Add encodings for i8 and i16 copy, spill, fill, ireduce.i8.i16 (#534)

* Add encodings for i8 and i16 copy, spill, fill, ireduce.i8.i16

Also adds legalization for srem, irsub_imm, {u,s}extend.i16.i8

Fixes #477 cc #466

* Legalize popcnt, clz and ctz for i8 and i16

* Fix bug in call_memset
This commit is contained in:
bjorn3
2018-10-03 23:43:59 +02:00
committed by Dan Gohman
parent ddf8fd23b5
commit b2a28d69e6
8 changed files with 139 additions and 12 deletions

View File

@@ -0,0 +1,8 @@
test compile
target x86_64
function u0:0(i16) -> i8 fast {
ebb0(v0: i16):
v1 = ireduce.i8 v0
return v1
}

View File

@@ -0,0 +1,13 @@
test compile
target x86_64
function u0:0(i8) -> i8 fast {
ebb0(v0: i8):
v1 = iconst.i8 0
v2 = isub v1, v0
; check: v4 = uextend.i32 v0
; nextln: v6 = iconst.i32 0
; nextln = isub v6, v4
; nextln = ireduce.i8 v5
return v2
}

View File

@@ -0,0 +1,25 @@
test compile
target x86_64
; regex: V=v\d+
function u0:0(i8) -> i8, i8 fast {
ebb0(v0: i8):
v1 = clz v0
; check: v3 = uextend.i32 v0
; nextln: v6 = iconst.i32 -1
; nextln: v7 = iconst.i32 31
; nextln: v8, v9 = x86_bsr v3
; nextln: v10 = selectif.i32 eq v9, v6, v8
; nextln: v4 = isub v7, v10
; nextln: v5 = iadd_imm v4, -24
; nextln: v1 = ireduce.i8 v5
v2 = ctz v0
; nextln: v11 = uextend.i32 v0
; nextln: v12 = bor_imm v11, 256
; nextln: v14 = iconst.i32 32
; nextln: v15, v16 = x86_bsf v12
; nextln: v13 = selectif.i32 eq v16, v14, v15
; nextln: v2 = ireduce.i8 v13
return v1, v2
}

View File

@@ -0,0 +1,9 @@
test compile
target x86_64
function u0:0(i8) -> i8 fast {
ebb0(v0: i8):
v1 = popcnt v0
; check-not: sextend.i32 v0
return v1
}

View File

@@ -0,0 +1,14 @@
test compile
target x86_64
function u0:0(i8) -> i16 fast {
ebb0(v0: i8):
v1 = uextend.i16 v0
return v1
}
function u0:1(i8) -> i16 fast {
ebb0(v0: i8):
v1 = sextend.i16 v0
return v1
}