v_mfma_f32_32x32x2bf16 GPU Native ISA AMD Matrix

Vector Packed Arithmetic

v_mfma_f32_32x32x2bf16

Multiply the 32x2 matrix in the first input by the 2x32 matrix in the second input and add the 32x32 matrix in the third input using fused multiply…

Encoding

Binary Layout (AMD CDNA 2)
BLGP
63:61
ACC
60:59
SRC2
58:50
SRC1
49:41
SRC0
40:32
110100111
31:23
1101000
22:16
ACC_CD
15
ABID
14:11
CBSZ
10:8
VDST
7:0
 

The VOP3P_MFMA layout from AMD's machine-readable ISA specification (AMD CDNA 2). Opcode 104 in OP rebuilds 0x00000000D3E80000, an identifier AMD lists for this encoding. Not the same in every generation: AMD CDNA 1 (field layout).

Format VOP3P
Width 64 bits
Opcode 104
Identifier 0x00000000D3E80000

Operands

  • VDST
    Written. N/A Data format: N/A (OPR_VGPR_OR_ACCVGPR, FMT_NUM_PK32_F32)
  • SRC0
    Read. N/A Data format: N/A (OPR_SRC_VGPR_OR_ACCVGPR, FMT_NUM_PK2_BF16)
  • SRC1
    Read. N/A Data format: N/A (OPR_SRC_VGPR_OR_ACCVGPR, FMT_NUM_PK2_BF16)
  • SRC2
    Read. N/A Data format: N/A (OPR_SRC_VGPR_OR_ACCVGPR_OR_CONST, FMT_NUM_PK32_F32)

In AMD's order, as its machine-readable ISA specification lists them for the VOP3P_MFMA encoding (AMD CDNA 2).

GFX Target Compatibility

Per-target GFX compatibility has not yet been verified for this instruction.

Related

More in Vector Packed Arithmetic

Reference

AMDGPU / GFX ISA

Description

Multiply the 32x2 matrix in the first input by the 2x32 matrix in the second input and add the 32x32 matrix in the third input using fused multiply add. Store the resulting matrix into vector registers.

Example

v_mfma_f32_32x32x2bf16 v[0:31], v0, v1, -2.0

A real instruction accepted by the LLVM assembler, taken verbatim from LLVM's own AMDGPU MC test suite (mai-gfx90a.s). Not from an AMD document, and not authored here.

Sources