Jump to content

Projects/knight/Dawn ISA

From ourproject.org wiki

Dawn implements floating point and SIMD extensions that are primarily used for scientific calculations.

Added with floating point extension:

Added to condition register spec

Condition Register flags
FI
FZ
FO
FU
FI
Category Name Description
Floating point Add FADD floating point add Round to nearest, ties to even
FADD.A floating point add Round to nearest, ties away from zero
FADD.Z floating point add Round toward 0
FADD.P floating point add Round toward +∞
FADD.N floating point add Round toward -∞
Floating point Subtract FSUB floating point subtract Round to nearest, ties to even
FSUB.A floating point subtract Round to nearest, ties away from zero
FSUB.Z floating point subtract Round toward 0
FSUB.P floating point subtract Round toward +∞
FSUB.N floating point subtract Round toward -∞
Floating point Multiply FMUL floating point multiply Round to nearest, ties to even
FMUL.A floating point multiply Round to nearest, ties away from zero
FMUL.Z floating point multiply Round toward 0
FMUL.P floating point multiply Round toward +∞
FMUL.N floating point multiply Round toward -∞
Floating point Divide FDIV floating point divide Round to nearest, ties to even
FDIV.A floating point divide Round to nearest, ties away from zero
FDIV.Z floating point divide Round toward 0
FDIV.P floating point divide Round toward +∞
FDIV.N floating point divide Round toward -∞
Other ABS floating point Absolute value
NABS Floating point Negative absolute value
CONVI floating point convert to integer Round to nearest, ties to even
CONVI.A floating point convert to integer Round to nearest, ties away from zero
CONVI.Z floating point convert to integer Round toward 0
CONVI.P floating point convert to integer Round toward +∞
CONVI.N floating point convert to integer Round toward -∞
CONVF Integer convert to floating point Round to nearest, ties to even
CONVF.A Integer convert to floating point Round to nearest, ties away from zero
CONVF.Z Integer convert to floating point Round toward 0
CONVF.P Integer convert to floating point Round toward +∞
CONVF.N Integer convert to floating point Round toward -∞
Floating point Pack PACKF Pack floating point into the bottom (16,32,64 or 128) bits of a register Round to nearest, ties to even
PACKF.A Pack floating point into the bottom (16,32,64 or 128) bits of a register Round to nearest, ties away from zero
PACKF.Z Pack floating point into the bottom (16,32,64 or 128) bits of a register Round toward 0
PACKF.P Pack floating point into the bottom (16,32,64 or 128) bits of a register Round toward +∞
PACKF.N Pack floating point into the bottom (16,32,64 or 128) bits of a register Round toward -∞
Floating point Jump JMPF.I Jumps to destination if Invalid operation bit set
JMPF.DZ Jumps to destination if Division by zero bit set
JMPF.O Jumps to destination if floating point Overflow bit set
JMPF.U Jumps to destination if floating point Underflow bit set
JMPF.I Jumps to destination if Inexact bit set
JMPF.F Jumps to destination if GPR1 is finite
JMPF.NAN Jumps to destination if GPR1 is NAN
JMPF.NI Jumps to destination if GPR1 is negative infinity
JMPF.PI Jumps to destination if GPR1 is positive infinity
JMPF.Z Jumps to destination if GPR1 is +0 or -0
Logical CMPF Floating point comparison
Load LOADF32 Load 32 bit floating point
LOADF32.U Load 32 bit floating point unaligned

Added with 16 bit floating extension

Category Name Description
Load LOADF16 Load 16 bit floating point
LOADF16.U Load 16 bit floating point unaligned

Added with 64 bit floating extension

Category Name Description
Load LOADF64 Load 64 bit floating point
LOADF64.U Load 64 bit floating point unaligned

Added with 128 bit floating extension

Category Name Description
Load LOADF128 Load 128 bit floating point
LOADF128.U Load 128 bit floating point unaligned