aboutsummaryrefslogtreecommitdiff
path: root/lib/Target/ARM
diff options
context:
space:
mode:
authorJim Grosbach <grosbach@apple.com>2010-03-25 23:32:19 +0000
committerJim Grosbach <grosbach@apple.com>2010-03-25 23:32:19 +0000
commitbd17bc96bf54cc58d91c2d20964c6c5e28bffa57 (patch)
treecab0b92000926f0dba8e5f471c7294d2009db045 /lib/Target/ARM
parent44424646ac9db5c4d3919462bd0831ec22783085 (diff)
switch the flag for using NEON for SP floating point to a subtarget 'feature'
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@99568 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'lib/Target/ARM')
-rw-r--r--lib/Target/ARM/ARM.td13
-rw-r--r--lib/Target/ARM/ARMSubtarget.cpp14
2 files changed, 11 insertions, 16 deletions
diff --git a/lib/Target/ARM/ARM.td b/lib/Target/ARM/ARM.td
index 50d54b436e..931c81c5b6 100644
--- a/lib/Target/ARM/ARM.td
+++ b/lib/Target/ARM/ARM.td
@@ -49,8 +49,14 @@ def FeatureFP16 : SubtargetFeature<"fp16", "HasFP16", "true",
// FIXME: Currently, this is only flagged for Cortex-A8. It may be true for
// others as well. We should do more benchmarking and confirm one way or
// the other.
-def HasSlowVMLx : SubtargetFeature<"vmlx", "SlowVMLx", "true",
- "Disable VFP MAC instructions">;
+def FeatureHasSlowVMLx : SubtargetFeature<"vmlx", "SlowVMLx", "true",
+ "Disable VFP MAC instructions">;
+// Some processors benefit from using NEON instructions for scalar
+// single-precision FP operations.
+def FeatureNEONForFP : SubtargetFeature<"neonfp", "UseNEONForSinglePrecisionFP",
+ "true",
+ "Use NEON for single precision FP">;
+
//===----------------------------------------------------------------------===//
// ARM Processors supported.
@@ -115,7 +121,8 @@ def : Processor<"arm1156t2f-s", ARMV6Itineraries,
// V7 Processors.
def : Processor<"cortex-a8", CortexA8Itineraries,
- [ArchV7A, FeatureThumb2, FeatureNEON, HasSlowVMLx]>;
+ [ArchV7A, FeatureThumb2, FeatureNEON, FeatureHasSlowVMLx,
+ FeatureNEONForFP]>;
def : ProcNoItin<"cortex-a9", [ArchV7A, FeatureThumb2, FeatureNEON]>;
//===----------------------------------------------------------------------===//
diff --git a/lib/Target/ARM/ARMSubtarget.cpp b/lib/Target/ARM/ARMSubtarget.cpp
index 76ed66fe7d..9e55cd8700 100644
--- a/lib/Target/ARM/ARMSubtarget.cpp
+++ b/lib/Target/ARM/ARMSubtarget.cpp
@@ -22,10 +22,6 @@ using namespace llvm;
static cl::opt<bool>
ReserveR9("arm-reserve-r9", cl::Hidden,
cl::desc("Reserve R9, making it unavailable as GPR"));
-static cl::opt<bool>
-UseNEONFP("arm-use-neon-fp",
- cl::desc("Use NEON for single-precision FP"),
- cl::init(false), cl::Hidden);
static cl::opt<bool>
UseMOVT("arm-use-movt",
@@ -35,7 +31,7 @@ ARMSubtarget::ARMSubtarget(const std::string &TT, const std::string &FS,
bool isT)
: ARMArchVersion(V4)
, ARMFPUType(None)
- , UseNEONForSinglePrecisionFP(UseNEONFP)
+ , UseNEONForSinglePrecisionFP(false)
, SlowVMLx(false)
, IsThumb(isT)
, ThumbMode(Thumb1)
@@ -116,14 +112,6 @@ ARMSubtarget::ARMSubtarget(const std::string &TT, const std::string &FS,
if (!isThumb() || hasThumb2())
PostRAScheduler = true;
-
- // Set CPU specific features.
- if (CPUString == "cortex-a8") {
- // On Cortex-a8, it's faster to perform some single-precision FP
- // operations with NEON instructions.
- if (UseNEONFP.getPosition() == 0)
- UseNEONForSinglePrecisionFP = true;
- }
}
/// GVIsIndirectSymbol - true if the GV will be accessed via an indirect symbol.