Store

Store Local

Represents ST_L instruction. Stores a scalar to scalar local memory.

Intrinsics Prototype and Arguments

Intrinsics Definition

void s_f32_st_l(uint32_t addr, float value, int switches=0, bool predicate=1, bool polarity=0)

void s_bf16_st_l(uint32_t addr, bf16 value, int switches=0, bool predicate=1, bool polarity=0)

void s_i32_st_l(uint32_t addr, int32_t value, int switches=0, bool predicate=1, bool polarity=0)

void s_u32_st_l(uint32_t addr, uint32_t value, int switches=0, bool predicate=1, bool polarity=0)

void s_i16_st_l(uint32_t addr, int16_t value, int switches=0, bool predicate=1, bool polarity=0)

void s_u16_st_l(uint32_t addr, uint16_t value, int switches=0, bool predicate=1, bool polarity=0)

void s_i8_st_l(uint32_t addr, int8_t value, int switches=0, bool predicate=1, bool polarity=0)

void s_u8_st_l(uint32_t addr, uint8_t value, int switches=0, bool predicate=1, bool polarity=0)

void s_i1_st_l(uint32_t addr, bool value, int switches=0, bool predicate=1, bool polarity=0)

Parameters

addr	Address to write to (SRC1).
value	Value to write (SRC2).
switches	Instruction switches.
predicate	Predicate value for the instruction.
polarity	True if polarity of the predicate is inverted.

Switches

Allowed switches are:

• SW_MMIO - store to MMIO.

• SW_SLM - store to SLM (the default).

Store Local vector

Represents ST_L_V instruction. Stores a vector to vector local memory.

Intrinsics Prototype and Arguments

Intrinsics Definition

void v_f32_st_l_v(uint32_t addr, float64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_bf16_st_l_v(uint32_t addr, bfloat128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i32_st_l_v(uint32_t addr, int64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u32_st_l_v(uint32_t addr, uint64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i16_st_l_v(uint32_t addr, short128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u16_st_l_v(uint32_t addr, ushort128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i8_st_l_v(uint32_t addr, char256 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u8_st_l_v(uint32_t addr, uchar256 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i1_st_l_v(uint32_t addr, bool256 value, int switches=0, bool predicate=1, bool polarity=0)

Parameters

addr	Address to write to (SRC1).
value	Value to write (SRC2).
switches	Instruction switches.
predicate	Predicate value for the instruction.
polarity	True if polarity of the predicate is inverted.

Store local vector low

Represents ST_L_V_LOW instruction. Stores half a vector to vector local memory.

Intrinsics Prototype and Arguments

Intrinsics Definition

void v_f32_st_l_v_low(uint32_t addr, float64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_bf16_st_l_v_low(uint32_t addr, bfloat128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i32_st_l_v_low(uint32_t addr, int64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u32_st_l_v_low(uint32_t addr, uint64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i16_st_l_v_low(uint32_t addr, short128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u16_st_l_v_low(uint32_t addr, ushort128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i8_st_l_v_low(uint32_t addr, char256 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u8_st_l_v_low(uint32_t addr, uchar256 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i1_st_l_v_low(uint32_t addr, bool256 value, int switches=0, bool predicate=1, bool polarity=0)

Parameters

addr	Address to write to (SRC1).
value	Value to write (SRC2).
switches	Instruction switches.
predicate	Predicate value for the instruction.
polarity	True if polarity of the predicate is inverted.

Store local vector high

Represents ST_L_V_HIGH instruction. Stores half a vector to vector local memory.

Intrinsics Prototype and Arguments

Intrinsics Definition

void v_f32_st_l_v_high(uint32_t addr, float64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_bf16_st_l_v_high(uint32_t addr, bfloat128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i32_st_l_v_high(uint32_t addr, int64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u32_st_l_v_high(uint32_t addr, uint64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i16_st_l_v_high(uint32_t addr, short128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u16_st_l_v_high(uint32_t addr, ushort128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i8_st_l_v_high(uint32_t addr, char256 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u8_st_l_v_high(uint32_t addr, uchar256 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i1_st_l_v_high(uint32_t addr, bool256 value, int switches=0, bool predicate=1, bool polarity=0)

Parameters

addr	Address to write to (SRC1).
value	Value to write (SRC2).
switches	Instruction switches.
predicate	Predicate value for the instruction.
polarity	True if polarity of the predicate is inverted.

Store tensor

Represents ST_TNSR instruction.

Intrinsics Prototype and Arguments

Intrinsics Definition

void v_f32_st_tnsr(int5 ndx, const int8_t tensor, float64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_bf16_st_tnsr(int5 ndx, const int8_t tensor, bfloat128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i32_st_tnsr(int5 ndx, const int8_t tensor, int64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u32_st_tnsr(int5 ndx, const int8_t tensor, uint64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i16_st_tnsr(int5 ndx, const int8_t tensor, short128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u16_st_tnsr(int5 ndx, const int8_t tensor, ushort128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i8_st_tnsr(int5 ndx, const int8_t tensor, char256 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u8_st_tnsr(int5 ndx, const int8_t tensor, uchar256 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i1_st_tnsr(int5 ndx, const int8_t tensor, bool256 value, int switches=0, bool predicate=1, bool polarity=0)

Parameters

ndx	Tensor coordinates (SRC1).
tensor	Tensor number.
value	Value to write (SRC3).
switches	Instruction switches.
predicate	Predicate value for the instruction.
polarity	True if polarity of the predicate is inverted.

Switches

Allowed switches are:

• SW_PACK - When set, store pipe will perform packing according to PACK_DT.

Store tensor using PARTIAL switch

Represents ST_TNSR instruction with PARTIAL switch.

Intrinsics Prototype and Arguments

Intrinsics Definition

void v_f32_st_tnsr_partial(int5 ndx, int8_t tensor, float64 value, int8_t size, int8_t offset, int switches=0, bool predicate=1, bool polarity=0)

void v_bf16_st_tnsr_partial(int5 ndx, int8_t tensor, bfloat128 value, int8_t size, int8_t offset, int switches=0, bool predicate=1, bool polarity=0)

void v_i32_st_tnsr_partial(int5 ndx, int8_t tensor, int64 value, int8_t size, int8_t offset, int switches=0, bool predicate=1, bool polarity=0)

void v_u32_st_tnsr_partial(int5 ndx, int8_t tensor, uint64 value, int8_t size, int8_t offset, int switches=0, bool predicate=1, bool polarity=0)

void v_i16_st_tnsr_partial(int5 ndx, int8_t tensor, short128 value, int8_t size, int8_t offset, int switches=0, bool predicate=1, bool polarity=0)

void v_u16_st_tnsr_partial(int5 ndx, int8_t tensor, ushort128 value, int8_t size, int8_t offset, int switches=0, bool predicate=1, bool polarity=0)

void v_i8_st_tnsr_partial(int5 ndx, int8_t tensor, char256 value, int8_t size, int8_t offset, int switches=0, bool predicate=1, bool polarity=0)

void v_u8_st_tnsr_partial(int5 ndx, int8_t tensor, uchar256 value, int8_t size, int8_t offset, int switches=0, bool predicate=1, bool polarity=0)

void v_i1_st_tnsr_partial(int5 ndx, int8_t tensor, bool256 value, int8_t size, int8_t offset, int switches=0, bool predicate=1, bool polarity=0)

Parameters

ndx	Tensor coordinates (SRC1).
tensor	Tensor number.
value	Value to write (SRC3).
size	Size in elements minus 1.
offset	Offset in elements.
switches	Instruction switches.
predicate	Predicate value for the instruction.
polarity	True if polarity of the predicate is inverted.

Switches

Allowed switches are:

• SW_PACK - When set, store pipe will perform packing according to PACK_DT.

Store tensor high

Represents ST_TNSR_HIGH instruction.

Intrinsics Prototype and Arguments

Intrinsics Definition

void v_f32_st_tnsr_high(int5 ndx, const int8_t tensor, float64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_bf16_st_tnsr_high(int5 ndx, const int8_t tensor, bfloat128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i32_st_tnsr_high(int5 ndx, const int8_t tensor, int64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u32_st_tnsr_high(int5 ndx, const int8_t tensor, uint64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i16_st_tnsr_high(int5 ndx, const int8_t tensor, short128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u16_st_tnsr_high(int5 ndx, const int8_t tensor, ushort128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i8_st_tnsr_high(int5 ndx, const int8_t tensor, char256 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u8_st_tnsr_high(int5 ndx, const int8_t tensor, uchar256 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i1_st_tnsr_high(int5 ndx, const int8_t tensor, bool256 value, int switches=0, bool predicate=1, bool polarity=0)

Parameters

ndx	Tensor coordinates (SRC1).
tensor	Tensor number.
value	Value to write (SRC3).
switches	Instruction switches.
predicate	Predicate value for the instruction.
polarity	True if polarity of the predicate is inverted.

Switches

Allowed switches are:

• SW_PACK - When set, store pipe will perform packing according to PACK_DT.

Store tensor low

Represents ST_TNSR_LOW instruction.

Intrinsics Prototype and Arguments

Intrinsics Definition

void v_f32_st_tnsr_low(int5 ndx, const int8_t tensor, float64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_bf16_st_tnsr_low(int5 ndx, const int8_t tensor, bfloat128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i32_st_tnsr_low(int5 ndx, const int8_t tensor, int64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u32_st_tnsr_low(int5 ndx, const int8_t tensor, uint64 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i16_st_tnsr_low(int5 ndx, const int8_t tensor, short128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u16_st_tnsr_low(int5 ndx, const int8_t tensor, ushort128 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i8_st_tnsr_low(int5 ndx, const int8_t tensor, char256 value, int switches=0, bool predicate=1, bool polarity=0)

void v_u8_st_tnsr_low(int5 ndx, const int8_t tensor, uchar256 value, int switches=0, bool predicate=1, bool polarity=0)

void v_i1_st_tnsr_low(int5 ndx, const int8_t tensor, bool256 value, int switches=0, bool predicate=1, bool polarity=0)

Parameters

ndx	Tensor coordinates (SRC1).
tensor	Tensor number.
value	Value to write (SRC3).
switches	Instruction switches.
predicate	Predicate value for the instruction.
polarity	True if polarity of the predicate is inverted.

Switches

Allowed switches are:

• SW_PACK - When set, store pipe will perform packing according to PACK_DT.