Instruction
TOP4BSUD
ACE outer product of a ZMM register of signed bytes with one of unsigned bytes, four per dword, accumulated into a tile of 32-bit integers.
- Extensions
- Forms
- 1 (1 EVEX)
- Data
- Intel XED v2026.08.23
TOP4BSUD is TOP4BSSD with unsigned bytes in its second source: for every row i and column j of the tile it adds the four products of the signed bytes of dword i of zmm2 with the unsigned bytes of dword j of zmm3 to the 32-bit integer at (i, j). TOP4BSSD’s page covers the outer product and how a matrix product is built from it; this page covers what the signedness changes. The order is that of AMX’s TDPBSUD and of VPDPBSUD in AVX-VNNI-INT8: signed first, unsigned second.
How far the sums reach
Each product lies between -32,640 (-128 times 255) and 32,385 (127 times 255), so one instruction moves an element by at most 130,560, twice as far as TOP4BSSD, whose four products stay between -65,024 and 65,536. The tile adds modulo 2^32, with no saturation. An accumulator that starts at zero therefore stays inside the signed 32-bit range for 16,448 instructions, a depth of 65,792 bytes along K, against 32,767 instructions for TOP4BSSD; a longer sum has to be split into blocks that are added in a wider type.
Encoding
EVEX opcode 5Eh in map 0F38 with W0 and the F3 prefix; F2, 66 and no
prefix give TOP4BSSD, TOP4BUSD and
TOP4BUUD, the same prefixes that tell TDPBSSD and its
siblings apart in VEX. The tile is in ModRM.reg, zmm2 in ModRM.rm and
zmm3 in EVEX.vvvv, so by the specification’s encoding top4bsud tmm0, zmm1, zmm2 is 62 F2 6E 48 5E C1. The exception class is
ACE-E4, as for every outer product.
In our builds
GNU as 2.44 and Clang 23.1.2 reject the mnemonic, neither GCC 14.2.0 nor
Clang 23.1.2 has ACE’s _tile_top4bsud intrinsic, and objdump 2.44
decodes those bytes as (bad). No processor at hand has ACE, so this
site’s recorded cases include none of TOP4BSUD.
Forms
| Instruction | Encoding | Requires |
|---|---|---|
TOP4BSUD tmm, zmm, zmm 64-bit mode only
TOP4BSUD_TMMu32_ZMMi8_ZMMu8 | EVEX.512.F3.0F38.W0 5E /r | ACE v1 |
Operation
What the forms do, in pseudocode that our tests run against results recorded on real processors. How to read.
on = xcr0 if cr4 >> 18 & 1 else 0 # XCR0, where CR4.OSXSAVE is set
if on >> 17 & 0xB != 0xB: # TILECFG, TILEDATA, SCALEDATA
fault(UD)
if on & 0xE6 != 0xE6: # AVX-512's state, AVX's and
fault(UD) # SSE's
if palette != 2: # ACE's palette alone
fault(UD)
if cr0 >> 3 & 1 or xfd >> 18 & 1: # CR0.TS, or the OS keeps the
fault(NM) # tiles from this thread
for i in range(16): # row i: dword i of zmm2
for j in range(16): # column j: dword j of zmm3
s = tmm1.row[i].d[j]
for k in range(4):
a = zmm2.sb[4 * i + k]
s += a * zmm3.b[4 * j + k]
tmm1.row[i].d[j] = s # modulo 2^32
Not run on a processor: none of these processors has ACE.
Operands
| Operand | Kind | Action |
|---|---|---|
TOP4BSUD tmm, zmm, zmm 64-bit mode only | ||
tmm | explicit | read, write |
zmm | explicit | read |
zmm | explicit | read |
Flags
None: XED records no flag that the forms test or change.
Exceptions
The exception class is XED's name for the set of exception conditions that a form follows. How to read.
- All forms
ACE-E4
Attributes
What else XED records about the forms. How to read.
- All forms
- may abort a TSX transaction
CPUID and processors
Each form belongs to an XED ISA set. A form can be used when the processor reports every CPUID bit of one of its ISA set's alternatives.
| ISA set | CPUID | Processors |
|---|---|---|
ACE_1 | ACE v1 | None in XED’s chip model, which lists these instructions only under its placeholder for future processors. |
More ACE instructions
Sources
- Intel XED v2026.08.23 (commit
0bcb6237345c): forms, encodings, operands, flags, exception classes, attributes, ISA sets, CPUID bits and chip model. - x86 Ecosystem Advisory Group: AI Compute Extensions (ACE) Specification, version 1.16.2 (July 2026)
The tables are derived from Intel XED, Copyright Intel Corporation, licensed under the Apache License 2.0; x86-64.net converted and reformatted them. The text is our own.