Add V extension pseudoinstructions #111
Conversation
Signed-off-by: Afonso Oliveira <[email protected]>
Signed-off-by: Afonso Oliveira <[email protected]>
|
@cmuellner I added the ones you mentioned, but I couldn't really find many more. |
|
@cmuellner just a quick ping, is there anything else I should enhance in the PR? |
|
👀 :) |
src/asm-manual.adoc
Outdated
| |vmsgeu.vi vd, va, i, vm | vmsgtu.vi vd, va, i-1, vm | Vector >= Immediate, unsigned| | ||
| |vmsge.vv vd, va, vb, vm | vmsle.vv vd, vb, va, vm | Vector >= Vector| | ||
| |vmsgeu.vv vd, va, vb, vm | vmsleu.vv vd, vb, va, vm | Vector >= Vector, unsigned | | ||
| |vmsge.vx vd, va, x, vm | vmsle.vx vd, x, va, vm | Vector >= scalar| |
There was a problem hiding this comment.
This is the not the correct expansion for vmsge.vx
From the spec.
Sequences to synthesize `vmsge{u}.vx` instruction
va >= x, x > minimum
addi t0, x, -1; vmsgt{u}.vx vd, va, t0, vm
unmasked va >= x
pseudoinstruction: vmsge{u}.vx vd, va, x
expansion: vmslt{u}.vx vd, va, x; vmnand.mm vd, vd, vd
masked va >= x, vd != v0
pseudoinstruction: vmsge{u}.vx vd, va, x, v0.t
expansion: vmslt{u}.vx vd, va, x, v0.t; vmxor.mm vd, vd, v0
masked va >= x, vd == v0
pseudoinstruction: vmsge{u}.vx vd, va, x, v0.t, vt
expansion: vmslt{u}.vx vt, va, x; vmandn.mm vd, vd, vt
masked va >= x, any vd
pseudoinstruction: vmsge{u}.vx vd, va, x, v0.t, vt
expansion: vmslt{u}.vx vt, va, x; vmandn.mm vt, v0, vt; vmandn.mm vd, vd,
v0; vmor.mm vd, vt, vd
The vt argument to the pseudoinstruction must name a temporary vector
register that is
not same as vd and which will be clobbered by the pseudoinstruction
There was a problem hiding this comment.
@topperc How can I define this different implementations and follow manual style? Shall I just do a * like in li?
There was a problem hiding this comment.
There are 3 different pseudoinstructions there that should each have their own row
vmsge{u}.vx vd, va, x
vmsge{u}.vx vd, va, x, v0.t // add a note in the last column that vd can't be v0
vmsge{u}.vx vd, va, x, v0.t, vt // you can write the 4 instruction expansion and a note that 2 instructions aren't needed if vd==v0.
There was a problem hiding this comment.
@topperc In the last commit I kept it exactly as described in the ISA Manual because I feel the 1:1 mapping is easier to understand, is this ok?
Signed-off-by: Afonso Oliveira <[email protected]>
| |vl4r.v v4,x0 | vl4re8.v v4, x0 | Equal to vl4re8.v | | ||
| |vl8r.v v8,x0 | vl8re8.v v8, x0 | Equal to vl8re8.v | | ||
|
|
||
| |vmsge{u}.vx vd, va, x| |
There was a problem hiding this comment.
This form isn't implemented in the assembler because it's impossible to know if x>minimum without doing more analysis than an assembler is capable of.
Closes #110 .
Preview: riscv-asm.pdf