXT: [[TMP10:%.*]] = icmp eq i64 [[TMP9]], 4096
+// CHECK-NEXT: [[TMP11:%.*]] = and i1 true, [[TMP10]]
+// CHECK-NEXT: br i1 [[TMP11]], label [[RESOLVER_RETURN2:%.*]], label [[RESOLVER_ELSE3]]
+// CHECK: resolver_return2:
+// CHECK-NEXT: ret ptr @"foo2.arch=+m"
+// CHECK: resolver_else3:
+// CHECK-NEXT: ret ptr @foo2.default
+//
+//
+// CHECK-LABEL: define dso_local signext i32 @foo3.default(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: ret i32 3
+//
+//
+// CHECK-LABEL: define weak_odr ptr @foo3.resolver() comdat {
+// CHECK-NEXT: resolver_entry:
+// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
+// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 12
+// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 12
+// CHECK-NEXT: [[TMP5:%.*]] = and i1 true, [[TMP4]]
+// CHECK-NEXT: br i1 [[TMP5]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE]]
+// CHECK: resolver_return:
+// CHECK-NEXT: ret ptr @"foo3.arch=+zbb,+c"
+// CHECK: resolver_else:
+// CHECK-NEXT: ret ptr @foo3.default
+//
+//
+// CHECK-LABEL: define dso_local signext i32 @foo4.default(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: ret i32 4
+//
+//
+// CHECK-LABEL: define weak_odr ptr @foo4.resolver() comdat {
+// CHECK-NEXT: resolver_entry:
+// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
+// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 69206016
+// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 69206016
+// CHECK-NEXT: [[TMP5:%.*]] = and i1 true, [[TMP4]]
+// CHECK-NEXT: br i1 [[TMP5]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE]]
+// CHECK: resolver_return:
+// CHECK-NEXT: ret ptr @"foo4.arch=+zbb,+v"
+// CHECK: resolver_else:
+// CHECK-NEXT: ret ptr @foo4.default
+//
+//
+// CHECK-LABEL: define dso_local signext i32 @foo5.default(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: ret i32 5
+//
+//
+// CHECK-LABEL: define weak_odr ptr @foo5.resolver() comdat {
+// CHECK-NEXT: resolver_entry:
+// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: ret ptr @foo5.default
+//
+//
+// CHECK-LABEL: define dso_local signext i32 @foo6.default(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: ret i32 2
+//
+//
+// CHECK-LABEL: define weak_odr ptr @foo6.resolver() comdat {
+// CHECK-NEXT: resolver_entry:
+// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
+// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 576460752303423488
+// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 576460752303423488
+// CHECK-NEXT: [[TMP5:%.*]] = and i1 true, [[TMP4]]
+// CHECK-NEXT: br i1 [[TMP5]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE]]
+// CHECK: resolver_return:
+// CHECK-NEXT: ret ptr @"foo6.arch=+zvkt"
+// CHECK: resolver_else:
+// CHECK-NEXT: ret ptr @foo6.default
+//
+//
+// CHECK-LABEL: define dso_local signext i32 @bar(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[CALL:%.*]] = call signext i32 @foo1()
+// CHECK-NEXT: [[CALL1:%.*]] = call signext i32 @foo2()
+// CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[CALL]], [[CALL1]]
+// CHECK-NEXT: [[CALL2:%.*]] = call signext i32 @foo3()
+// CHECK-NEXT: [[ADD3:%.*]] = add nsw i32 [[ADD]], [[CALL2]]
+// CHECK-NEXT: [[CALL4:%.*]] = call signext i32 @foo4()
+// CHECK-NEXT: [[ADD5:%.*]] = add nsw i32 [[ADD3]], [[CALL4]]
+// CHECK-NEXT: [[CALL6:%.*]] = call signext i32 @foo5()
+// CHECK-NEXT: [[ADD7:%.*]] = add nsw i32 [[ADD5]], [[CALL6]]
+// CHECK-NEXT: ret i32 [[ADD7]]
+//
+//.
+// CHECK: attributes #[[ATTR0]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+64bit,+i" }
+// CHECK: attributes #[[ATTR1:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+64bit,+i,+m,+zmmul" }
+// CHECK: attributes #[[ATTR2:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+64bit,+i,+zbb" }
+// CHECK: attributes #[[ATTR3:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+64bit,+c,+i,+zbb" }
+// CHECK: attributes #[[ATTR4:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+64bit,+d,+f,+i,+v,+zbb,+zicsr,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl32b,+zvl64b" }
+// CHECK: attributes #[[ATTR5:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+64bit,+i,+zvkt" }
+//.
+// CHECK: [[META0:![0-9]+]] = !{i32 1, !"wchar_size", i32 4}
+// CHECK: [[META1:![0-9]+]] = !{i32 1, !"target-abi", !"lp64"}
+// CHECK: [[META2:![0-9]+]] = !{i32 6, !"riscv-isa", [[META3:![0-9]+]]}
+// CHECK: [[META3]] = !{!"rv64i2p1"}
+// CHECK: [[META4:![0-9]+]] = !{i32 8, !"SmallDataLimit", i32 0}
+// CHECK: [[META5:![0-9]+]] = !{!"{{.*}}clang version {{.*}}"}
+//.
diff --git a/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp b/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp
new file mode 100644
index 0000000000000..73bbfcd8b9652
--- /dev/null
+++ b/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp
@@ -0,0 +1,204 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --check-globals all --include-generated-funcs --version 4
+// RUN: %clang_cc1 -std=c++11 -triple riscv64-linux-gnu -target-feature +i -target-feature +m -emit-llvm %s -o - | FileCheck %s
+
+__attribute__((target_clones("default", "arch=+m"))) int foo1(void) {
+ return 1;
+}
+__attribute__((target_clones("default", "arch=+zbb", "arch=+m"))) int foo2(void) { return 2; }
+__attribute__((target_clones("default", "arch=+zbb,+c"))) int foo3(void) { return 3; }
+__attribute__((target_clones("default", "arch=+zbb,+v"))) int
+foo4(void) {
+ return 4;
+}
+__attribute__((target_clones("default"))) int foo5(void) { return 5; }
+__attribute__((target_clones("default", "arch=+zvkt"))) int foo6(void) { return 2; }
+
+int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
+
+//.
+// CHECK: @__riscv_feature_bits = external dso_local global { i32, [1 x i64] }
+// CHECK: @_Z4foo1v.ifunc = weak_odr alias i32 (), ptr @_Z4foo1v
+// CHECK: @_Z4foo2v.ifunc = weak_odr alias i32 (), ptr @_Z4foo2v
+// CHECK: @_Z4foo3v.ifunc = weak_odr alias i32 (), ptr @_Z4foo3v
+// CHECK: @_Z4foo4v.ifunc = weak_odr alias i32 (), ptr @_Z4foo4v
+// CHECK: @_Z4foo5v.ifunc = weak_odr alias i32 (), ptr @_Z4foo5v
+// CHECK: @_Z4foo6v.ifunc = weak_odr alias i32 (), ptr @_Z4foo6v
+// CHECK: @_Z4foo1v = weak_odr ifunc i32 (), ptr @_Z4foo1v.resolver
+// CHECK: @_Z4foo2v = weak_odr ifunc i32 (), ptr @_Z4foo2v.resolver
+// CHECK: @_Z4foo3v = weak_odr ifunc i32 (), ptr @_Z4foo3v.resolver
+// CHECK: @_Z4foo4v = weak_odr ifunc i32 (), ptr @_Z4foo4v.resolver
+// CHECK: @_Z4foo5v = weak_odr ifunc i32 (), ptr @_Z4foo5v.resolver
+// CHECK: @_Z4foo6v = weak_odr ifunc i32 (), ptr @_Z4foo6v.resolver
+//.
+// CHECK-LABEL: define dso_local noundef signext i32 @_Z4foo1v.default(
+// CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: ret i32 1
+//
+//
+// CHECK-LABEL: define weak_odr ptr @_Z4foo1v.resolver() comdat {
+// CHECK-NEXT: resolver_entry:
+// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
+// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 4096
+// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 4096
+// CHECK-NEXT: [[TMP5:%.*]] = and i1 true, [[TMP4]]
+// CHECK-NEXT: br i1 [[TMP5]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE]]
+// CHECK: resolver_return:
+// CHECK-NEXT: ret ptr @"_Z4foo1v.arch=+m"
+// CHECK: resolver_else:
+// CHECK-NEXT: ret ptr @_Z4foo1v.default
+//
+//
+// CHECK-LABEL: define dso_local noundef signext i32 @_Z4foo2v.default(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: ret i32 2
+//
+//
+// CHECK-LABEL: define weak_odr ptr @_Z4foo2v.resolver() comdat {
+// CHECK-NEXT: resolver_entry:
+// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
+// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 268435456
+// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 268435456
+// CHECK-NEXT: [[TMP5:%.*]] = and i1 true, [[TMP4]]
+// CHECK-NEXT: br i1 [[TMP5]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE]]
+// CHECK: resolver_return:
+// CHECK-NEXT: ret ptr @"_Z4foo2v.arch=+zbb"
+// CHECK: resolver_else:
+// CHECK-NEXT: [[TMP6:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
+// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i64 0, [[TMP6]]
+// CHECK-NEXT: br i1 [[TMP7]], label [[RESOVLER_COND1:%.*]], label [[RESOLVER_ELSE3:%.*]]
+// CHECK: resovler_cond1:
+// CHECK-NEXT: [[TMP8:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
+// CHECK-NEXT: [[TMP9:%.*]] = and i64 [[TMP8]], 4096
+// CHECK-NEXT: [[TMP10:%.*]] = icmp eq i64 [[TMP9]], 4096
+// CHECK-NEXT: [[TMP11:%.*]] = and i1 true, [[TMP10]]
+// CHECK-NEXT: br i1 [[TMP11]], label [[RESOLVER_RETURN2:%.*]], label [[RESOLVER_ELSE3]]
+// CHECK: resolver_return2:
+// CHECK-NEXT: ret ptr @"_Z4foo2v.arch=+m"
+// CHECK: resolver_else3:
+// CHECK-NEXT: ret ptr @_Z4foo2v.default
+//
+//
+// CHECK-LABEL: define dso_local noundef signext i32 @_Z4foo3v.default(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: ret i32 3
+//
+//
+// CHECK-LABEL: define weak_odr ptr @_Z4foo3v.resolver() comdat {
+// CHECK-NEXT: resolver_entry:
+// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
+// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 12
+// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 12
+// CHECK-NEXT: [[TMP5:%.*]] = and i1 true, [[TMP4]]
+// CHECK-NEXT: br i1 [[TMP5]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE]]
+// CHECK: resolver_return:
+// CHECK-NEXT: ret ptr @"_Z4foo3v.arch=+zbb,+c"
+// CHECK: resolver_else:
+// CHECK-NEXT: ret ptr @_Z4foo3v.default
+//
+//
+// CHECK-LABEL: define dso_local noundef signext i32 @_Z4foo4v.default(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: ret i32 4
+//
+//
+// CHECK-LABEL: define weak_odr ptr @_Z4foo4v.resolver() comdat {
+// CHECK-NEXT: resolver_entry:
+// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
+// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 69206016
+// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 69206016
+// CHECK-NEXT: [[TMP5:%.*]] = and i1 true, [[TMP4]]
+// CHECK-NEXT: br i1 [[TMP5]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE]]
+// CHECK: resolver_return:
+// CHECK-NEXT: ret ptr @"_Z4foo4v.arch=+zbb,+v"
+// CHECK: resolver_else:
+// CHECK-NEXT: ret ptr @_Z4foo4v.default
+//
+//
+// CHECK-LABEL: define dso_local noundef signext i32 @_Z4foo5v.default(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: ret i32 5
+//
+//
+// CHECK-LABEL: define weak_odr ptr @_Z4foo5v.resolver() comdat {
+// CHECK-NEXT: resolver_entry:
+// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: ret ptr @_Z4foo5v.default
+//
+//
+// CHECK-LABEL: define dso_local noundef signext i32 @_Z4foo6v.default(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: ret i32 2
+//
+//
+// CHECK-LABEL: define weak_odr ptr @_Z4foo6v.resolver() comdat {
+// CHECK-NEXT: resolver_entry:
+// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
+// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 576460752303423488
+// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 576460752303423488
+// CHECK-NEXT: [[TMP5:%.*]] = and i1 true, [[TMP4]]
+// CHECK-NEXT: br i1 [[TMP5]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE]]
+// CHECK: resolver_return:
+// CHECK-NEXT: ret ptr @"_Z4foo6v.arch=+zvkt"
+// CHECK: resolver_else:
+// CHECK-NEXT: ret ptr @_Z4foo6v.default
+//
+//
+// CHECK-LABEL: define dso_local noundef signext i32 @_Z3barv(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[CALL:%.*]] = call noundef signext i32 @_Z4foo1v()
+// CHECK-NEXT: [[CALL1:%.*]] = call noundef signext i32 @_Z4foo2v()
+// CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[CALL]], [[CALL1]]
+// CHECK-NEXT: [[CALL2:%.*]] = call noundef signext i32 @_Z4foo3v()
+// CHECK-NEXT: [[ADD3:%.*]] = add nsw i32 [[ADD]], [[CALL2]]
+// CHECK-NEXT: [[CALL4:%.*]] = call noundef signext i32 @_Z4foo4v()
+// CHECK-NEXT: [[ADD5:%.*]] = add nsw i32 [[ADD3]], [[CALL4]]
+// CHECK-NEXT: [[CALL6:%.*]] = call noundef signext i32 @_Z4foo5v()
+// CHECK-NEXT: [[ADD7:%.*]] = add nsw i32 [[ADD5]], [[CALL6]]
+// CHECK-NEXT: ret i32 [[ADD7]]
+//
+//.
+// CHECK: attributes #[[ATTR0]] = { mustprogress noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+64bit,+i,+m,+zmmul" }
+// CHECK: attributes #[[ATTR1:[0-9]+]] = { mustprogress noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+64bit,+i,+m,+zbb,+zmmul" }
+// CHECK: attributes #[[ATTR2:[0-9]+]] = { mustprogress noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+64bit,+c,+i,+m,+zbb,+zmmul" }
+// CHECK: attributes #[[ATTR3:[0-9]+]] = { mustprogress noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+64bit,+d,+f,+i,+m,+v,+zbb,+zicsr,+zmmul,+zve32f,+zve32x,+zve64d,+zve64f,+zve64x,+zvl128b,+zvl32b,+zvl64b" }
+// CHECK: attributes #[[ATTR4:[0-9]+]] = { mustprogress noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+64bit,+i,+m,+zmmul,+zvkt" }
+//.
+// CHECK: [[META0:![0-9]+]] = !{i32 1, !"wchar_size", i32 4}
+// CHECK: [[META1:![0-9]+]] = !{i32 1, !"target-abi", !"lp64"}
+// CHECK: [[META2:![0-9]+]] = !{i32 6, !"riscv-isa", [[META3:![0-9]+]]}
+// CHECK: [[META3]] = !{!"rv64i2p1_m2p0_zmmul1p0"}
+// CHECK: [[META4:![0-9]+]] = !{i32 8, !"SmallDataLimit", i32 0}
+// CHECK: [[META5:![0-9]+]] = !{!"{{.*}}clang version {{.*}}"}
+//.
diff --git a/clang/test/SemaCXX/attr-target-clones-riscv.cpp b/clang/test/SemaCXX/attr-target-clones-riscv.cpp
new file mode 100644
index 0000000000000..edee5f7c231e9
--- /dev/null
+++ b/clang/test/SemaCXX/attr-target-clones-riscv.cpp
@@ -0,0 +1,32 @@
+// RUN: %clang_cc1 -triple riscv64-linux-gnu -fsyntax-only -verify -fexceptions -fcxx-exceptions %s -std=c++14
+
+// expected-warning at +1 {{unsupported 'mcpu=sifive-u74' in the 'target_clones' attribute string; 'target_clones' attribute ignored}}
+void __attribute__((target_clones("default", "mcpu=sifive-u74"))) mcpu() {}
+
+// expected-warning at +1 {{unsupported 'mtune=sifive-u74' in the 'target_clones' attribute string; 'target_clones' attribute ignored}}
+void __attribute__((target_clones("default", "mtune=sifive-u74"))) mtune() {}
+
+// expected-warning at +1 {{version list contains duplicate entries}}
+void __attribute__((target_clones("default", "arch=+c", "arch=+c"))) dupVersion() {}
+
+// expected-warning at +1 {{unsupported '' in the 'target_clones' attribute string; 'target_clones' attribute ignored}}
+void __attribute__((target_clones("default", ""))) emptyVersion() {}
+
+// expected-error at +1 {{'target_clones' multiversioning requires a default target}}
+void __attribute__((target_clones("arch=+c"))) withoutDefault() {}
+
+// expected-warning at +1 {{unsupported '+c' in the 'target_clones' attribute string; 'target_clones' attribute ignored}}
+void __attribute__((target_clones("default", "+c"))) invaildVersion() {}
+
+// expected-warning at +1 {{unsupported 'arch=rv64g' in the 'target_clones' attribute string; 'target_clones' attribute ignored}}
+void __attribute__((target_clones("default", "arch=rv64g"))) fullArchString() {}
+
+
+void lambda() {
+ // expected-error at +1 {{attribute 'target_clones' multiversioned functions do not yet support lambdas}}
+ auto x = []() __attribute__((target_clones("default"))){};
+ x();
+ // expected-error at +1 {{attribute 'target_clones' multiversioned functions do not yet support lambdas}}
+ auto y = []() __attribute__((target_clones("arch=+v", "default"))){};
+ y();
+}
diff --git a/llvm/include/llvm/TargetParser/RISCVTargetParser.h b/llvm/include/llvm/TargetParser/RISCVTargetParser.h
index e998bc4ca59ee..59f4bc6c80f97 100644
--- a/llvm/include/llvm/TargetParser/RISCVTargetParser.h
+++ b/llvm/include/llvm/TargetParser/RISCVTargetParser.h
@@ -32,6 +32,8 @@ struct RISCVExtensionBitmask {
};
} // namespace RISCVExtensionBitmaskTable
+static constexpr unsigned RISCVFeatureBitSize = 1;
+
// We use 64 bits as the known part in the scalable vector types.
static constexpr unsigned RVVBitsPerBlock = 64;
@@ -44,6 +46,8 @@ StringRef getMArchFromMcpu(StringRef CPU);
void fillValidCPUArchList(SmallVectorImpl<StringRef> &Values, bool IsRV64);
void fillValidTuneCPUArchList(SmallVectorImpl<StringRef> &Values, bool IsRV64);
bool hasFastUnalignedAccess(StringRef CPU);
+llvm::SmallVector<unsigned long long>
+ getRequireFeatureBitMask(ArrayRef<StringRef>);
} // namespace RISCV
diff --git a/llvm/lib/TargetParser/RISCVTargetParser.cpp b/llvm/lib/TargetParser/RISCVTargetParser.cpp
index adb594777be80..2f9fef594e894 100644
--- a/llvm/lib/TargetParser/RISCVTargetParser.cpp
+++ b/llvm/lib/TargetParser/RISCVTargetParser.cpp
@@ -140,7 +140,7 @@ struct LessExtName {
};
} // namespace
-static RISCVExtensionBitmaskTable::RISCVExtensionBitmask
+static Expected<RISCVExtensionBitmaskTable::RISCVExtensionBitmask>
getExtensionBitmask(StringRef ExtName) {
ArrayRef<RISCVExtensionBitmaskTable::RISCVExtensionBitmask> ExtBitmasks =
RISCVExtensionBitmaskTable::ExtensionBitmask;
@@ -149,7 +149,21 @@ getExtensionBitmask(StringRef ExtName) {
if (I != ExtBitmasks.end())
return *I;
- return RISCVExtensionBitmaskTable::RISCVExtensionBitmask();
+ return createStringError("Unsupport extension");
+}
+
+llvm::SmallVector<unsigned long long>
+getRequireFeatureBitMask(ArrayRef<StringRef> Exts) {
+ llvm::SmallVector<unsigned long long> BitMasks(RISCV::RISCVFeatureBitSize);
+
+ for (auto Ext : Exts) {
+ Expected<RISCVExtensionBitmaskTable::RISCVExtensionBitmask> ExtBitmask =
+ getExtensionBitmask(Ext);
+ assert(ExtBitmask && "This extension doesn't has bitmask.");
+ BitMasks[ExtBitmask->GroupID] |= (1ULL << ExtBitmask->BitPosition);
+ }
+
+ return BitMasks;
}
} // namespace RISCV
>From 10f8f50f67a381153f8a1b49834cfcf93a94dcd1 Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Fri, 12 Jul 2024 02:17:34 -0700
Subject: [PATCH 04/19] Fixup format
---
clang/lib/CodeGen/CodeGenFunction.cpp | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp
index 27305e0f4b7b4..f2b2915878a90 100644
--- a/clang/lib/CodeGen/CodeGenFunction.cpp
+++ b/clang/lib/CodeGen/CodeGenFunction.cpp
@@ -2906,9 +2906,9 @@ void CodeGenFunction::EmitRISCVMultiVersionResolver(
unsigned MaxGroupIDUsed = 0;
llvm::SmallVector<StringRef, 8> CurrTargetAttrFeats;
- for (auto Feat: TargetAttrFeats)
+ for (auto Feat : TargetAttrFeats)
CurrTargetAttrFeats.push_back(StringRef(Feat).substr(1));
-
+
llvm::Value *SecondCondition =
EmitRISCVCpuSupports(CurrTargetAttrFeats, MaxGroupIDUsed);
>From 85264d19c988144745df77c9578a6c96287be0e4 Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Tue, 16 Jul 2024 06:15:27 -0700
Subject: [PATCH 05/19] __init_riscv_features_bit -> __init_riscv_features_bits
---
clang/lib/CodeGen/CGBuiltin.cpp | 2 +-
clang/test/CodeGen/attr-target-clones-riscv.c | 12 ++++++------
clang/test/CodeGenCXX/attr-target-clones-riscv.cpp | 12 ++++++------
3 files changed, 13 insertions(+), 13 deletions(-)
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 42647224a88a4..83ad64cb1cc3c 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -14218,7 +14218,7 @@ Value *CodeGenFunction::EmitAArch64CpuInit() {
Value *CodeGenFunction::EmitRISCVCpuInit() {
llvm::FunctionType *FTy = llvm::FunctionType::get(VoidTy, false);
llvm::FunctionCallee Func =
- CGM.CreateRuntimeFunction(FTy, "__init_riscv_features_bit");
+ CGM.CreateRuntimeFunction(FTy, "__init_riscv_features_bits");
cast<llvm::GlobalValue>(Func.getCallee())->setDSOLocal(true);
cast<llvm::GlobalValue>(Func.getCallee())
->setDLLStorageClass(llvm::GlobalValue::DefaultStorageClass);
diff --git a/clang/test/CodeGen/attr-target-clones-riscv.c b/clang/test/CodeGen/attr-target-clones-riscv.c
index 1088dd56a4d83..38adba7eac312 100644
--- a/clang/test/CodeGen/attr-target-clones-riscv.c
+++ b/clang/test/CodeGen/attr-target-clones-riscv.c
@@ -38,7 +38,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @foo1.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: call void @__init_riscv_features_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
@@ -62,7 +62,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @foo2.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: call void @__init_riscv_features_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
@@ -98,7 +98,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @foo3.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: call void @__init_riscv_features_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
@@ -122,7 +122,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @foo4.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: call void @__init_riscv_features_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
@@ -146,7 +146,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @foo5.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: call void @__init_riscv_features_bits()
// CHECK-NEXT: ret ptr @foo5.default
//
//
@@ -158,7 +158,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @foo6.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: call void @__init_riscv_features_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
diff --git a/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp b/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp
index 73bbfcd8b9652..024c1c715c617 100644
--- a/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp
+++ b/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp
@@ -38,7 +38,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @_Z4foo1v.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: call void @__init_riscv_features_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
@@ -62,7 +62,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @_Z4foo2v.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: call void @__init_riscv_features_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
@@ -98,7 +98,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @_Z4foo3v.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: call void @__init_riscv_features_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
@@ -122,7 +122,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @_Z4foo4v.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: call void @__init_riscv_features_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
@@ -146,7 +146,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @_Z4foo5v.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: call void @__init_riscv_features_bits()
// CHECK-NEXT: ret ptr @_Z4foo5v.default
//
//
@@ -158,7 +158,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @_Z4foo6v.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bit()
+// CHECK-NEXT: call void @__init_riscv_features_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
>From 265973a60ec446f2b0f2ce292f6f045a1744dd6c Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Tue, 16 Jul 2024 19:06:17 -0700
Subject: [PATCH 06/19] Update __init_riscv_feature_bits naming
---
clang/lib/CodeGen/CGBuiltin.cpp | 2 +-
clang/test/CodeGen/attr-target-clones-riscv.c | 12 ++++++------
clang/test/CodeGenCXX/attr-target-clones-riscv.cpp | 12 ++++++------
3 files changed, 13 insertions(+), 13 deletions(-)
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 83ad64cb1cc3c..aebd826c8077f 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -14218,7 +14218,7 @@ Value *CodeGenFunction::EmitAArch64CpuInit() {
Value *CodeGenFunction::EmitRISCVCpuInit() {
llvm::FunctionType *FTy = llvm::FunctionType::get(VoidTy, false);
llvm::FunctionCallee Func =
- CGM.CreateRuntimeFunction(FTy, "__init_riscv_features_bits");
+ CGM.CreateRuntimeFunction(FTy, "__init_riscv_feature_bits");
cast<llvm::GlobalValue>(Func.getCallee())->setDSOLocal(true);
cast<llvm::GlobalValue>(Func.getCallee())
->setDLLStorageClass(llvm::GlobalValue::DefaultStorageClass);
diff --git a/clang/test/CodeGen/attr-target-clones-riscv.c b/clang/test/CodeGen/attr-target-clones-riscv.c
index 38adba7eac312..27e8d20f7ba17 100644
--- a/clang/test/CodeGen/attr-target-clones-riscv.c
+++ b/clang/test/CodeGen/attr-target-clones-riscv.c
@@ -38,7 +38,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @foo1.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bits()
+// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
@@ -62,7 +62,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @foo2.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bits()
+// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
@@ -98,7 +98,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @foo3.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bits()
+// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
@@ -122,7 +122,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @foo4.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bits()
+// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
@@ -146,7 +146,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @foo5.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bits()
+// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: ret ptr @foo5.default
//
//
@@ -158,7 +158,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @foo6.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bits()
+// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
diff --git a/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp b/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp
index 024c1c715c617..6bac679cf1425 100644
--- a/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp
+++ b/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp
@@ -38,7 +38,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @_Z4foo1v.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bits()
+// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
@@ -62,7 +62,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @_Z4foo2v.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bits()
+// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
@@ -98,7 +98,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @_Z4foo3v.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bits()
+// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
@@ -122,7 +122,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @_Z4foo4v.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bits()
+// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
@@ -146,7 +146,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @_Z4foo5v.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bits()
+// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: ret ptr @_Z4foo5v.default
//
//
@@ -158,7 +158,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
//
// CHECK-LABEL: define weak_odr ptr @_Z4foo6v.resolver() comdat {
// CHECK-NEXT: resolver_entry:
-// CHECK-NEXT: call void @__init_riscv_features_bits()
+// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
>From fe95aa277eba098a3a29d916417e1e97f8e1332b Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Tue, 16 Jul 2024 23:11:38 -0700
Subject: [PATCH 07/19] Rename variable
---
clang/lib/CodeGen/CodeGenFunction.cpp | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp
index f2b2915878a90..c79e9747a5193 100644
--- a/clang/lib/CodeGen/CodeGenFunction.cpp
+++ b/clang/lib/CodeGen/CodeGenFunction.cpp
@@ -2909,7 +2909,7 @@ void CodeGenFunction::EmitRISCVMultiVersionResolver(
for (auto Feat : TargetAttrFeats)
CurrTargetAttrFeats.push_back(StringRef(Feat).substr(1));
- llvm::Value *SecondCondition =
+ llvm::Value *FeatsCondition =
EmitRISCVCpuSupports(CurrTargetAttrFeats, MaxGroupIDUsed);
Builder.SetInsertPoint(CurBlock);
@@ -2926,7 +2926,7 @@ void CodeGenFunction::EmitRISCVMultiVersionResolver(
Builder.CreateCondBr(FirstCondition, SecondCond, ElseBlock);
Builder.SetInsertPoint(SecondCond);
- Builder.CreateCondBr(SecondCondition, RetBlock, ElseBlock);
+ Builder.CreateCondBr(FeatsCondition, RetBlock, ElseBlock);
CurBlock = ElseBlock;
}
>From 73115f80be1609529d800243964c72bebeefd5a1 Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Tue, 16 Jul 2024 23:17:05 -0700
Subject: [PATCH 08/19] SecondCond -> FeatsCondBB
---
clang/lib/CodeGen/CodeGenFunction.cpp | 13 ++++++-------
1 file changed, 6 insertions(+), 7 deletions(-)
diff --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp
index c79e9747a5193..d4dc80ee2a347 100644
--- a/clang/lib/CodeGen/CodeGenFunction.cpp
+++ b/clang/lib/CodeGen/CodeGenFunction.cpp
@@ -2898,17 +2898,16 @@ void CodeGenFunction::EmitRISCVMultiVersionResolver(
.Features;
if (!TargetAttrFeats.empty()) {
-
- llvm::BasicBlock *SecondCond =
- createBasicBlock("resovler_cond", Resolver);
-
- Builder.SetInsertPoint(SecondCond);
unsigned MaxGroupIDUsed = 0;
llvm::SmallVector<StringRef, 8> CurrTargetAttrFeats;
for (auto Feat : TargetAttrFeats)
CurrTargetAttrFeats.push_back(StringRef(Feat).substr(1));
+ llvm::BasicBlock *FeatsCondBB =
+ createBasicBlock("resovler_cond", Resolver);
+
+ Builder.SetInsertPoint(FeatsCondBB);
llvm::Value *FeatsCondition =
EmitRISCVCpuSupports(CurrTargetAttrFeats, MaxGroupIDUsed);
@@ -2923,9 +2922,9 @@ void CodeGenFunction::EmitRISCVMultiVersionResolver(
llvm::BasicBlock *ElseBlock = createBasicBlock("resolver_else", Resolver);
Builder.SetInsertPoint(CurBlock);
- Builder.CreateCondBr(FirstCondition, SecondCond, ElseBlock);
+ Builder.CreateCondBr(FirstCondition, FeatsCondBB, ElseBlock);
- Builder.SetInsertPoint(SecondCond);
+ Builder.SetInsertPoint(FeatsCondBB);
Builder.CreateCondBr(FeatsCondition, RetBlock, ElseBlock);
CurBlock = ElseBlock;
>From ac73857747f618c165a57f61f8580a491fdd8b47 Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Tue, 16 Jul 2024 23:44:28 -0700
Subject: [PATCH 09/19] Add comment
---
clang/lib/CodeGen/CodeGenFunction.cpp | 28 +++++++++++++++++++++++++--
1 file changed, 26 insertions(+), 2 deletions(-)
diff --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp
index d4dc80ee2a347..ed1826ca26c64 100644
--- a/clang/lib/CodeGen/CodeGenFunction.cpp
+++ b/clang/lib/CodeGen/CodeGenFunction.cpp
@@ -2897,6 +2897,29 @@ void CodeGenFunction::EmitRISCVMultiVersionResolver(
.parseTargetAttr(Options[Index].Conditions.Features[0])
.Features;
+ // Two conditions need to be checked for the current version:
+ //
+ // 1. LengthCondition: The maximum group ID of the required extension
+ // does not exceed the runtime object's length.
+ // __riscv_feature_bits.length > MAX_USED_GROUPID
+ //
+ // 2. FeaturesCondition: The bitmask of the required extension has been
+ // enabled by the runtime object.
+ // (__riscv_feature_bits.features[i] & REQUIRED_BITMASK) ==
+ // REQUIRED_BITMASK
+ //
+ // When both conditions are met, return this version of the function.
+ // Otherwise, try the next version.
+ //
+ // if (LengthConditionVersion1 && FeaturesConditionVersion1)
+ // return Version1;
+ // else if (LengthConditionVersion2 && FeaturesConditionVersion2)
+ // return Version2;
+ // else if (LengthConditionVersion3 && FeaturesConditionVersion3)
+ // return Version3;
+ // ...
+ // else
+ // return DefaultVersion;
if (!TargetAttrFeats.empty()) {
unsigned MaxGroupIDUsed = 0;
llvm::SmallVector<StringRef, 8> CurrTargetAttrFeats;
@@ -2912,7 +2935,8 @@ void CodeGenFunction::EmitRISCVMultiVersionResolver(
EmitRISCVCpuSupports(CurrTargetAttrFeats, MaxGroupIDUsed);
Builder.SetInsertPoint(CurBlock);
- llvm::Value *FirstCondition = EmitRISCVFeatureBitsLength(MaxGroupIDUsed);
+ llvm::Value *MaxGroupLengthCondition =
+ EmitRISCVFeatureBitsLength(MaxGroupIDUsed);
llvm::BasicBlock *RetBlock =
createBasicBlock("resolver_return", Resolver);
@@ -2922,7 +2946,7 @@ void CodeGenFunction::EmitRISCVMultiVersionResolver(
llvm::BasicBlock *ElseBlock = createBasicBlock("resolver_else", Resolver);
Builder.SetInsertPoint(CurBlock);
- Builder.CreateCondBr(FirstCondition, FeatsCondBB, ElseBlock);
+ Builder.CreateCondBr(MaxGroupLengthCondition, FeatsCondBB, ElseBlock);
Builder.SetInsertPoint(FeatsCondBB);
Builder.CreateCondBr(FeatsCondition, RetBlock, ElseBlock);
>From 89e72e8428de924a9c91e3ef8ed578c888f1a8aa Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Tue, 16 Jul 2024 23:46:23 -0700
Subject: [PATCH 10/19] Move MaxGroupIDUsed closer to use
---
clang/lib/CodeGen/CodeGenFunction.cpp | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp
index ed1826ca26c64..db43abfcd8f58 100644
--- a/clang/lib/CodeGen/CodeGenFunction.cpp
+++ b/clang/lib/CodeGen/CodeGenFunction.cpp
@@ -2921,7 +2921,6 @@ void CodeGenFunction::EmitRISCVMultiVersionResolver(
// else
// return DefaultVersion;
if (!TargetAttrFeats.empty()) {
- unsigned MaxGroupIDUsed = 0;
llvm::SmallVector<StringRef, 8> CurrTargetAttrFeats;
for (auto Feat : TargetAttrFeats)
@@ -2931,6 +2930,7 @@ void CodeGenFunction::EmitRISCVMultiVersionResolver(
createBasicBlock("resovler_cond", Resolver);
Builder.SetInsertPoint(FeatsCondBB);
+ unsigned MaxGroupIDUsed = 0;
llvm::Value *FeatsCondition =
EmitRISCVCpuSupports(CurrTargetAttrFeats, MaxGroupIDUsed);
>From 82228b7843b91d92f23170d64ea25a30407535be Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Wed, 17 Jul 2024 18:58:34 -0700
Subject: [PATCH 11/19] invaild -> invalid
---
...-clones-riscv-invaild.c => attr-target-clones-riscv-invalid.c} | 0
1 file changed, 0 insertions(+), 0 deletions(-)
rename clang/test/CodeGen/{attr-target-clones-riscv-invaild.c => attr-target-clones-riscv-invalid.c} (100%)
diff --git a/clang/test/CodeGen/attr-target-clones-riscv-invaild.c b/clang/test/CodeGen/attr-target-clones-riscv-invalid.c
similarity index 100%
rename from clang/test/CodeGen/attr-target-clones-riscv-invaild.c
rename to clang/test/CodeGen/attr-target-clones-riscv-invalid.c
>From 3ccd54d477c2900fd89d3c0bbce170fdecce03c4 Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Wed, 17 Jul 2024 19:12:01 -0700
Subject: [PATCH 12/19] unsigned long long -> uint64
---
clang/lib/CodeGen/CGBuiltin.cpp | 2 +-
llvm/include/llvm/TargetParser/RISCVTargetParser.h | 3 +--
llvm/lib/TargetParser/RISCVTargetParser.cpp | 5 ++---
3 files changed, 4 insertions(+), 6 deletions(-)
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index aebd826c8077f..f2e250d6c5852 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -14301,7 +14301,7 @@ Value *CodeGenFunction::EmitRISCVCpuSupports(ArrayRef<StringRef> FeaturesStrs,
return FeaturesBit;
};
- SmallVector<unsigned long long> RequireFeatureBits =
+ SmallVector<uint64_t> RequireFeatureBits =
llvm::RISCV::getRequireFeatureBitMask(FeaturesStrs);
Value *Result = Builder.getTrue();
for (unsigned i = 0; i < RequireFeatureBits.size(); i++) {
diff --git a/llvm/include/llvm/TargetParser/RISCVTargetParser.h b/llvm/include/llvm/TargetParser/RISCVTargetParser.h
index 59f4bc6c80f97..e8a842f40f81b 100644
--- a/llvm/include/llvm/TargetParser/RISCVTargetParser.h
+++ b/llvm/include/llvm/TargetParser/RISCVTargetParser.h
@@ -46,8 +46,7 @@ StringRef getMArchFromMcpu(StringRef CPU);
void fillValidCPUArchList(SmallVectorImpl<StringRef> &Values, bool IsRV64);
void fillValidTuneCPUArchList(SmallVectorImpl<StringRef> &Values, bool IsRV64);
bool hasFastUnalignedAccess(StringRef CPU);
-llvm::SmallVector<unsigned long long>
- getRequireFeatureBitMask(ArrayRef<StringRef>);
+llvm::SmallVector<uint64_t> getRequireFeatureBitMask(ArrayRef<StringRef>);
} // namespace RISCV
diff --git a/llvm/lib/TargetParser/RISCVTargetParser.cpp b/llvm/lib/TargetParser/RISCVTargetParser.cpp
index 2f9fef594e894..c23c3238693b0 100644
--- a/llvm/lib/TargetParser/RISCVTargetParser.cpp
+++ b/llvm/lib/TargetParser/RISCVTargetParser.cpp
@@ -152,9 +152,8 @@ getExtensionBitmask(StringRef ExtName) {
return createStringError("Unsupport extension");
}
-llvm::SmallVector<unsigned long long>
-getRequireFeatureBitMask(ArrayRef<StringRef> Exts) {
- llvm::SmallVector<unsigned long long> BitMasks(RISCV::RISCVFeatureBitSize);
+llvm::SmallVector<uint64_t> getRequireFeatureBitMask(ArrayRef<StringRef> Exts) {
+ llvm::SmallVector<uint64_t> BitMasks(RISCV::RISCVFeatureBitSize);
for (auto Ext : Exts) {
Expected<RISCVExtensionBitmaskTable::RISCVExtensionBitmask> ExtBitmask =
>From 37b113cc5a28c75d364f5aed8111e7b40d85fe88 Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Wed, 17 Jul 2024 19:17:12 -0700
Subject: [PATCH 13/19] Update GEPIndices
---
clang/lib/CodeGen/CGBuiltin.cpp | 8 +++-----
1 file changed, 3 insertions(+), 5 deletions(-)
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index f2e250d6c5852..1cf745f93a4cc 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -14291,9 +14291,8 @@ Value *CodeGenFunction::EmitRISCVCpuSupports(ArrayRef<StringRef> FeaturesStrs,
auto LoadFeatureBit = [&](unsigned Index) {
// Create GEP then load.
Value *IndexVal = llvm::ConstantInt::get(Int32Ty, Index);
- std::vector<llvm::Value *> GEPIndices = {llvm::ConstantInt::get(Int32Ty, 0),
- llvm::ConstantInt::get(Int32Ty, 1),
- IndexVal};
+ llvm::Value *GEPIndices[] = {Builder.getInt32(0), Builder.getInt32(1),
+ IndexVal};
Value *Ptr =
Builder.CreateInBoundsGEP(StructTy, RISCVFeaturesBits, GEPIndices);
Value *FeaturesBit =
@@ -14328,8 +14327,7 @@ Value *CodeGenFunction::EmitRISCVFeatureBitsLength(unsigned MaxGroupIDUsed) {
cast<llvm::GlobalValue>(RISCVFeaturesBits)->setDSOLocal(true);
auto LoadMaxGroupID = [&]() {
- std::vector<llvm::Value *> GEPIndices = {
- llvm::ConstantInt::get(Int32Ty, 0), llvm::ConstantInt::get(Int32Ty, 0)};
+ llvm::Value *GEPIndices[] = {Builder.getInt32(0), Builder.getInt32(0)};
llvm::Value *Ptr =
Builder.CreateInBoundsGEP(StructTy, RISCVFeaturesBits, GEPIndices);
Value *Length =
>From 03c8927805b1bf71f7da8297290d742b8239d534 Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Wed, 17 Jul 2024 19:22:05 -0700
Subject: [PATCH 14/19] Early exit
---
clang/lib/CodeGen/CodeGenFunction.cpp | 49 +++++++++++++--------------
1 file changed, 24 insertions(+), 25 deletions(-)
diff --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp
index db43abfcd8f58..691c630a9f2b0 100644
--- a/clang/lib/CodeGen/CodeGenFunction.cpp
+++ b/clang/lib/CodeGen/CodeGenFunction.cpp
@@ -2897,6 +2897,9 @@ void CodeGenFunction::EmitRISCVMultiVersionResolver(
.parseTargetAttr(Options[Index].Conditions.Features[0])
.Features;
+ if (TargetAttrFeats.empty())
+ continue;
+
// Two conditions need to be checked for the current version:
//
// 1. LengthCondition: The maximum group ID of the required extension
@@ -2920,39 +2923,35 @@ void CodeGenFunction::EmitRISCVMultiVersionResolver(
// ...
// else
// return DefaultVersion;
- if (!TargetAttrFeats.empty()) {
- llvm::SmallVector<StringRef, 8> CurrTargetAttrFeats;
+ llvm::SmallVector<StringRef, 8> CurrTargetAttrFeats;
- for (auto Feat : TargetAttrFeats)
- CurrTargetAttrFeats.push_back(StringRef(Feat).substr(1));
+ for (auto Feat : TargetAttrFeats)
+ CurrTargetAttrFeats.push_back(StringRef(Feat).substr(1));
- llvm::BasicBlock *FeatsCondBB =
- createBasicBlock("resovler_cond", Resolver);
+ llvm::BasicBlock *FeatsCondBB = createBasicBlock("resovler_cond", Resolver);
- Builder.SetInsertPoint(FeatsCondBB);
- unsigned MaxGroupIDUsed = 0;
- llvm::Value *FeatsCondition =
- EmitRISCVCpuSupports(CurrTargetAttrFeats, MaxGroupIDUsed);
+ Builder.SetInsertPoint(FeatsCondBB);
+ unsigned MaxGroupIDUsed = 0;
+ llvm::Value *FeatsCondition =
+ EmitRISCVCpuSupports(CurrTargetAttrFeats, MaxGroupIDUsed);
- Builder.SetInsertPoint(CurBlock);
- llvm::Value *MaxGroupLengthCondition =
- EmitRISCVFeatureBitsLength(MaxGroupIDUsed);
+ Builder.SetInsertPoint(CurBlock);
+ llvm::Value *MaxGroupLengthCondition =
+ EmitRISCVFeatureBitsLength(MaxGroupIDUsed);
- llvm::BasicBlock *RetBlock =
- createBasicBlock("resolver_return", Resolver);
- CGBuilderTy RetBuilder(*this, RetBlock);
- CreateMultiVersionResolverReturn(CGM, Resolver, RetBuilder,
- Options[Index].Function, SupportsIFunc);
- llvm::BasicBlock *ElseBlock = createBasicBlock("resolver_else", Resolver);
+ llvm::BasicBlock *RetBlock = createBasicBlock("resolver_return", Resolver);
+ CGBuilderTy RetBuilder(*this, RetBlock);
+ CreateMultiVersionResolverReturn(CGM, Resolver, RetBuilder,
+ Options[Index].Function, SupportsIFunc);
+ llvm::BasicBlock *ElseBlock = createBasicBlock("resolver_else", Resolver);
- Builder.SetInsertPoint(CurBlock);
- Builder.CreateCondBr(MaxGroupLengthCondition, FeatsCondBB, ElseBlock);
+ Builder.SetInsertPoint(CurBlock);
+ Builder.CreateCondBr(MaxGroupLengthCondition, FeatsCondBB, ElseBlock);
- Builder.SetInsertPoint(FeatsCondBB);
- Builder.CreateCondBr(FeatsCondition, RetBlock, ElseBlock);
+ Builder.SetInsertPoint(FeatsCondBB);
+ Builder.CreateCondBr(FeatsCondition, RetBlock, ElseBlock);
- CurBlock = ElseBlock;
- }
+ CurBlock = ElseBlock;
}
// Finally, emit the default one.
>From 5b02bbbb7369554a3a6f9e0692af812015e5cf45 Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Wed, 17 Jul 2024 19:22:48 -0700
Subject: [PATCH 15/19] Fix typo
---
clang/lib/CodeGen/CodeGenFunction.cpp | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp
index 691c630a9f2b0..1507dfc469d40 100644
--- a/clang/lib/CodeGen/CodeGenFunction.cpp
+++ b/clang/lib/CodeGen/CodeGenFunction.cpp
@@ -2928,7 +2928,7 @@ void CodeGenFunction::EmitRISCVMultiVersionResolver(
for (auto Feat : TargetAttrFeats)
CurrTargetAttrFeats.push_back(StringRef(Feat).substr(1));
- llvm::BasicBlock *FeatsCondBB = createBasicBlock("resovler_cond", Resolver);
+ llvm::BasicBlock *FeatsCondBB = createBasicBlock("resolver_cond", Resolver);
Builder.SetInsertPoint(FeatsCondBB);
unsigned MaxGroupIDUsed = 0;
>From 09dec12193c577a79802bcf539ca6302924e83f2 Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Wed, 17 Jul 2024 21:41:34 -0700
Subject: [PATCH 16/19] Change ULT operand order and update testcase
---
clang/lib/CodeGen/CGBuiltin.cpp | 2 +-
clang/test/CodeGen/attr-target-clones-riscv.c | 36 +++++++++----------
.../CodeGenCXX/attr-target-clones-riscv.cpp | 36 +++++++++----------
3 files changed, 37 insertions(+), 37 deletions(-)
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 1cf745f93a4cc..b4d896cd09f33 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -14337,7 +14337,7 @@ Value *CodeGenFunction::EmitRISCVFeatureBitsLength(unsigned MaxGroupIDUsed) {
Value *UsedMaxGroupID = Builder.getInt64(MaxGroupIDUsed);
Value *GroupIDResult =
- Builder.CreateICmpULT(UsedMaxGroupID, LoadMaxGroupID());
+ Builder.CreateICmpULT(LoadMaxGroupID(), UsedMaxGroupID);
return GroupIDResult;
}
diff --git a/clang/test/CodeGen/attr-target-clones-riscv.c b/clang/test/CodeGen/attr-target-clones-riscv.c
index 27e8d20f7ba17..81474ed7724a2 100644
--- a/clang/test/CodeGen/attr-target-clones-riscv.c
+++ b/clang/test/CodeGen/attr-target-clones-riscv.c
@@ -40,9 +40,9 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
-// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
-// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 4096
// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 4096
@@ -64,9 +64,9 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
-// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
-// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 268435456
// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 268435456
@@ -76,9 +76,9 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: ret ptr @"foo2.arch=+zbb"
// CHECK: resolver_else:
// CHECK-NEXT: [[TMP6:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i64 0, [[TMP6]]
-// CHECK-NEXT: br i1 [[TMP7]], label [[RESOVLER_COND1:%.*]], label [[RESOLVER_ELSE3:%.*]]
-// CHECK: resovler_cond1:
+// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i64 [[TMP6]], 0
+// CHECK-NEXT: br i1 [[TMP7]], label [[RESOLVER_COND1:%.*]], label [[RESOLVER_ELSE3:%.*]]
+// CHECK: resolver_cond1:
// CHECK-NEXT: [[TMP8:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
// CHECK-NEXT: [[TMP9:%.*]] = and i64 [[TMP8]], 4096
// CHECK-NEXT: [[TMP10:%.*]] = icmp eq i64 [[TMP9]], 4096
@@ -100,9 +100,9 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
-// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
-// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 12
// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 12
@@ -124,9 +124,9 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
-// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
-// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 69206016
// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 69206016
@@ -160,9 +160,9 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
-// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
-// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 576460752303423488
// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 576460752303423488
diff --git a/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp b/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp
index 6bac679cf1425..63ba1836056a7 100644
--- a/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp
+++ b/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp
@@ -40,9 +40,9 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
-// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
-// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 4096
// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 4096
@@ -64,9 +64,9 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
-// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
-// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 268435456
// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 268435456
@@ -76,9 +76,9 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: ret ptr @"_Z4foo2v.arch=+zbb"
// CHECK: resolver_else:
// CHECK-NEXT: [[TMP6:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i64 0, [[TMP6]]
-// CHECK-NEXT: br i1 [[TMP7]], label [[RESOVLER_COND1:%.*]], label [[RESOLVER_ELSE3:%.*]]
-// CHECK: resovler_cond1:
+// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i64 [[TMP6]], 0
+// CHECK-NEXT: br i1 [[TMP7]], label [[RESOLVER_COND1:%.*]], label [[RESOLVER_ELSE3:%.*]]
+// CHECK: resolver_cond1:
// CHECK-NEXT: [[TMP8:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
// CHECK-NEXT: [[TMP9:%.*]] = and i64 [[TMP8]], 4096
// CHECK-NEXT: [[TMP10:%.*]] = icmp eq i64 [[TMP9]], 4096
@@ -100,9 +100,9 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
-// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
-// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 12
// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 12
@@ -124,9 +124,9 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
-// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
-// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 69206016
// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 69206016
@@ -160,9 +160,9 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 0, [[TMP0]]
-// CHECK-NEXT: br i1 [[TMP1]], label [[RESOVLER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
-// CHECK: resovler_cond:
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
+// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
// CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], 576460752303423488
// CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[TMP3]], 576460752303423488
>From e5c0c8d91107e7e0a7eceb1168e9a77cc838a06a Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Wed, 17 Jul 2024 22:11:40 -0700
Subject: [PATCH 17/19] less than -> greater than
---
clang/lib/CodeGen/CGBuiltin.cpp | 2 +-
clang/test/CodeGen/attr-target-clones-riscv.c | 12 ++++++------
clang/test/CodeGenCXX/attr-target-clones-riscv.cpp | 12 ++++++------
3 files changed, 13 insertions(+), 13 deletions(-)
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index b4d896cd09f33..08df92d09eb6c 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -14337,7 +14337,7 @@ Value *CodeGenFunction::EmitRISCVFeatureBitsLength(unsigned MaxGroupIDUsed) {
Value *UsedMaxGroupID = Builder.getInt64(MaxGroupIDUsed);
Value *GroupIDResult =
- Builder.CreateICmpULT(LoadMaxGroupID(), UsedMaxGroupID);
+ Builder.CreateICmpUGT(LoadMaxGroupID(), UsedMaxGroupID);
return GroupIDResult;
}
diff --git a/clang/test/CodeGen/attr-target-clones-riscv.c b/clang/test/CodeGen/attr-target-clones-riscv.c
index 81474ed7724a2..d3e0c63baa8a2 100644
--- a/clang/test/CodeGen/attr-target-clones-riscv.c
+++ b/clang/test/CodeGen/attr-target-clones-riscv.c
@@ -40,7 +40,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i64 [[TMP0]], 0
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
@@ -64,7 +64,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i64 [[TMP0]], 0
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
@@ -76,7 +76,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: ret ptr @"foo2.arch=+zbb"
// CHECK: resolver_else:
// CHECK-NEXT: [[TMP6:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i64 [[TMP6]], 0
+// CHECK-NEXT: [[TMP7:%.*]] = icmp ugt i64 [[TMP6]], 0
// CHECK-NEXT: br i1 [[TMP7]], label [[RESOLVER_COND1:%.*]], label [[RESOLVER_ELSE3:%.*]]
// CHECK: resolver_cond1:
// CHECK-NEXT: [[TMP8:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
@@ -100,7 +100,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i64 [[TMP0]], 0
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
@@ -124,7 +124,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i64 [[TMP0]], 0
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
@@ -160,7 +160,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i64 [[TMP0]], 0
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
diff --git a/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp b/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp
index 63ba1836056a7..2b8c1650fd1b1 100644
--- a/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp
+++ b/clang/test/CodeGenCXX/attr-target-clones-riscv.cpp
@@ -40,7 +40,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i64 [[TMP0]], 0
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
@@ -64,7 +64,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i64 [[TMP0]], 0
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
@@ -76,7 +76,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: ret ptr @"_Z4foo2v.arch=+zbb"
// CHECK: resolver_else:
// CHECK-NEXT: [[TMP6:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i64 [[TMP6]], 0
+// CHECK-NEXT: [[TMP7:%.*]] = icmp ugt i64 [[TMP6]], 0
// CHECK-NEXT: br i1 [[TMP7]], label [[RESOLVER_COND1:%.*]], label [[RESOLVER_ELSE3:%.*]]
// CHECK: resolver_cond1:
// CHECK-NEXT: [[TMP8:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
@@ -100,7 +100,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i64 [[TMP0]], 0
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
@@ -124,7 +124,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i64 [[TMP0]], 0
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
@@ -160,7 +160,7 @@ int bar() { return foo1() + foo2() + foo3() + foo4() + foo5(); }
// CHECK-NEXT: resolver_entry:
// CHECK-NEXT: call void @__init_riscv_feature_bits()
// CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__riscv_feature_bits, align 8
-// CHECK-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 0
+// CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i64 [[TMP0]], 0
// CHECK-NEXT: br i1 [[TMP1]], label [[RESOLVER_COND:%.*]], label [[RESOLVER_ELSE:%.*]]
// CHECK: resolver_cond:
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr getelementptr inbounds ({ i32, [1 x i64] }, ptr @__riscv_feature_bits, i32 0, i32 1, i32 0), align 8
>From f639d5bbb77d1fdf0298f6b155ef35174d070d1b Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Wed, 17 Jul 2024 22:35:25 -0700
Subject: [PATCH 18/19] Update attr-target-clones-riscv-invalid.c
---
clang/test/CodeGen/attr-target-clones-riscv-invalid.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/clang/test/CodeGen/attr-target-clones-riscv-invalid.c b/clang/test/CodeGen/attr-target-clones-riscv-invalid.c
index f5e8f40d7a8f3..a84a0608044b5 100644
--- a/clang/test/CodeGen/attr-target-clones-riscv-invalid.c
+++ b/clang/test/CodeGen/attr-target-clones-riscv-invalid.c
@@ -1,8 +1,8 @@
// RUN: not %clang_cc1 -triple riscv64 -target-feature +i -emit-llvm -o - %s 2>&1 | FileCheck %s --check-prefix=CHECK-UNSUPPORT-OS
// CHECK-UNSUPPORT-OS: error: target_clones is currently only supported on Linux
-__attribute__((target_clones("default", "arch=+c"))) int foo2(void) {
+__attribute__((target_clones("default", "arch=+c"))) int foo(void) {
return 2;
}
-int bar() { return foo1()+foo2(); }
+int bar() { return foo(); }
>From e52d5793bb173fe207ee0548c7dd2f3ca2fbde71 Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen at sifive.com>
Date: Wed, 17 Jul 2024 23:06:40 -0700
Subject: [PATCH 19/19] Use unsigned instead of int
---
clang/lib/CodeGen/CodeGenFunction.cpp | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp
index 1507dfc469d40..1a54ab1fa317e 100644
--- a/clang/lib/CodeGen/CodeGenFunction.cpp
+++ b/clang/lib/CodeGen/CodeGenFunction.cpp
@@ -2879,7 +2879,7 @@ void CodeGenFunction::EmitRISCVMultiVersionResolver(
bool SupportsIFunc = getContext().getTargetInfo().supportsIFunc();
bool HasDefault = false;
- int DefaultIndex = 0;
+ unsigned DefaultIndex = 0;
// Check the each candidate function.
for (unsigned Index = 0; Index < Options.size(); Index++) {
More information about the cfe-commits
mailing list