blob: 0592cc0ae0b39bb2a8afac981eb391fb7b41367f [file] [log] [blame]
/*
* Copyright (C) 2019 Connor Abbott <cwabbott0@gmail.com>
* Copyright (C) 2019 Lyude Paul <thatslyude@gmail.com>
* Copyright (C) 2019 Ryan Houdek <Sonicadvance1@gmail.com>
*
* 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
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* 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 NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS 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 __bifrost_h__
#define __bifrost_h__
#include <stdint.h>
#include <stdbool.h>
#define BIFROST_DBG_MSGS 0x0001
#define BIFROST_DBG_SHADERS 0x0002
extern int bifrost_debug;
enum bifrost_message_type {
BIFROST_MESSAGE_NONE = 0,
BIFROST_MESSAGE_VARYING = 1,
BIFROST_MESSAGE_ATTRIBUTE = 2,
BIFROST_MESSAGE_TEX = 3,
BIFROST_MESSAGE_VARTEX = 4,
BIFROST_MESSAGE_LOAD = 5,
BIFROST_MESSAGE_STORE = 6,
BIFROST_MESSAGE_ATOMIC = 7,
BIFROST_MESSAGE_BARRIER = 8,
BIFROST_MESSAGE_BLEND = 9,
BIFROST_MESSAGE_TILE = 10,
/* type 11 reserved */
BIFROST_MESSAGE_Z_STENCIL = 12,
BIFROST_MESSAGE_ATEST = 13,
BIFROST_MESSAGE_JOB = 14,
BIFROST_MESSAGE_64BIT = 15
};
enum bifrost_ftz {
BIFROST_FTZ_DISABLE = 0,
BIFROST_FTZ_DX11 = 1,
BIFROST_FTZ_ALWAYS = 2,
BIFROST_FTZ_ABRUPT = 3
};
enum bifrost_exceptions {
BIFROST_EXCEPTIONS_ENABLED = 0,
BIFROST_EXCEPTIONS_DISABLED = 1,
BIFROST_EXCEPTIONS_PRECISE_DIVISION = 2,
BIFROST_EXCEPTIONS_PRECISE_SQRT = 3,
};
/* Describes clause flow control, with respect to control flow and branch
* reconvergence.
*
* Control flow may be considered back-to-back (execute clauses back-to-back),
* non-back-to-back (switch warps after clause before the next clause), write
* elision (back-to-back and elide register slot #3 write from the clause), or
* end of shader.
*
* Branch reconvergence may be disabled, enabled unconditionally, or enabled
* based on the program counter. A clause requires reconvergence if it has a
* successor that can be executed without first executing the clause itself.
* Separate iterations of a loop are treated separately here, so it is also the
* case for a loop exit where the iteration count is not warp-invariant.
*
*/
enum bifrost_flow {
/* End-of-shader */
BIFROST_FLOW_END = 0,
/* Non back-to-back, PC-encoded reconvergence */
BIFROST_FLOW_NBTB_PC = 1,
/* Non back-to-back, unconditional reconvergence */
BIFROST_FLOW_NBTB_UNCONDITIONAL = 2,
/* Non back-to-back, no reconvergence */
BIFROST_FLOW_NBTB = 3,
/* Back-to-back, unconditional reconvergence */
BIFROST_FLOW_BTB_UNCONDITIONAL = 4,
/* Back-to-back, no reconvergence */
BIFROST_FLOW_BTB_NONE = 5,
/* Write elision, unconditional reconvergence */
BIFROST_FLOW_WE_UNCONDITIONAL = 6,
/* Write elision, no reconvergence */
BIFROST_FLOW_WE = 7,
};
struct bifrost_header {
/* Reserved */
unsigned zero1 : 5;
/* Flush-to-zero mode, leave zero for GL */
enum bifrost_ftz flush_to_zero : 2;
/* Convert any infinite result of any floating-point operation to the
* biggest representable number */
unsigned suppress_inf: 1;
/* Convert NaN to +0.0 */
unsigned suppress_nan : 1;
/* Floating-point excception handling mode */
enum bifrost_exceptions float_exceptions : 2;
/* Enum describing the flow control, which matters for handling
* divergence and reconvergence efficiently */
enum bifrost_flow flow_control : 3;
/* Reserved */
unsigned zero2 : 1;
/* Terminate discarded threads, rather than continuing execution. Set
* for fragment shaders for standard GL behaviour of DISCARD. Also in a
* fragment shader, this disables helper invocations, so cannot be used
* in a shader that requires derivatives or texture LOD computation */
unsigned terminate_discarded_threads : 1;
/* If set, the hardware may prefetch the next clause. If false, the
* hardware may not. Clear for unconditional branches. */
unsigned next_clause_prefetch : 1;
/* If set, a barrier will be inserted after the clause waiting for all
* message passing instructions to read their staging registers, such
* that it is safe for the next clause to write them. */
unsigned staging_barrier: 1;
unsigned staging_register : 6;
/* Slots to wait on and slot to be used for message passing
* instructions respectively */
unsigned dependency_wait : 8;
unsigned dependency_slot : 3;
enum bifrost_message_type message_type : 5;
enum bifrost_message_type next_message_type : 5;
} __attribute__((packed));
enum bifrost_packed_src {
BIFROST_SRC_PORT0 = 0,
BIFROST_SRC_PORT1 = 1,
BIFROST_SRC_PORT2 = 2,
BIFROST_SRC_STAGE = 3,
BIFROST_SRC_CONST_LO = 4,
BIFROST_SRC_CONST_HI = 5,
BIFROST_SRC_PASS_FMA = 6,
BIFROST_SRC_PASS_ADD = 7,
};
struct bifrost_fma_inst {
unsigned src0 : 3;
unsigned op : 20;
} __attribute__((packed));
struct bifrost_add_inst {
unsigned src0 : 3;
unsigned op : 17;
} __attribute__((packed));
enum bifrost_outmod {
BIFROST_NONE = 0x0,
BIFROST_POS = 0x1,
BIFROST_SAT_SIGNED = 0x2,
BIFROST_SAT = 0x3,
};
enum bifrost_roundmode {
BIFROST_RTE = 0x0, /* round to even */
BIFROST_RTP = 0x1, /* round to positive */
BIFROST_RTN = 0x2, /* round to negative */
BIFROST_RTZ = 0x3 /* round to zero */
};
/* NONE: Same as fmax() and fmin() -- return the other
* number if any number is NaN. Also always return +0 if
* one argument is +0 and the other is -0.
*
* NAN_WINS: Instead of never returning a NaN, always return
* one. The "greater"/"lesser" NaN is always returned, first
* by checking the sign and then the mantissa bits.
*
* SRC1_WINS: For max, implement src0 > src1 ? src0 : src1.
* For min, implement src0 < src1 ? src0 : src1. This
* includes handling NaN's and signedness of 0 differently
* from above, since +0 and -0 compare equal and comparisons
* always return false for NaN's. As a result, this mode is
* *not* commutative.
*
* SRC0_WINS: For max, implement src0 < src1 ? src1 : src0
* For min, implement src0 > src1 ? src1 : src0
*/
enum bifrost_minmax_mode {
BIFROST_MINMAX_NONE = 0x0,
BIFROST_NAN_WINS = 0x1,
BIFROST_SRC1_WINS = 0x2,
BIFROST_SRC0_WINS = 0x3,
};
enum bifrost_interp_mode {
BIFROST_INTERP_PER_FRAG = 0x0,
BIFROST_INTERP_CENTROID = 0x1,
BIFROST_INTERP_DEFAULT = 0x2,
BIFROST_INTERP_EXPLICIT = 0x3
};
/* Fixed location for gl_FragCoord.zw */
#define BIFROST_FRAGZ (23)
#define BIFROST_FRAGW (22)
enum branch_bit_size {
BR_SIZE_32 = 0,
BR_SIZE_16XX = 1,
BR_SIZE_16YY = 2,
// For the above combinations of bitsize and location, an extra bit is
// encoded via comparing the sources. The only possible source of ambiguity
// would be if the sources were the same, but then the branch condition
// would be always true or always false anyways, so we can ignore it. But
// this no longer works when comparing the y component to the x component,
// since it's valid to compare the y component of a source against its own
// x component. Instead, the extra bit is encoded via an extra bitsize.
BR_SIZE_16YX0 = 3,
BR_SIZE_16YX1 = 4,
BR_SIZE_32_AND_16X = 5,
BR_SIZE_32_AND_16Y = 6,
// Used for comparisons with zero and always-true, see below. I think this
// only works for integer comparisons.
BR_SIZE_ZERO = 7,
};
struct bifrost_regs {
unsigned uniform_const : 8;
unsigned reg3 : 6;
unsigned reg2 : 6;
unsigned reg0 : 5;
unsigned reg1 : 6;
unsigned ctrl : 4;
} __attribute__((packed));
enum bifrost_branch_cond {
BR_COND_LT = 0,
BR_COND_LE = 1,
BR_COND_GE = 2,
BR_COND_GT = 3,
// Equal vs. not-equal determined by src0/src1 comparison
BR_COND_EQ = 4,
// floating-point comparisons
// Becomes UNE when you flip the arguments
BR_COND_OEQ = 5,
// TODO what happens when you flip the arguments?
BR_COND_OGT = 6,
BR_COND_OLT = 7,
};
enum bifrost_branch_code {
BR_ALWAYS = 63,
};
#define BIFROST_ADD_OP_BRANCH (0x0d000 >> 12)
struct bifrost_branch {
unsigned src0 : 3;
/* For BR_SIZE_ZERO, upper two bits become ctrl */
unsigned src1 : 3;
/* Offset source -- always uniform/const but
* theoretically could support indirect jumps? */
unsigned src2 : 3;
enum bifrost_branch_cond cond : 3;
enum branch_bit_size size : 3;
unsigned op : 5;
};
/* Clause packing */
struct bifrost_fmt1 {
unsigned ins_0 : 3;
unsigned tag : 5;
uint64_t ins_1 : 64;
unsigned ins_2 : 11;
uint64_t header : 45;
} __attribute__((packed));
#define BIFROST_FMT1_INSTRUCTIONS 0b00101
#define BIFROST_FMT1_FINAL 0b01001
#define BIFROST_FMT1_CONSTANTS 0b00001
#define BIFROST_FMTC_CONSTANTS 0b0011
#define BIFROST_FMTC_FINAL 0b0111
struct bifrost_fmt_constant {
unsigned pos : 4;
unsigned tag : 4;
uint64_t imm_1 : 60;
uint64_t imm_2 : 60;
} __attribute__((packed));
/* 32-bit modes for slots 2/3, as encoded in the register block. Other values
* are reserved. First part specifies behaviour of slot 2 (Idle, Read, Write
* Full, Write Low, Write High), second part behaviour of slot 3, and the last
* part specifies the source for the write (FMA, ADD, or MIX for FMA/ADD).
*
* IDLE is a special mode disabling both slots, except for the first
* instruction in the clause which uses IDLE_1 for the same purpose.
*
* All fields 0 used as sentinel for reserved encoding, so IDLE(_1) have FMA
* set (and ignored) as a placeholder to differentiate from reserved.
*/
enum bifrost_reg_mode {
BIFROST_R_WL_FMA = 1,
BIFROST_R_WH_FMA = 2,
BIFROST_R_W_FMA = 3,
BIFROST_R_WL_ADD = 4,
BIFROST_R_WH_ADD = 5,
BIFROST_R_W_ADD = 6,
BIFROST_WL_WL_ADD = 7,
BIFROST_WL_WH_ADD = 8,
BIFROST_WL_W_ADD = 9,
BIFROST_WH_WL_ADD = 10,
BIFROST_WH_WH_ADD = 11,
BIFROST_WH_W_ADD = 12,
BIFROST_W_WL_ADD = 13,
BIFROST_W_WH_ADD = 14,
BIFROST_W_W_ADD = 15,
BIFROST_IDLE_1 = 16,
BIFROST_I_W_FMA = 17,
BIFROST_I_WL_FMA = 18,
BIFROST_I_WH_FMA = 19,
BIFROST_R_I = 20,
BIFROST_I_W_ADD = 21,
BIFROST_I_WL_ADD = 22,
BIFROST_I_WH_ADD = 23,
BIFROST_WL_WH_MIX = 24,
BIFROST_WH_WL_MIX = 26,
BIFROST_IDLE = 27,
};
enum bifrost_reg_op {
BIFROST_OP_IDLE = 0,
BIFROST_OP_READ = 1,
BIFROST_OP_WRITE = 2,
BIFROST_OP_WRITE_LO = 3,
BIFROST_OP_WRITE_HI = 4,
};
struct bifrost_reg_ctrl_23 {
enum bifrost_reg_op slot2;
enum bifrost_reg_op slot3;
bool slot3_fma;
};
static const struct bifrost_reg_ctrl_23 bifrost_reg_ctrl_lut[32] = {
[BIFROST_R_WL_FMA] = { BIFROST_OP_READ, BIFROST_OP_WRITE_LO, true },
[BIFROST_R_WH_FMA] = { BIFROST_OP_READ, BIFROST_OP_WRITE_HI, true },
[BIFROST_R_W_FMA] = { BIFROST_OP_READ, BIFROST_OP_WRITE, true },
[BIFROST_R_WL_ADD] = { BIFROST_OP_READ, BIFROST_OP_WRITE_LO, false },
[BIFROST_R_WH_ADD] = { BIFROST_OP_READ, BIFROST_OP_WRITE_HI, false },
[BIFROST_R_W_ADD] = { BIFROST_OP_READ, BIFROST_OP_WRITE, false },
[BIFROST_WL_WL_ADD] = { BIFROST_OP_WRITE_LO, BIFROST_OP_WRITE_LO, false },
[BIFROST_WL_WH_ADD] = { BIFROST_OP_WRITE_LO, BIFROST_OP_WRITE_HI, false },
[BIFROST_WL_W_ADD] = { BIFROST_OP_WRITE_LO, BIFROST_OP_WRITE, false },
[BIFROST_WH_WL_ADD] = { BIFROST_OP_WRITE_HI, BIFROST_OP_WRITE_LO, false },
[BIFROST_WH_WH_ADD] = { BIFROST_OP_WRITE_HI, BIFROST_OP_WRITE_HI, false },
[BIFROST_WH_W_ADD] = { BIFROST_OP_WRITE_HI, BIFROST_OP_WRITE, false },
[BIFROST_W_WL_ADD] = { BIFROST_OP_WRITE, BIFROST_OP_WRITE_LO, false },
[BIFROST_W_WH_ADD] = { BIFROST_OP_WRITE, BIFROST_OP_WRITE_HI, false },
[BIFROST_W_W_ADD] = { BIFROST_OP_WRITE, BIFROST_OP_WRITE, false },
[BIFROST_IDLE_1] = { BIFROST_OP_IDLE, BIFROST_OP_IDLE, true },
[BIFROST_I_W_FMA] = { BIFROST_OP_IDLE, BIFROST_OP_WRITE, true },
[BIFROST_I_WL_FMA] = { BIFROST_OP_IDLE, BIFROST_OP_WRITE_LO, true },
[BIFROST_I_WH_FMA] = { BIFROST_OP_IDLE, BIFROST_OP_WRITE_HI, true },
[BIFROST_R_I] = { BIFROST_OP_READ, BIFROST_OP_IDLE, false },
[BIFROST_I_W_ADD] = { BIFROST_OP_IDLE, BIFROST_OP_WRITE, false },
[BIFROST_I_WL_ADD] = { BIFROST_OP_IDLE, BIFROST_OP_WRITE_LO, false },
[BIFROST_I_WH_ADD] = { BIFROST_OP_IDLE, BIFROST_OP_WRITE_HI, false },
[BIFROST_WL_WH_MIX] = { BIFROST_OP_WRITE_LO, BIFROST_OP_WRITE_HI, false },
[BIFROST_WH_WL_MIX] = { BIFROST_OP_WRITE_HI, BIFROST_OP_WRITE_LO, false },
[BIFROST_IDLE] = { BIFROST_OP_IDLE, BIFROST_OP_IDLE, true },
};
/* Texture operator descriptors in various states. Usually packed in the
* compiler and stored as a constant */
enum bifrost_index {
/* Both texture/sampler index immediate */
BIFROST_INDEX_IMMEDIATE_SHARED = 0,
/* Sampler index immediate, texture index from staging */
BIFROST_INDEX_IMMEDIATE_SAMPLER = 1,
/* Texture index immediate, sampler index from staging */
BIFROST_INDEX_IMMEDIATE_TEXTURE = 2,
/* Both indices from (separate) staging registers */
BIFROST_INDEX_REGISTER = 3,
};
enum bifrost_tex_op {
/* Given explicit derivatives, compute a gradient descriptor */
BIFROST_TEX_OP_GRDESC_DER = 4,
/* Given implicit derivatives (texture coordinates in a fragment
* shader), compute a gradient descriptor */
BIFROST_TEX_OP_GRDESC = 5,
/* Fetch a texel. Takes a staging register with LOD level / face index
* packed 16:16 */
BIFROST_TEX_OP_FETCH = 6,
/* Filtered texture */
BIFROST_TEX_OP_TEX = 7,
};
enum bifrost_lod_mode {
/* Takes two staging registers forming a 64-bit gradient descriptor
* (computed by a previous GRDESC or GRDESC_DER operation) */
BIFROST_LOD_MODE_GRDESC = 3,
/* Take a staging register with 8:8 fixed-point in bottom 16-bits
* specifying an explicit LOD */
BIFROST_LOD_MODE_EXPLICIT = 4,
/* Takes a staging register with bottom 16-bits as 8:8 fixed-point LOD
* bias and top 16-bit as 8:8 fixed-point lower bound (generally left
* zero), added and clamped to a computed LOD */
BIFROST_LOD_MODE_BIAS = 5,
/* Set LOD to zero */
BIFROST_LOD_MODE_ZERO = 6,
/* Compute LOD */
BIFROST_LOD_MODE_COMPUTE = 7,
};
enum bifrost_texture_format {
/* 16-bit floating point, with optional clamping */
BIFROST_TEXTURE_FORMAT_F16 = 0,
BIFROST_TEXTURE_FORMAT_F16_POS = 1,
BIFROST_TEXTURE_FORMAT_F16_PM1 = 2,
BIFROST_TEXTURE_FORMAT_F16_1 = 3,
/* 32-bit floating point, with optional clamping */
BIFROST_TEXTURE_FORMAT_F32 = 4,
BIFROST_TEXTURE_FORMAT_F32_POS = 5,
BIFROST_TEXTURE_FORMAT_F32_PM1 = 6,
BIFROST_TEXTURE_FORMAT_F32_1 = 7,
};
enum bifrost_texture_format_full {
/* Transclude bifrost_texture_format from above */
/* Integers, unclamped */
BIFROST_TEXTURE_FORMAT_U16 = 12,
BIFROST_TEXTURE_FORMAT_S16 = 13,
BIFROST_TEXTURE_FORMAT_U32 = 14,
BIFROST_TEXTURE_FORMAT_S32 = 15,
};
enum bifrost_texture_fetch {
/* Default texelFetch */
BIFROST_TEXTURE_FETCH_TEXEL = 1,
/* Deprecated, fetches 4x U32 of a U8 x 4 texture. Do not use. */
BIFROST_TEXTURE_FETCH_GATHER4_RGBA = 3,
/* Gathers */
BIFROST_TEXTURE_FETCH_GATHER4_R = 4,
BIFROST_TEXTURE_FETCH_GATHER4_G = 5,
BIFROST_TEXTURE_FETCH_GATHER4_B = 6,
BIFROST_TEXTURE_FETCH_GATHER4_A = 7
};
struct bifrost_texture_operation {
/* If immediate_indices is set:
* - immediate sampler index
* - index used as texture index
* Otherwise:
* - bifrost_single_index in lower 2 bits
* - 0x3 in upper 2 bits (single-texturing)
*/
unsigned sampler_index_or_mode : 4;
unsigned index : 7;
bool immediate_indices : 1;
enum bifrost_tex_op op : 3;
/* If set for TEX/FETCH, loads texel offsets and multisample index from
* a staging register containing offset_x:offset_y:offset_z:ms_index
* packed 8:8:8:8. Offsets must be in [-31, +31]. If set for
* GRDESC(_DER), disable LOD bias. */
bool offset_or_bias_disable : 1;
/* If set for TEX/FETCH, loads fp32 shadow comparison value from a
* staging register. Implies fetch_component = gather4_r. If set for
* GRDESC(_DER), disables LOD clamping. */
bool shadow_or_clamp_disable : 1;
/* If set, loads an uint32 array index from a staging register. */
bool array : 1;
/* Texture dimension, or 0 for a cubemap */
unsigned dimension : 2;
/* Method to compute LOD value or for a FETCH, the
* bifrost_texture_fetch component specification */
enum bifrost_lod_mode lod_or_fetch : 3;
/* Reserved */
unsigned zero : 1;
/* Register format for the result */
enum bifrost_texture_format_full format : 4;
/* Write mask for the result */
unsigned mask : 4;
} __attribute__((packed));
#endif