fix(asm): scale the FLDPQ and FSTPQ pair offsets by sixteen
The pair encoder derived the imm7 divisor from the width suffix alone, so the 128-bit FP pairs divided their offsets by eight and encoded twice the distance. The Q spellings scale by sixteen like every other 128-bit access; the differential kernel carries them now. Assisted-by: GLM 5.3 Flash
This commit is contained in:
1 parent
daf7fad5b9
commit
2bd52eb7ad
5 files changed
+14
-5
No files matched your search
@@ -625,10 +625,12 @@ func TestArm64ADR(t *testing.T) {
|
||||
}
|
||||
}
|
||||
|
||||
// TestArm64PairLoadStore pins LDP/STP/LDPW/FLDPD/FSTPD.
|
||||
// TestArm64PairLoadStore pins LDP/STP/LDPW/FLDPD/FSTPD and the 128-bit FP
|
||||
// pairs, whose offsets scale by sixteen.
|
||||
func TestArm64PairLoadStore(t *testing.T) {
|
||||
got := arm64Words(t, "\tSTP (R2, R3), 8(R5)\n\tLDP -8(R5), (R2, R3)\n\tLDPW 4(R0), (R1, R2)\n\tSTPW (R1, R2), 4(R0)\n"+
|
||||
"\tFLDPD 8(R0), (F1, F2)\n\tFSTPD (F3, F4), -8(R5)\n")
|
||||
"\tFLDPD 8(R0), (F1, F2)\n\tFSTPD (F3, F4), -8(R5)\n"+
|
||||
"\tFLDPQ 16(R0), (F1, F2)\n\tFSTPQ (F1, F2), 16(R0)\n")
|
||||
want := []uint32{
|
||||
0xa9008ca2, // STP (R2, R3), 8(R5)
|
||||
0xa97f8ca2, // LDP -8(R5), (R2, R3)
|
||||
@@ -636,6 +638,8 @@ func TestArm64PairLoadStore(t *testing.T) {
|
||||
0x29008801, // STPW (R1, R2), 4(R0)
|
||||
0x6d408801, // FLDPD 8(R0), (F1, F2)
|
||||
0x6d3f90a3, // FSTPD (F3, F4), -8(R5)
|
||||
0xad408801, // FLDPQ 16(R0), (F1, F2)
|
||||
0xad008801, // FSTPQ (F1, F2), 16(R0)
|
||||
0xd65f03c0,
|
||||
}
|
||||
if len(got) != len(want) {
|
||||
|
||||
Reference in new issue
Block a user