fhahn created this revision.
fhahn added reviewers: aemerson, ab, dmgreen, SjoerdMeijer, scanon.
Herald added subscribers: dang, kristof.beyls.
fhahn requested review of this revision.
Herald added a project: clang.
This patch adds support for `-m[no]fpf16`, ` `-m[no]fpf16fml` and
`-m[no]dotprod` feature flags for ARM.
The flags behave similar to specifying an -march flag with
+/-{fp16,fp16fml,dotprod}, but is more convenient for users that do not
want to provide a specific architecture version. There are similar flags
alread, like `-mcrc`.
Repository:
rG LLVM Github Monorepo
https://reviews.llvm.org/D113779
Files:
clang/include/clang/Driver/Options.td
clang/lib/Driver/ToolChains/Arch/AArch64.cpp
clang/lib/Driver/ToolChains/Arch/ARM.cpp
clang/test/Preprocessor/aarch64-target-features.c
clang/test/Preprocessor/arm-target-features.c
Index: clang/test/Preprocessor/arm-target-features.c
===================================================================
--- clang/test/Preprocessor/arm-target-features.c
+++ clang/test/Preprocessor/arm-target-features.c
@@ -25,6 +25,7 @@
// CHECK-V8A-ALLOW-FP-INSTR-V8A-NOT: #define __ARM_FEATURE_DOTPROD
// RUN: %clang -target arm-none-linux-gnueabi -march=armv8.2-a+nofp16fml+fp16 -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
+// RUN: %clang -target arm-none-linux-gnueabi -march=armv8.2-a -mnofp16fml -mfp16 -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
// RUN: %clang -target arm-none-linux-gnueabi -march=armv8.2-a+nofp16+fp16fml -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
// RUN: %clang -target arm-none-linux-gnueabi -march=armv8.2-a+fp16+nofp16fml -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
// RUN: %clang -target arm-none-linux-gnueabi -march=armv8-a+fp16fml -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
@@ -77,6 +78,7 @@
// CHECK-FULLFP16-SOFT-NOT: #define __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
// RUN: %clang -target arm-none-linux-gnueabi -march=armv8.2a+dotprod -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-DOTPROD %s
+// RUN: %clang -target arm-none-linux-gnueabi -march=armv8.2a -mdotprod -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-DOTPROD %s
// CHECK-DOTPROD: #define __ARM_FEATURE_DOTPROD 1
// RUN: %clang -target armv8r-none-linux-gnu -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-V8R %s
Index: clang/test/Preprocessor/aarch64-target-features.c
===================================================================
--- clang/test/Preprocessor/aarch64-target-features.c
+++ clang/test/Preprocessor/aarch64-target-features.c
@@ -199,20 +199,29 @@
// CHECK-SVE2BITPERM: __ARM_FEATURE_SVE2_BITPERM 1
// RUN: %clang -target aarch64-none-linux-gnu -march=armv8.2a+dotprod -x c -E -dM %s -o - | FileCheck --check-prefix=CHECK-DOTPROD %s
+// RUN: %clang -target aarch64-none-linux-gnu -march=armv8.2a -mdotprod -x c -E -dM %s -o - | FileCheck --check-prefix=CHECK-DOTPROD %s
// CHECK-DOTPROD: __ARM_FEATURE_DOTPROD 1
// On ARMv8.2-A and above, +fp16fml implies +fp16.
// On ARMv8.4-A and above, +fp16 implies +fp16fml.
// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8.2-a+nofp16fml+fp16 -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-NOFML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
+// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8.2-a -mnofp16fml -mfp16 -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-NOFML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8.2-a+nofp16+fp16fml -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-FML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
+// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8.2-a -mnofp16 -mfp16fml -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-FML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8.2-a+fp16+nofp16fml -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-NOFML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
+// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8.2-a -mfp16 -mnofp16fml -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-NOFML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8-a+fp16fml -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-FML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
+// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8-a -mfp16fml -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-FML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8-a+fp16 -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-NOFML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
+// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8-a -mfp16 -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-NOFML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8.4-a+nofp16fml+fp16 -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-FML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8.4-a+nofp16+fp16fml -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-FML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
+// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8.4-a -mnofp16 -mfp16fml -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-FML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8.4-a+fp16+nofp16fml -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-NOFML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8.4-a+fp16fml -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-FML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
+// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8.4-a -mfp16fml -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-FML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8.4-a+fp16 -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-FML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
+// RUN: %clang -target aarch64-none-linux-gnueabi -march=armv8.4-a -mfp16 -x c -E -dM %s -o - | FileCheck -match-full-lines --check-prefix=CHECK-FULLFP16-FML --check-prefix=CHECK-FULLFP16-VECTOR-SCALAR %s
// CHECK-FULLFP16-FML: #define __ARM_FEATURE_FP16_FML 1
// CHECK-FULLFP16-NOFML-NOT: #define __ARM_FEATURE_FP16_FML 1
// CHECK-FULLFP16-VECTOR-SCALAR: #define __ARM_FEATURE_FP16_SCALAR_ARITHMETIC 1
Index: clang/lib/Driver/ToolChains/Arch/ARM.cpp
===================================================================
--- clang/lib/Driver/ToolChains/Arch/ARM.cpp
+++ clang/lib/Driver/ToolChains/Arch/ARM.cpp
@@ -638,6 +638,29 @@
Features.push_back("-crc");
}
+ if (Arg *A = Args.getLastArg(options::OPT_mfp16, options::OPT_mnofp16)) {
+ if (A->getOption().matches(options::OPT_mfp16))
+ Features.push_back("+fullfp16");
+ else
+ Features.push_back("-fullfp16");
+ }
+
+ if (Arg *A =
+ Args.getLastArg(options::OPT_mfp16fml, options::OPT_mnofp16fml)) {
+ if (A->getOption().matches(options::OPT_mfp16fml))
+ Features.push_back("+fp16fml");
+ else
+ Features.push_back("-fp16fml");
+ }
+
+ if (Arg *A =
+ Args.getLastArg(options::OPT_mdotprod, options::OPT_mnodotprod)) {
+ if (A->getOption().matches(options::OPT_mdotprod))
+ Features.push_back("+dotprod");
+ else
+ Features.push_back("-dotprod");
+ }
+
// For Arch >= ARMv8.0 && A or R profile: crypto = sha2 + aes
// Rather than replace within the feature vector, determine whether each
// algorithm is enabled and append this to the end of the vector.
Index: clang/lib/Driver/ToolChains/Arch/AArch64.cpp
===================================================================
--- clang/lib/Driver/ToolChains/Arch/AArch64.cpp
+++ clang/lib/Driver/ToolChains/Arch/AArch64.cpp
@@ -304,6 +304,29 @@
Features.push_back("-crc");
}
+ if (Arg *A = Args.getLastArg(options::OPT_mfp16, options::OPT_mnofp16)) {
+ if (A->getOption().matches(options::OPT_mfp16))
+ Features.push_back("+fullfp16");
+ else
+ Features.push_back("-fullfp16");
+ }
+
+ if (Arg *A =
+ Args.getLastArg(options::OPT_mfp16fml, options::OPT_mnofp16fml)) {
+ if (A->getOption().matches(options::OPT_mfp16fml))
+ Features.push_back("+fp16fml");
+ else
+ Features.push_back("-fp16fml");
+ }
+
+ if (Arg *A =
+ Args.getLastArg(options::OPT_mdotprod, options::OPT_mnodotprod)) {
+ if (A->getOption().matches(options::OPT_mdotprod))
+ Features.push_back("+dotprod");
+ else
+ Features.push_back("-dotprod");
+ }
+
// Handle (arch-dependent) fp16fml/fullfp16 relationship.
// FIXME: this fp16fml option handling will be reimplemented after the
// TargetParser rewrite.
Index: clang/include/clang/Driver/Options.td
===================================================================
--- clang/include/clang/Driver/Options.td
+++ clang/include/clang/Driver/Options.td
@@ -3308,6 +3308,18 @@
BothFlags<[NoXarchOption, CC1Option], " the AAPCS standard requirement stating that"
" volatile bit-field width is dictated by the field container type. (ARM only).">>,
Group<m_arm_Features_Group>;
+def mfp16: Flag<["-"], "mfp16">, Group<m_arm_Features_Group>,
+ HelpText<"Allow use of FP16 instructions.">;
+def mnofp16: Flag<["-"], "mnofp16">, Group<m_arm_Features_Group>,
+ HelpText<"Disallow use of FP16 instructions.">;
+def mfp16fml: Flag<["-"], "mfp16fml">, Group<m_arm_Features_Group>,
+ HelpText<"Allow use of FP16 machine learning instructions.">;
+def mnofp16fml: Flag<["-"], "mnofp16fml">, Group<m_arm_Features_Group>,
+ HelpText<"Disallow use of FP16 FML instructions.">;
+def mdotprod: Flag<["-"], "mdotprod">, Group<m_arm_Features_Group>,
+ HelpText<"Allow use of dot product instructions.">;
+def mnodotprod: Flag<["-"], "mnodotprod">, Group<m_arm_Features_Group>,
+ HelpText<"Disallow use of dot product instructions.">;
def mgeneral_regs_only : Flag<["-"], "mgeneral-regs-only">, Group<m_Group>,
HelpText<"Generate code which only uses the general purpose registers (AArch64/x86 only)">;
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
