Module amdgpu

Available on target_arch=amdgpu only.

Expand description

amdgpu intrinsics

The reference is the LLVM amdgpu guide and the LLVM implementation. The order of intrinsics here follows the order in the LLVM implementation.

Functions§

llvm_ballot 🔒
llvm_dispatch_id 🔒
llvm_ds_bpermute 🔒 ^⚠
llvm_ds_permute 🔒 ^⚠
llvm_endpgm 🔒
llvm_groupstaticsize 🔒
llvm_inverse_ballot 🔒
llvm_mbcnt_hi 🔒
llvm_mbcnt_lo 🔒
llvm_perm 🔒 ^⚠
llvm_permlane16_swap 🔒 ^⚠
llvm_permlane16_u32 🔒 ^⚠
llvm_permlane16_var 🔒 ^⚠
llvm_permlane32_swap 🔒 ^⚠
llvm_permlane64_u32 🔒 ^⚠
llvm_permlanex16_u32 🔒 ^⚠
llvm_permlanex16_var 🔒 ^⚠
llvm_readfirstlane_u32 🔒
llvm_readfirstlane_u64 🔒
llvm_readlane_u32 🔒 ^⚠
llvm_readlane_u64 🔒 ^⚠
llvm_s_barrier 🔒
llvm_s_barrier_signal 🔒 ^⚠
llvm_s_barrier_signal_isfirst 🔒 ^⚠
llvm_s_barrier_wait 🔒 ^⚠
llvm_s_get_barrier_state 🔒 ^⚠
llvm_s_get_waveid_in_workgroup 🔒
llvm_s_getpc 🔒
llvm_s_memrealtime 🔒
llvm_s_sethalt 🔒
llvm_s_sleep 🔒
llvm_sched_barrier 🔒 ^⚠
llvm_sched_group_barrier 🔒 ^⚠
llvm_update_dpp 🔒 ^⚠
llvm_wave_barrier 🔒
llvm_wave_id 🔒
llvm_wave_reduce_add 🔒
llvm_wave_reduce_and 🔒
llvm_wave_reduce_max 🔒
llvm_wave_reduce_min 🔒
llvm_wave_reduce_or 🔒
llvm_wave_reduce_umax 🔒
llvm_wave_reduce_umin 🔒
llvm_wave_reduce_xor 🔒
llvm_wavefrontsize 🔒
llvm_workgroup_id_x 🔒
llvm_workgroup_id_y 🔒
llvm_workgroup_id_z 🔒
llvm_workitem_id_x 🔒
llvm_workitem_id_y 🔒
llvm_workitem_id_z 🔒
llvm_writelane_u32 🔒 ^⚠
llvm_writelane_u64 🔒 ^⚠
ballotExperimental: Returns a bitfield (u32 or u64) containing the result of its i1 argument in all active lanes, and zero in all inactive lanes.
dispatch_idExperimental: Returns the id of the dispatch that is currently executed.
ds_bpermute^⚠Experimental: Gather data across all lanes in a wavefront.
ds_permute^⚠Experimental: Scatter data across all lanes in a wavefront.
endpgmExperimental: Stop execution of the wavefront.
groupstaticsizeExperimental: Returns the size of statically allocated shared memory for this program in bytes.
inverse_ballotExperimental: Indexes into the value with the current lane id and returns for each lane if the corresponding bit is set.
mbcnt_hiExperimental: Masked bit count, high 32 lanes.
mbcnt_loExperimental: Masked bit count, low 32 lanes.
perm^⚠Experimental: Permute a 64-bit value.
permlane16_swap^⚠Experimental: Provide direct access to v_permlane16_swap_b32 instruction on supported targets.
permlane16_u32^⚠Experimental: Performs arbitrary gather-style operation within a row (16 contiguous lanes) of the second input operand.
permlane16_var^⚠Experimental: Performs arbitrary gather-style operation within a row (16 contiguous lanes) of the second input operand.
permlane32_swap^⚠Experimental: Provide direct access to v_permlane32_swap_b32 instruction on supported targets.
permlane64_u32^⚠Experimental: Swap value between upper and lower 32 lanes in a wavefront.
permlanex16_u32^⚠Experimental: Performs arbitrary gather-style operation across two rows (16 contiguous lanes) of the second input operand.
permlanex16_var^⚠Experimental: Performs arbitrary gather-style operation across two rows (16 contiguous lanes) of the second input operand.
readfirstlane_u32Experimental: Get value from the first active lane in the wavefront.
readfirstlane_u64Experimental: Get value from the first active lane in the wavefront.
readlane_u32^⚠Experimental: Get value from the lane at index lane in the wavefront.
readlane_u64^⚠Experimental: Get value from the lane at index lane in the wavefront.
s_barrierExperimental: Synchronize all wavefronts in a workgroup.
s_barrier_signal^⚠Experimental: Signal a specific barrier type.
s_barrier_signal_isfirst^⚠Experimental: Signal a specific barrier type.
s_barrier_wait^⚠Experimental: Wait for a specific barrier type.
s_get_barrier_state^⚠Experimental: Get the state of a specific barrier type.
s_get_waveid_in_workgroupExperimental: Get the index of the current wavefront in the workgroup.
s_getpcExperimental: Returns the current process counter.
s_memrealtimeExperimental: Measures time based on a fixed frequency.
s_sethaltExperimental: Stop execution of the kernel.
s_sleepExperimental: Sleeps for approximately COUNT * 64 cycles.
sched_barrier^⚠Experimental: Prevent movement of some instruction types.
sched_group_barrier^⚠Experimental: Creates schedule groups with specific properties to create custom scheduling pipelines.
update_dpp^⚠Experimental: The update_dpp intrinsic represents the update.dpp operation in AMDGPU. It takes an old value, a source operand, a DPP control operand, a row mask, a bank mask, and a bound control. This operation is equivalent to a sequence of v_mov_b32 operations.
wave_barrierExperimental: A barrier for only the threads within the current wavefront.
wave_idExperimental: Get the index of the current wavefront in the workgroup.
wave_reduce_addExperimental: Performs an arithmetic add reduction on the values provided by each lane in the wavefront.
wave_reduce_andExperimental: Performs a logical and reduction on the unsigned values provided by each lane in the wavefront.
wave_reduce_maxExperimental: Performs an arithmetic max reduction on the signed values provided by each lane in the wavefront.
wave_reduce_minExperimental: Performs an arithmetic min reduction on the signed values provided by each lane in the wavefront.
wave_reduce_orExperimental: Performs a logical or reduction on the unsigned values provided by each lane in the wavefront.
wave_reduce_umaxExperimental: Performs an arithmetic max reduction on the unsigned values provided by each lane in the wavefront.
wave_reduce_uminExperimental: Performs an arithmetic min reduction on the unsigned values provided by each lane in the wavefront.
wave_reduce_xorExperimental: Performs a logical xor reduction on the unsigned values provided by each lane in the wavefront.
wavefrontsizeExperimental: Returns the number of threads in a wavefront.
workgroup_id_xExperimental: Returns the x coordinate of the workgroup index within the dispatch.
workgroup_id_yExperimental: Returns the y coordinate of the workgroup index within the dispatch.
workgroup_id_zExperimental: Returns the z coordinate of the workgroup index within the dispatch.
workitem_id_xExperimental: Returns the x coordinate of the workitem index within the workgroup.
workitem_id_yExperimental: Returns the y coordinate of the workitem index within the workgroup.
workitem_id_zExperimental: Returns the z coordinate of the workitem index within the workgroup.
writelane_u32^⚠Experimental: Return value for the lane at index lane in the wavefront. Return default for all other lanes.
writelane_u64^⚠Experimental: Return value for the lane at index lane in the wavefront. Return default for all other lanes.

Module amdgpu

Module amdgpu Copy item path

Functions§

Module amdgpu