r254250 - ARM v8.1a adds Advanced SIMD instructions for Rounding Double Multiply
Alexandros Lamprineas via cfe-commits
cfe-commits at lists.llvm.org
Sun Nov 29 02:44:01 PST 2015
Author: alelab01
Date: Sun Nov 29 04:43:59 2015
New Revision: 254250
URL: http://llvm.org/viewvc/llvm-project?rev=254250&view=rev
Log:
ARM v8.1a adds Advanced SIMD instructions for Rounding Double Multiply
Add/Subtract.
The following instructions are added to AArch32 instruction set:
- VQRDMLAH: Vector Saturating Rounding Doubling Multiply Accumulate
Returning High Half
- VQRDMLSH: Vector Saturating Rounding Doubling Multiply Subtract
Returning High Half
The following instructions are added to AArch64 instruction set:
- SQRDMLAH: Signed Saturating Rounding Doubling Multiply Accumulate
Returning High Half
- SQRDMLSH: Signed Saturating Rounding Doubling Multiply Subtract
Returning High Half
This patch adds intrinsic and ACLE macro support for these instructions,
as well as corresponding tests.
Differential Revision: http://reviews.llvm.org/D14982
Modified:
cfe/trunk/include/clang/Basic/arm_neon.td
cfe/trunk/lib/Basic/Targets.cpp
cfe/trunk/test/Preprocessor/aarch64-target-features.c
cfe/trunk/test/Preprocessor/arm-target-features.c
Modified: cfe/trunk/include/clang/Basic/arm_neon.td
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/arm_neon.td?rev=254250&r1=254249&r2=254250&view=diff
==============================================================================
--- cfe/trunk/include/clang/Basic/arm_neon.td (original)
+++ cfe/trunk/include/clang/Basic/arm_neon.td Sun Nov 29 04:43:59 2015
@@ -373,6 +373,10 @@ def OP_QDMLSLHi_LN : Op<(call "vqdmlsl",
(splat $p2, $p3))>;
def OP_QDMULH_LN : Op<(call "vqdmulh", $p0, (splat $p1, $p2))>;
def OP_QRDMULH_LN : Op<(call "vqrdmulh", $p0, (splat $p1, $p2))>;
+def OP_QRDMLAH : Op<(call "vqadd", $p0, (call "vqrdmulh", $p1, $p2))>;
+def OP_QRDMLSH : Op<(call "vqsub", $p0, (call "vqrdmulh", $p1, $p2))>;
+def OP_QRDMLAH_LN : Op<(call "vqadd", $p0, (call "vqrdmulh", $p1, (splat $p2, $p3)))>;
+def OP_QRDMLSH_LN : Op<(call "vqsub", $p0, (call "vqrdmulh", $p1, (splat $p2, $p3)))>;
def OP_FMS_LN : Op<(call "vfma_lane", $p0, $p1, (op "-", $p2), $p3)>;
def OP_FMS_LNQ : Op<(call "vfma_laneq", $p0, $p1, (op "-", $p2), $p3)>;
def OP_TRN1 : Op<(shuffle $p0, $p1, (interleave (decimate mask0, 2),
@@ -473,6 +477,11 @@ def OP_SCALAR_QDMULL_LN : ScalarMulOp<"v
def OP_SCALAR_QDMULH_LN : ScalarMulOp<"vqdmulh">;
def OP_SCALAR_QRDMULH_LN : ScalarMulOp<"vqrdmulh">;
+def OP_SCALAR_QRDMLAH_LN : Op<(call "vqadd", $p0, (call "vqrdmulh", $p1,
+ (call "vget_lane", $p2, $p3)))>;
+def OP_SCALAR_QRDMLSH_LN : Op<(call "vqsub", $p0, (call "vqrdmulh", $p1,
+ (call "vget_lane", $p2, $p3)))>;
+
def OP_SCALAR_HALF_GET_LN : Op<(bitcast "float16_t",
(call "vget_lane",
(bitcast "int16x4_t", $p0), $p1))>;
@@ -514,6 +523,12 @@ def VMLS : IOpInst<"vmls", "dddd", "
def VMLSL : SOpInst<"vmlsl", "wwdd", "csiUcUsUi", OP_MLSL>;
def VQDMULH : SInst<"vqdmulh", "ddd", "siQsQi">;
def VQRDMULH : SInst<"vqrdmulh", "ddd", "siQsQi">;
+
+let ArchGuard = "defined(__ARM_FEATURE_QRDMX)" in {
+def VQRDMLAH : SOpInst<"vqrdmlah", "dddd", "siQsQi", OP_QRDMLAH>;
+def VQRDMLSH : SOpInst<"vqrdmlsh", "dddd", "siQsQi", OP_QRDMLSH>;
+}
+
def VQDMLAL : SInst<"vqdmlal", "wwdd", "si">;
def VQDMLSL : SInst<"vqdmlsl", "wwdd", "si">;
def VMULL : SInst<"vmull", "wdd", "csiUcUsUiPc">;
@@ -741,6 +756,12 @@ def VQDMULH_N : SInst<"vqdmulh_n", "
def VQDMULH_LANE : SOpInst<"vqdmulh_lane", "ddgi", "siQsQi", OP_QDMULH_LN>;
def VQRDMULH_N : SInst<"vqrdmulh_n", "dda", "siQsQi">;
def VQRDMULH_LANE : SOpInst<"vqrdmulh_lane", "ddgi", "siQsQi", OP_QRDMULH_LN>;
+
+let ArchGuard = "defined(__ARM_FEATURE_QRDMX)" in {
+def VQRDMLAH_LANE : SOpInst<"vqrdmlah_lane", "dddgi", "siQsQi", OP_QRDMLAH_LN>;
+def VQRDMLSH_LANE : SOpInst<"vqrdmlsh_lane", "dddgi", "siQsQi", OP_QRDMLSH_LN>;
+}
+
def VMLA_N : IOpInst<"vmla_n", "ddda", "siUsUifQsQiQUsQUiQf", OP_MLA_N>;
def VMLAL_N : SOpInst<"vmlal_n", "wwda", "siUsUi", OP_MLAL_N>;
def VQDMLAL_N : SInst<"vqdmlal_n", "wwda", "si">;
@@ -1160,6 +1181,11 @@ def VQDMULL_HIGH_LANEQ : SOpInst<"vqdmu
def VQDMULH_LANEQ : SOpInst<"vqdmulh_laneq", "ddji", "siQsQi", OP_QDMULH_LN>;
def VQRDMULH_LANEQ : SOpInst<"vqrdmulh_laneq", "ddji", "siQsQi", OP_QRDMULH_LN>;
+let ArchGuard = "defined(__ARM_FEATURE_QRDMX) && defined(__aarch64__)" in {
+def VQRDMLAH_LANEQ : SOpInst<"vqrdmlah_laneq", "dddji", "siQsQi", OP_QRDMLAH_LN>;
+def VQRDMLSH_LANEQ : SOpInst<"vqrdmlsh_laneq", "dddji", "siQsQi", OP_QRDMLSH_LN>;
+}
+
// Note: d type implemented by SCALAR_VMULX_LANE
def VMULX_LANE : IOpInst<"vmulx_lane", "ddgi", "fQfQd", OP_MULX_LN>;
// Note: d type is implemented by SCALAR_VMULX_LANEQ
@@ -1405,6 +1431,16 @@ def SCALAR_SQDMULH : SInst<"vqdmulh", "s
// Scalar Integer Saturating Rounding Doubling Multiply Half High
def SCALAR_SQRDMULH : SInst<"vqrdmulh", "sss", "SsSi">;
+let ArchGuard = "defined(__ARM_FEATURE_QRDMX) && defined(__aarch64__)" in {
+////////////////////////////////////////////////////////////////////////////////
+// Signed Saturating Rounding Doubling Multiply Accumulate Returning High Half
+def SCALAR_SQRDMLAH : SOpInst<"vqrdmlah", "ssss", "SsSi", OP_QRDMLAH>;
+
+////////////////////////////////////////////////////////////////////////////////
+// Signed Saturating Rounding Doubling Multiply Subtract Returning High Half
+def SCALAR_SQRDMLSH : SOpInst<"vqrdmlsh", "ssss", "SsSi", OP_QRDMLSH>;
+}
+
////////////////////////////////////////////////////////////////////////////////
// Scalar Floating-point Multiply Extended
def SCALAR_FMULX : IInst<"vmulx", "sss", "SfSd">;
@@ -1606,6 +1642,16 @@ def SCALAR_SQDMULH_LANEQ : SOpInst<"vqdm
def SCALAR_SQRDMULH_LANE : SOpInst<"vqrdmulh_lane", "ssdi", "SsSi", OP_SCALAR_QRDMULH_LN>;
def SCALAR_SQRDMULH_LANEQ : SOpInst<"vqrdmulh_laneq", "ssji", "SsSi", OP_SCALAR_QRDMULH_LN>;
+let ArchGuard = "defined(__ARM_FEATURE_QRDMX) && defined(__aarch64__)" in {
+// Signed Saturating Rounding Doubling Multiply Accumulate Returning High Half
+def SCALAR_SQRDMLAH_LANE : SOpInst<"vqrdmlah_lane", "sssdi", "SsSi", OP_SCALAR_QRDMLAH_LN>;
+def SCALAR_SQRDMLAH_LANEQ : SOpInst<"vqrdmlah_laneq", "sssji", "SsSi", OP_SCALAR_QRDMLAH_LN>;
+
+// Signed Saturating Rounding Doubling Multiply Subtract Returning High Half
+def SCALAR_SQRDMLSH_LANE : SOpInst<"vqrdmlsh_lane", "sssdi", "SsSi", OP_SCALAR_QRDMLSH_LN>;
+def SCALAR_SQRDMLSH_LANEQ : SOpInst<"vqrdmlsh_laneq", "sssji", "SsSi", OP_SCALAR_QRDMLSH_LN>;
+}
+
def SCALAR_VDUP_LANE : IInst<"vdup_lane", "sdi", "ScSsSiSlSfSdSUcSUsSUiSUlSPcSPs">;
def SCALAR_VDUP_LANEQ : IInst<"vdup_laneq", "sji", "ScSsSiSlSfSdSUcSUsSUiSUlSPcSPs">;
}
Modified: cfe/trunk/lib/Basic/Targets.cpp
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Basic/Targets.cpp?rev=254250&r1=254249&r2=254250&view=diff
==============================================================================
--- cfe/trunk/lib/Basic/Targets.cpp (original)
+++ cfe/trunk/lib/Basic/Targets.cpp Sun Nov 29 04:43:59 2015
@@ -4869,6 +4869,9 @@ public:
if (Opts.UnsafeFPMath)
Builder.defineMacro("__ARM_FP_FAST", "1");
+
+ if (ArchKind == llvm::ARM::AK_ARMV8_1A)
+ Builder.defineMacro("__ARM_FEATURE_QRDMX", "1");
}
ArrayRef<Builtin::Info> getTargetBuiltins() const override {
@@ -5250,6 +5253,7 @@ class AArch64TargetInfo : public TargetI
unsigned CRC;
unsigned Crypto;
unsigned Unaligned;
+ unsigned V8_1A;
static const Builtin::Info BuiltinInfo[];
@@ -5372,6 +5376,9 @@ public:
if (Unaligned)
Builder.defineMacro("__ARM_FEATURE_UNALIGNED", "1");
+ if (V8_1A)
+ Builder.defineMacro("__ARM_FEATURE_QRDMX", "1");
+
// All of the __sync_(bool|val)_compare_and_swap_(1|2|4|8) builtins work.
Builder.defineMacro("__GCC_HAVE_SYNC_COMPARE_AND_SWAP_1");
Builder.defineMacro("__GCC_HAVE_SYNC_COMPARE_AND_SWAP_2");
@@ -5397,6 +5404,7 @@ public:
CRC = 0;
Crypto = 0;
Unaligned = 1;
+ V8_1A = 0;
for (const auto &Feature : Features) {
if (Feature == "+neon")
@@ -5407,6 +5415,8 @@ public:
Crypto = 1;
if (Feature == "+strict-align")
Unaligned = 0;
+ if (Feature == "+v8.1a")
+ V8_1A = 1;
}
setDataLayoutString();
Modified: cfe/trunk/test/Preprocessor/aarch64-target-features.c
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/Preprocessor/aarch64-target-features.c?rev=254250&r1=254249&r2=254250&view=diff
==============================================================================
--- cfe/trunk/test/Preprocessor/aarch64-target-features.c (original)
+++ cfe/trunk/test/Preprocessor/aarch64-target-features.c Sun Nov 29 04:43:59 2015
@@ -71,6 +71,9 @@
// CHECK-NEON: __ARM_NEON 1
// CHECK-NEON: __ARM_NEON_FP 0xE
+// RUN: %clang -target aarch64-none-eabi -march=armv8.1-a -x c -E -dM %s -o - | FileCheck --check-prefix=CHECK-V81A %s
+// CHECK-V81A: __ARM_FEATURE_QRDMX 1
+
// RUN: %clang -target aarch64 -march=arm64 -### -c %s 2>&1 | FileCheck -check-prefix=CHECK-ARCH-NOT-ACCEPT %s
// RUN: %clang -target aarch64 -march=aarch64 -### -c %s 2>&1 | FileCheck -check-prefix=CHECK-ARCH-NOT-ACCEPT %s
// CHECK-ARCH-NOT-ACCEPT: error: the clang compiler does not support
Modified: cfe/trunk/test/Preprocessor/arm-target-features.c
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/Preprocessor/arm-target-features.c?rev=254250&r1=254249&r2=254250&view=diff
==============================================================================
--- cfe/trunk/test/Preprocessor/arm-target-features.c (original)
+++ cfe/trunk/test/Preprocessor/arm-target-features.c Sun Nov 29 04:43:59 2015
@@ -407,4 +407,5 @@
// CHECK-V81A: __ARM_ARCH 8
// CHECK-V81A: __ARM_ARCH_8_1A__ 1
// CHECK-V81A: #define __ARM_ARCH_PROFILE 'A'
+// CHECK-V81A: __ARM_FEATURE_QRDMX 1
// CHECK-V81A: #define __ARM_FP 0xE
More information about the cfe-commits
mailing list