Beaver.MLIR.Dialect.ROCDL (beaver v0.4.7)

Ballot provides a bit mask containing the 1-bit predicate value from each lane. The nth bit of the result contains the 1 bit contributed by the nth warp lane.

barrier(ssa)

rocdl.barrier

cvt_f32_bf8(ssa)

rocdl.cvt.f32.bf8 - Convert bf8 to f32

Attributes

byteSel - Single, I32Attr, 32-bit signless integer attribute

Operands

srcA - Single, I32, 32-bit signless integer

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

Description

Convert 8-bit bf8 value from the byteSelth bit of srcA to fp32.

cvt_f32_fp8(ssa)

rocdl.cvt.f32.fp8 - Convert fp8 to f32

Attributes

byteSel - Single, I32Attr, 32-bit signless integer attribute

Operands

srcA - Single, I32, 32-bit signless integer

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

Description

Convert 8-bit fp8 value from the byteSelth bit of srcA to fp32.

cvt_pk_bf8_f32(ssa)

rocdl.cvt.pk.bf8.f32 - Convert two f32's to bf8

Attributes

wordSel - Single, I1Attr, 1-bit signless integer attribute

Operands

srcA - Single, F32, 32-bit float
srcB - Single, F32, 32-bit float
old - Single, I32, 32-bit signless integer

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

Description

Convert srcA and srcB to bf8 and store into the low/high word of old, preserving the other word.

cvt_pk_f32_bf8(ssa)

rocdl.cvt.pk.f32.bf8 - Convert packed bf8 to packed f32

Attributes

wordSel - Single, I1Attr, 1-bit signless integer attribute

Operands

src - Single, I32, 32-bit signless integer

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

Description

Convert src based on $wordSel to packed fp32,

cvt_pk_f32_fp8(ssa)

rocdl.cvt.pk.f32.fp8 - Convert packed fp8 to packed f32

Attributes

wordSel - Single, I1Attr, 1-bit signless integer attribute

Operands

src - Single, I32, 32-bit signless integer

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

Description

Convert src based on $wordSel to packed fp32.

cvt_pk_fp8_f32(ssa)

rocdl.cvt.pk.fp8.f32 - Convert two f32's to fp8

Attributes

wordSel - Single, I1Attr, 1-bit signless integer attribute

Operands

srcA - Single, F32, 32-bit float
srcB - Single, F32, 32-bit float
old - Single, I32, 32-bit signless integer

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

Description

Convert srcA and srcB to fp8 and store into the low/high word of old, preserving the other word.

cvt_pkrtz(ssa)

rocdl.cvt.pkrtz - Convert two f32 input into a vector<2xf16>

Operands

srcA - Single, F32, 32-bit float
srcB - Single, F32, 32-bit float

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

Description

Convert two f32 values into a packed vector<2xf16>.

cvt_scale_pk8_bf16_bf8(ssa)

rocdl.cvt.scale.pk8.bf16.bf8 - Scales 8 bf8 and converts them to 8 bf16.

Attributes

scaleSel - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, ROCDL_V2I32Type, fixed-length vector of 32-bit signless integer values of length 2
scale - Single, I32, 32-bit signless integer

Results

res - Single, ROCDL_V8BF16Type, fixed-length vector of bfloat16 type values of length 8

Description

Available on gfx1250+.

cvt_scale_pk8_bf16_fp4(ssa)

rocdl.cvt.scale.pk8.bf16.fp4 - Scales 8 fp4 and converts them to 8 bf16.

Attributes

scaleSel - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, I32, 32-bit signless integer
scale - Single, I32, 32-bit signless integer

Results

res - Single, ROCDL_V8BF16Type, fixed-length vector of bfloat16 type values of length 8

Description

Available on gfx1250+.

cvt_scale_pk8_bf16_fp8(ssa)

rocdl.cvt.scale.pk8.bf16.fp8 - Scales 8 fp8 and converts them to 8 bf16.

Attributes

scaleSel - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, ROCDL_V2I32Type, fixed-length vector of 32-bit signless integer values of length 2
scale - Single, I32, 32-bit signless integer

Results

res - Single, ROCDL_V8BF16Type, fixed-length vector of bfloat16 type values of length 8

Description

Available on gfx1250+.

cvt_scale_pk8_f16_bf8(ssa)

rocdl.cvt.scale.pk8.f16.bf8 - Scales 8 bf8 and converts them to 8 f16.

Attributes

scaleSel - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, ROCDL_V2I32Type, fixed-length vector of 32-bit signless integer values of length 2
scale - Single, I32, 32-bit signless integer

Results

res - Single, ROCDL_V8F16Type, fixed-length vector of 16-bit float values of length 8

Description

Available on gfx1250+.

cvt_scale_pk8_f16_fp4(ssa)

rocdl.cvt.scale.pk8.f16.fp4 - Scales 8 fp4 and converts them to 8 f16.

Attributes

scaleSel - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, I32, 32-bit signless integer
scale - Single, I32, 32-bit signless integer

Results

res - Single, ROCDL_V8F16Type, fixed-length vector of 16-bit float values of length 8

Description

Available on gfx1250+.

cvt_scale_pk8_f16_fp8(ssa)

rocdl.cvt.scale.pk8.f16.fp8 - Scales 8 fp8 and converts them to 8 f16.

Attributes

scaleSel - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, ROCDL_V2I32Type, fixed-length vector of 32-bit signless integer values of length 2
scale - Single, I32, 32-bit signless integer

Results

res - Single, ROCDL_V8F16Type, fixed-length vector of 16-bit float values of length 8

Description

Available on gfx1250+.

cvt_scale_pk8_f32_bf8(ssa)

rocdl.cvt.scale.pk8.f32.bf8 - Scales 8 bf8 and converts them to 8 f32.

Attributes

scaleSel - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, ROCDL_V2I32Type, fixed-length vector of 32-bit signless integer values of length 2
scale - Single, I32, 32-bit signless integer

Results

res - Single, ROCDL_V8F32Type, fixed-length vector of 32-bit float values of length 8

Description

Available on gfx1250+.

cvt_scale_pk8_f32_fp4(ssa)

rocdl.cvt.scale.pk8.f32.fp4 - Scales 8 fp4 and converts them to 8 f32.

Attributes

scaleSel - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, I32, 32-bit signless integer
scale - Single, I32, 32-bit signless integer

Results

res - Single, ROCDL_V8F32Type, fixed-length vector of 32-bit float values of length 8

Description

Available on gfx1250+.

cvt_scale_pk8_f32_fp8(ssa)

rocdl.cvt.scale.pk8.f32.fp8 - Scales 8 fp8 and converts them to 8 f32.

Attributes

scaleSel - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, ROCDL_V2I32Type, fixed-length vector of 32-bit signless integer values of length 2
scale - Single, I32, 32-bit signless integer

Results

res - Single, ROCDL_V8F32Type, fixed-length vector of 32-bit float values of length 8

Description

Available on gfx1250+.

cvt_scale_pk16_bf16_bf6(ssa)

rocdl.cvt.scale.pk16.bf16.bf6 - Scales 16 bf6 and converts them to 16 bf16.

Attributes

scaleSel - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, ROCDL_V3I32Type, fixed-length vector of 32-bit signless integer values of length 3
scale - Single, I32, 32-bit signless integer

Results

res - Single, ROCDL_V16BF16Type, fixed-length vector of bfloat16 type values of length 16

Description

Available on gfx1250+.

cvt_scale_pk16_bf16_fp6(ssa)

rocdl.cvt.scale.pk16.bf16.fp6 - Scales 16 fp6 and converts them to 16 bf16.

Attributes

scaleSel - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, ROCDL_V3I32Type, fixed-length vector of 32-bit signless integer values of length 3
scale - Single, I32, 32-bit signless integer

Results

res - Single, ROCDL_V16BF16Type, fixed-length vector of bfloat16 type values of length 16

Description

Available on gfx1250+.

cvt_scale_pk16_f16_bf6(ssa)

rocdl.cvt.scale.pk16.f16.bf6 - Scales 16 bf6 and converts them to 16 f16.

Attributes

scaleSel - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, ROCDL_V3I32Type, fixed-length vector of 32-bit signless integer values of length 3
scale - Single, I32, 32-bit signless integer

Results

res - Single, ROCDL_V16F16Type, fixed-length vector of 16-bit float values of length 16

Description

Available on gfx1250+.

cvt_scale_pk16_f16_fp6(ssa)

rocdl.cvt.scale.pk16.f16.fp6 - Scales 16 fp6 and converts them to 16 f16.

Attributes

scaleSel - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, ROCDL_V3I32Type, fixed-length vector of 32-bit signless integer values of length 3
scale - Single, I32, 32-bit signless integer

Results

res - Single, ROCDL_V16F16Type, fixed-length vector of 16-bit float values of length 16

Description

Available on gfx1250+.

cvt_scale_pk16_f32_bf6(ssa)

rocdl.cvt.scale.pk16.f32.bf6 - Scales 16 bf6 and converts them to 16 f32.

Attributes

scaleSel - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, ROCDL_V3I32Type, fixed-length vector of 32-bit signless integer values of length 3
scale - Single, I32, 32-bit signless integer

Results

res - Single, ROCDL_V16F32Type, fixed-length vector of 32-bit float values of length 16

Description

Available on gfx1250+.

cvt_scale_pk16_f32_fp6(ssa)

rocdl.cvt.scale.pk16.f32.fp6 - Scales 16 fp6 and converts them to 16 f32.

Attributes

scaleSel - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, ROCDL_V3I32Type, fixed-length vector of 32-bit signless integer values of length 3
scale - Single, I32, 32-bit signless integer

Results

res - Single, ROCDL_V16F32Type, fixed-length vector of 32-bit float values of length 16

Description

Available on gfx1250+.

cvt_scalef32_2xpk16_bf6_f32(ssa)

rocdl.cvt.scalef32.2xpk16.bf6.f32 - Scale and convert two vector<16xf32> to 32 packed bf6

Operands

src0 - Single, ROCDL_V16F32Type, fixed-length vector of 32-bit float values of length 16
src1 - Single, ROCDL_V16F32Type, fixed-length vector of 32-bit float values of length 16
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6

Description

Convert 32 single-precision float values, packed into two length-16 vectors that will be logically concanenated, to packed bf6, dividing by the exponent part of scale before doing so.

cvt_scalef32_2xpk16_fp6_f32(ssa)

rocdl.cvt.scalef32.2xpk16.fp6.f32 - Scale and convert two vector<16xf32> to 32 packed fp6

Operands

src0 - Single, ROCDL_V16F32Type, fixed-length vector of 32-bit float values of length 16
src1 - Single, ROCDL_V16F32Type, fixed-length vector of 32-bit float values of length 16
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6

Description

Convert 32 single-precision float values, packed into two length-16 vectors that will be logically concanenated, to packed fp6, dividing by the exponent part of scale before doing so.

cvt_scalef32_f16_bf8(ssa)

rocdl.cvt.scalef32.f16.bf8 - Scaled convert bf8 from packed vector to f16, updating tied result

Attributes

srcSelIndex - Single, I32Attr, 32-bit signless integer attribute
dstLoHiSel - Single, I1Attr, 1-bit signless integer attribute

Operands

oldVdst - Single, ROCDL_V2F16Type, fixed-length vector of 16-bit float values of length 2
src - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V2F16Type, fixed-length vector of 16-bit float values of length 2

Description

Convert a bf8 byte from src, selected by srcSelIndex, to f16 while multiplying it by the expontent of scale, and place it into the dstLoHiSelth bit of oldVdst preserving the other element of that vector in the return value.

The bytes are stored as an i32 and not a <4 x i8>.

cvt_scalef32_f16_fp8(ssa)

rocdl.cvt.scalef32.f16.fp8 - Scaled convert fp8 from packed vector to f16, updating tied result

Attributes

srcSelIndex - Single, I32Attr, 32-bit signless integer attribute
dstLoHiSel - Single, I1Attr, 1-bit signless integer attribute

Operands

oldVdst - Single, ROCDL_V2F16Type, fixed-length vector of 16-bit float values of length 2
src - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V2F16Type, fixed-length vector of 16-bit float values of length 2

Description

Convert a fp8 byte from src, selected by srcSelIndex, to f16 while multiplying it by the expontent of scale, and place it into the dstLoHiSelth bit of oldVdst preserving the other element of that vector in the return value.

The bytes are stored as an i32 and not a <4 x i8>.

cvt_scalef32_f32_bf8(ssa)

rocdl.cvt.scalef32.f32.bf8 - Scaled convert bf8 from packed vector to f32

Attributes

srcSelIndex - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, F32, 32-bit float

Description

Convert a bf8 byte from src, selected by srcSelIndex, to f32, multiplying it by the exponent of scale.

The bytes are stored in an i32, not a <4 x i8>.

cvt_scalef32_f32_fp8(ssa)

rocdl.cvt.scalef32.f32.fp8 - Scaled convert fp8 from packed vector to f32

Attributes

srcSelIndex - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, F32, 32-bit float

Description

Convert a fp8 byte from src, selected by srcSelIndex, to f32, multiplying it by the exponent of scale.

The bytes are stored in an i32, not a <4 x i8>.

cvt_scalef32_pk32_bf6_bf16(ssa)

rocdl.cvt.scalef32.pk32.bf6.bf16 - Scale and convert packed bf16 to packed bf6

Operands

src - Single, ROCDL_V32BF16Type, fixed-length vector of bfloat16 type values of length 32
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6

Description

Convert 32 packed bf16 values to packed bf6, dividing by the exponent part of scale before doing so.

cvt_scalef32_pk32_bf6_f16(ssa)

rocdl.cvt.scalef32.pk32.bf6.f16 - Scale and convert packed f16 to packed bf6

Operands

src - Single, ROCDL_V32F16Type, fixed-length vector of 16-bit float values of length 32
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6

Description

Convert 32 packed f16 values to packed bf6, dividing by the exponent part of scale before doing so.

cvt_scalef32_pk32_bf16_bf6(ssa)

rocdl.cvt.scalef32.pk32.bf16.bf6 - Scale and convert packed bf6 to packed bf16

Operands

src - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V32BF16Type, fixed-length vector of bfloat16 type values of length 32

Description

Convert 32 packed bf6 values to packed bf16, multiplying by the exponent part of scale before doing so.

cvt_scalef32_pk32_bf16_fp6(ssa)

rocdl.cvt.scalef32.pk32.bf16.fp6 - Scale and convert packed fp6 to packed bf16

Operands

src - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V32BF16Type, fixed-length vector of bfloat16 type values of length 32

Description

Convert 32 packed fp6 values to packed bf16, multiplying by the exponent part of scale before doing so.

cvt_scalef32_pk32_f16_bf6(ssa)

rocdl.cvt.scalef32.pk32.f16.bf6 - Scale and convert packed bf6 to packed f16

Operands

src - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V32F16Type, fixed-length vector of 16-bit float values of length 32

Description

Convert 32 packed bf6 values to packed f16, multiplying by the exponent part of scale before doing so.

cvt_scalef32_pk32_f16_fp6(ssa)

rocdl.cvt.scalef32.pk32.f16.fp6 - Scale and convert packed fp6 to packed f16

Operands

src - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V32F16Type, fixed-length vector of 16-bit float values of length 32

Description

Convert 32 packed fp6 values to packed f16, multiplying by the exponent part of scale before doing so.

cvt_scalef32_pk32_f32_bf6(ssa)

rocdl.cvt.scalef32.pk32.f32.bf6 - Scale and convert packed bf6 to packed f32

Operands

src - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V32F32Type, fixed-length vector of 32-bit float values of length 32

Description

Convert 32 packed bf6 values to packed f32, multiplying by the exponent part of scale before doing so.

cvt_scalef32_pk32_f32_fp6(ssa)

rocdl.cvt.scalef32.pk32.f32.fp6 - Scale and convert packed fp6 to packed f32

Operands

src - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V32F32Type, fixed-length vector of 32-bit float values of length 32

Description

Convert 32 packed fp6 values to packed f32, multiplying by the exponent part of scale before doing so.

cvt_scalef32_pk32_fp6_bf16(ssa)

rocdl.cvt.scalef32.pk32.fp6.bf16 - Scale and convert packed bf16 to packed fp6

Operands

src - Single, ROCDL_V32BF16Type, fixed-length vector of bfloat16 type values of length 32
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6

Description

Convert 32 packed bf16 values to packed fp6, dividing by the exponent part of scale before doing so.

cvt_scalef32_pk32_fp6_f16(ssa)

rocdl.cvt.scalef32.pk32.fp6.f16 - Scale and convert packed f16 to packed fp6

Operands

src - Single, ROCDL_V32F16Type, fixed-length vector of 16-bit float values of length 32
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6

Description

Convert 32 packed f16 values to packed fp6, dividing by the exponent part of scale before doing so.

cvt_scalef32_pk_bf8_bf16(ssa)

rocdl.cvt.scalef32.pk.bf8.bf16 - Scaled convert two bf16to two bf8, updating packed vector

Attributes

dstLoHiSel - Single, I1Attr, 1-bit signless integer attribute

Operands

oldVdst - Single, ROCDL_V2I16Type, fixed-length vector of 16-bit signless integer values of length 2
src0 - Single, ROCDL_V2BF16Type, fixed-length vector of bfloat16 type values of length 2
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V2I16Type, fixed-length vector of 16-bit signless integer values of length 2

Description

Convert two bf16 values in src0 to two bf8 bytes, dividing by the exponent in scale. The bytes are packed into a 16-bit value which is inserted into oldVdst at the dstLoHiSel position, with the entire updated vector being returned.

cvt_scalef32_pk_bf8_f16(ssa)

rocdl.cvt.scalef32.pk.bf8.f16 - Scaled convert two f16to two bf8, updating packed vector

Attributes

dstLoHiSel - Single, I1Attr, 1-bit signless integer attribute

Operands

oldVdst - Single, ROCDL_V2I16Type, fixed-length vector of 16-bit signless integer values of length 2
src0 - Single, ROCDL_V2F16Type, fixed-length vector of 16-bit float values of length 2
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V2I16Type, fixed-length vector of 16-bit signless integer values of length 2

Description

Convert two f16 values in src0 to two bf8 bytes, dividing by the exponent in scale. The bytes are packed into a 16-bit value which is inserted into oldVdst at the dstLoHiSel position, with the entire updated vector being returned.

cvt_scalef32_pk_bf8_f32(ssa)

rocdl.cvt.scalef32.pk.bf8.f32 - Scaled convert two f32 to two bf8, updating packed vector

Attributes

dstLoHiSel - Single, I1Attr, 1-bit signless integer attribute

Operands

oldVdst - Single, ROCDL_V2I16Type, fixed-length vector of 16-bit signless integer values of length 2
src0 - Single, F32, 32-bit float
src1 - Single, F32, 32-bit float
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V2I16Type, fixed-length vector of 16-bit signless integer values of length 2

Description

Convert two f32 values in src0 and src1 to two bf8 bytes, dividing by the exponent in scale. The bytes are packed into a 16-bit value which is inserted into oldVdst at the dstLoHiSel position, with the entire updated vector being returned.

cvt_scalef32_pk_bf16_bf8(ssa)

rocdl.cvt.scalef32.pk.bf16.bf8 - Scaled convert two bf8to two bf16

Attributes

srcLoHiSel - Single, I1Attr, 1-bit signless integer attribute

Operands

src - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V2BF16Type, fixed-length vector of bfloat16 type values of length 2

Description

Convert two packed bf8 values in src0 to two bf16 values, multiplying by the exponent in scale. The two values to be converted are selected from the low or high half of src (a packed vector represented as an i32) on the basis of srcLoHiSel.

cvt_scalef32_pk_bf16_fp4(ssa)

rocdl.cvt.scalef32.pk.bf16.fp4 - Scale and convert two packed fp4 to packed bf16

Attributes

srcSelIndex - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V2BF16Type, fixed-length vector of bfloat16 type values of length 2

Description

Convert two packed fp4 (f4E2M1) values stored as one byte of a 32-bit integer to packed bf16, multiplying by the exponent part of scale before doing so.

The byte to convert is chosen by srcSelIndex.

cvt_scalef32_pk_bf16_fp8(ssa)

rocdl.cvt.scalef32.pk.bf16.fp8 - Scaled convert two fp8to two bf16

Attributes

srcLoHiSel - Single, I1Attr, 1-bit signless integer attribute

Operands

src - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V2BF16Type, fixed-length vector of bfloat16 type values of length 2

Description

Convert two packed fp8 values in src0 to two bf16 values, multiplying by the exponent in scale. The two values to be converted are selected from the low or high half of src (a packed vector represented as an i32) on the basis of srcLoHiSel.

cvt_scalef32_pk_f16_bf8(ssa)

rocdl.cvt.scalef32.pk.f16.bf8 - Scaled convert two bf8to two f16

Attributes

srcLoHiSel - Single, I1Attr, 1-bit signless integer attribute

Operands

src - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V2F16Type, fixed-length vector of 16-bit float values of length 2

Description

Convert two packed bf8 values in src0 to two f16 values, multiplying by the exponent in scale. The two values to be converted are selected from the low or high half of src (a packed vector represented as an i32) on the basis of srcLoHiSel.

cvt_scalef32_pk_f16_fp4(ssa)

rocdl.cvt.scalef32.pk.f16.fp4 - Scale and convert two packed fp4 to packed f16

Attributes

srcSelIndex - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V2F16Type, fixed-length vector of 16-bit float values of length 2

Description

Convert two packed fp4 (f4E2M1) values stored as one byte of a 32-bit integer to packed f16, multiplying by the exponent part of scale before doing so.

The byte to convert is chosen by srcSelIndex.

cvt_scalef32_pk_f16_fp8(ssa)

rocdl.cvt.scalef32.pk.f16.fp8 - Scaled convert two fp8to two f16

Attributes

srcLoHiSel - Single, I1Attr, 1-bit signless integer attribute

Operands

src - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V2F16Type, fixed-length vector of 16-bit float values of length 2

Description

Convert two packed fp8 values in src0 to two f16 values, multiplying by the exponent in scale. The two values to be converted are selected from the low or high half of src (a packed vector represented as an i32) on the basis of srcLoHiSel.

cvt_scalef32_pk_f32_bf8(ssa)

rocdl.cvt.scalef32.pk.f32.bf8 - Scaled convert two bf8to two f32

Attributes

srcLoHiSel - Single, I1Attr, 1-bit signless integer attribute

Operands

src - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V2F32Type, fixed-length vector of 32-bit float values of length 2

Description

Convert two packed bf8 values in src0 to two f32 values, multiplying by the exponent in scale. The two values to be converted are selected from the low or high half of src (a packed vector represented as an i32) on the basis of srcLoHiSel.

cvt_scalef32_pk_f32_fp4(ssa)

rocdl.cvt.scalef32.pk.f32.fp4 - Scale and convert two packed fp4 to packed f32

Attributes

srcSelIndex - Single, I32Attr, 32-bit signless integer attribute

Operands

src - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V2F32Type, fixed-length vector of 32-bit float values of length 2

Description

Convert two packed fp4 (f4E2M1) values stored as one byte of a 32-bit integer to packed f32, multiplying by the exponent part of scale before doing so.

The byte to convert is chosen by srcSelIndex.

cvt_scalef32_pk_f32_fp8(ssa)

rocdl.cvt.scalef32.pk.f32.fp8 - Scaled convert two fp8to two f32

Attributes

srcLoHiSel - Single, I1Attr, 1-bit signless integer attribute

Operands

src - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V2F32Type, fixed-length vector of 32-bit float values of length 2

Description

Convert two packed fp8 values in src0 to two f32 values, multiplying by the exponent in scale. The two values to be converted are selected from the low or high half of src (a packed vector represented as an i32) on the basis of srcLoHiSel.

cvt_scalef32_pk_fp4_bf16(ssa)

rocdl.cvt.scalef32.pk.fp4.bf16 - Scale and convert two bf16 to packed fp4, updating tied vector

Attributes

dstSelIndex - Single, I32Attr, 32-bit signless integer attribute

Operands

oldVdst - Single, I32, 32-bit signless integer
src - Single, ROCDL_V2BF16Type, fixed-length vector of bfloat16 type values of length 2
scale - Single, F32, 32-bit float

Results

res - Single, I32, 32-bit signless integer

Description

Convert two packed bf16 values to packed fp4, dividing by the exponent part of scale before doing so.

The two scaled values are packed into a byte. That byte is used to update the dstSelIndexth byte of oldVdst, which is returned in its entirity.

cvt_scalef32_pk_fp4_f16(ssa)

rocdl.cvt.scalef32.pk.fp4.f16 - Scale and convert two f16 to packed fp4, updating tied vector

Attributes

dstSelIndex - Single, I32Attr, 32-bit signless integer attribute

Operands

oldVdst - Single, I32, 32-bit signless integer
src - Single, ROCDL_V2F16Type, fixed-length vector of 16-bit float values of length 2
scale - Single, F32, 32-bit float

Results

res - Single, I32, 32-bit signless integer

Description

Convert two packed f16 values to packed fp4, dividing by the exponent part of scale before doing so.

The two scaled values are packed into a byte. That byte is used to update the dstSelIndexth byte of oldVdst, which is returned in its entirity.

cvt_scalef32_pk_fp4_f32(ssa)

rocdl.cvt.scalef32.pk.fp4.f32 - Scale and convert two f32 values to two packed fp4, updating tied vector

Attributes

dstSelIndex - Single, I32Attr, 32-bit signless integer attribute

Operands

oldVdst - Single, I32, 32-bit signless integer
src0 - Single, F32, 32-bit float
src1 - Single, F32, 32-bit float
scale - Single, F32, 32-bit float

Results

res - Single, I32, 32-bit signless integer

Description

Convert two single-precision float values, passed in src0 and src1 into two fp4 values, dividing them by the expontent part of scale before doing so.

The two scaled values are packed into a byte. That byte is used to update the dstSelIndexth byte of oldVdst, which is returned in its entirity.

cvt_scalef32_pk_fp8_bf16(ssa)

rocdl.cvt.scalef32.pk.fp8.bf16 - Scaled convert two bf16to two fp8, updating packed vector

Attributes

dstLoHiSel - Single, I1Attr, 1-bit signless integer attribute

Operands

oldVdst - Single, ROCDL_V2I16Type, fixed-length vector of 16-bit signless integer values of length 2
src0 - Single, ROCDL_V2BF16Type, fixed-length vector of bfloat16 type values of length 2
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V2I16Type, fixed-length vector of 16-bit signless integer values of length 2

Description

Convert two bf16 values in src0 to two fp8 bytes, dividing by the exponent in scale. The bytes are packed into a 16-bit value which is inserted into oldVdst at the dstLoHiSel position, with the entire updated vector being returned.

cvt_scalef32_pk_fp8_f16(ssa)

rocdl.cvt.scalef32.pk.fp8.f16 - Scaled convert two f16to two fp8, updating packed vector

Attributes

dstLoHiSel - Single, I1Attr, 1-bit signless integer attribute

Operands

oldVdst - Single, ROCDL_V2I16Type, fixed-length vector of 16-bit signless integer values of length 2
src0 - Single, ROCDL_V2F16Type, fixed-length vector of 16-bit float values of length 2
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V2I16Type, fixed-length vector of 16-bit signless integer values of length 2

Description

Convert two f16 values in src0 to two fp8 bytes, dividing by the exponent in scale. The bytes are packed into a 16-bit value which is inserted into oldVdst at the dstLoHiSel position, with the entire updated vector being returned.

cvt_scalef32_pk_fp8_f32(ssa)

rocdl.cvt.scalef32.pk.fp8.f32 - Scaled convert two f32 to two fp8, updating packed vector

Attributes

dstLoHiSel - Single, I1Attr, 1-bit signless integer attribute

Operands

oldVdst - Single, ROCDL_V2I16Type, fixed-length vector of 16-bit signless integer values of length 2
src0 - Single, F32, 32-bit float
src1 - Single, F32, 32-bit float
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V2I16Type, fixed-length vector of 16-bit signless integer values of length 2

Description

Convert two f32 values in src0 and src1 to two fp8 bytes, dividing by the exponent in scale. The bytes are packed into a 16-bit value which is inserted into oldVdst at the dstLoHiSel position, with the entire updated vector being returned.

cvt_scalef32_sr_bf8_bf16(ssa)

rocdl.cvt.scalef32.sr.bf8.bf16 - Scaled convert bf16to bf8 with stochiastic rounding, updating packed vector

Attributes

dstSelIndex - Single, I32Attr, 32-bit signless integer attribute

Operands

oldVdst - Single, I32, 32-bit signless integer
src0 - Single, BF16, bfloat16 type
seed - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, I32, 32-bit signless integer

Description

Convert a bf16 value in src0 to a bf8 bytes, dividing by the exponent in scale and using seed for stochiastic rounding. Place the resulting byte in the dstSelIndexth bit of oldVdst and return the entire packed vector, which is stored as an i32.

cvt_scalef32_sr_bf8_f16(ssa)

rocdl.cvt.scalef32.sr.bf8.f16 - Scaled convert f16to bf8 with stochiastic rounding, updating packed vector

Attributes

dstSelIndex - Single, I32Attr, 32-bit signless integer attribute

Operands

oldVdst - Single, I32, 32-bit signless integer
src0 - Single, F16, 16-bit float
seed - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, I32, 32-bit signless integer

Description

Convert a f16 value in src0 to a bf8 bytes, dividing by the exponent in scale and using seed for stochiastic rounding. Place the resulting byte in the dstSelIndexth bit of oldVdst and return the entire packed vector, which is stored as an i32.

cvt_scalef32_sr_bf8_f32(ssa)

rocdl.cvt.scalef32.sr.bf8.f32 - Scaled convert f32to bf8 with stochiastic rounding, updating packed vector

Attributes

dstSelIndex - Single, I32Attr, 32-bit signless integer attribute

Operands

oldVdst - Single, I32, 32-bit signless integer
src0 - Single, F32, 32-bit float
seed - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, I32, 32-bit signless integer

Description

Convert a f32 value in src0 to a bf8 bytes, dividing by the exponent in scale and using seed for stochiastic rounding. Place the resulting byte in the dstSelIndexth bit of oldVdst and return the entire packed vector, which is stored as an i32.

cvt_scalef32_sr_fp8_bf16(ssa)

rocdl.cvt.scalef32.sr.fp8.bf16 - Scaled convert bf16to fp8 with stochiastic rounding, updating packed vector

Attributes

dstSelIndex - Single, I32Attr, 32-bit signless integer attribute

Operands

oldVdst - Single, I32, 32-bit signless integer
src0 - Single, BF16, bfloat16 type
seed - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, I32, 32-bit signless integer

Description

Convert a bf16 value in src0 to a fp8 bytes, dividing by the exponent in scale and using seed for stochiastic rounding. Place the resulting byte in the dstSelIndexth bit of oldVdst and return the entire packed vector, which is stored as an i32.

cvt_scalef32_sr_fp8_f16(ssa)

rocdl.cvt.scalef32.sr.fp8.f16 - Scaled convert f16to fp8 with stochiastic rounding, updating packed vector

Attributes

dstSelIndex - Single, I32Attr, 32-bit signless integer attribute

Operands

oldVdst - Single, I32, 32-bit signless integer
src0 - Single, F16, 16-bit float
seed - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, I32, 32-bit signless integer

Description

Convert a f16 value in src0 to a fp8 bytes, dividing by the exponent in scale and using seed for stochiastic rounding. Place the resulting byte in the dstSelIndexth bit of oldVdst and return the entire packed vector, which is stored as an i32.

cvt_scalef32_sr_fp8_f32(ssa)

rocdl.cvt.scalef32.sr.fp8.f32 - Scaled convert f32to fp8 with stochiastic rounding, updating packed vector

Attributes

dstSelIndex - Single, I32Attr, 32-bit signless integer attribute

Operands

oldVdst - Single, I32, 32-bit signless integer
src0 - Single, F32, 32-bit float
seed - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, I32, 32-bit signless integer

Description

Convert a f32 value in src0 to a fp8 bytes, dividing by the exponent in scale and using seed for stochiastic rounding. Place the resulting byte in the dstSelIndexth bit of oldVdst and return the entire packed vector, which is stored as an i32.

cvt_scalef32_sr_pk32_bf6_bf16(ssa)

rocdl.cvt.scalef32.sr.pk32.bf6.bf16 - Scale and convert packed bf16 to packed bf6 with stochiastic rounding

Operands

src - Single, ROCDL_V32BF16Type, fixed-length vector of bfloat16 type values of length 32
seed - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6

Description

Convert 32 packed bf16 values to packed bf6, dividing by the exponent part of scale before doing so and applying random rounding derived from seed.

cvt_scalef32_sr_pk32_bf6_f16(ssa)

rocdl.cvt.scalef32.sr.pk32.bf6.f16 - Scale and convert packed f16 to packed bf6 with stochiastic rounding

Operands

src - Single, ROCDL_V32F16Type, fixed-length vector of 16-bit float values of length 32
seed - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6

Description

Convert 32 packed f16 values to packed bf6, dividing by the exponent part of scale before doing so and applying random rounding derived from seed.

cvt_scalef32_sr_pk32_bf6_f32(ssa)

rocdl.cvt.scalef32.sr.pk32.bf6.f32 - Scale and convert packed f32 to packed bf6 with stochiastic rounding

Operands

src - Single, ROCDL_V32F32Type, fixed-length vector of 32-bit float values of length 32
seed - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6

Description

Convert 32 packed f32 values to packed bf6, dividing by the exponent part of scale before doing so and applying random rounding derived from seed.

cvt_scalef32_sr_pk32_fp6_bf16(ssa)

rocdl.cvt.scalef32.sr.pk32.fp6.bf16 - Scale and convert packed bf16 to packed fp6 with stochiastic rounding

Operands

src - Single, ROCDL_V32BF16Type, fixed-length vector of bfloat16 type values of length 32
seed - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6

Description

Convert 32 packed bf16 values to packed fp6, dividing by the exponent part of scale before doing so and applying random rounding derived from seed.

cvt_scalef32_sr_pk32_fp6_f16(ssa)

rocdl.cvt.scalef32.sr.pk32.fp6.f16 - Scale and convert packed f16 to packed fp6 with stochiastic rounding

Operands

src - Single, ROCDL_V32F16Type, fixed-length vector of 16-bit float values of length 32
seed - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6

Description

Convert 32 packed f16 values to packed fp6, dividing by the exponent part of scale before doing so and applying random rounding derived from seed.

cvt_scalef32_sr_pk32_fp6_f32(ssa)

rocdl.cvt.scalef32.sr.pk32.fp6.f32 - Scale and convert packed f32 to packed fp6 with stochiastic rounding

Operands

src - Single, ROCDL_V32F32Type, fixed-length vector of 32-bit float values of length 32
seed - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, ROCDL_V6I32Type, fixed-length vector of 32-bit signless integer values of length 6

Description

Convert 32 packed f32 values to packed fp6, dividing by the exponent part of scale before doing so and applying random rounding derived from seed.

cvt_scalef32_sr_pk_fp4_bf16(ssa)

rocdl.cvt.scalef32.sr.pk.fp4.bf16 - Scale and convert two bf16 to packed fp4 with stochiastic rounding, updating tied vector

Attributes

dstSelIndex - Single, I32Attr, 32-bit signless integer attribute

Operands

oldVdst - Single, I32, 32-bit signless integer
src - Single, ROCDL_V2BF16Type, fixed-length vector of bfloat16 type values of length 2
seed - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, I32, 32-bit signless integer

Description

Convert two packed bf16 values to packed fp4, dividing by the exponent part of scale before doing so and using seed as the random seed for stochiastic rounding.

The two scaled values are packed (little-endian) into a byte. That byte is used to update the dstSelIndexth byte of oldVdst, which is returned in its entirity.

cvt_scalef32_sr_pk_fp4_f16(ssa)

rocdl.cvt.scalef32.sr.pk.fp4.f16 - Scale and convert two f16 to packed fp4 with stochiastic rounding, updating tied vector

Attributes

dstSelIndex - Single, I32Attr, 32-bit signless integer attribute

Operands

oldVdst - Single, I32, 32-bit signless integer
src - Single, ROCDL_V2F16Type, fixed-length vector of 16-bit float values of length 2
seed - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, I32, 32-bit signless integer

Description

Convert two packed f16 values to packed fp4, dividing by the exponent part of scale before doing so and using seed as the random seed for stochiastic rounding.

The two scaled values are packed (little-endian) into a byte. That byte is used to update the dstSelIndexth byte of oldVdst, which is returned in its entirity.

cvt_scalef32_sr_pk_fp4_f32(ssa)

rocdl.cvt.scalef32.sr.pk.fp4.f32 - Scale and convert two f32 to packed fp4 with stochiastic rounding, updating tied vector

Attributes

dstSelIndex - Single, I32Attr, 32-bit signless integer attribute

Operands

oldVdst - Single, I32, 32-bit signless integer
src - Single, ROCDL_V2F32Type, fixed-length vector of 32-bit float values of length 2
seed - Single, I32, 32-bit signless integer
scale - Single, F32, 32-bit float

Results

res - Single, I32, 32-bit signless integer

Description

Convert two packed f32 values to packed fp4, dividing by the exponent part of scale before doing so and using seed as the random seed for stochiastic rounding.

The two scaled values are packed (little-endian) into a byte. That byte is used to update the dstSelIndexth byte of oldVdst, which is returned in its entirity.

cvt_sr_bf8_f32(ssa)

rocdl.cvt.sr.bf8.f32 - Convert f32 to bf8, stochiastic rounding

Attributes

byteSel - Single, I32Attr, 32-bit signless integer attribute

Operands

srcA - Single, F32, 32-bit float
srcB - Single, I32, 32-bit signless integer
old - Single, I32, 32-bit signless integer

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

Description

Convert srcA to bf8, adding the rounding factor from srcB, and store into the byteSelth byte of old, preserving the others.

cvt_sr_fp8_f32(ssa)

rocdl.cvt.sr.fp8.f32 - Convert f32 to fp8, stochiastic rounding

Attributes

byteSel - Single, I32Attr, 32-bit signless integer attribute

Operands

srcA - Single, F32, 32-bit float
srcB - Single, I32, 32-bit signless integer
old - Single, I32, 32-bit signless integer

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

Description

Convert srcA to fp8, adding the rounding factor from srcB, and store into the byteSelth byte of old, preserving the others.

ds_bpermute(ssa)

rocdl.ds_bpermute

Operands

index - Single, I32, 32-bit signless integer
src - Single, I32, 32-bit signless integer

Results

res - Single, I32, 32-bit signless integer

ds_read_tr4_b64(ssa)

rocdl.ds.read.tr4.b64

Attributes

alias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
noalias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
tbaa - Optional, LLVM_TBAATagArrayAttr, LLVM dialect TBAA tag metadata array

Operands

ptr - Single, ROCDLBufferLDS, LLVM pointer in address space 3

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

ds_read_tr6_b96(ssa)

rocdl.ds.read.tr6.b96

Attributes

alias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
noalias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
tbaa - Optional, LLVM_TBAATagArrayAttr, LLVM dialect TBAA tag metadata array

Operands

ptr - Single, ROCDLBufferLDS, LLVM pointer in address space 3

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

ds_read_tr8_b64(ssa)

rocdl.ds.read.tr8.b64

Attributes

alias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
noalias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
tbaa - Optional, LLVM_TBAATagArrayAttr, LLVM dialect TBAA tag metadata array

Operands

ptr - Single, ROCDLBufferLDS, LLVM pointer in address space 3

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

ds_read_tr16_b64(ssa)

rocdl.ds.read.tr16.b64

Attributes

alias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
noalias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
tbaa - Optional, LLVM_TBAATagArrayAttr, LLVM dialect TBAA tag metadata array

Operands

ptr - Single, ROCDLBufferLDS, LLVM pointer in address space 3

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

ds_swizzle(ssa)

rocdl.ds_swizzle

Operands

src - Single, I32, 32-bit signless integer
offset - Single, I32, 32-bit signless integer

Results

res - Single, I32, 32-bit signless integer

fmed3(ssa)

rocdl.fmed3 - Median of three float/half values

Operands

src0 - Single, anonymous/composite constraint, floating point LLVM type or LLVM dialect-compatible vector of floating point LLVM type
src1 - Single, anonymous/composite constraint, floating point LLVM type or LLVM dialect-compatible vector of floating point LLVM type
src2 - Single, anonymous/composite constraint, floating point LLVM type or LLVM dialect-compatible vector of floating point LLVM type

Results

res - Single, anonymous/composite constraint, floating point LLVM type or LLVM dialect-compatible vector of floating point LLVM type

Description

Computes the median of three floating-point values using the AMDGPU fmed3 intrinsic. This operation is equivalent to max(min(a, b), min(max(a, b), c)) but uses the hardware-accelerated V_MED3_F16/V_MED3_F32 instruction for better performance.

The operation supports both scalar and vector floating-point types (f16, f32).

Example:

// Scalar f32 median
%result = rocdl.fmed3 %a, %b, %c : f32

// Vector f16 median
%result = rocdl.fmed3 %va, %vb, %vc : vector<4xf16>

global_load_lds(ssa)

rocdl.global.load.lds

Attributes

size - Single, I32Attr, 32-bit signless integer attribute
offset - Single, I32Attr, 32-bit signless integer attribute
aux - Single, I32Attr, 32-bit signless integer attribute
alias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
noalias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
tbaa - Optional, LLVM_TBAATagArrayAttr, LLVM dialect TBAA tag metadata array

Operands

globalPtr - Single, ROCDLGlobalBuffer, LLVM pointer in address space 1
ldsPtr - Single, ROCDLBufferLDS, LLVM pointer in address space 3

grid_dim_x(ssa)

rocdl.grid.dim.x

grid_dim_y(ssa)

rocdl.grid.dim.y

grid_dim_z(ssa)

rocdl.grid.dim.z

iglp_opt(ssa)

rocdl.iglp.opt

load_to_lds(ssa)

rocdl.load.to.lds

Attributes

size - Single, I32Attr, 32-bit signless integer attribute
offset - Single, I32Attr, 32-bit signless integer attribute
aux - Single, I32Attr, 32-bit signless integer attribute
alias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
noalias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
tbaa - Optional, LLVM_TBAATagArrayAttr, LLVM dialect TBAA tag metadata array

Operands

globalPtr - Single, LLVM_AnyPointer, LLVM pointer type
ldsPtr - Single, ROCDLBufferLDS, LLVM pointer in address space 3

make_buffer_rsrc(ssa)

rocdl.make.buffer.rsrc

Operands

base - Single, LLVM_AnyPointer, LLVM pointer type
stride - Single, I16, 16-bit signless integer
numRecords - Single, I64, 64-bit signless integer
flags - Single, I32, 32-bit signless integer

Results

res - Single, LLVM_AnyPointer, LLVM pointer type

mbcnt_hi(ssa)

rocdl.mbcnt.hi

Attributes

arg_attrs - Optional, DictArrayAttr, Array of dictionary attributes
res_attrs - Optional, DictArrayAttr, Array of dictionary attributes

Operands

in0 - Single, I32, 32-bit signless integer
in1 - Single, I32, 32-bit signless integer

Results

res - Single, I32, 32-bit signless integer

mbcnt_lo(ssa)

rocdl.mbcnt.lo

Attributes

arg_attrs - Optional, DictArrayAttr, Array of dictionary attributes
res_attrs - Optional, DictArrayAttr, Array of dictionary attributes

Operands

in0 - Single, I32, 32-bit signless integer
in1 - Single, I32, 32-bit signless integer

Results

res - Single, I32, 32-bit signless integer

mfma_f32_4x4x1f32(ssa)

rocdl.mfma.f32.4x4x1f32

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_4x4x2bf16(ssa)

rocdl.mfma.f32.4x4x2bf16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_4x4x4bf16_1k(ssa)

rocdl.mfma.f32.4x4x4bf16.1k

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_4x4x4f16(ssa)

rocdl.mfma.f32.4x4x4f16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_16x16x1f32(ssa)

rocdl.mfma.f32.16x16x1f32

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_16x16x2bf16(ssa)

rocdl.mfma.f32.16x16x2bf16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_16x16x4bf16_1k(ssa)

rocdl.mfma.f32.16x16x4bf16.1k

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_16x16x4f16(ssa)

rocdl.mfma.f32.16x16x4f16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_16x16x4f32(ssa)

rocdl.mfma.f32.16x16x4f32

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_16x16x8_xf32(ssa)

rocdl.mfma.f32.16x16x8.xf32

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_16x16x8bf16(ssa)

rocdl.mfma.f32.16x16x8bf16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_16x16x16bf16_1k(ssa)

rocdl.mfma.f32.16x16x16bf16.1k

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_16x16x16f16(ssa)

rocdl.mfma.f32.16x16x16f16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_16x16x32_bf8_bf8(ssa)

rocdl.mfma.f32.16x16x32.bf8.bf8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_16x16x32_bf8_fp8(ssa)

rocdl.mfma.f32.16x16x32.bf8.fp8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_16x16x32_bf16(ssa)

rocdl.mfma.f32.16x16x32.bf16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_16x16x32_f16(ssa)

rocdl.mfma.f32.16x16x32.f16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_16x16x32_fp8_bf8(ssa)

rocdl.mfma.f32.16x16x32.fp8.bf8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_16x16x32_fp8_fp8(ssa)

rocdl.mfma.f32.16x16x32.fp8.fp8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_32x32x1f32(ssa)

rocdl.mfma.f32.32x32x1f32

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_32x32x2bf16(ssa)

rocdl.mfma.f32.32x32x2bf16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_32x32x2f32(ssa)

rocdl.mfma.f32.32x32x2f32

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_32x32x4_xf32(ssa)

rocdl.mfma.f32.32x32x4.xf32

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_32x32x4bf16(ssa)

rocdl.mfma.f32.32x32x4bf16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_32x32x4bf16_1k(ssa)

rocdl.mfma.f32.32x32x4bf16.1k

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_32x32x4f16(ssa)

rocdl.mfma.f32.32x32x4f16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_32x32x8bf16_1k(ssa)

rocdl.mfma.f32.32x32x8bf16.1k

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_32x32x8f16(ssa)

rocdl.mfma.f32.32x32x8f16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_32x32x16_bf8_bf8(ssa)

rocdl.mfma.f32.32x32x16.bf8.bf8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_32x32x16_bf8_fp8(ssa)

rocdl.mfma.f32.32x32x16.bf8.fp8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_32x32x16_bf16(ssa)

rocdl.mfma.f32.32x32x16.bf16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_32x32x16_f16(ssa)

rocdl.mfma.f32.32x32x16.f16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_32x32x16_fp8_bf8(ssa)

rocdl.mfma.f32.32x32x16.fp8.bf8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f32_32x32x16_fp8_fp8(ssa)

rocdl.mfma.f32.32x32x16.fp8.fp8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f64_4x4x4f64(ssa)

rocdl.mfma.f64.4x4x4f64

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_f64_16x16x4f64(ssa)

rocdl.mfma.f64.16x16x4f64

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_i32_4x4x4i8(ssa)

rocdl.mfma.i32.4x4x4i8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_i32_16x16x4i8(ssa)

rocdl.mfma.i32.16x16x4i8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_i32_16x16x16i8(ssa)

rocdl.mfma.i32.16x16x16i8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_i32_16x16x32_i8(ssa)

rocdl.mfma.i32.16x16x32.i8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_i32_16x16x64_i8(ssa)

rocdl.mfma.i32.16x16x64.i8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_i32_32x32x4i8(ssa)

rocdl.mfma.i32.32x32x4i8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_i32_32x32x8i8(ssa)

rocdl.mfma.i32.32x32x8i8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_i32_32x32x16_i8(ssa)

rocdl.mfma.i32.32x32x16.i8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_i32_32x32x32_i8(ssa)

rocdl.mfma.i32.32x32x32.i8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_scale_f32_16x16x128_f8f6f4(ssa)

rocdl.mfma.scale.f32.16x16x128.f8f6f4

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

mfma_scale_f32_32x32x64_f8f6f4(ssa)

rocdl.mfma.scale.f32.32x32x64.f8f6f4

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

permlane16_swap(ssa)

rocdl.permlane16.swap

Attributes

fi - Single, I1Attr, 1-bit signless integer attribute
boundControl - Single, I1Attr, 1-bit signless integer attribute

Operands

old - Single, I32, 32-bit signless integer
src - Single, I32, 32-bit signless integer

Results

res - Single, anonymous/composite constraint, LLVM dialect-compatible struct of 32-bit signless integerand32-bit signless integer

Description

Performs a permlane16.swap operation with the given operands, applying the permutation specified by $fi to the provided inputs.

permlane32_swap(ssa)

rocdl.permlane32.swap

Attributes

fi - Single, I1Attr, 1-bit signless integer attribute
boundControl - Single, I1Attr, 1-bit signless integer attribute

Operands

old - Single, I32, 32-bit signless integer
src - Single, I32, 32-bit signless integer

Results

res - Single, anonymous/composite constraint, LLVM dialect-compatible struct of 32-bit signless integerand32-bit signless integer

Description

Performs a permlane32.swap operation with the given operands, applying the permutation specified by $fi to the provided inputs.

permlanex16(ssa)

rocdl.permlanex16

Attributes

fi - Single, I1Attr, 1-bit signless integer attribute
boundControl - Single, I1Attr, 1-bit signless integer attribute

Operands

old - Single, LLVM_Type, LLVM dialect-compatible type
src0 - Single, LLVM_Type, LLVM dialect-compatible type
src1 - Single, LLVM_Type, LLVM dialect-compatible type
src2 - Single, LLVM_Type, LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

Description

Performs a permlanex16 operation with the given operands, applying the permutation specified by $fi to the provided inputs.

raw_buffer_atomic_cmpswap(ssa)

rocdl.raw.buffer.atomic.cmpswap

Operands

src - Single, LLVM_Type, LLVM dialect-compatible type
cmp - Single, LLVM_Type, LLVM dialect-compatible type
rsrc - Single, LLVM_Type, LLVM dialect-compatible type
offset - Single, I32, 32-bit signless integer
soffset - Single, I32, 32-bit signless integer
aux - Single, I32, 32-bit signless integer

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

raw_buffer_atomic_fadd(ssa)

rocdl.raw.buffer.atomic.fadd

Operands

vdata - Single, LLVM_Type, LLVM dialect-compatible type
rsrc - Single, LLVM_Type, LLVM dialect-compatible type
offset - Single, LLVM_Type, LLVM dialect-compatible type
soffset - Single, LLVM_Type, LLVM dialect-compatible type
aux - Single, LLVM_Type, LLVM dialect-compatible type

raw_buffer_atomic_fmax(ssa)

rocdl.raw.buffer.atomic.fmax

Operands

vdata - Single, LLVM_Type, LLVM dialect-compatible type
rsrc - Single, LLVM_Type, LLVM dialect-compatible type
offset - Single, LLVM_Type, LLVM dialect-compatible type
soffset - Single, LLVM_Type, LLVM dialect-compatible type
aux - Single, LLVM_Type, LLVM dialect-compatible type

raw_buffer_atomic_smax(ssa)

rocdl.raw.buffer.atomic.smax

Operands

vdata - Single, LLVM_Type, LLVM dialect-compatible type
rsrc - Single, LLVM_Type, LLVM dialect-compatible type
offset - Single, LLVM_Type, LLVM dialect-compatible type
soffset - Single, LLVM_Type, LLVM dialect-compatible type
aux - Single, LLVM_Type, LLVM dialect-compatible type

raw_buffer_atomic_umin(ssa)

rocdl.raw.buffer.atomic.umin

Operands

vdata - Single, LLVM_Type, LLVM dialect-compatible type
rsrc - Single, LLVM_Type, LLVM dialect-compatible type
offset - Single, LLVM_Type, LLVM dialect-compatible type
soffset - Single, LLVM_Type, LLVM dialect-compatible type
aux - Single, LLVM_Type, LLVM dialect-compatible type

raw_buffer_load(ssa)

rocdl.raw.buffer.load

Operands

rsrc - Single, LLVM_Type, LLVM dialect-compatible type
offset - Single, LLVM_Type, LLVM dialect-compatible type
soffset - Single, LLVM_Type, LLVM dialect-compatible type
aux - Single, LLVM_Type, LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

raw_buffer_store(ssa)

rocdl.raw.buffer.store

Operands

vdata - Single, LLVM_Type, LLVM dialect-compatible type
rsrc - Single, LLVM_Type, LLVM dialect-compatible type
offset - Single, LLVM_Type, LLVM dialect-compatible type
soffset - Single, LLVM_Type, LLVM dialect-compatible type
aux - Single, LLVM_Type, LLVM dialect-compatible type

raw_ptr_buffer_atomic_cmpswap(ssa)

rocdl.raw.ptr.buffer.atomic.cmpswap

Attributes

alias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
noalias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
tbaa - Optional, LLVM_TBAATagArrayAttr, LLVM dialect TBAA tag metadata array

Operands

src - Single, LLVM_Type, LLVM dialect-compatible type
cmp - Single, LLVM_Type, LLVM dialect-compatible type
rsrc - Single, ROCDLBufferRsrc, LLVM pointer in address space 8
offset - Single, I32, 32-bit signless integer
soffset - Single, I32, 32-bit signless integer
aux - Single, I32, 32-bit signless integer

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

raw_ptr_buffer_atomic_fadd(ssa)

rocdl.raw.ptr.buffer.atomic.fadd

Attributes

alias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
noalias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
tbaa - Optional, LLVM_TBAATagArrayAttr, LLVM dialect TBAA tag metadata array

Operands

vdata - Single, LLVM_Type, LLVM dialect-compatible type
rsrc - Single, ROCDLBufferRsrc, LLVM pointer in address space 8
offset - Single, I32, 32-bit signless integer
soffset - Single, I32, 32-bit signless integer
aux - Single, I32, 32-bit signless integer

raw_ptr_buffer_atomic_fmax(ssa)

rocdl.raw.ptr.buffer.atomic.fmax

Attributes

alias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
noalias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
tbaa - Optional, LLVM_TBAATagArrayAttr, LLVM dialect TBAA tag metadata array

Operands

vdata - Single, LLVM_Type, LLVM dialect-compatible type
rsrc - Single, ROCDLBufferRsrc, LLVM pointer in address space 8
offset - Single, I32, 32-bit signless integer
soffset - Single, I32, 32-bit signless integer
aux - Single, I32, 32-bit signless integer

raw_ptr_buffer_atomic_smax(ssa)

rocdl.raw.ptr.buffer.atomic.smax

Attributes

alias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
noalias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
tbaa - Optional, LLVM_TBAATagArrayAttr, LLVM dialect TBAA tag metadata array

Operands

vdata - Single, LLVM_Type, LLVM dialect-compatible type
rsrc - Single, ROCDLBufferRsrc, LLVM pointer in address space 8
offset - Single, I32, 32-bit signless integer
soffset - Single, I32, 32-bit signless integer
aux - Single, I32, 32-bit signless integer

raw_ptr_buffer_atomic_umin(ssa)

rocdl.raw.ptr.buffer.atomic.umin

Attributes

alias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
noalias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
tbaa - Optional, LLVM_TBAATagArrayAttr, LLVM dialect TBAA tag metadata array

Operands

vdata - Single, LLVM_Type, LLVM dialect-compatible type
rsrc - Single, ROCDLBufferRsrc, LLVM pointer in address space 8
offset - Single, I32, 32-bit signless integer
soffset - Single, I32, 32-bit signless integer
aux - Single, I32, 32-bit signless integer

raw_ptr_buffer_load(ssa)

rocdl.raw.ptr.buffer.load

Attributes

alias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
noalias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
tbaa - Optional, LLVM_TBAATagArrayAttr, LLVM dialect TBAA tag metadata array

Operands

rsrc - Single, ROCDLBufferRsrc, LLVM pointer in address space 8
offset - Single, I32, 32-bit signless integer
soffset - Single, I32, 32-bit signless integer
aux - Single, I32, 32-bit signless integer

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

raw_ptr_buffer_load_lds(ssa)

rocdl.raw.ptr.buffer.load.lds

Attributes

alias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
noalias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
tbaa - Optional, LLVM_TBAATagArrayAttr, LLVM dialect TBAA tag metadata array

Operands

rsrc - Single, ROCDLBufferRsrc, LLVM pointer in address space 8
ldsPtr - Single, ROCDLBufferLDS, LLVM pointer in address space 3
size - Single, I32, 32-bit signless integer
voffset - Single, I32, 32-bit signless integer
soffset - Single, I32, 32-bit signless integer
offset - Single, I32, 32-bit signless integer
aux - Single, I32, 32-bit signless integer

raw_ptr_buffer_store(ssa)

rocdl.raw.ptr.buffer.store

Attributes

alias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
noalias_scopes - Optional, LLVM_AliasScopeArrayAttr, LLVM dialect alias scope array
tbaa - Optional, LLVM_TBAATagArrayAttr, LLVM dialect TBAA tag metadata array

Operands

vdata - Single, LLVM_Type, LLVM dialect-compatible type
rsrc - Single, ROCDLBufferRsrc, LLVM pointer in address space 8
offset - Single, I32, 32-bit signless integer
soffset - Single, I32, 32-bit signless integer
aux - Single, I32, 32-bit signless integer

readfirstlane(ssa)

rocdl.readfirstlane - Get the value in first active lane.

Operands

src - Single, LLVM_Type, LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

Description

Returns the value in the lowest active lane of the input operand.

readlane(ssa)

rocdl.readlane - Get the value in the specific lane.

Operands

src0 - Single, LLVM_Type, LLVM dialect-compatible type
src1 - Single, I32, 32-bit signless integer

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

Description

Get the value in lane src1 from input src0.

s_barrier(ssa)

rocdl.s.barrier

s_barrier_signal(ssa)

rocdl.s.barrier.signal

s_barrier_wait(ssa)

rocdl.s.barrier.wait

s_setprio(ssa)

rocdl.s.setprio

s_sleep(ssa)

rocdl.s.sleep

s_wait_dscnt(ssa)

rocdl.s.wait.dscnt

s_wait_expcnt(ssa)

rocdl.s.wait.expcnt

s_wait_loadcnt(ssa)

rocdl.s.wait.loadcnt

s_wait_storecnt(ssa)

rocdl.s.wait.storecnt

s_waitcnt(ssa)

rocdl.s.waitcnt

sched_barrier(ssa)

rocdl.sched.barrier

sched_group_barrier(ssa)

rocdl.sched.group.barrier

smfmac_f32_16x16x32_bf16(ssa)

rocdl.smfmac.f32.16x16x32.bf16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

smfmac_f32_16x16x32_f16(ssa)

rocdl.smfmac.f32.16x16x32.f16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

smfmac_f32_16x16x64_bf8_bf8(ssa)

rocdl.smfmac.f32.16x16x64.bf8.bf8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

smfmac_f32_16x16x64_bf8_fp8(ssa)

rocdl.smfmac.f32.16x16x64.bf8.fp8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

smfmac_f32_16x16x64_fp8_bf8(ssa)

rocdl.smfmac.f32.16x16x64.fp8.bf8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

smfmac_f32_16x16x64_fp8_fp8(ssa)

rocdl.smfmac.f32.16x16x64.fp8.fp8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

smfmac_f32_32x32x16_bf16(ssa)

rocdl.smfmac.f32.32x32x16.bf16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

smfmac_f32_32x32x16_f16(ssa)

rocdl.smfmac.f32.32x32x16.f16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

smfmac_f32_32x32x32_bf8_bf8(ssa)

rocdl.smfmac.f32.32x32x32.bf8.bf8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

smfmac_f32_32x32x32_bf8_fp8(ssa)

rocdl.smfmac.f32.32x32x32.bf8.fp8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

smfmac_f32_32x32x32_fp8_bf8(ssa)

rocdl.smfmac.f32.32x32x32.fp8.bf8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

smfmac_f32_32x32x32_fp8_fp8(ssa)

rocdl.smfmac.f32.32x32x32.fp8.fp8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

smfmac_i32_16x16x64_i8(ssa)

rocdl.smfmac.i32.16x16x64.i8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

smfmac_i32_32x32x32_i8(ssa)

rocdl.smfmac.i32.32x32x32.i8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

update_dpp(ssa)

rocdl.update.dpp

Attributes

dppCtrl - Single, I32Attr, 32-bit signless integer attribute
rowMask - Single, I32Attr, 32-bit signless integer attribute
bankMask - Single, I32Attr, 32-bit signless integer attribute
boundCtrl - Single, I1Attr, 1-bit signless integer attribute

Operands

old - Single, LLVM_Type, LLVM dialect-compatible type
src - Single, LLVM_Type, LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

wavefrontsize(ssa)

rocdl.wavefrontsize

wmma_bf16_16x16x16_bf16(ssa)

rocdl.wmma.bf16.16x16x16.bf16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

wmma_f16_16x16x16_f16(ssa)

rocdl.wmma.f16.16x16x16.f16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

wmma_f32_16x16x16_bf8_bf8(ssa)

rocdl.wmma.f32.16x16x16.bf8_bf8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

wmma_f32_16x16x16_bf8_fp8(ssa)

rocdl.wmma.f32.16x16x16.bf8_fp8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

wmma_f32_16x16x16_bf16(ssa)

rocdl.wmma.f32.16x16x16.bf16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

wmma_f32_16x16x16_f16(ssa)

rocdl.wmma.f32.16x16x16.f16

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

wmma_f32_16x16x16_fp8_bf8(ssa)

rocdl.wmma.f32.16x16x16.fp8_bf8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

wmma_f32_16x16x16_fp8_fp8(ssa)

rocdl.wmma.f32.16x16x16.fp8_fp8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

wmma_i32_16x16x16_iu4(ssa)

rocdl.wmma.i32.16x16x16.iu4

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

wmma_i32_16x16x16_iu8(ssa)

rocdl.wmma.i32.16x16x16.iu8

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

wmma_i32_16x16x32_iu4(ssa)

rocdl.wmma.i32.16x16x32.iu4

Operands

args - Variadic, LLVM_Type, variadic of LLVM dialect-compatible type

Results

res - Single, LLVM_Type, LLVM dialect-compatible type

workgroup_dim_x(ssa)

rocdl.workgroup.dim.x

workgroup_dim_y(ssa)

rocdl.workgroup.dim.y

workgroup_dim_z(ssa)

rocdl.workgroup.dim.z

workgroup_id_x(ssa)

rocdl.workgroup.id.x

workgroup_id_y(ssa)

rocdl.workgroup.id.y

workgroup_id_z(ssa)

rocdl.workgroup.id.z

workitem_id_x(ssa)

rocdl.workitem.id.x

workitem_id_y(ssa)

rocdl.workitem.id.y

workitem_id_z(ssa)

rocdl.workitem.id.z