The hand-written parts of the manual still described how the retired xenia-rs interpreter behaved: its snapshots, Rust casts and helpers. Each of those 490 statements is now either restated as what Canary's emitters and x64 backend actually do (at the pinned canary_experimental commit), or dropped where it only made sense for xenia-rs. Checking them turned up claims that were wrong, not just outdated: - VSCR[SAT] is never modelled in Canary (DID_SATURATE is a stub and mfvscr cannot see it); the pages said saturating ops set it stickily. - Canary does not implement lswi/lswx/stswi/stswx, dcbi, mtfsb0/mtfsb1, vmsum*, vmhaddshs, vupkhpx/vupklpx, and most SPRs; pages described them as working. - Traps evaluate TO in Canary; stvebx/stvehx/stvewx store one element, not 16 bytes; mtmsrd writes only EE; fres/frsqrte/vrsqrtefp precision claims and the stfs "rounds under RN / sets FPSCR" claim contradicted the spec. - Reservations are a 64 KiB block bitmap plus a value compare, not per-address tracking. Claims that neither Canary's source nor a public spec settles are marked unverified (NI at boot, vmaddcfp128 operand order, estimate bit-exactness). Generated regions are untouched; re-running the generator changes nothing. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
6.4 KiB
6.4 KiB
cmp — Compare
Category: Integer ALU · Form: X · Opcode:
0x7c000000
Assembler Mnemonics
| Mnemonic | XML entry | Flags | Description |
|---|---|---|---|
cmp |
cmp |
— | Compare |
Syntax
cmp [CRFD], [L], [RA], [RB]
Encoding
cmp — form X
- Opcode word:
0x7c000000 - Primary opcode (bits 0–5):
31 - Extended opcode:
0 - Synchronising: no
| Bits | Field | Meaning |
|---|---|---|
| 0–5 | OPCD |
primary opcode |
| 6–10 | RT/FRT/VRT |
destination |
| 11–15 | RA/FRA/VRA |
source A |
| 16–20 | RB/FRB/VRB |
source B |
| 21–30 | XO |
extended opcode (10 bits) |
| 31 | Rc |
record-form flag |
Operands
| Field | Role | Description |
|---|---|---|
L |
cmp: read | Operand-length bit for compare instructions (0 ⇒ 32-bit, 1 ⇒ 64-bit). |
RA |
cmp: read | Source GPR (r0–r31). |
RB |
cmp: read | Source GPR. |
CRFD |
cmp: write | CR destination field (crf, 0–7). |
Register Effects
cmp
- Reads (always):
L,RA,RB - Reads (conditional): none
- Writes (always):
CRFD - Writes (conditional): none
Status-Register Effects
No condition-register or status-register effects.
Operation (pseudocode)
if L = 0 then a,b <- EXTS((RA)[32:63]), EXTS((RB)[32:63])
else a,b <- (RA), (RB)
CR[BF] <- signed_compare(a, b) || XER[SO]
C Translation Example
/* No hand-written C yet. Translate the Canary emitter snapshot */
/* under Implementation References; its HIR maps directly: */
/* f.LoadGPR(n) / f.StoreGPR(n, v) -> r[n] / r[n] = v */
/* f.LoadFPR / StoreFPR, f.LoadVR / StoreVR -> f[n], v[n] */
/* f.Load(ea, T), f.Store(ea, v) -> raw read / write; emitters */
/* wrap them in f.ByteSwap for the big-endian guest value */
/* f.UpdateCR(n, v) -> CR field n from v's LOW 32 BITS vs 0 */
/* f.LoadCA / f.StoreCA -> xer.CA; f.StoreSAT -> vscr.SAT */
/* i.XO.RA, i.D.DS, ... -> the bit-fields listed under Operands */
/* The Register Effects and Status-Register Effects tables above */
/* enumerate every side effect a faithful translation must emit. */
Implementation References
cmp
- Canary XML:
tools/ppc-instructions.xml— search formnem="cmp" - Canary emitter:
src/xenia/cpu/ppc/ppc_emit_alu.cc:523 - Sylpheed opcode:
crates/sylpheed-ppc/src/opcode.rs:27 - Sylpheed decoder:
crates/sylpheed-ppc/src/decoder.rs:864
Canary emitter (frozen snapshot @ f21ebd49e9)
int InstrEmit_cmp(PPCHIRBuilder& f, const InstrData& i) {
// if L = 0 then
// a <- EXTS((RA)[32:63])
// b <- EXTS((RB)[32:63])
// else
// a <- (RA)
// b <- (RB)
// if a < b then
// c <- 0b100
// else if a > b then
// c <- 0b010
// else
// c <- 0b001
// CR[4×BF+32:4×BF+35] <- c || XER[SO]
uint32_t BF = i.X.RT >> 2;
uint32_t L = i.X.RT & 1;
Value* lhs;
Value* rhs;
if (L) {
lhs = f.LoadGPR(i.X.RA);
rhs = f.LoadGPR(i.X.RB);
} else {
lhs = f.Truncate(f.LoadGPR(i.X.RA), INT32_TYPE);
rhs = f.Truncate(f.LoadGPR(i.X.RB), INT32_TYPE);
}
f.UpdateCR(BF, lhs, rhs);
return 0;
}
Extended Pseudocode
if L = 0 then ; 32-bit compare
a <- EXTS((RA)[32:63]) ; sign-extend low word to 64
b <- EXTS((RB)[32:63])
else ; 64-bit compare
a <- (RA)
b <- (RB)
CR[BF] <- { LT: a <s b, GT: a >s b, EQ: a = b, SO: XER[SO] } ; signed
Special Cases & Edge Conditions
BFis a CR field (0–7), not a bit. ThecrfDoperand encodes which of the eight 4-bit CR fields is updated. Assemblers write it ascrNwhereN ∈ 0..7. The simplified mnemoniccmpw RA, RB≡cmp cr0, 0, RA, RBis universal in Xbox 360 code.Lbit selects width.L = 0(the usualcmpw/cmpd-is-rare path) performs a 32-bit signed compare ofRA[32:63]andRB[32:63], both sign-extended to 64 bits.L = 1(cmpd) performs a full 64-bit signed compare.- Signed. Use
cmpl/cmplifor unsigned comparisons. Confusing signed/unsigned is the most common compare-family bug in hand-written asm. - SO is always copied from
XER[SO]. This makes overflow observable across arithmetic/compare sequences: anaddo.followed bybeqcan branch on the record-form flag whilebsocan inspect the sticky overflow. cr0is the default for record-form ALU; by convention assemblers and generators reservecr0for the chain ofRc=1instructions and usecr1..cr7(orcmpto an explicit field) for standalone compares. Don't assumecmpwritescr0unless theBFoperand says so.- No register is written beyond the 4-bit CR field.
cmphas noRcorOEbit. - Canary is one direct compare.
Lselects either the full 64-bit operands or both truncated toINT32, thenf.UpdateCR(BF, lhs, rhs)setsLT/GT/EQfrom a single signed comparison. ⚠️ It never writes the field'sSObit — spec copiesXER[SO]into it — so Canary leavesSOholding whatever was there before.
Related Instructions
cmpi— signed compare against a 16-bit immediate.cmpl,cmpli— unsigned versions.cmpw,cmpd— simplified mnemonics selectingL.mcrxr— moveXER[SO..CA]into a CR field and clear them; used to reset sticky overflow.- Every
Rc=1ALU instruction (addx,subfx,andx, …) — these implicitly perform a signed-compare-to-zero intocr0; use explicitcmponly when comparing two non-zero values or using a non-zero CR field.