Refactor vector packing.
x86 vector packing instructions always treat the input as having signed
integer components, but can perform signed or unsigned saturation on
the output. In Reactor the Pack() intrinsic has overloads which
differentiate between them based on the signedness of the input, but
this is confusing.
Also simplify emulation of saturating add/subtract.
Bug b/37496082
Change-Id: I0625fff429ffb40f42baf9600c7760d9858b5d89
Reviewed-on: https://swiftshader-review.googlesource.com/12548
Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
diff --git a/src/Main/FrameBuffer.cpp b/src/Main/FrameBuffer.cpp
index e95f766..d3e6383 100644
--- a/src/Main/FrameBuffer.cpp
+++ b/src/Main/FrameBuffer.cpp
@@ -253,10 +253,10 @@
case FORMAT_A16B16G16R16:
For(, x < width - 1, x += 2)
{
- UShort4 c0 = As<UShort4>(Swizzle(*Pointer<Short4>(s + 0), 0xC6)) >> 8;
- UShort4 c1 = As<UShort4>(Swizzle(*Pointer<Short4>(s + 8), 0xC6)) >> 8;
+ Short4 c0 = As<UShort4>(Swizzle(*Pointer<Short4>(s + 0), 0xC6)) >> 8;
+ Short4 c1 = As<UShort4>(Swizzle(*Pointer<Short4>(s + 8), 0xC6)) >> 8;
- *Pointer<Int2>(d) = As<Int2>(Pack(c0, c1));
+ *Pointer<Int2>(d) = As<Int2>(PackUnsigned(c0, c1));
s += 2 * sBytes;
d += 2 * dBytes;
@@ -300,9 +300,9 @@
break;
case FORMAT_A16B16G16R16:
{
- UShort4 c = As<UShort4>(Swizzle(*Pointer<Short4>(s), 0xC6)) >> 8;
+ Short4 c = As<UShort4>(Swizzle(*Pointer<Short4>(s), 0xC6)) >> 8;
- *Pointer<Int>(d) = Int(As<Int2>(Pack(c, c)));
+ *Pointer<Int>(d) = Int(As<Int2>(PackUnsigned(c, c)));
}
break;
case FORMAT_R5G6B5:
@@ -361,10 +361,10 @@
case FORMAT_A16B16G16R16:
For(, x < width - 1, x += 2)
{
- UShort4 c0 = *Pointer<UShort4>(s + 0) >> 8;
- UShort4 c1 = *Pointer<UShort4>(s + 8) >> 8;
+ Short4 c0 = *Pointer<UShort4>(s + 0) >> 8;
+ Short4 c1 = *Pointer<UShort4>(s + 8) >> 8;
- *Pointer<Int2>(d) = As<Int2>(Pack(c0, c1));
+ *Pointer<Int2>(d) = As<Int2>(PackUnsigned(c0, c1));
s += 2 * sBytes;
d += 2 * dBytes;
@@ -408,9 +408,9 @@
break;
case FORMAT_A16B16G16R16:
{
- UShort4 c = *Pointer<UShort4>(s) >> 8;
+ Short4 c = *Pointer<UShort4>(s) >> 8;
- *Pointer<Int>(d) = Int(As<Int2>(Pack(c, c)));
+ *Pointer<Int>(d) = Int(As<Int2>(PackUnsigned(c, c)));
}
break;
case FORMAT_R5G6B5:
@@ -503,8 +503,8 @@
break;
case FORMAT_A16B16G16R16:
{
- UShort4 cc = *Pointer<UShort4>(s) >> 8;
- Int c = Int(As<Int2>(Pack(cc, cc)));
+ Short4 cc = *Pointer<UShort4>(s) >> 8;
+ Int c = Int(As<Int2>(PackUnsigned(cc, cc)));
*Pointer<Short>(d) = Short((c & 0x00F80000) >> 19 |
(c & 0x0000FC00) >> 5 |
@@ -615,7 +615,7 @@
{
case FORMAT_X8R8G8B8:
case FORMAT_A8R8G8B8:
- *Pointer<Byte4>(d) = Byte4(Pack(As<UShort4>(c1), As<UShort4>(c1)));
+ *Pointer<Byte4>(d) = Byte4(PackUnsigned(c1, c1));
break;
case FORMAT_X8B8G8R8:
case FORMAT_A8B8G8R8:
@@ -624,12 +624,12 @@
{
c1 = Swizzle(c1, 0xC6);
- *Pointer<Byte4>(d) = Byte4(Pack(As<UShort4>(c1), As<UShort4>(c1)));
+ *Pointer<Byte4>(d) = Byte4(PackUnsigned(c1, c1));
}
break;
case FORMAT_R8G8B8:
{
- Int c = Int(As<Int2>(Pack(As<UShort4>(c1), As<UShort4>(c1))));
+ Int c = Int(As<Int2>(PackUnsigned(c1, c1)));
*Pointer<Byte>(d + 0) = Byte(c >> 0);
*Pointer<Byte>(d + 1) = Byte(c >> 8);
@@ -638,7 +638,7 @@
break;
case FORMAT_R5G6B5:
{
- Int c = Int(As<Int2>(Pack(As<UShort4>(c1), As<UShort4>(c1))));
+ Int c = Int(As<Int2>(PackUnsigned(c1, c1)));
*Pointer<Short>(d) = Short((c & 0x00F80000) >> 8 |
(c & 0x0000FC00) >> 5 |
