Skip to content

S
swiftshader
Commit 2aa852f4 by Alexis Hetu Committed by Alexis Hétu
Options

New Short->Int and UShort->Int constructors

Added 2 new constructors for the Int4 class in Nucleus, which are for conversion to Int4 from Short4 or UShort4. Some of the required code was extracted from Float4 constructors, which were cleaned up to remove the duplicate code. All 4 constructors (2 new Int4 and 2 modified Float4) behaved properly in dEQP tests. Change-Id: Ib400ea27a867d31901f4a9ca08bf6093f186b277 Reviewed-on: https://swiftshader-review.googlesource.com/4060Tested-by: 's avatarAlexis Hétu <sugoi@google.com> Reviewed-by: 's avatarNicolas Capens <capn@google.com>
parent 034dabf3
Showing with 70 additions and 118 deletions
src/Reactor/Nucleus.cpp
......@@ -5142,6 +5142,70 @@ namespace sw
storeValue(xyzw);
}
Int4::Int4(RValue<Short4> cast)
{
Value *long2 = UndefValue::get(Long2::getType());
Value *element = Nucleus::createBitCast(cast.value, Long::getType());
long2 = Nucleus::createInsertElement(long2, element, 0);
RValue<Int4> vector = RValue<Int4>(Nucleus::createBitCast(long2, Int4::getType()));
if(CPUID::supportsSSE4_1())
{
storeValue(x86::pmovsxwd(vector).value);
}
else
{
Value *b = Nucleus::createBitCast(vector.value, Short8::getType());
Constant *swizzle[8];
swizzle[0] = Nucleus::createConstantInt(0);
swizzle[1] = Nucleus::createConstantInt(0);
swizzle[2] = Nucleus::createConstantInt(1);
swizzle[3] = Nucleus::createConstantInt(1);
swizzle[4] = Nucleus::createConstantInt(2);
swizzle[5] = Nucleus::createConstantInt(2);
swizzle[6] = Nucleus::createConstantInt(3);
swizzle[7] = Nucleus::createConstantInt(3);
storeValue(Nucleus::createShuffleVector(b, b, Nucleus::createConstantVector(swizzle, 8)));
// Each Short is packed into each Int in the (Short | Short) format.
// Shifting by 16 will retrieve the original Short value.
// Shitfing an Int will propagate the sign bit, which will work
// for both positive and negative values of a Short.
*this >>= 16;
}
}
Int4::Int4(RValue<UShort4> cast)
{
Value *long2 = UndefValue::get(Long2::getType());
Value *element = Nucleus::createBitCast(cast.value, Long::getType());
long2 = Nucleus::createInsertElement(long2, element, 0);
RValue<Int4> vector = RValue<Int4>(Nucleus::createBitCast(long2, Int4::getType()));
if(CPUID::supportsSSE4_1())
{
storeValue(x86::pmovzxwd(RValue<Int4>(vector)).value);
}
else
{
Value *b = Nucleus::createBitCast(vector.value, Short8::getType());
Constant *swizzle[8];
swizzle[0] = Nucleus::createConstantInt(0);
swizzle[1] = Nucleus::createConstantInt(8);
swizzle[2] = Nucleus::createConstantInt(1);
swizzle[3] = Nucleus::createConstantInt(9);
swizzle[4] = Nucleus::createConstantInt(2);
swizzle[5] = Nucleus::createConstantInt(10);
swizzle[6] = Nucleus::createConstantInt(3);
swizzle[7] = Nucleus::createConstantInt(11);
storeValue(Nucleus::createShuffleVector(b, Nucleus::createNullValue(Short8::getType()), Nucleus::createConstantVector(swizzle, 8)));
}
}
Int4::Int4()
{
// xyzw.parent = this;
......@@ -6210,130 +6274,16 @@ namespace sw
{
xyzw.parent = this;
#if 0
Value *xyzw = Nucleus::createSIToFP(cast.value, Float4::getType()); // FIXME: Crashes
#elif 0
Value *vector = loadValue();
Value *i16x = Nucleus::createExtractElement(cast.value, 0);
Value *f32x = Nucleus::createSIToFP(i16x, Float::getType());
Value *x = Nucleus::createInsertElement(vector, f32x, 0);
Value *i16y = Nucleus::createExtractElement(cast.value, Nucleus::createConstantInt(1));
Value *f32y = Nucleus::createSIToFP(i16y, Float::getType());
Value *xy = Nucleus::createInsertElement(x, f32y, Nucleus::createConstantInt(1));
Value *i16z = Nucleus::createExtractElement(cast.value, Nucleus::createConstantInt(2));
Value *f32z = Nucleus::createSIToFP(i16z, Float::getType());
Value *xyz = Nucleus::createInsertElement(xy, f32z, Nucleus::createConstantInt(2));
Value *i16w = Nucleus::createExtractElement(cast.value, Nucleus::createConstantInt(3));
Value *f32w = Nucleus::createSIToFP(i16w, Float::getType());
Value *xyzw = Nucleus::createInsertElement(xyz, f32w, Nucleus::createConstantInt(3));
#else
Value *long2 = UndefValue::get(Long2::getType());
Value *element = Nucleus::createBitCast(cast.value, Long::getType());
long2 = Nucleus::createInsertElement(long2, element, 0);
RValue<Int4> vector = RValue<Int4>(Nucleus::createBitCast(long2, Int4::getType()));
Value *xyzw;
if(CPUID::supportsSSE4_1())
{
Value *c = x86::pmovsxwd(vector).value;
xyzw = Nucleus::createSIToFP(c, Float4::getType());
}
else
{
Value *b = Nucleus::createBitCast(vector.value, Short8::getType());
Constant *swizzle[8];
swizzle[0] = Nucleus::createConstantInt(0);
swizzle[1] = Nucleus::createConstantInt(0);
swizzle[2] = Nucleus::createConstantInt(1);
swizzle[3] = Nucleus::createConstantInt(1);
swizzle[4] = Nucleus::createConstantInt(2);
swizzle[5] = Nucleus::createConstantInt(2);
swizzle[6] = Nucleus::createConstantInt(3);
swizzle[7] = Nucleus::createConstantInt(3);
Value *c = Nucleus::createShuffleVector(b, b, Nucleus::createConstantVector(swizzle, 8));
Value *d = Nucleus::createBitCast(c, Int4::getType());
Value *e = Nucleus::createSIToFP(d, Float4::getType());
Constant *constantVector[4];
constantVector[0] = Nucleus::createConstantFloat(1.0f / (1 << 16));
constantVector[1] = Nucleus::createConstantFloat(1.0f / (1 << 16));
constantVector[2] = Nucleus::createConstantFloat(1.0f / (1 << 16));
constantVector[3] = Nucleus::createConstantFloat(1.0f / (1 << 16));
xyzw = Nucleus::createFMul(e, Nucleus::createConstantVector(constantVector, 4));
}
#endif
storeValue(xyzw);
Int4 c(cast);
storeValue(Nucleus::createSIToFP(RValue<Int4>(c).value, Float4::getType()));
}
Float4::Float4(RValue<UShort4> cast)
{
xyzw.parent = this;
#if 0
Value *xyzw = Nucleus::createUIToFP(cast.value, Float4::getType()); // FIXME: Crashes
#elif 0
Value *vector = loadValue();
Value *i16x = Nucleus::createExtractElement(cast.value, 0);
Value *f32x = Nucleus::createUIToFP(i16x, Float::getType());
Value *x = Nucleus::createInsertElement(vector, f32x, 0);
Value *i16y = Nucleus::createExtractElement(cast.value, Nucleus::createConstantInt(1));
Value *f32y = Nucleus::createUIToFP(i16y, Float::getType());
Value *xy = Nucleus::createInsertElement(x, f32y, Nucleus::createConstantInt(1));
Value *i16z = Nucleus::createExtractElement(cast.value, Nucleus::createConstantInt(2));
Value *f32z = Nucleus::createUIToFP(i16z, Float::getType());
Value *xyz = Nucleus::createInsertElement(xy, f32z, Nucleus::createConstantInt(2));
Value *i16w = Nucleus::createExtractElement(cast.value, Nucleus::createConstantInt(3));
Value *f32w = Nucleus::createUIToFP(i16w, Float::getType());
Value *xyzw = Nucleus::createInsertElement(xyz, f32w, Nucleus::createConstantInt(3));
#else
Value *long2 = UndefValue::get(Long2::getType());
Value *element = Nucleus::createBitCast(cast.value, Long::getType());
long2 = Nucleus::createInsertElement(long2, element, 0);
RValue<Int4> vector = RValue<Int4>(Nucleus::createBitCast(long2, Int4::getType()));
Value *c;
if(CPUID::supportsSSE4_1())
{
c = x86::pmovzxwd(RValue<Int4>(vector)).value;
}
else
{
Value *b = Nucleus::createBitCast(vector.value, Short8::getType());
Constant *swizzle[8];
swizzle[0] = Nucleus::createConstantInt(0);
swizzle[1] = Nucleus::createConstantInt(8);
swizzle[2] = Nucleus::createConstantInt(1);
swizzle[3] = Nucleus::createConstantInt(9);
swizzle[4] = Nucleus::createConstantInt(2);
swizzle[5] = Nucleus::createConstantInt(10);
swizzle[6] = Nucleus::createConstantInt(3);
swizzle[7] = Nucleus::createConstantInt(11);
c = Nucleus::createShuffleVector(b, Nucleus::createNullValue(Short8::getType()), Nucleus::createConstantVector(swizzle, 8));
}
Value *d = Nucleus::createBitCast(c, Int4::getType());
Value *e = Nucleus::createSIToFP(d, Float4::getType());
Value *xyzw = e;
#endif
storeValue(xyzw);
Int4 c(cast);
storeValue(Nucleus::createSIToFP(RValue<Int4>(c).value, Float4::getType()));
}
Float4::Float4(RValue<Int4> cast)
......
src/Reactor/Nucleus.hpp
......@@ -1619,6 +1619,8 @@ namespace sw
{
public:
explicit Int4(RValue<Float4> cast);
explicit Int4(RValue<Short4> cast);
explicit Int4(RValue<UShort4> cast);
Int4();
Int4(int xyzw);
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment