From 91ec6e5664979382f3cb18ba91edc4af99f93471 Mon Sep 17 00:00:00 2001 From: Tom Stellard Date: Wed, 20 Jul 2016 14:32:59 +0000 Subject: [PATCH] radeonsi/compute: Use the HSA abi for non-TGSI compute shaders v3 MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit This patch switches non-TGSI compute shaders over to using the HSA ABI described here: https://github.com/RadeonOpenCompute/ROCm-Docs/blob/master/AMDGPU-ABI.md The HSA ABI provides a much cleaner interface for compute shaders and allows us to share more code in the compiler with the HSA stack. The main changes in this patch are: - We now pass the scratch buffer resource into the shader via user sgprs rather than using relocations. - Grid/Block sizes are now passed to the shader via the dispatch packet rather than at the beginning of the kernel arguments. Typically for HSA, the CP firmware will create the dispatch packet and set up the user sgprs automatically. However, in Mesa we let the driver do this work. The main reason for this is that I haven't researched how to get the CP to do all these things, and I'm not sure if it is supported for all GPUs. v2: - Add comments explaining why we are setting certain bits of the scratch resource descriptor. v3: - Use amdgcn-mesa-mesa3d triple instead of amdgcn--mesa3d. Reviewed-by: Nicolai Hähnle --- src/amd/common/amd_kernel_code_t.h | 534 ++++++++++++++++++ src/gallium/drivers/radeon/r600_pipe_common.c | 6 +- src/gallium/drivers/radeonsi/si_compute.c | 239 +++++++- 3 files changed, 761 insertions(+), 18 deletions(-) create mode 100644 src/amd/common/amd_kernel_code_t.h diff --git a/src/amd/common/amd_kernel_code_t.h b/src/amd/common/amd_kernel_code_t.h new file mode 100644 index 00000000000..d0d7809da10 --- /dev/null +++ b/src/amd/common/amd_kernel_code_t.h @@ -0,0 +1,534 @@ +/* + * Copyright 2015,2016 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef AMDKERNELCODET_H +#define AMDKERNELCODET_H + +//---------------------------------------------------------------------------// +// AMD Kernel Code, and its dependencies // +//---------------------------------------------------------------------------// + +// Sets val bits for specified mask in specified dst packed instance. +#define AMD_HSA_BITS_SET(dst, mask, val) \ + dst &= (~(1 << mask ## _SHIFT) & ~mask); \ + dst |= (((val) << mask ## _SHIFT) & mask) + +// Gets bits for specified mask from specified src packed instance. +#define AMD_HSA_BITS_GET(src, mask) \ + ((src & mask) >> mask ## _SHIFT) \ + +/* Every amd_*_code_t has the following properties, which are composed of + * a number of bit fields. Every bit field has a mask (AMD_CODE_PROPERTY_*), + * bit width (AMD_CODE_PROPERTY_*_WIDTH, and bit shift amount + * (AMD_CODE_PROPERTY_*_SHIFT) for convenient access. Unused bits must be 0. + * + * (Note that bit fields cannot be used as their layout is + * implementation defined in the C standard and so cannot be used to + * specify an ABI) + */ +enum amd_code_property_mask_t { + + /* Enable the setup of the SGPR user data registers + * (AMD_CODE_PROPERTY_ENABLE_SGPR_*), see documentation of amd_kernel_code_t + * for initial register state. + * + * The total number of SGPRuser data registers requested must not + * exceed 16. Any requests beyond 16 will be ignored. + * + * Used to set COMPUTE_PGM_RSRC2.USER_SGPR (set to total count of + * SGPR user data registers enabled up to 16). + */ + + AMD_CODE_PROPERTY_ENABLE_SGPR_PRIVATE_SEGMENT_BUFFER_SHIFT = 0, + AMD_CODE_PROPERTY_ENABLE_SGPR_PRIVATE_SEGMENT_BUFFER_WIDTH = 1, + AMD_CODE_PROPERTY_ENABLE_SGPR_PRIVATE_SEGMENT_BUFFER = ((1 << AMD_CODE_PROPERTY_ENABLE_SGPR_PRIVATE_SEGMENT_BUFFER_WIDTH) - 1) << AMD_CODE_PROPERTY_ENABLE_SGPR_PRIVATE_SEGMENT_BUFFER_SHIFT, + + AMD_CODE_PROPERTY_ENABLE_SGPR_DISPATCH_PTR_SHIFT = 1, + AMD_CODE_PROPERTY_ENABLE_SGPR_DISPATCH_PTR_WIDTH = 1, + AMD_CODE_PROPERTY_ENABLE_SGPR_DISPATCH_PTR = ((1 << AMD_CODE_PROPERTY_ENABLE_SGPR_DISPATCH_PTR_WIDTH) - 1) << AMD_CODE_PROPERTY_ENABLE_SGPR_DISPATCH_PTR_SHIFT, + + AMD_CODE_PROPERTY_ENABLE_SGPR_QUEUE_PTR_SHIFT = 2, + AMD_CODE_PROPERTY_ENABLE_SGPR_QUEUE_PTR_WIDTH = 1, + AMD_CODE_PROPERTY_ENABLE_SGPR_QUEUE_PTR = ((1 << AMD_CODE_PROPERTY_ENABLE_SGPR_QUEUE_PTR_WIDTH) - 1) << AMD_CODE_PROPERTY_ENABLE_SGPR_QUEUE_PTR_SHIFT, + + AMD_CODE_PROPERTY_ENABLE_SGPR_KERNARG_SEGMENT_PTR_SHIFT = 3, + AMD_CODE_PROPERTY_ENABLE_SGPR_KERNARG_SEGMENT_PTR_WIDTH = 1, + AMD_CODE_PROPERTY_ENABLE_SGPR_KERNARG_SEGMENT_PTR = ((1 << AMD_CODE_PROPERTY_ENABLE_SGPR_KERNARG_SEGMENT_PTR_WIDTH) - 1) << AMD_CODE_PROPERTY_ENABLE_SGPR_KERNARG_SEGMENT_PTR_SHIFT, + + AMD_CODE_PROPERTY_ENABLE_SGPR_DISPATCH_ID_SHIFT = 4, + AMD_CODE_PROPERTY_ENABLE_SGPR_DISPATCH_ID_WIDTH = 1, + AMD_CODE_PROPERTY_ENABLE_SGPR_DISPATCH_ID = ((1 << AMD_CODE_PROPERTY_ENABLE_SGPR_DISPATCH_ID_WIDTH) - 1) << AMD_CODE_PROPERTY_ENABLE_SGPR_DISPATCH_ID_SHIFT, + + AMD_CODE_PROPERTY_ENABLE_SGPR_FLAT_SCRATCH_INIT_SHIFT = 5, + AMD_CODE_PROPERTY_ENABLE_SGPR_FLAT_SCRATCH_INIT_WIDTH = 1, + AMD_CODE_PROPERTY_ENABLE_SGPR_FLAT_SCRATCH_INIT = ((1 << AMD_CODE_PROPERTY_ENABLE_SGPR_FLAT_SCRATCH_INIT_WIDTH) - 1) << AMD_CODE_PROPERTY_ENABLE_SGPR_FLAT_SCRATCH_INIT_SHIFT, + + AMD_CODE_PROPERTY_ENABLE_SGPR_PRIVATE_SEGMENT_SIZE_SHIFT = 6, + AMD_CODE_PROPERTY_ENABLE_SGPR_PRIVATE_SEGMENT_SIZE_WIDTH = 1, + AMD_CODE_PROPERTY_ENABLE_SGPR_PRIVATE_SEGMENT_SIZE = ((1 << AMD_CODE_PROPERTY_ENABLE_SGPR_PRIVATE_SEGMENT_SIZE_WIDTH) - 1) << AMD_CODE_PROPERTY_ENABLE_SGPR_PRIVATE_SEGMENT_SIZE_SHIFT, + + AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_X_SHIFT = 7, + AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_X_WIDTH = 1, + AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_X = ((1 << AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_X_WIDTH) - 1) << AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_X_SHIFT, + + AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_Y_SHIFT = 8, + AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_Y_WIDTH = 1, + AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_Y = ((1 << AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_Y_WIDTH) - 1) << AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_Y_SHIFT, + + AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_Z_SHIFT = 9, + AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_Z_WIDTH = 1, + AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_Z = ((1 << AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_Z_WIDTH) - 1) << AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_Z_SHIFT, + + AMD_CODE_PROPERTY_RESERVED1_SHIFT = 10, + AMD_CODE_PROPERTY_RESERVED1_WIDTH = 6, + AMD_CODE_PROPERTY_RESERVED1 = ((1 << AMD_CODE_PROPERTY_RESERVED1_WIDTH) - 1) << AMD_CODE_PROPERTY_RESERVED1_SHIFT, + + /* Control wave ID base counter for GDS ordered-append. Used to set + * COMPUTE_DISPATCH_INITIATOR.ORDERED_APPEND_ENBL. (Not sure if + * ORDERED_APPEND_MODE also needs to be settable) + */ + AMD_CODE_PROPERTY_ENABLE_ORDERED_APPEND_GDS_SHIFT = 16, + AMD_CODE_PROPERTY_ENABLE_ORDERED_APPEND_GDS_WIDTH = 1, + AMD_CODE_PROPERTY_ENABLE_ORDERED_APPEND_GDS = ((1 << AMD_CODE_PROPERTY_ENABLE_ORDERED_APPEND_GDS_WIDTH) - 1) << AMD_CODE_PROPERTY_ENABLE_ORDERED_APPEND_GDS_SHIFT, + + /* The interleave (swizzle) element size in bytes required by the + * code for private memory. This must be 2, 4, 8 or 16. This value + * is provided to the finalizer when it is invoked and is recorded + * here. The hardware will interleave the memory requests of each + * lane of a wavefront by this element size to ensure each + * work-item gets a distinct memory memory location. Therefore, the + * finalizer ensures that all load and store operations done to + * private memory do not exceed this size. For example, if the + * element size is 4 (32-bits or dword) and a 64-bit value must be + * loaded, the finalizer will generate two 32-bit loads. This + * ensures that the interleaving will get the work-item + * specific dword for both halves of the 64-bit value. If it just + * did a 64-bit load then it would get one dword which belonged to + * its own work-item, but the second dword would belong to the + * adjacent lane work-item since the interleaving is in dwords. + * + * The value used must match the value that the runtime configures + * the GPU flat scratch (SH_STATIC_MEM_CONFIG.ELEMENT_SIZE). This + * is generally DWORD. + * + * USE VALUES FROM THE AMD_ELEMENT_BYTE_SIZE_T ENUM. + */ + AMD_CODE_PROPERTY_PRIVATE_ELEMENT_SIZE_SHIFT = 17, + AMD_CODE_PROPERTY_PRIVATE_ELEMENT_SIZE_WIDTH = 2, + AMD_CODE_PROPERTY_PRIVATE_ELEMENT_SIZE = ((1 << AMD_CODE_PROPERTY_PRIVATE_ELEMENT_SIZE_WIDTH) - 1) << AMD_CODE_PROPERTY_PRIVATE_ELEMENT_SIZE_SHIFT, + + /* Are global memory addresses 64 bits. Must match + * amd_kernel_code_t.hsail_machine_model == + * HSA_MACHINE_LARGE. Must also match + * SH_MEM_CONFIG.PTR32 (GFX6 (SI)/GFX7 (CI)), + * SH_MEM_CONFIG.ADDRESS_MODE (GFX8 (VI)+). + */ + AMD_CODE_PROPERTY_IS_PTR64_SHIFT = 19, + AMD_CODE_PROPERTY_IS_PTR64_WIDTH = 1, + AMD_CODE_PROPERTY_IS_PTR64 = ((1 << AMD_CODE_PROPERTY_IS_PTR64_WIDTH) - 1) << AMD_CODE_PROPERTY_IS_PTR64_SHIFT, + + /* Indicate if the generated ISA is using a dynamically sized call + * stack. This can happen if calls are implemented using a call + * stack and recursion, alloca or calls to indirect functions are + * present. In these cases the Finalizer cannot compute the total + * private segment size at compile time. In this case the + * workitem_private_segment_byte_size only specifies the statically + * know private segment size, and additional space must be added + * for the call stack. + */ + AMD_CODE_PROPERTY_IS_DYNAMIC_CALLSTACK_SHIFT = 20, + AMD_CODE_PROPERTY_IS_DYNAMIC_CALLSTACK_WIDTH = 1, + AMD_CODE_PROPERTY_IS_DYNAMIC_CALLSTACK = ((1 << AMD_CODE_PROPERTY_IS_DYNAMIC_CALLSTACK_WIDTH) - 1) << AMD_CODE_PROPERTY_IS_DYNAMIC_CALLSTACK_SHIFT, + + /* Indicate if code generated has support for debugging. */ + AMD_CODE_PROPERTY_IS_DEBUG_SUPPORTED_SHIFT = 21, + AMD_CODE_PROPERTY_IS_DEBUG_SUPPORTED_WIDTH = 1, + AMD_CODE_PROPERTY_IS_DEBUG_SUPPORTED = ((1 << AMD_CODE_PROPERTY_IS_DEBUG_SUPPORTED_WIDTH) - 1) << AMD_CODE_PROPERTY_IS_DEBUG_SUPPORTED_SHIFT, + + AMD_CODE_PROPERTY_IS_XNACK_SUPPORTED_SHIFT = 22, + AMD_CODE_PROPERTY_IS_XNACK_SUPPORTED_WIDTH = 1, + AMD_CODE_PROPERTY_IS_XNACK_SUPPORTED = ((1 << AMD_CODE_PROPERTY_IS_XNACK_SUPPORTED_WIDTH) - 1) << AMD_CODE_PROPERTY_IS_XNACK_SUPPORTED_SHIFT, + + AMD_CODE_PROPERTY_RESERVED2_SHIFT = 23, + AMD_CODE_PROPERTY_RESERVED2_WIDTH = 9, + AMD_CODE_PROPERTY_RESERVED2 = ((1 << AMD_CODE_PROPERTY_RESERVED2_WIDTH) - 1) << AMD_CODE_PROPERTY_RESERVED2_SHIFT +}; + +/* AMD Kernel Code Object (amd_kernel_code_t). GPU CP uses the AMD Kernel + * Code Object to set up the hardware to execute the kernel dispatch. + * + * Initial Kernel Register State. + * + * Initial kernel register state will be set up by CP/SPI prior to the start + * of execution of every wavefront. This is limited by the constraints of the + * current hardware. + * + * The order of the SGPR registers is defined, but the Finalizer can specify + * which ones are actually setup in the amd_kernel_code_t object using the + * enable_sgpr_* bit fields. The register numbers used for enabled registers + * are dense starting at SGPR0: the first enabled register is SGPR0, the next + * enabled register is SGPR1 etc.; disabled registers do not have an SGPR + * number. + * + * The initial SGPRs comprise up to 16 User SRGPs that are set up by CP and + * apply to all waves of the grid. It is possible to specify more than 16 User + * SGPRs using the enable_sgpr_* bit fields, in which case only the first 16 + * are actually initialized. These are then immediately followed by the System + * SGPRs that are set up by ADC/SPI and can have different values for each wave + * of the grid dispatch. + * + * SGPR register initial state is defined as follows: + * + * Private Segment Buffer (enable_sgpr_private_segment_buffer): + * Number of User SGPR registers: 4. V# that can be used, together with + * Scratch Wave Offset as an offset, to access the Private/Spill/Arg + * segments using a segment address. It must be set as follows: + * - Base address: of the scratch memory area used by the dispatch. It + * does not include the scratch wave offset. It will be the per process + * SH_HIDDEN_PRIVATE_BASE_VMID plus any offset from this dispatch (for + * example there may be a per pipe offset, or per AQL Queue offset). + * - Stride + data_format: Element Size * Index Stride (???) + * - Cache swizzle: ??? + * - Swizzle enable: SH_STATIC_MEM_CONFIG.SWIZZLE_ENABLE (must be 1 for + * scratch) + * - Num records: Flat Scratch Work Item Size / Element Size (???) + * - Dst_sel_*: ??? + * - Num_format: ??? + * - Element_size: SH_STATIC_MEM_CONFIG.ELEMENT_SIZE (will be DWORD, must + * agree with amd_kernel_code_t.privateElementSize) + * - Index_stride: SH_STATIC_MEM_CONFIG.INDEX_STRIDE (will be 64 as must + * be number of wavefront lanes for scratch, must agree with + * amd_kernel_code_t.wavefrontSize) + * - Add tid enable: 1 + * - ATC: from SH_MEM_CONFIG.PRIVATE_ATC, + * - Hash_enable: ??? + * - Heap: ??? + * - Mtype: from SH_STATIC_MEM_CONFIG.PRIVATE_MTYPE + * - Type: 0 (a buffer) (???) + * + * Dispatch Ptr (enable_sgpr_dispatch_ptr): + * Number of User SGPR registers: 2. 64 bit address of AQL dispatch packet + * for kernel actually executing. + * + * Queue Ptr (enable_sgpr_queue_ptr): + * Number of User SGPR registers: 2. 64 bit address of AmdQueue object for + * AQL queue on which the dispatch packet was queued. + * + * Kernarg Segment Ptr (enable_sgpr_kernarg_segment_ptr): + * Number of User SGPR registers: 2. 64 bit address of Kernarg segment. This + * is directly copied from the kernargPtr in the dispatch packet. Having CP + * load it once avoids loading it at the beginning of every wavefront. + * + * Dispatch Id (enable_sgpr_dispatch_id): + * Number of User SGPR registers: 2. 64 bit Dispatch ID of the dispatch + * packet being executed. + * + * Flat Scratch Init (enable_sgpr_flat_scratch_init): + * Number of User SGPR registers: 2. This is 2 SGPRs. + * + * For CI/VI: + * The first SGPR is a 32 bit byte offset from SH_MEM_HIDDEN_PRIVATE_BASE + * to base of memory for scratch for this dispatch. This is the same offset + * used in computing the Scratch Segment Buffer base address. The value of + * Scratch Wave Offset must be added by the kernel code and moved to + * SGPRn-4 for use as the FLAT SCRATCH BASE in flat memory instructions. + * + * The second SGPR is 32 bit byte size of a single work-item's scratch + * memory usage. This is directly loaded from the dispatch packet Private + * Segment Byte Size and rounded up to a multiple of DWORD. + * + * \todo [Does CP need to round this to >4 byte alignment?] + * + * The kernel code must move to SGPRn-3 for use as the FLAT SCRATCH SIZE in + * flat memory instructions. Having CP load it once avoids loading it at + * the beginning of every wavefront. + * + * Private Segment Size (enable_sgpr_private_segment_size): + * Number of User SGPR registers: 1. The 32 bit byte size of a single + * work-item's scratch memory allocation. This is the value from the dispatch + * packet. Private Segment Byte Size rounded up by CP to a multiple of DWORD. + * + * \todo [Does CP need to round this to >4 byte alignment?] + * + * Having CP load it once avoids loading it at the beginning of every + * wavefront. + * + * \todo [This will not be used for CI/VI since it is the same value as + * the second SGPR of Flat Scratch Init. + * + * Grid Work-Group Count X (enable_sgpr_grid_workgroup_count_x): + * Number of User SGPR registers: 1. 32 bit count of the number of + * work-groups in the X dimension for the grid being executed. Computed from + * the fields in the HsaDispatchPacket as + * ((gridSize.x+workgroupSize.x-1)/workgroupSize.x). + * + * Grid Work-Group Count Y (enable_sgpr_grid_workgroup_count_y): + * Number of User SGPR registers: 1. 32 bit count of the number of + * work-groups in the Y dimension for the grid being executed. Computed from + * the fields in the HsaDispatchPacket as + * ((gridSize.y+workgroupSize.y-1)/workgroupSize.y). + * + * Only initialized if <16 previous SGPRs initialized. + * + * Grid Work-Group Count Z (enable_sgpr_grid_workgroup_count_z): + * Number of User SGPR registers: 1. 32 bit count of the number of + * work-groups in the Z dimension for the grid being executed. Computed + * from the fields in the HsaDispatchPacket as + * ((gridSize.z+workgroupSize.z-1)/workgroupSize.z). + * + * Only initialized if <16 previous SGPRs initialized. + * + * Work-Group Id X (enable_sgpr_workgroup_id_x): + * Number of System SGPR registers: 1. 32 bit work group id in X dimension + * of grid for wavefront. Always present. + * + * Work-Group Id Y (enable_sgpr_workgroup_id_y): + * Number of System SGPR registers: 1. 32 bit work group id in Y dimension + * of grid for wavefront. + * + * Work-Group Id Z (enable_sgpr_workgroup_id_z): + * Number of System SGPR registers: 1. 32 bit work group id in Z dimension + * of grid for wavefront. If present then Work-group Id Y will also be + * present + * + * Work-Group Info (enable_sgpr_workgroup_info): + * Number of System SGPR registers: 1. {first_wave, 14'b0000, + * ordered_append_term[10:0], threadgroup_size_in_waves[5:0]} + * + * Private Segment Wave Byte Offset + * (enable_sgpr_private_segment_wave_byte_offset): + * Number of System SGPR registers: 1. 32 bit byte offset from base of + * dispatch scratch base. Must be used as an offset with Private/Spill/Arg + * segment address when using Scratch Segment Buffer. It must be added to + * Flat Scratch Offset if setting up FLAT SCRATCH for flat addressing. + * + * + * The order of the VGPR registers is defined, but the Finalizer can specify + * which ones are actually setup in the amd_kernel_code_t object using the + * enableVgpr* bit fields. The register numbers used for enabled registers + * are dense starting at VGPR0: the first enabled register is VGPR0, the next + * enabled register is VGPR1 etc.; disabled registers do not have an VGPR + * number. + * + * VGPR register initial state is defined as follows: + * + * Work-Item Id X (always initialized): + * Number of registers: 1. 32 bit work item id in X dimension of work-group + * for wavefront lane. + * + * Work-Item Id X (enable_vgpr_workitem_id > 0): + * Number of registers: 1. 32 bit work item id in Y dimension of work-group + * for wavefront lane. + * + * Work-Item Id X (enable_vgpr_workitem_id > 0): + * Number of registers: 1. 32 bit work item id in Z dimension of work-group + * for wavefront lane. + * + * + * The setting of registers is being done by existing GPU hardware as follows: + * 1) SGPRs before the Work-Group Ids are set by CP using the 16 User Data + * registers. + * 2) Work-group Id registers X, Y, Z are set by SPI which supports any + * combination including none. + * 3) Scratch Wave Offset is also set by SPI which is why its value cannot + * be added into the value Flat Scratch Offset which would avoid the + * Finalizer generated prolog having to do the add. + * 4) The VGPRs are set by SPI which only supports specifying either (X), + * (X, Y) or (X, Y, Z). + * + * Flat Scratch Dispatch Offset and Flat Scratch Size are adjacent SGRRs so + * they can be moved as a 64 bit value to the hardware required SGPRn-3 and + * SGPRn-4 respectively using the Finalizer ?FLAT_SCRATCH? Register. + * + * The global segment can be accessed either using flat operations or buffer + * operations. If buffer operations are used then the Global Buffer used to + * access HSAIL Global/Readonly/Kernarg (which are combine) segments using a + * segment address is not passed into the kernel code by CP since its base + * address is always 0. Instead the Finalizer generates prolog code to + * initialize 4 SGPRs with a V# that has the following properties, and then + * uses that in the buffer instructions: + * - base address of 0 + * - no swizzle + * - ATC=1 + * - MTYPE set to support memory coherence specified in + * amd_kernel_code_t.globalMemoryCoherence + * + * When the Global Buffer is used to access the Kernarg segment, must add the + * dispatch packet kernArgPtr to a kernarg segment address before using this V#. + * Alternatively scalar loads can be used if the kernarg offset is uniform, as + * the kernarg segment is constant for the duration of the kernel execution. + */ + +typedef struct amd_kernel_code_s { + uint32_t amd_kernel_code_version_major; + uint32_t amd_kernel_code_version_minor; + uint16_t amd_machine_kind; + uint16_t amd_machine_version_major; + uint16_t amd_machine_version_minor; + uint16_t amd_machine_version_stepping; + + /* Byte offset (possibly negative) from start of amd_kernel_code_t + * object to kernel's entry point instruction. The actual code for + * the kernel is required to be 256 byte aligned to match hardware + * requirements (SQ cache line is 16). The code must be position + * independent code (PIC) for AMD devices to give runtime the + * option of copying code to discrete GPU memory or APU L2 + * cache. The Finalizer should endeavour to allocate all kernel + * machine code in contiguous memory pages so that a device + * pre-fetcher will tend to only pre-fetch Kernel Code objects, + * improving cache performance. + */ + int64_t kernel_code_entry_byte_offset; + + /* Range of bytes to consider prefetching expressed as an offset + * and size. The offset is from the start (possibly negative) of + * amd_kernel_code_t object. Set both to 0 if no prefetch + * information is available. + */ + int64_t kernel_code_prefetch_byte_offset; + uint64_t kernel_code_prefetch_byte_size; + + /* Number of bytes of scratch backing memory required for full + * occupancy of target chip. This takes into account the number of + * bytes of scratch per work-item, the wavefront size, the maximum + * number of wavefronts per CU, and the number of CUs. This is an + * upper limit on scratch. If the grid being dispatched is small it + * may only need less than this. If the kernel uses no scratch, or + * the Finalizer has not computed this value, it must be 0. + */ + uint64_t max_scratch_backing_memory_byte_size; + + /* Shader program settings for CS. Contains COMPUTE_PGM_RSRC1 and + * COMPUTE_PGM_RSRC2 registers. + */ + uint64_t compute_pgm_resource_registers; + + /* Code properties. See amd_code_property_mask_t for a full list of + * properties. + */ + uint32_t code_properties; + + /* The amount of memory required for the combined private, spill + * and arg segments for a work-item in bytes. If + * is_dynamic_callstack is 1 then additional space must be added to + * this value for the call stack. + */ + uint32_t workitem_private_segment_byte_size; + + /* The amount of group segment memory required by a work-group in + * bytes. This does not include any dynamically allocated group + * segment memory that may be added when the kernel is + * dispatched. + */ + uint32_t workgroup_group_segment_byte_size; + + /* Number of byte of GDS required by kernel dispatch. Must be 0 if + * not using GDS. + */ + uint32_t gds_segment_byte_size; + + /* The size in bytes of the kernarg segment that holds the values + * of the arguments to the kernel. This could be used by CP to + * prefetch the kernarg segment pointed to by the dispatch packet. + */ + uint64_t kernarg_segment_byte_size; + + /* Number of fbarrier's used in the kernel and all functions it + * calls. If the implementation uses group memory to allocate the + * fbarriers then that amount must already be included in the + * workgroup_group_segment_byte_size total. + */ + uint32_t workgroup_fbarrier_count; + + /* Number of scalar registers used by a wavefront. This includes + * the special SGPRs for VCC, Flat Scratch Base, Flat Scratch Size + * and XNACK (for GFX8 (VI)). It does not include the 16 SGPR added if a + * trap handler is enabled. Used to set COMPUTE_PGM_RSRC1.SGPRS. + */ + uint16_t wavefront_sgpr_count; + + /* Number of vector registers used by each work-item. Used to set + * COMPUTE_PGM_RSRC1.VGPRS. + */ + uint16_t workitem_vgpr_count; + + /* If reserved_vgpr_count is 0 then must be 0. Otherwise, this is the + * first fixed VGPR number reserved. + */ + uint16_t reserved_vgpr_first; + + /* The number of consecutive VGPRs reserved by the client. If + * is_debug_supported then this count includes VGPRs reserved + * for debugger use. + */ + uint16_t reserved_vgpr_count; + + /* If reserved_sgpr_count is 0 then must be 0. Otherwise, this is the + * first fixed SGPR number reserved. + */ + uint16_t reserved_sgpr_first; + + /* The number of consecutive SGPRs reserved by the client. If + * is_debug_supported then this count includes SGPRs reserved + * for debugger use. + */ + uint16_t reserved_sgpr_count; + + /* If is_debug_supported is 0 then must be 0. Otherwise, this is the + * fixed SGPR number used to hold the wave scratch offset for the + * entire kernel execution, or uint16_t(-1) if the register is not + * used or not known. + */ + uint16_t debug_wavefront_private_segment_offset_sgpr; + + /* If is_debug_supported is 0 then must be 0. Otherwise, this is the + * fixed SGPR number of the first of 4 SGPRs used to hold the + * scratch V# used for the entire kernel execution, or uint16_t(-1) + * if the registers are not used or not known. + */ + uint16_t debug_private_segment_buffer_sgpr; + + /* The maximum byte alignment of variables used by the kernel in + * the specified memory segment. Expressed as a power of two. Must + * be at least HSA_POWERTWO_16. + */ + uint8_t kernarg_segment_alignment; + uint8_t group_segment_alignment; + uint8_t private_segment_alignment; + + /* Wavefront size expressed as a power of two. Must be a power of 2 + * in range 1..64 inclusive. Used to support runtime query that + * obtains wavefront size, which may be used by application to + * allocated dynamic group memory and set the dispatch work-group + * size. + */ + uint8_t wavefront_size; + + int32_t call_convention; + uint8_t reserved3[12]; + uint64_t runtime_loader_kernel_symbol; + uint64_t control_directives[16]; +} amd_kernel_code_t; + +#endif // AMDKERNELCODET_H diff --git a/src/gallium/drivers/radeon/r600_pipe_common.c b/src/gallium/drivers/radeon/r600_pipe_common.c index f0fdc9b904c..b0d981331c8 100644 --- a/src/gallium/drivers/radeon/r600_pipe_common.c +++ b/src/gallium/drivers/radeon/r600_pipe_common.c @@ -822,7 +822,11 @@ static int r600_get_compute_param(struct pipe_screen *screen, if (rscreen->family <= CHIP_ARUBA) { triple = "r600--"; } else { - triple = "amdgcn--"; + if (HAVE_LLVM < 0x0400) { + triple = "amdgcn--"; + } else { + triple = "amdgcn-mesa-mesa3d"; + } } switch(rscreen->family) { /* Clang < 3.6 is missing Hainan in its list of diff --git a/src/gallium/drivers/radeonsi/si_compute.c b/src/gallium/drivers/radeonsi/si_compute.c index 56b511848c9..ad9cb7ac1fc 100644 --- a/src/gallium/drivers/radeonsi/si_compute.c +++ b/src/gallium/drivers/radeonsi/si_compute.c @@ -28,6 +28,7 @@ #include "radeon/r600_pipe_common.h" #include "radeon/radeon_elf_util.h" +#include "amd_kernel_code_t.h" #include "radeon/r600_cs.h" #include "si_pipe.h" #include "si_shader.h" @@ -43,8 +44,52 @@ struct si_compute { struct si_shader shader; struct pipe_resource *global_buffers[MAX_GLOBAL_BUFFERS]; + bool use_code_object_v2; }; +struct dispatch_packet { + uint16_t header; + uint16_t setup; + uint16_t workgroup_size_x; + uint16_t workgroup_size_y; + uint16_t workgroup_size_z; + uint16_t reserved0; + uint32_t grid_size_x; + uint32_t grid_size_y; + uint32_t grid_size_z; + uint32_t private_segment_size; + uint32_t group_segment_size; + uint64_t kernel_object; + uint64_t kernarg_address; + uint64_t reserved2; +}; + +static const amd_kernel_code_t *si_compute_get_code_object( + const struct si_compute *program, + uint64_t symbol_offset) +{ + if (!program->use_code_object_v2) { + return NULL; + } + return (const amd_kernel_code_t*) + (program->shader.binary.code + symbol_offset); +} + +static void code_object_to_config(const amd_kernel_code_t *code_object, + struct si_shader_config *out_config) { + + uint32_t rsrc1 = code_object->compute_pgm_resource_registers; + uint32_t rsrc2 = code_object->compute_pgm_resource_registers >> 32; + out_config->num_sgprs = code_object->wavefront_sgpr_count; + out_config->num_vgprs = code_object->workitem_vgpr_count; + out_config->float_mode = G_00B028_FLOAT_MODE(rsrc1); + out_config->rsrc1 = rsrc1; + out_config->lds_size = MAX2(out_config->lds_size, G_00B84C_LDS_SIZE(rsrc2)); + out_config->rsrc2 = rsrc2; + out_config->scratch_bytes_per_wave = + align(code_object->workitem_private_segment_byte_size * 64, 1024); +} + static void *si_create_compute_state( struct pipe_context *ctx, const struct pipe_compute_state *cso) @@ -59,6 +104,8 @@ static void *si_create_compute_state( program->local_size = cso->req_local_mem; program->private_size = cso->req_private_mem; program->input_size = cso->req_input_mem; + program->use_code_object_v2 = HAVE_LLVM >= 0x0400 && + cso->ir_type == PIPE_SHADER_IR_NATIVE; if (cso->ir_type == PIPE_SHADER_IR_TGSI) { @@ -110,8 +157,14 @@ static void *si_create_compute_state( code = cso->prog + sizeof(struct pipe_llvm_program_header); radeon_elf_read(code, header->num_bytes, &program->shader.binary); - si_shader_binary_read_config(&program->shader.binary, - &program->shader.config, 0); + if (program->use_code_object_v2) { + const amd_kernel_code_t *code_object = + si_compute_get_code_object(program, 0); + code_object_to_config(code_object, &program->shader.config); + } else { + si_shader_binary_read_config(&program->shader.binary, + &program->shader.config, 0); + } si_shader_dump(sctx->screen, &program->shader, &sctx->b.debug, PIPE_SHADER_COMPUTE, stderr); si_shader_binary_upload(sctx->screen, &program->shader); @@ -233,7 +286,9 @@ static bool si_setup_compute_scratch_buffer(struct si_context *sctx, static bool si_switch_compute_shader(struct si_context *sctx, struct si_compute *program, - struct si_shader *shader, unsigned offset) + struct si_shader *shader, + const amd_kernel_code_t *code_object, + unsigned offset) { struct radeon_winsys_cs *cs = sctx->b.gfx.cs; struct si_shader_config inline_config = {0}; @@ -250,7 +305,11 @@ static bool si_switch_compute_shader(struct si_context *sctx, unsigned lds_blocks; config = &inline_config; - si_shader_binary_read_config(&shader->binary, config, offset); + if (code_object) { + code_object_to_config(code_object, config); + } else { + si_shader_binary_read_config(&shader->binary, config, offset); + } lds_blocks = config->lds_size; /* XXX: We are over allocating LDS. For SI, the shader reports @@ -286,6 +345,11 @@ static bool si_switch_compute_shader(struct si_context *sctx, } shader_va = shader->bo->gpu_address + offset; + if (program->use_code_object_v2) { + /* Shader code is placed after the amd_kernel_code_t + * struct. */ + shader_va += sizeof(amd_kernel_code_t); + } radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx, shader->bo, RADEON_USAGE_READ, RADEON_PRIO_SHADER_BINARY); @@ -313,14 +377,145 @@ static bool si_switch_compute_shader(struct si_context *sctx, return true; } +static void setup_scratch_rsrc_user_sgprs(struct si_context *sctx, + const amd_kernel_code_t *code_object, + unsigned user_sgpr) +{ + struct radeon_winsys_cs *cs = sctx->b.gfx.cs; + uint64_t scratch_va = sctx->compute_scratch_buffer->gpu_address; + + unsigned max_private_element_size = AMD_HSA_BITS_GET( + code_object->code_properties, + AMD_CODE_PROPERTY_PRIVATE_ELEMENT_SIZE); + + uint32_t scratch_dword0 = scratch_va & 0xffffffff; + uint32_t scratch_dword1 = + S_008F04_BASE_ADDRESS_HI(scratch_va >> 32) | + S_008F04_SWIZZLE_ENABLE(1); + + /* Disable address clamping */ + uint32_t scratch_dword2 = 0xffffffff; + uint32_t scratch_dword3 = + S_008F0C_ELEMENT_SIZE(max_private_element_size) | + S_008F0C_INDEX_STRIDE(3) | + S_008F0C_ADD_TID_ENABLE(1); + + + if (sctx->screen->b.chip_class < VI) { + /* BUF_DATA_FORMAT is ignored, but it cannot be + BUF_DATA_FORMAT_INVALID. */ + scratch_dword3 |= + S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_8); + } + + radeon_set_sh_reg_seq(cs, R_00B900_COMPUTE_USER_DATA_0 + + (user_sgpr * 4), 4); + radeon_emit(cs, scratch_dword0); + radeon_emit(cs, scratch_dword1); + radeon_emit(cs, scratch_dword2); + radeon_emit(cs, scratch_dword3); +} + +static void si_setup_user_sgprs_co_v2(struct si_context *sctx, + const amd_kernel_code_t *code_object, + const struct pipe_grid_info *info, + uint64_t kernel_args_va) +{ + struct si_compute *program = sctx->cs_shader_state.program; + struct radeon_winsys_cs *cs = sctx->b.gfx.cs; + + static const enum amd_code_property_mask_t workgroup_count_masks [] = { + AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_X, + AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_Y, + AMD_CODE_PROPERTY_ENABLE_SGPR_GRID_WORKGROUP_COUNT_Z + }; + + unsigned i, user_sgpr = 0; + if (AMD_HSA_BITS_GET(code_object->code_properties, + AMD_CODE_PROPERTY_ENABLE_SGPR_PRIVATE_SEGMENT_BUFFER)) { + if (code_object->workitem_private_segment_byte_size > 0) { + setup_scratch_rsrc_user_sgprs(sctx, code_object, + user_sgpr); + } + user_sgpr += 4; + } + + if (AMD_HSA_BITS_GET(code_object->code_properties, + AMD_CODE_PROPERTY_ENABLE_SGPR_DISPATCH_PTR)) { + struct dispatch_packet dispatch; + unsigned dispatch_offset; + struct r600_resource *dispatch_buf = NULL; + uint64_t dispatch_va; + + /* Upload dispatch ptr */ + memset(&dispatch, 0, sizeof(dispatch)); + + dispatch.workgroup_size_x = info->block[0]; + dispatch.workgroup_size_y = info->block[1]; + dispatch.workgroup_size_z = info->block[2]; + + dispatch.grid_size_x = info->grid[0] * info->block[0]; + dispatch.grid_size_y = info->grid[1] * info->block[1]; + dispatch.grid_size_z = info->grid[2] * info->block[2]; + + dispatch.private_segment_size = program->private_size; + dispatch.group_segment_size = program->local_size; + + dispatch.kernarg_address = kernel_args_va; + + u_upload_data(sctx->b.uploader, 0, sizeof(dispatch), 256, + &dispatch, &dispatch_offset, + (struct pipe_resource**)&dispatch_buf); + + if (!dispatch_buf) { + fprintf(stderr, "Error: Failed to allocate dispatch " + "packet."); + } + radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx, dispatch_buf, + RADEON_USAGE_READ, RADEON_PRIO_CONST_BUFFER); + + dispatch_va = dispatch_buf->gpu_address + dispatch_offset; + + radeon_set_sh_reg_seq(cs, R_00B900_COMPUTE_USER_DATA_0 + + (user_sgpr * 4), 2); + radeon_emit(cs, dispatch_va); + radeon_emit(cs, S_008F04_BASE_ADDRESS_HI(dispatch_va >> 32) | + S_008F04_STRIDE(0)); + + r600_resource_reference(&dispatch_buf, NULL); + user_sgpr += 2; + } + + if (AMD_HSA_BITS_GET(code_object->code_properties, + AMD_CODE_PROPERTY_ENABLE_SGPR_KERNARG_SEGMENT_PTR)) { + radeon_set_sh_reg_seq(cs, R_00B900_COMPUTE_USER_DATA_0 + + (user_sgpr * 4), 2); + radeon_emit(cs, kernel_args_va); + radeon_emit(cs, S_008F04_BASE_ADDRESS_HI (kernel_args_va >> 32) | + S_008F04_STRIDE(0)); + user_sgpr += 2; + } + + for (i = 0; i < 3 && user_sgpr < 16; i++) { + if (code_object->code_properties & workgroup_count_masks[i]) { + radeon_set_sh_reg_seq(cs, + R_00B900_COMPUTE_USER_DATA_0 + + (user_sgpr * 4), 1); + radeon_emit(cs, info->grid[i]); + user_sgpr += 1; + } + } +} + static void si_upload_compute_input(struct si_context *sctx, - const struct pipe_grid_info *info) + const amd_kernel_code_t *code_object, + const struct pipe_grid_info *info) { struct radeon_winsys_cs *cs = sctx->b.gfx.cs; struct si_compute *program = sctx->cs_shader_state.program; struct r600_resource *input_buffer = NULL; unsigned kernel_args_size; - unsigned num_work_size_bytes = 36; + unsigned num_work_size_bytes = program->use_code_object_v2 ? 0 : 36; uint32_t kernel_args_offset = 0; uint32_t *kernel_args; void *kernel_args_ptr; @@ -335,10 +530,14 @@ static void si_upload_compute_input(struct si_context *sctx, (struct pipe_resource**)&input_buffer, &kernel_args_ptr); kernel_args = (uint32_t*)kernel_args_ptr; - for (i = 0; i < 3; i++) { - kernel_args[i] = info->grid[i]; - kernel_args[i + 3] = info->grid[i] * info->block[i]; - kernel_args[i + 6] = info->block[i]; + kernel_args_va = input_buffer->gpu_address + kernel_args_offset; + + if (!code_object) { + for (i = 0; i < 3; i++) { + kernel_args[i] = info->grid[i]; + kernel_args[i + 3] = info->grid[i] * info->block[i]; + kernel_args[i + 6] = info->block[i]; + } } memcpy(kernel_args + (num_work_size_bytes / 4), info->input, @@ -350,15 +549,18 @@ static void si_upload_compute_input(struct si_context *sctx, kernel_args[i]); } - kernel_args_va = input_buffer->gpu_address + kernel_args_offset; radeon_add_to_buffer_list(&sctx->b, &sctx->b.gfx, input_buffer, RADEON_USAGE_READ, RADEON_PRIO_CONST_BUFFER); - radeon_set_sh_reg_seq(cs, R_00B900_COMPUTE_USER_DATA_0, 2); - radeon_emit(cs, kernel_args_va); - radeon_emit(cs, S_008F04_BASE_ADDRESS_HI (kernel_args_va >> 32) | - S_008F04_STRIDE(0)); + if (code_object) { + si_setup_user_sgprs_co_v2(sctx, code_object, info, kernel_args_va); + } else { + radeon_set_sh_reg_seq(cs, R_00B900_COMPUTE_USER_DATA_0, 2); + radeon_emit(cs, kernel_args_va); + radeon_emit(cs, S_008F04_BASE_ADDRESS_HI (kernel_args_va >> 32) | + S_008F04_STRIDE(0)); + } r600_resource_reference(&input_buffer, NULL); } @@ -446,6 +648,8 @@ static void si_launch_grid( { struct si_context *sctx = (struct si_context*)ctx; struct si_compute *program = sctx->cs_shader_state.program; + const amd_kernel_code_t *code_object = + si_compute_get_code_object(program, info->pc); int i; /* HW bug workaround when CS threadgroups > 256 threads and async * compute isn't used, i.e. only one compute job can run at a time. @@ -487,7 +691,8 @@ static void si_launch_grid( if (sctx->b.flags) si_emit_cache_flush(sctx); - if (!si_switch_compute_shader(sctx, program, &program->shader, info->pc)) + if (!si_switch_compute_shader(sctx, program, &program->shader, + code_object, info->pc)) return; si_upload_compute_shader_descriptors(sctx); @@ -500,7 +705,7 @@ static void si_launch_grid( } if (program->input_size || program->ir_type == PIPE_SHADER_IR_NATIVE) - si_upload_compute_input(sctx, info); + si_upload_compute_input(sctx, code_object, info); /* Global buffers */ for (i = 0; i < MAX_GLOBAL_BUFFERS; i++) { -- 2.30.2