Files
sim f3c512f2ab docs(ppc-manual): check every xenia-rs claim against Canary's source
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Checking them turned up claims that were wrong, not just outdated: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>
2026-09-16 21:52:38 +02:00

11 KiB
Raw Permalink Blame History

lfs — Load Floating-Point Single

Category: Memory · Form: D · Opcode: 0xc0000000

Assembler Mnemonics

Mnemonic XML entry Flags Description
lfs lfs Load Floating-Point Single
lfsu lfsu Load Floating-Point Single with Update
lfsux lfsux Load Floating-Point Single with Update Indexed
lfsx lfsx Load Floating-Point Single Indexed

Syntax

lfs [FD], [d]([RA0])
lfsu [FD], [d]([RA])
lfsux [FD], [RA], [RB]
lfsx [FD], [RA0], [RB]

Encoding

lfs — form D

  • Opcode word: 0xc0000000
  • Primary opcode (bits 05): 48
  • Extended opcode:
  • Synchronising: no
Bits Field Meaning
05 OPCD primary opcode
610 RT destination GPR (or RS when storing)
1115 RA source GPR (0 ⇒ literal 0 for RA0 forms)
1631 D/SI/UI 16-bit signed or unsigned immediate

lfsu — form D

  • Opcode word: 0xc4000000
  • Primary opcode (bits 05): 49
  • Extended opcode:
  • Synchronising: no
Bits Field Meaning
05 OPCD primary opcode
610 RT destination GPR (or RS when storing)
1115 RA source GPR (0 ⇒ literal 0 for RA0 forms)
1631 D/SI/UI 16-bit signed or unsigned immediate

lfsux — form X

  • Opcode word: 0x7c00046e
  • Primary opcode (bits 05): 31
  • Extended opcode: 567
  • Synchronising: no
Bits Field Meaning
05 OPCD primary opcode
610 RT/FRT/VRT destination
1115 RA/FRA/VRA source A
1620 RB/FRB/VRB source B
2130 XO extended opcode (10 bits)
31 Rc record-form flag

lfsx — form X

  • Opcode word: 0x7c00042e
  • Primary opcode (bits 05): 31
  • Extended opcode: 535
  • Synchronising: no
Bits Field Meaning
05 OPCD primary opcode
610 RT/FRT/VRT destination
1115 RA/FRA/VRA source A
1620 RB/FRB/VRB source B
2130 XO extended opcode (10 bits)
31 Rc record-form flag

Operands

Field Role Description
RA0 lfs: read; lfsx: read Source GPR; when the encoded register number is 0 the operand is the literal 64-bit zero, not r0.
d lfs: read; lfsu: read 16-bit signed displacement (d) added to the base address register.
FD lfs: write; lfsu: write; lfsux: write; lfsx: write Destination floating-point register.
RA lfsu: read; lfsu: write; lfsux: read; lfsux: write Source GPR (r0r31).
RB lfsux: read; lfsx: read Source GPR.

Register Effects

lfs

  • Reads (always): RA0, d
  • Reads (conditional): none
  • Writes (always): FD
  • Writes (conditional): none

lfsu

  • Reads (always): RA, d
  • Reads (conditional): none
  • Writes (always): FD, RA
  • Writes (conditional): none

lfsux

  • Reads (always): RA, RB
  • Reads (conditional): none
  • Writes (always): FD, RA
  • Writes (conditional): none

lfsx

  • Reads (always): RA0, RB
  • Reads (conditional): none
  • Writes (always): FD
  • Writes (conditional): none

Status-Register Effects

No condition-register or status-register effects.

Operation (pseudocode)

EA <- (RA|0) + EXTS(d)
FRT <- DoubleFromSingle(MEM(EA, 4))

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

lfs

Canary emitter (frozen snapshot @ f21ebd49e9)
int InstrEmit_lfs(PPCHIRBuilder& f, const InstrData& i) {
  // if RA = 0 then
  //   b <- 0
  // else
  //   b <- (RA)
  // EA <- b + EXTS(D)
  // FRT <- DOUBLE(MEM(EA, 4))
  Value* ea = CalculateEA_0_i(f, i.D.RA, XEEXTS16(i.D.DS));
  Value* rt = f.Convert(
      f.Cast(f.ByteSwap(f.Load(ea, INT32_TYPE)), FLOAT32_TYPE), FLOAT64_TYPE);
  f.StoreFPR(i.D.RT, rt);
  return 0;
}

lfsu

Canary emitter (frozen snapshot @ f21ebd49e9)
int InstrEmit_lfsu(PPCHIRBuilder& f, const InstrData& i) {
  // EA <- (RA) + EXTS(D)
  // FRT <- DOUBLE(MEM(EA, 4))
  // RA <- EA
  Value* ea = CalculateEA_i(f, i.D.RA, XEEXTS16(i.D.DS));
  Value* rt = f.Convert(
      f.Cast(f.ByteSwap(f.Load(ea, INT32_TYPE)), FLOAT32_TYPE), FLOAT64_TYPE);
  f.StoreFPR(i.D.RT, rt);
  StoreEA(f, i.D.RA, ea);
  return 0;
}

lfsux

Canary emitter (frozen snapshot @ f21ebd49e9)
int InstrEmit_lfsux(PPCHIRBuilder& f, const InstrData& i) {
  // EA <- (RA) + (RB)
  // FRT <- DOUBLE(MEM(EA, 4))
  // RA <- EA
  Value* ea = CalculateEA(f, i.X.RA, i.X.RB);
  Value* rt = f.Convert(
      f.Cast(f.ByteSwap(f.Load(ea, INT32_TYPE)), FLOAT32_TYPE), FLOAT64_TYPE);
  f.StoreFPR(i.X.RT, rt);
  StoreEA(f, i.X.RA, ea);
  return 0;
}

lfsx

Canary emitter (frozen snapshot @ f21ebd49e9)
int InstrEmit_lfsx(PPCHIRBuilder& f, const InstrData& i) {
  // if RA = 0 then
  //   b <- 0
  // else
  //   b <- (RA)
  // EA <- b + (RB)
  // FRT <- DOUBLE(MEM(EA, 4))
  Value* ea = CalculateEA_0(f, i.X.RA, i.X.RB);
  Value* rt = f.Convert(
      f.Cast(f.ByteSwap(f.Load(ea, INT32_TYPE)), FLOAT32_TYPE), FLOAT64_TYPE);
  f.StoreFPR(i.X.RT, rt);
  return 0;
}

Special Cases & Edge Conditions

  • Single → double in-register. Reads 4 bytes as IEEE binary32, then exactly converts to binary64 (every binary32 has a representation in binary64). The result occupies all 64 bits of the FPR; subsequent FP arithmetic operates in double regardless of the value's origin.
  • No FPSCR side effects. The single→double widening is exact, so lfs cannot raise inexact, overflow, underflow, or invalid. A signalling NaN passes through unchanged into the FPR — it will signal at the next FP arithmetic instruction.
  • Subnormals. A binary32 subnormal expands to a binary64 normal — lfs quietly normalises. There is no "FPSCR[NI] non-IEEE mode" subnormal-to-zero behaviour applied at this stage on Xenon (NI affects arithmetic, not loads).
  • RA0 semantics. In lfs / lfsx, RA = 0 selects literal zero. Update forms lfsu / lfsux are invalid with RA = 0.
  • Alignment. Xenon tolerates unaligned 4-byte loads; PowerISA permits implementations to raise alignment exceptions for FP loads on cache-inhibited storage.
  • Big-endian read. Bytes EA..EA+3 form the binary32 pattern, sign bit at EA[7]. Canary byte-swaps the loaded word, reinterprets it as binary32 and widens it with vcvtss2sd, carrying a NaN's quiet/signalling bit across unchanged.
  • MSR[FP] required. Disabled FP unit raises Floating-Point Unavailable.
  • Pair with stfs. Store-single performs the inverse double→single rounding (which can raise FPSCR exceptions because that direction may be inexact).
  • lfd — double-precision load (no format conversion).
  • stfs, stfsu, stfsx, stfsux — corresponding stores; these can round.
  • stfiwx — store-FP-as-integer-word.
  • lwz — integer word load (same width, GPR target).

IBM Reference