* USE OR OTHER DEALINGS IN THE SOFTWARE.
*/
#include "r600_sq.h"
-#include "r600_llvm.h"
#include "r600_formats.h"
#include "r600_opcodes.h"
#include "r600_shader.h"
#include "tgsi/tgsi_parse.h"
#include "tgsi/tgsi_scan.h"
#include "tgsi/tgsi_dump.h"
+#include "util/u_bitcast.h"
#include "util/u_memory.h"
#include "util/u_math.h"
#include <stdio.h>
#include <errno.h>
-/* CAYMAN notes
+/* CAYMAN notes
Why CAYMAN got loops for lots of instructions is explained here.
-These 8xx t-slot only ops are implemented in all vector slots.
MUL_LIT, FLT_TO_UINT, INT_TO_FLT, UINT_TO_FLT
-These 8xx t-slot only opcodes become vector ops, with all four
-slots expecting the arguments on sources a and b. Result is
+These 8xx t-slot only opcodes become vector ops, with all four
+slots expecting the arguments on sources a and b. Result is
broadcast to all channels.
MULLO_INT, MULHI_INT, MULLO_UINT, MULHI_UINT, MUL_64
-These 8xx t-slot only opcodes become vector ops in the z, y, and
+These 8xx t-slot only opcodes become vector ops in the z, y, and
x slots.
EXP_IEEE, LOG_IEEE/CLAMPED, RECIP_IEEE/CLAMPED/FF/INT/UINT/_64/CLAMPED_64
RECIPSQRT_IEEE/CLAMPED/FF/_64/CLAMPED_64
SQRT_IEEE/_64
SIN/COS
-The w slot may have an independent co-issued operation, or if the
-result is required to be in the w slot, the opcode above may be
+The w slot may have an independent co-issued operation, or if the
+result is required to be in the w slot, the opcode above may be
issued in the w slot as well.
The compiler must issue the source argument to slots z, y, and x
*/
if (shader->bo == NULL) {
shader->bo = (struct r600_resource*)
- pipe_buffer_create(ctx->screen, PIPE_BIND_CUSTOM, PIPE_USAGE_IMMUTABLE, shader->shader.bc.ndw * 4);
+ pipe_buffer_create(ctx->screen, 0, PIPE_USAGE_IMMUTABLE, shader->shader.bc.ndw * 4);
if (shader->bo == NULL) {
return -ENOMEM;
}
R600_ERR("translation from TGSI failed !\n");
goto error;
}
- if (shader->shader.processor_type == TGSI_PROCESSOR_VERTEX) {
+ if (shader->shader.processor_type == PIPE_SHADER_VERTEX) {
/* only disable for vertex shaders in tess paths */
if (key.vs.as_ls)
use_sb = 0;
}
- use_sb &= (shader->shader.processor_type != TGSI_PROCESSOR_TESS_CTRL);
- use_sb &= (shader->shader.processor_type != TGSI_PROCESSOR_TESS_EVAL);
+ use_sb &= (shader->shader.processor_type != PIPE_SHADER_TESS_CTRL);
+ use_sb &= (shader->shader.processor_type != PIPE_SHADER_TESS_EVAL);
/* disable SB for shaders using doubles */
use_sb &= !shader->shader.uses_doubles;
- /* Check if the bytecode has already been built. When using the llvm
- * backend, r600_shader_from_tgsi() will take care of building the
- * bytecode.
- */
+ use_sb &= !shader->shader.uses_atomics;
+
+ /* Check if the bytecode has already been built. */
if (!shader->shader.bc.bytecode) {
r = r600_bytecode_build(&shader->shader.bc);
if (r) {
/* Build state. */
switch (shader->shader.processor_type) {
- case TGSI_PROCESSOR_TESS_CTRL:
+ case PIPE_SHADER_TESS_CTRL:
evergreen_update_hs_state(ctx, shader);
break;
- case TGSI_PROCESSOR_TESS_EVAL:
+ case PIPE_SHADER_TESS_EVAL:
if (key.tes.as_es)
evergreen_update_es_state(ctx, shader);
else
evergreen_update_vs_state(ctx, shader);
break;
- case TGSI_PROCESSOR_GEOMETRY:
+ case PIPE_SHADER_GEOMETRY:
if (rctx->b.chip_class >= EVERGREEN) {
evergreen_update_gs_state(ctx, shader);
evergreen_update_vs_state(ctx, shader->gs_copy_shader);
r600_update_vs_state(ctx, shader->gs_copy_shader);
}
break;
- case TGSI_PROCESSOR_VERTEX:
+ case PIPE_SHADER_VERTEX:
export_shader = key.vs.as_es;
if (rctx->b.chip_class >= EVERGREEN) {
if (key.vs.as_ls)
r600_update_vs_state(ctx, shader);
}
break;
- case TGSI_PROCESSOR_FRAGMENT:
+ case PIPE_SHADER_FRAGMENT:
if (rctx->b.chip_class >= EVERGREEN) {
evergreen_update_ps_state(ctx, shader);
} else {
void r600_pipe_shader_destroy(struct pipe_context *ctx, struct r600_pipe_shader *shader)
{
- pipe_resource_reference((struct pipe_resource**)&shader->bo, NULL);
+ r600_resource_reference(&shader->bo, NULL);
r600_bytecode_clear(&shader->shader.bc);
r600_release_command_buffer(&shader->command_buffer);
}
uint32_t *literals;
uint32_t nliterals;
uint32_t max_driver_temp_used;
- boolean use_llvm;
/* needed for evergreen interpolation */
struct eg_interp eg_interpolators[6]; // indexed by Persp/Linear * 3 + sample/center/centroid
/* evergreen/cayman also store sample mask in face register */
static int tgsi_is_supported(struct r600_shader_ctx *ctx)
{
struct tgsi_full_instruction *i = &ctx->parse.FullToken.FullInstruction;
- int j;
+ unsigned j;
if (i->Instruction.NumDstRegs > 1 && i->Instruction.Opcode != TGSI_OPCODE_DFRACEXP) {
R600_ERR("too many dst (%d)\n", i->Instruction.NumDstRegs);
return -EINVAL;
}
- if (i->Instruction.Predicate) {
- R600_ERR("predicate unsupported\n");
- return -EINVAL;
- }
#if 0
if (i->Instruction.Label) {
R600_ERR("label unsupported\n");
if (i->Src[j].Register.Dimension) {
switch (i->Src[j].Register.File) {
case TGSI_FILE_CONSTANT:
+ case TGSI_FILE_HW_ATOMIC:
break;
case TGSI_FILE_INPUT:
- if (ctx->type == TGSI_PROCESSOR_GEOMETRY ||
- ctx->type == TGSI_PROCESSOR_TESS_CTRL ||
- ctx->type == TGSI_PROCESSOR_TESS_EVAL)
+ if (ctx->type == PIPE_SHADER_GEOMETRY ||
+ ctx->type == PIPE_SHADER_TESS_CTRL ||
+ ctx->type == PIPE_SHADER_TESS_EVAL)
break;
case TGSI_FILE_OUTPUT:
- if (ctx->type == TGSI_PROCESSOR_TESS_CTRL)
+ if (ctx->type == PIPE_SHADER_TESS_CTRL)
break;
default:
R600_ERR("unsupported src %d (file %d, dimension %d)\n", j,
}
for (j = 0; j < i->Instruction.NumDstRegs; j++) {
if (i->Dst[j].Register.Dimension) {
- if (ctx->type == TGSI_PROCESSOR_TESS_CTRL)
+ if (ctx->type == PIPE_SHADER_TESS_CTRL)
continue;
R600_ERR("unsupported dst (dimension)\n");
return -EINVAL;
ctx->shader->input[index].lds_pos = ctx->shader->nlds++;
if (ctx->shader->input[index].interpolate > 0) {
evergreen_interp_assign_ij_index(ctx, index);
- if (!ctx->use_llvm)
- r = evergreen_interp_alu(ctx, index);
+ r = evergreen_interp_alu(ctx, index);
} else {
- if (!ctx->use_llvm)
- r = evergreen_interp_flat(ctx, index);
+ r = evergreen_interp_flat(ctx, index);
}
}
return r;
case TGSI_FILE_INPUT:
for (j = 0; j < count; j++) {
i = ctx->shader->ninput + j;
- assert(i < Elements(ctx->shader->input));
+ assert(i < ARRAY_SIZE(ctx->shader->input));
ctx->shader->input[i].name = d->Semantic.Name;
ctx->shader->input[i].sid = d->Semantic.Index + j;
ctx->shader->input[i].interpolate = d->Interp.Interpolate;
ctx->shader->input[i].interpolate_location = d->Interp.Location;
ctx->shader->input[i].gpr = ctx->file_offset[TGSI_FILE_INPUT] + d->Range.First + j;
- if (ctx->type == TGSI_PROCESSOR_FRAGMENT) {
+ if (ctx->type == PIPE_SHADER_FRAGMENT) {
ctx->shader->input[i].spi_sid = r600_spi_sid(&ctx->shader->input[i]);
switch (ctx->shader->input[i].name) {
case TGSI_SEMANTIC_FACE:
if ((r = evergreen_interp_input(ctx, i)))
return r;
}
- } else if (ctx->type == TGSI_PROCESSOR_GEOMETRY) {
+ } else if (ctx->type == PIPE_SHADER_GEOMETRY) {
/* FIXME probably skip inputs if they aren't passed in the ring */
ctx->shader->input[i].ring_offset = ctx->next_ring_offset;
ctx->next_ring_offset += 16;
case TGSI_FILE_OUTPUT:
for (j = 0; j < count; j++) {
i = ctx->shader->noutput + j;
- assert(i < Elements(ctx->shader->output));
+ assert(i < ARRAY_SIZE(ctx->shader->output));
ctx->shader->output[i].name = d->Semantic.Name;
ctx->shader->output[i].sid = d->Semantic.Index + j;
ctx->shader->output[i].gpr = ctx->file_offset[TGSI_FILE_OUTPUT] + d->Range.First + j;
ctx->shader->output[i].interpolate = d->Interp.Interpolate;
ctx->shader->output[i].write_mask = d->Declaration.UsageMask;
- if (ctx->type == TGSI_PROCESSOR_VERTEX ||
- ctx->type == TGSI_PROCESSOR_GEOMETRY ||
- ctx->type == TGSI_PROCESSOR_TESS_EVAL) {
+ if (ctx->type == PIPE_SHADER_VERTEX ||
+ ctx->type == PIPE_SHADER_GEOMETRY ||
+ ctx->type == PIPE_SHADER_TESS_EVAL) {
ctx->shader->output[i].spi_sid = r600_spi_sid(&ctx->shader->output[i]);
switch (d->Semantic.Name) {
case TGSI_SEMANTIC_CLIPDIST:
ctx->cv_output = i;
break;
}
- if (ctx->type == TGSI_PROCESSOR_GEOMETRY) {
+ if (ctx->type == PIPE_SHADER_GEOMETRY) {
ctx->gs_out_ring_offset += 16;
}
- } else if (ctx->type == TGSI_PROCESSOR_FRAGMENT) {
+ } else if (ctx->type == PIPE_SHADER_FRAGMENT) {
switch (d->Semantic.Name) {
case TGSI_SEMANTIC_COLOR:
ctx->shader->nr_ps_max_color_exports++;
case TGSI_FILE_ADDRESS:
break;
+ case TGSI_FILE_HW_ATOMIC:
+ i = ctx->shader->nhwatomic_ranges;
+ ctx->shader->atomics[i].start = d->Range.First;
+ ctx->shader->atomics[i].end = d->Range.Last;
+ ctx->shader->atomics[i].hw_idx = ctx->shader->atomic_base + ctx->shader->nhwatomic;
+ ctx->shader->atomics[i].array_id = d->Array.ArrayID;
+ ctx->shader->atomics[i].buffer_id = d->Dim.Index2D;
+ ctx->shader->nhwatomic_ranges++;
+ ctx->shader->nhwatomic += count;
+ break;
+
case TGSI_FILE_SYSTEM_VALUE:
if (d->Semantic.Name == TGSI_SEMANTIC_SAMPLEMASK ||
d->Semantic.Name == TGSI_SEMANTIC_SAMPLEID ||
interpolate = ctx->info.input_interpolate[inst->Src[0].Register.Index];
k = eg_get_interpolator_index(interpolate, location);
- ctx->eg_interpolators[k].enabled = true;
+ if (k >= 0)
+ ctx->eg_interpolators[k].enabled = true;
}
} else if (parse.FullToken.Token.Type == TGSI_TOKEN_TYPE_DECLARATION) {
struct tgsi_full_declaration *d = &parse.FullToken.FullDeclaration;
if (d->Declaration.File == TGSI_FILE_SYSTEM_VALUE) {
- for (k = 0; k < Elements(inputs); k++) {
+ for (k = 0; k < ARRAY_SIZE(inputs); k++) {
if (d->Semantic.Name == inputs[k].name ||
d->Semantic.Name == inputs[k].alternate_name) {
inputs[k].enabled = true;
tgsi_parse_free(&parse);
- for (i = 0; i < Elements(inputs); i++) {
+ for (i = 0; i < ARRAY_SIZE(inputs); i++) {
boolean enabled = inputs[i].enabled;
int *reg = inputs[i].reg;
unsigned name = inputs[i].name;
if (enabled) {
int gpr = gpr_offset + num_regs++;
+ ctx->shader->nsys_inputs++;
// add to inputs, allocate a gpr
- k = ctx->shader->ninput ++;
+ k = ctx->shader->ninput++;
ctx->shader->input[k].name = name;
ctx->shader->input[k].sid = 0;
ctx->shader->input[k].interpolate = TGSI_INTERPOLATE_CONSTANT;
*/
static int evergreen_gpr_count(struct r600_shader_ctx *ctx)
{
- int i;
+ unsigned i;
int num_baryc;
struct tgsi_parse_context parse;
interpolate = ctx->info.input_interpolate[inst->Src[0].Register.Index];
k = eg_get_interpolator_index(interpolate, location);
- ctx->eg_interpolators[k].enabled = true;
+ if (k >= 0)
+ ctx->eg_interpolators[k].enabled = true;
}
}
}
/* assign gpr to each interpolator according to priority */
num_baryc = 0;
- for (i = 0; i < Elements(ctx->eg_interpolators); i++) {
+ for (i = 0; i < ARRAY_SIZE(ctx->eg_interpolators); i++) {
if (ctx->eg_interpolators[i].enabled) {
ctx->eg_interpolators[i].ij_index = num_baryc;
num_baryc ++;
r600_src->swizzle[2] = 0;
r600_src->swizzle[3] = 0;
r600_src->sel = 0;
- } else if (ctx->type != TGSI_PROCESSOR_TESS_CTRL && ctx->info.system_value_semantic_name[tgsi_src->Register.Index] == TGSI_SEMANTIC_INVOCATIONID) {
+ } else if (ctx->type != PIPE_SHADER_TESS_CTRL && ctx->info.system_value_semantic_name[tgsi_src->Register.Index] == TGSI_SEMANTIC_INVOCATIONID) {
r600_src->swizzle[0] = 3;
r600_src->swizzle[1] = 3;
r600_src->swizzle[2] = 3;
} else if (ctx->info.system_value_semantic_name[tgsi_src->Register.Index] == TGSI_SEMANTIC_TESSOUTER) {
r600_src->sel = 2;
} else if (ctx->info.system_value_semantic_name[tgsi_src->Register.Index] == TGSI_SEMANTIC_VERTICESIN) {
- if (ctx->type == TGSI_PROCESSOR_TESS_CTRL) {
+ if (ctx->type == PIPE_SHADER_TESS_CTRL) {
r600_src->sel = ctx->tess_input_info;
r600_src->swizzle[0] = 2;
r600_src->swizzle[1] = 2;
r600_src->swizzle[2] = 3;
r600_src->swizzle[3] = 3;
}
- } else if (ctx->type == TGSI_PROCESSOR_TESS_CTRL && ctx->info.system_value_semantic_name[tgsi_src->Register.Index] == TGSI_SEMANTIC_PRIMID) {
+ } else if (ctx->type == PIPE_SHADER_TESS_CTRL && ctx->info.system_value_semantic_name[tgsi_src->Register.Index] == TGSI_SEMANTIC_PRIMID) {
r600_src->sel = 0;
r600_src->swizzle[0] = 0;
r600_src->swizzle[1] = 0;
r600_src->swizzle[2] = 0;
r600_src->swizzle[3] = 0;
- } else if (ctx->type == TGSI_PROCESSOR_TESS_EVAL && ctx->info.system_value_semantic_name[tgsi_src->Register.Index] == TGSI_SEMANTIC_PRIMID) {
+ } else if (ctx->type == PIPE_SHADER_TESS_EVAL && ctx->info.system_value_semantic_name[tgsi_src->Register.Index] == TGSI_SEMANTIC_PRIMID) {
r600_src->sel = 0;
r600_src->swizzle[0] = 3;
r600_src->swizzle[1] = 3;
static int tgsi_split_gs_inputs(struct r600_shader_ctx *ctx)
{
struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
- int i;
+ unsigned i;
for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
struct tgsi_full_src_register *src = &inst->Src[i];
static int tgsi_split_lds_inputs(struct r600_shader_ctx *ctx)
{
struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
- int i;
+ unsigned i;
for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
struct tgsi_full_src_register *src = &inst->Src[i];
- if (ctx->type == TGSI_PROCESSOR_TESS_EVAL && src->Register.File == TGSI_FILE_INPUT) {
+ if (ctx->type == PIPE_SHADER_TESS_EVAL && src->Register.File == TGSI_FILE_INPUT) {
int treg = r600_get_temp(ctx);
fetch_tes_input(ctx, src, treg);
ctx->src[i].sel = treg;
ctx->src[i].rel = 0;
}
- if (ctx->type == TGSI_PROCESSOR_TESS_CTRL && src->Register.File == TGSI_FILE_INPUT) {
+ if (ctx->type == PIPE_SHADER_TESS_CTRL && src->Register.File == TGSI_FILE_INPUT) {
int treg = r600_get_temp(ctx);
fetch_tcs_input(ctx, src, treg);
ctx->src[i].sel = treg;
ctx->src[i].rel = 0;
}
- if (ctx->type == TGSI_PROCESSOR_TESS_CTRL && src->Register.File == TGSI_FILE_OUTPUT) {
+ if (ctx->type == PIPE_SHADER_TESS_CTRL && src->Register.File == TGSI_FILE_OUTPUT) {
int treg = r600_get_temp(ctx);
fetch_tcs_output(ctx, src, treg);
ctx->src[i].sel = treg;
ctx.shader = &cshader->shader;
ctx.bc = &ctx.shader->bc;
- ctx.type = ctx.bc->type = TGSI_PROCESSOR_VERTEX;
+ ctx.type = ctx.bc->type = PIPE_SHADER_VERTEX;
r600_bytecode_init(ctx.bc, rctx->b.chip_class, rctx->b.family,
rctx->screen->has_compressed_msaa_texturing);
static int r600_emit_tess_factor(struct r600_shader_ctx *ctx)
{
- int i;
+ unsigned i;
int stride, outer_comps, inner_comps;
int tessinner_idx = -1, tessouter_idx = -1;
int r;
int out_idx = i >= outer_comps ? tessinner_idx : tessouter_idx;
int out_comp = i >= outer_comps ? i - outer_comps : i;
+ if (ctx->shader->tcs_prim_mode == PIPE_PRIM_LINES) {
+ if (out_comp == 1)
+ out_comp = 0;
+ else if (out_comp == 0)
+ out_comp = 1;
+ }
+
r = single_alu_op2(ctx, ALU_OP2_ADD_INT,
treg[i / 2], (2 * (i % 2)),
temp_reg, 0,
struct pipe_stream_output_info so = pipeshader->selector->so;
struct tgsi_full_immediate *immediate;
struct r600_shader_ctx ctx;
- struct r600_bytecode_output output[32];
+ struct r600_bytecode_output output[ARRAY_SIZE(shader->output)];
unsigned output_done, noutput;
unsigned opcode;
int i, j, k, r = 0;
int next_param_base = 0, next_clip_base;
int max_color_exports = MAX2(key.ps.nr_cbufs, 1);
- /* Declarations used by llvm code */
- bool use_llvm = false;
bool indirect_gprs;
bool ring_outputs = false;
bool lds_outputs = false;
bool lds_inputs = false;
bool pos_emitted = false;
-#ifdef R600_USE_LLVM
- use_llvm = rscreen->b.debug_flags & DBG_LLVM;
-#endif
ctx.bc = &shader->bc;
ctx.shader = shader;
ctx.native_integers = true;
-
r600_bytecode_init(ctx.bc, rscreen->b.chip_class, rscreen->b.family,
rscreen->has_compressed_msaa_texturing);
ctx.tokens = tokens;
shader->indirect_files = ctx.info.indirect_files;
shader->uses_doubles = ctx.info.uses_doubles;
+ shader->uses_atomics = ctx.info.file_mask[TGSI_FILE_HW_ATOMIC];
+ shader->nsys_inputs = 0;
indirect_gprs = ctx.info.indirect_files & ~((1 << TGSI_FILE_CONSTANT) | (1 << TGSI_FILE_SAMPLER));
tgsi_parse_init(&ctx.parse, tokens);
ctx.bc->type = shader->processor_type;
switch (ctx.type) {
- case TGSI_PROCESSOR_VERTEX:
+ case PIPE_SHADER_VERTEX:
shader->vs_as_gs_a = key.vs.as_gs_a;
shader->vs_as_es = key.vs.as_es;
shader->vs_as_ls = key.vs.as_ls;
+ shader->atomic_base = key.vs.first_atomic_counter;
if (shader->vs_as_es)
ring_outputs = true;
if (shader->vs_as_ls)
lds_outputs = true;
break;
- case TGSI_PROCESSOR_GEOMETRY:
+ case PIPE_SHADER_GEOMETRY:
ring_outputs = true;
+ shader->atomic_base = key.gs.first_atomic_counter;
break;
- case TGSI_PROCESSOR_TESS_CTRL:
+ case PIPE_SHADER_TESS_CTRL:
shader->tcs_prim_mode = key.tcs.prim_mode;
+ shader->atomic_base = key.tcs.first_atomic_counter;
lds_outputs = true;
lds_inputs = true;
break;
- case TGSI_PROCESSOR_TESS_EVAL:
+ case PIPE_SHADER_TESS_EVAL:
shader->tes_as_es = key.tes.as_es;
+ shader->atomic_base = key.tes.first_atomic_counter;
lds_inputs = true;
if (shader->tes_as_es)
ring_outputs = true;
break;
- case TGSI_PROCESSOR_FRAGMENT:
+ case PIPE_SHADER_FRAGMENT:
shader->two_side = key.ps.color_two_side;
+ shader->atomic_base = key.ps.first_atomic_counter;
break;
default:
break;
ctx.file_offset[i] = 0;
}
-#ifdef R600_USE_LLVM
- if (use_llvm && ctx.info.indirect_files && (ctx.info.indirect_files & (1 << TGSI_FILE_CONSTANT)) != ctx.info.indirect_files) {
- fprintf(stderr, "Warning: R600 LLVM backend does not support "
- "indirect adressing. Falling back to TGSI "
- "backend.\n");
- use_llvm = 0;
- }
-#endif
- if (ctx.type == TGSI_PROCESSOR_VERTEX) {
+ if (ctx.type == PIPE_SHADER_VERTEX) {
+
ctx.file_offset[TGSI_FILE_INPUT] = 1;
- if (!use_llvm) {
+ if (ctx.info.num_inputs)
r600_bytecode_add_cfinst(ctx.bc, CF_OP_CALL_FS);
- }
}
- if (ctx.type == TGSI_PROCESSOR_FRAGMENT) {
+ if (ctx.type == PIPE_SHADER_FRAGMENT) {
if (ctx.bc->chip_class >= EVERGREEN)
ctx.file_offset[TGSI_FILE_INPUT] = evergreen_gpr_count(&ctx);
else
ctx.file_offset[TGSI_FILE_INPUT] = allocate_system_value_inputs(&ctx, ctx.file_offset[TGSI_FILE_INPUT]);
}
- if (ctx.type == TGSI_PROCESSOR_GEOMETRY) {
+ if (ctx.type == PIPE_SHADER_GEOMETRY) {
/* FIXME 1 would be enough in some cases (3 or less input vertices) */
ctx.file_offset[TGSI_FILE_INPUT] = 2;
}
- if (ctx.type == TGSI_PROCESSOR_TESS_CTRL)
+ if (ctx.type == PIPE_SHADER_TESS_CTRL)
ctx.file_offset[TGSI_FILE_INPUT] = 1;
- if (ctx.type == TGSI_PROCESSOR_TESS_EVAL) {
+ if (ctx.type == PIPE_SHADER_TESS_EVAL) {
bool add_tesscoord = false, add_tess_inout = false;
ctx.file_offset[TGSI_FILE_INPUT] = 1;
for (i = 0; i < PIPE_MAX_SHADER_INPUTS; i++) {
if (add_tess_inout)
ctx.file_offset[TGSI_FILE_INPUT]+=2;
}
- ctx.use_llvm = use_llvm;
- if (use_llvm) {
- ctx.file_offset[TGSI_FILE_OUTPUT] =
- ctx.file_offset[TGSI_FILE_INPUT];
- } else {
- ctx.file_offset[TGSI_FILE_OUTPUT] =
+ ctx.file_offset[TGSI_FILE_OUTPUT] =
ctx.file_offset[TGSI_FILE_INPUT] +
ctx.info.file_max[TGSI_FILE_INPUT] + 1;
- }
ctx.file_offset[TGSI_FILE_TEMPORARY] = ctx.file_offset[TGSI_FILE_OUTPUT] +
ctx.info.file_max[TGSI_FILE_OUTPUT] + 1;
ctx.bc->index_reg[0] = ctx.bc->ar_reg + 1;
ctx.bc->index_reg[1] = ctx.bc->ar_reg + 2;
- if (ctx.type == TGSI_PROCESSOR_TESS_CTRL) {
+ if (ctx.type == PIPE_SHADER_TESS_CTRL) {
ctx.tess_input_info = ctx.bc->ar_reg + 3;
ctx.tess_output_info = ctx.bc->ar_reg + 4;
ctx.temp_reg = ctx.bc->ar_reg + 5;
- } else if (ctx.type == TGSI_PROCESSOR_TESS_EVAL) {
+ } else if (ctx.type == PIPE_SHADER_TESS_EVAL) {
ctx.tess_input_info = 0;
ctx.tess_output_info = ctx.bc->ar_reg + 3;
ctx.temp_reg = ctx.bc->ar_reg + 4;
- } else if (ctx.type == TGSI_PROCESSOR_GEOMETRY) {
+ } else if (ctx.type == PIPE_SHADER_GEOMETRY) {
ctx.gs_export_gpr_tregs[0] = ctx.bc->ar_reg + 3;
ctx.gs_export_gpr_tregs[1] = ctx.bc->ar_reg + 4;
ctx.gs_export_gpr_tregs[2] = ctx.bc->ar_reg + 5;
if (shader->vs_as_gs_a)
vs_add_primid_output(&ctx, key.vs.prim_id_out);
- if (ctx.type == TGSI_PROCESSOR_TESS_EVAL)
+ if (ctx.type == PIPE_SHADER_TESS_EVAL)
r600_fetch_tess_io_info(&ctx);
while (!tgsi_parse_end_of_tokens(&ctx.parse)) {
goto out_err;
}
}
-
+
shader->ring_item_sizes[0] = ctx.next_ring_offset;
shader->ring_item_sizes[1] = 0;
shader->ring_item_sizes[2] = 0;
}
}
-/* LLVM backend setup */
-#ifdef R600_USE_LLVM
- if (use_llvm) {
- struct radeon_llvm_context radeon_llvm_ctx;
- LLVMModuleRef mod;
- bool dump = r600_can_dump_shader(&rscreen->b,
- tgsi_get_processor_type(tokens));
- boolean use_kill = false;
-
- memset(&radeon_llvm_ctx, 0, sizeof(radeon_llvm_ctx));
- radeon_llvm_ctx.type = ctx.type;
- radeon_llvm_ctx.two_side = shader->two_side;
- radeon_llvm_ctx.face_gpr = ctx.face_gpr;
- radeon_llvm_ctx.inputs_count = ctx.shader->ninput + 1;
- radeon_llvm_ctx.r600_inputs = ctx.shader->input;
- radeon_llvm_ctx.r600_outputs = ctx.shader->output;
- radeon_llvm_ctx.color_buffer_count = max_color_exports;
- radeon_llvm_ctx.chip_class = ctx.bc->chip_class;
- radeon_llvm_ctx.fs_color_all = shader->fs_write_all && (rscreen->b.chip_class >= EVERGREEN);
- radeon_llvm_ctx.stream_outputs = &so;
- radeon_llvm_ctx.alpha_to_one = key.ps.alpha_to_one;
- radeon_llvm_ctx.has_compressed_msaa_texturing =
- ctx.bc->has_compressed_msaa_texturing;
- mod = r600_tgsi_llvm(&radeon_llvm_ctx, tokens);
- ctx.shader->has_txq_cube_array_z_comp = radeon_llvm_ctx.has_txq_cube_array_z_comp;
- ctx.shader->uses_tex_buffers = radeon_llvm_ctx.uses_tex_buffers;
-
- if (r600_llvm_compile(mod, rscreen->b.family, ctx.bc, &use_kill,
- dump, &rctx->b.debug)) {
- radeon_llvm_dispose(&radeon_llvm_ctx);
- use_llvm = 0;
- fprintf(stderr, "R600 LLVM backend failed to compile "
- "shader. Falling back to TGSI\n");
- } else {
- ctx.file_offset[TGSI_FILE_OUTPUT] =
- ctx.file_offset[TGSI_FILE_INPUT];
- }
- if (use_kill)
- ctx.shader->uses_kill = use_kill;
- radeon_llvm_dispose(&radeon_llvm_ctx);
- }
-#endif
-/* End of LLVM backend setup */
-
if (shader->fs_write_all && rscreen->b.chip_class >= EVERGREEN)
shader->nr_ps_max_color_exports = 8;
- if (!use_llvm) {
- if (ctx.fragcoord_input >= 0) {
- if (ctx.bc->chip_class == CAYMAN) {
- for (j = 0 ; j < 4; j++) {
- struct r600_bytecode_alu alu;
- memset(&alu, 0, sizeof(struct r600_bytecode_alu));
- alu.op = ALU_OP1_RECIP_IEEE;
- alu.src[0].sel = shader->input[ctx.fragcoord_input].gpr;
- alu.src[0].chan = 3;
-
- alu.dst.sel = shader->input[ctx.fragcoord_input].gpr;
- alu.dst.chan = j;
- alu.dst.write = (j == 3);
- alu.last = 1;
- if ((r = r600_bytecode_add_alu(ctx.bc, &alu)))
- return r;
- }
- } else {
+ if (ctx.fragcoord_input >= 0) {
+ if (ctx.bc->chip_class == CAYMAN) {
+ for (j = 0 ; j < 4; j++) {
struct r600_bytecode_alu alu;
memset(&alu, 0, sizeof(struct r600_bytecode_alu));
alu.op = ALU_OP1_RECIP_IEEE;
alu.src[0].chan = 3;
alu.dst.sel = shader->input[ctx.fragcoord_input].gpr;
- alu.dst.chan = 3;
- alu.dst.write = 1;
+ alu.dst.chan = j;
+ alu.dst.write = (j == 3);
alu.last = 1;
if ((r = r600_bytecode_add_alu(ctx.bc, &alu)))
return r;
}
- }
-
- if (ctx.type == TGSI_PROCESSOR_GEOMETRY) {
+ } else {
struct r600_bytecode_alu alu;
- int r;
-
- /* GS thread with no output workaround - emit a cut at start of GS */
- if (ctx.bc->chip_class == R600)
- r600_bytecode_add_cfinst(ctx.bc, CF_OP_CUT_VERTEX);
+ memset(&alu, 0, sizeof(struct r600_bytecode_alu));
+ alu.op = ALU_OP1_RECIP_IEEE;
+ alu.src[0].sel = shader->input[ctx.fragcoord_input].gpr;
+ alu.src[0].chan = 3;
- for (j = 0; j < 4; j++) {
- memset(&alu, 0, sizeof(struct r600_bytecode_alu));
- alu.op = ALU_OP1_MOV;
- alu.src[0].sel = V_SQ_ALU_SRC_LITERAL;
- alu.src[0].value = 0;
- alu.dst.sel = ctx.gs_export_gpr_tregs[j];
- alu.dst.write = 1;
- alu.last = 1;
- r = r600_bytecode_add_alu(ctx.bc, &alu);
- if (r)
- return r;
- }
+ alu.dst.sel = shader->input[ctx.fragcoord_input].gpr;
+ alu.dst.chan = 3;
+ alu.dst.write = 1;
+ alu.last = 1;
+ if ((r = r600_bytecode_add_alu(ctx.bc, &alu)))
+ return r;
}
+ }
- if (ctx.type == TGSI_PROCESSOR_TESS_CTRL)
- r600_fetch_tess_io_info(&ctx);
+ if (ctx.type == PIPE_SHADER_GEOMETRY) {
+ struct r600_bytecode_alu alu;
+ int r;
- if (shader->two_side && ctx.colors_used) {
- if ((r = process_twoside_color_inputs(&ctx)))
+ /* GS thread with no output workaround - emit a cut at start of GS */
+ if (ctx.bc->chip_class == R600)
+ r600_bytecode_add_cfinst(ctx.bc, CF_OP_CUT_VERTEX);
+
+ for (j = 0; j < 4; j++) {
+ memset(&alu, 0, sizeof(struct r600_bytecode_alu));
+ alu.op = ALU_OP1_MOV;
+ alu.src[0].sel = V_SQ_ALU_SRC_LITERAL;
+ alu.src[0].value = 0;
+ alu.dst.sel = ctx.gs_export_gpr_tregs[j];
+ alu.dst.write = 1;
+ alu.last = 1;
+ r = r600_bytecode_add_alu(ctx.bc, &alu);
+ if (r)
return r;
}
+ }
- tgsi_parse_init(&ctx.parse, tokens);
- while (!tgsi_parse_end_of_tokens(&ctx.parse)) {
- tgsi_parse_token(&ctx.parse);
- switch (ctx.parse.FullToken.Token.Type) {
- case TGSI_TOKEN_TYPE_INSTRUCTION:
- r = tgsi_is_supported(&ctx);
- if (r)
- goto out_err;
- ctx.max_driver_temp_used = 0;
- /* reserve first tmp for everyone */
- r600_get_temp(&ctx);
+ if (ctx.type == PIPE_SHADER_TESS_CTRL)
+ r600_fetch_tess_io_info(&ctx);
+
+ if (shader->two_side && ctx.colors_used) {
+ if ((r = process_twoside_color_inputs(&ctx)))
+ return r;
+ }
+
+ tgsi_parse_init(&ctx.parse, tokens);
+ while (!tgsi_parse_end_of_tokens(&ctx.parse)) {
+ tgsi_parse_token(&ctx.parse);
+ switch (ctx.parse.FullToken.Token.Type) {
+ case TGSI_TOKEN_TYPE_INSTRUCTION:
+ r = tgsi_is_supported(&ctx);
+ if (r)
+ goto out_err;
+ ctx.max_driver_temp_used = 0;
+ /* reserve first tmp for everyone */
+ r600_get_temp(&ctx);
- opcode = ctx.parse.FullToken.FullInstruction.Instruction.Opcode;
- if ((r = tgsi_split_constant(&ctx)))
+ opcode = ctx.parse.FullToken.FullInstruction.Instruction.Opcode;
+ if ((r = tgsi_split_constant(&ctx)))
+ goto out_err;
+ if ((r = tgsi_split_literal_constant(&ctx)))
+ goto out_err;
+ if (ctx.type == PIPE_SHADER_GEOMETRY) {
+ if ((r = tgsi_split_gs_inputs(&ctx)))
goto out_err;
- if ((r = tgsi_split_literal_constant(&ctx)))
+ } else if (lds_inputs) {
+ if ((r = tgsi_split_lds_inputs(&ctx)))
goto out_err;
- if (ctx.type == TGSI_PROCESSOR_GEOMETRY) {
- if ((r = tgsi_split_gs_inputs(&ctx)))
- goto out_err;
- } else if (lds_inputs) {
- if ((r = tgsi_split_lds_inputs(&ctx)))
- goto out_err;
- }
- if (ctx.bc->chip_class == CAYMAN)
- ctx.inst_info = &cm_shader_tgsi_instruction[opcode];
- else if (ctx.bc->chip_class >= EVERGREEN)
- ctx.inst_info = &eg_shader_tgsi_instruction[opcode];
- else
- ctx.inst_info = &r600_shader_tgsi_instruction[opcode];
- r = ctx.inst_info->process(&ctx);
+ }
+ if (ctx.bc->chip_class == CAYMAN)
+ ctx.inst_info = &cm_shader_tgsi_instruction[opcode];
+ else if (ctx.bc->chip_class >= EVERGREEN)
+ ctx.inst_info = &eg_shader_tgsi_instruction[opcode];
+ else
+ ctx.inst_info = &r600_shader_tgsi_instruction[opcode];
+ r = ctx.inst_info->process(&ctx);
+ if (r)
+ goto out_err;
+
+ if (ctx.type == PIPE_SHADER_TESS_CTRL) {
+ r = r600_store_tcs_output(&ctx);
if (r)
goto out_err;
-
- if (ctx.type == TGSI_PROCESSOR_TESS_CTRL) {
- r = r600_store_tcs_output(&ctx);
- if (r)
- goto out_err;
- }
- break;
- default:
- break;
}
+ break;
+ default:
+ break;
}
}
alu.dst.write = (j == ochan);
if (j == 3)
alu.last = 1;
- if (!use_llvm)
- r = r600_bytecode_add_alu(ctx.bc, &alu);
+ r = r600_bytecode_add_alu(ctx.bc, &alu);
if (r)
return r;
}
}
/* Add stream outputs. */
- if (!use_llvm && so.num_outputs) {
+ if (so.num_outputs) {
bool emit = false;
- if (!lds_outputs && !ring_outputs && ctx.type == TGSI_PROCESSOR_VERTEX)
+ if (!lds_outputs && !ring_outputs && ctx.type == PIPE_SHADER_VERTEX)
emit = true;
- if (!ring_outputs && ctx.type == TGSI_PROCESSOR_TESS_EVAL)
+ if (!ring_outputs && ctx.type == PIPE_SHADER_TESS_EVAL)
emit = true;
if (emit)
emit_streamout(&ctx, &so, -1, NULL);
pipeshader->enabled_stream_buffers_mask = ctx.enabled_stream_buffers_mask;
convert_edgeflag_to_int(&ctx);
- if (ctx.type == TGSI_PROCESSOR_TESS_CTRL)
+ if (ctx.type == PIPE_SHADER_TESS_CTRL)
r600_emit_tess_factor(&ctx);
if (lds_outputs) {
- if (ctx.type == TGSI_PROCESSOR_VERTEX) {
+ if (ctx.type == PIPE_SHADER_VERTEX) {
if (ctx.shader->noutput)
emit_lds_vs_writes(&ctx);
}
output[j].type = -1;
output[j].op = CF_OP_EXPORT;
switch (ctx.type) {
- case TGSI_PROCESSOR_VERTEX:
- case TGSI_PROCESSOR_TESS_EVAL:
+ case PIPE_SHADER_VERTEX:
+ case PIPE_SHADER_TESS_EVAL:
switch (shader->output[i].name) {
case TGSI_SEMANTIC_POSITION:
output[j].array_base = 60;
}
break;
- case TGSI_PROCESSOR_FRAGMENT:
+ case PIPE_SHADER_FRAGMENT:
if (shader->output[i].name == TGSI_SEMANTIC_COLOR) {
/* never export more colors than the number of CBs */
if (shader->output[i].sid >= max_color_exports) {
goto out_err;
}
break;
- case TGSI_PROCESSOR_TESS_CTRL:
+ case PIPE_SHADER_TESS_CTRL:
break;
default:
R600_ERR("unsupported processor type %d\n", ctx.type);
}
/* add fake position export */
- if ((ctx.type == TGSI_PROCESSOR_VERTEX || ctx.type == TGSI_PROCESSOR_TESS_EVAL) && pos_emitted == false) {
+ if ((ctx.type == PIPE_SHADER_VERTEX || ctx.type == PIPE_SHADER_TESS_EVAL) && pos_emitted == false) {
memset(&output[j], 0, sizeof(struct r600_bytecode_output));
output[j].gpr = 0;
output[j].elem_size = 3;
}
/* add fake param output for vertex shader if no param is exported */
- if ((ctx.type == TGSI_PROCESSOR_VERTEX || ctx.type == TGSI_PROCESSOR_TESS_EVAL) && next_param_base == 0) {
+ if ((ctx.type == PIPE_SHADER_VERTEX || ctx.type == PIPE_SHADER_TESS_EVAL) && next_param_base == 0) {
memset(&output[j], 0, sizeof(struct r600_bytecode_output));
output[j].gpr = 0;
output[j].elem_size = 3;
}
/* add fake pixel export */
- if (ctx.type == TGSI_PROCESSOR_FRAGMENT && shader->nr_ps_color_exports == 0) {
+ if (ctx.type == PIPE_SHADER_FRAGMENT && shader->nr_ps_color_exports == 0) {
memset(&output[j], 0, sizeof(struct r600_bytecode_output));
output[j].gpr = 0;
output[j].elem_size = 3;
}
}
/* add output to bytecode */
- if (!use_llvm) {
- for (i = 0; i < noutput; i++) {
- r = r600_bytecode_add_output(ctx.bc, &output[i]);
- if (r)
- goto out_err;
- }
+ for (i = 0; i < noutput; i++) {
+ r = r600_bytecode_add_output(ctx.bc, &output[i]);
+ if (r)
+ goto out_err;
}
}
/* add program end */
- if (!use_llvm) {
- if (ctx.bc->chip_class == CAYMAN)
- cm_bytecode_add_cf_end(ctx.bc);
- else {
- const struct cf_op_info *last = NULL;
+ if (ctx.bc->chip_class == CAYMAN)
+ cm_bytecode_add_cf_end(ctx.bc);
+ else {
+ const struct cf_op_info *last = NULL;
- if (ctx.bc->cf_last)
- last = r600_isa_cf(ctx.bc->cf_last->op);
+ if (ctx.bc->cf_last)
+ last = r600_isa_cf(ctx.bc->cf_last->op);
- /* alu clause instructions don't have EOP bit, so add NOP */
- if (!last || last->flags & CF_ALU || ctx.bc->cf_last->op == CF_OP_LOOP_END || ctx.bc->cf_last->op == CF_OP_CALL_FS || ctx.bc->cf_last->op == CF_OP_POP || ctx.bc->cf_last->op == CF_OP_GDS)
- r600_bytecode_add_cfinst(ctx.bc, CF_OP_NOP);
+ /* alu clause instructions don't have EOP bit, so add NOP */
+ if (!last || last->flags & CF_ALU || ctx.bc->cf_last->op == CF_OP_LOOP_END || ctx.bc->cf_last->op == CF_OP_CALL_FS || ctx.bc->cf_last->op == CF_OP_POP || ctx.bc->cf_last->op == CF_OP_GDS)
+ r600_bytecode_add_cfinst(ctx.bc, CF_OP_NOP);
- ctx.bc->cf_last->end_of_program = 1;
- }
+ ctx.bc->cf_last->end_of_program = 1;
}
/* check GPR limit - we have 124 = 128 - 4
goto out_err;
}
- if (ctx.type == TGSI_PROCESSOR_GEOMETRY) {
+ if (ctx.type == PIPE_SHADER_GEOMETRY) {
if ((r = generate_gs_copy_shader(rctx, pipeshader, &so)))
return r;
}
if (inst->Instruction.Saturate) {
r600_dst->clamp = 1;
}
- if (ctx->type == TGSI_PROCESSOR_TESS_CTRL) {
+ if (ctx->type == PIPE_SHADER_TESS_CTRL) {
if (tgsi_dst->Register.File == TGSI_FILE_OUTPUT) {
return;
}
/* handle some special cases */
if (i == 1 || i == 3) {
switch (ctx->parse.FullToken.FullInstruction.Instruction.Opcode) {
- case TGSI_OPCODE_SUB:
- r600_bytecode_src_toggle_neg(&alu.src[1]);
- break;
case TGSI_OPCODE_DABS:
r600_bytecode_src_set_abs(&alu.src[0]);
break;
int i, j, r, lasti = tgsi_last_instruction(write_mask);
/* use temp register if trans_only and more than one dst component */
int use_tmp = trans_only && (write_mask ^ (1 << lasti));
+ unsigned op = ctx->inst_info->op;
+
+ if (op == ALU_OP2_MUL_IEEE &&
+ ctx->info.properties[TGSI_PROPERTY_MUL_ZERO_WINS])
+ op = ALU_OP2_MUL;
for (i = 0; i <= lasti; i++) {
if (!(write_mask & (1 << i)))
} else
tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
- alu.op = ctx->inst_info->op;
+ alu.op = op;
if (!swap) {
for (j = 0; j < inst->Instruction.NumSrcRegs; j++) {
r600_bytecode_src(&alu.src[j], &ctx->src[j], i);
r600_bytecode_src(&alu.src[0], &ctx->src[1], i);
r600_bytecode_src(&alu.src[1], &ctx->src[0], i);
}
- /* handle some special cases */
- switch (inst->Instruction.Opcode) {
- case TGSI_OPCODE_SUB:
- r600_bytecode_src_toggle_neg(&alu.src[1]);
- break;
- case TGSI_OPCODE_ABS:
- r600_bytecode_src_set_abs(&alu.src[0]);
- break;
- default:
- break;
- }
if (i == lasti || trans_only) {
alu.last = 1;
}
struct r600_bytecode_alu alu;
unsigned write_mask = inst->Dst[0].Register.WriteMask;
int i, j, r;
- int firsti = write_mask == 0xc ? 2 : 0;
for (i = 0; i <= 3; i++) {
memset(&alu, 0, sizeof(struct r600_bytecode_alu));
return r;
}
- /* MOV first two channels to writemask dst0 */
- for (i = 0; i <= 1; i++) {
+ /* Replicate significand result across channels. */
+ for (i = 0; i <= 3; i++) {
+ if (!(write_mask & (1 << i)))
+ continue;
+
memset(&alu, 0, sizeof(struct r600_bytecode_alu));
alu.op = ALU_OP1_MOV;
- alu.src[0].chan = i + 2;
+ alu.src[0].chan = (i & 1) + 2;
alu.src[0].sel = ctx->temp_reg;
- tgsi_dst(ctx, &inst->Dst[0], firsti + i, &alu.dst);
- alu.dst.write = (inst->Dst[0].Register.WriteMask >> (firsti + i)) & 1;
+ tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
+ alu.dst.write = 1;
alu.last = 1;
r = r600_bytecode_add_alu(ctx->bc, &alu);
if (r)
return 0;
}
-static int cayman_emit_double_instr(struct r600_shader_ctx *ctx)
+static int cayman_emit_unary_double_raw(struct r600_bytecode *bc,
+ unsigned op,
+ int dst_reg,
+ struct r600_shader_src *src,
+ bool abs)
{
- struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
- int i, r;
struct r600_bytecode_alu alu;
- int last_slot = 3;
- int lasti = tgsi_last_instruction(inst->Dst[0].Register.WriteMask);
- int t1 = ctx->temp_reg;
+ const int last_slot = 3;
+ int r;
/* these have to write the result to X/Y by the looks of it */
- for (i = 0 ; i < last_slot; i++) {
+ for (int i = 0 ; i < last_slot; i++) {
memset(&alu, 0, sizeof(struct r600_bytecode_alu));
- alu.op = ctx->inst_info->op;
+ alu.op = op;
- /* should only be one src regs */
- assert (inst->Instruction.NumSrcRegs == 1);
+ r600_bytecode_src(&alu.src[0], src, 1);
+ r600_bytecode_src(&alu.src[1], src, 0);
- r600_bytecode_src(&alu.src[0], &ctx->src[0], 1);
- r600_bytecode_src(&alu.src[1], &ctx->src[0], 0);
-
- /* RSQ should take the absolute value of src */
- if (ctx->parse.FullToken.FullInstruction.Instruction.Opcode == TGSI_OPCODE_DRSQ ||
- ctx->parse.FullToken.FullInstruction.Instruction.Opcode == TGSI_OPCODE_DSQRT) {
+ if (abs)
r600_bytecode_src_set_abs(&alu.src[1]);
- }
- alu.dst.sel = t1;
+
+ alu.dst.sel = dst_reg;
alu.dst.chan = i;
alu.dst.write = (i == 0 || i == 1);
- if (ctx->bc->chip_class != CAYMAN || i == last_slot - 1)
+ if (bc->chip_class != CAYMAN || i == last_slot - 1)
alu.last = 1;
- r = r600_bytecode_add_alu(ctx->bc, &alu);
+ r = r600_bytecode_add_alu(bc, &alu);
if (r)
return r;
}
+ return 0;
+}
+
+static int cayman_emit_double_instr(struct r600_shader_ctx *ctx)
+{
+ struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+ int i, r;
+ struct r600_bytecode_alu alu;
+ int lasti = tgsi_last_instruction(inst->Dst[0].Register.WriteMask);
+ int t1 = ctx->temp_reg;
+
+ /* should only be one src regs */
+ assert(inst->Instruction.NumSrcRegs == 1);
+
+ /* only support one double at a time */
+ assert(inst->Dst[0].Register.WriteMask == TGSI_WRITEMASK_XY ||
+ inst->Dst[0].Register.WriteMask == TGSI_WRITEMASK_ZW);
+
+ r = cayman_emit_unary_double_raw(
+ ctx->bc, ctx->inst_info->op, t1,
+ &ctx->src[0],
+ ctx->parse.FullToken.FullInstruction.Instruction.Opcode == TGSI_OPCODE_DRSQ ||
+ ctx->parse.FullToken.FullInstruction.Instruction.Opcode == TGSI_OPCODE_DSQRT);
+ if (r)
+ return r;
+
for (i = 0 ; i <= lasti; i++) {
if (!(inst->Dst[0].Register.WriteMask & (1 << i)))
continue;
int i, j, r;
struct r600_bytecode_alu alu;
int last_slot = (inst->Dst[0].Register.WriteMask & 0x8) ? 4 : 3;
-
+
for (i = 0 ; i < last_slot; i++) {
memset(&alu, 0, sizeof(struct r600_bytecode_alu));
alu.op = ctx->inst_info->op;
int lasti = tgsi_last_instruction(inst->Dst[0].Register.WriteMask);
int t1 = ctx->temp_reg;
- for (k = 0; k < 2; k++) {
- if (!(inst->Dst[0].Register.WriteMask & (0x3 << (k * 2))))
- continue;
+ /* t1 would get overwritten below if we actually tried to
+ * multiply two pairs of doubles at a time. */
+ assert(inst->Dst[0].Register.WriteMask == TGSI_WRITEMASK_XY ||
+ inst->Dst[0].Register.WriteMask == TGSI_WRITEMASK_ZW);
- for (i = 0; i < 4; i++) {
- memset(&alu, 0, sizeof(struct r600_bytecode_alu));
- alu.op = ctx->inst_info->op;
- for (j = 0; j < inst->Instruction.NumSrcRegs; j++) {
- r600_bytecode_src(&alu.src[j], &ctx->src[j], k * 2 + ((i == 3) ? 0 : 1));
- }
- alu.dst.sel = t1;
- alu.dst.chan = i;
- alu.dst.write = 1;
- if (i == 3)
- alu.last = 1;
- r = r600_bytecode_add_alu(ctx->bc, &alu);
- if (r)
- return r;
+ k = inst->Dst[0].Register.WriteMask == TGSI_WRITEMASK_XY ? 0 : 1;
+
+ for (i = 0; i < 4; i++) {
+ memset(&alu, 0, sizeof(struct r600_bytecode_alu));
+ alu.op = ctx->inst_info->op;
+ for (j = 0; j < inst->Instruction.NumSrcRegs; j++) {
+ r600_bytecode_src(&alu.src[j], &ctx->src[j], k * 2 + ((i == 3) ? 0 : 1));
}
+ alu.dst.sel = t1;
+ alu.dst.chan = i;
+ alu.dst.write = 1;
+ if (i == 3)
+ alu.last = 1;
+ r = r600_bytecode_add_alu(ctx->bc, &alu);
+ if (r)
+ return r;
}
for (i = 0; i <= lasti; i++) {
return 0;
}
+/*
+ * Emit RECIP_64 + MUL_64 to implement division.
+ */
+static int cayman_ddiv_instr(struct r600_shader_ctx *ctx)
+{
+ struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+ int r;
+ struct r600_bytecode_alu alu;
+ int t1 = ctx->temp_reg;
+ int k;
+
+ /* Only support one double at a time. This is the same constraint as
+ * in DMUL lowering. */
+ assert(inst->Dst[0].Register.WriteMask == TGSI_WRITEMASK_XY ||
+ inst->Dst[0].Register.WriteMask == TGSI_WRITEMASK_ZW);
+
+ k = inst->Dst[0].Register.WriteMask == TGSI_WRITEMASK_XY ? 0 : 1;
+
+ r = cayman_emit_unary_double_raw(ctx->bc, ALU_OP2_RECIP_64, t1, &ctx->src[1], false);
+ if (r)
+ return r;
+
+ for (int i = 0; i < 4; i++) {
+ memset(&alu, 0, sizeof(struct r600_bytecode_alu));
+ alu.op = ALU_OP2_MUL_64;
+
+ r600_bytecode_src(&alu.src[0], &ctx->src[0], k * 2 + ((i == 3) ? 0 : 1));
+
+ alu.src[1].sel = t1;
+ alu.src[1].chan = (i == 3) ? 0 : 1;
+
+ alu.dst.sel = t1;
+ alu.dst.chan = i;
+ alu.dst.write = 1;
+ if (i == 3)
+ alu.last = 1;
+ r = r600_bytecode_add_alu(ctx->bc, &alu);
+ if (r)
+ return r;
+ }
+
+ for (int i = 0; i < 2; i++) {
+ memset(&alu, 0, sizeof(struct r600_bytecode_alu));
+ alu.op = ALU_OP1_MOV;
+ alu.src[0].sel = t1;
+ alu.src[0].chan = i;
+ tgsi_dst(ctx, &inst->Dst[0], k * 2 + i, &alu.dst);
+ alu.dst.write = 1;
+ if (i == 1)
+ alu.last = 1;
+ r = r600_bytecode_add_alu(ctx->bc, &alu);
+ if (r)
+ return r;
+ }
+ return 0;
+}
+
/*
* r600 - trunc to -PI..PI range
* r700 - normalize by dividing by 2PI
*/
static int tgsi_setup_trig(struct r600_shader_ctx *ctx)
{
- static float half_inv_pi = 1.0 /(3.1415926535 * 2);
- static float double_pi = 3.1415926535 * 2;
- static float neg_pi = -3.1415926535;
-
int r;
struct r600_bytecode_alu alu;
alu.src[1].sel = V_SQ_ALU_SRC_LITERAL;
alu.src[1].chan = 0;
- alu.src[1].value = *(uint32_t *)&half_inv_pi;
+ alu.src[1].value = u_bitcast_f2u(0.5f * M_1_PI);
alu.src[2].sel = V_SQ_ALU_SRC_0_5;
alu.src[2].chan = 0;
alu.last = 1;
alu.src[2].chan = 0;
if (ctx->bc->chip_class == R600) {
- alu.src[1].value = *(uint32_t *)&double_pi;
- alu.src[2].value = *(uint32_t *)&neg_pi;
+ alu.src[1].value = u_bitcast_f2u(2.0f * M_PI);
+ alu.src[2].value = u_bitcast_f2u(-M_PI);
} else {
alu.src[1].sel = V_SQ_ALU_SRC_1;
alu.src[2].sel = V_SQ_ALU_SRC_0_5;
return 0;
}
-static int tgsi_scs(struct r600_shader_ctx *ctx)
+static int tgsi_kill(struct r600_shader_ctx *ctx)
{
- struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+ const struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
struct r600_bytecode_alu alu;
int i, r;
- /* We'll only need the trig stuff if we are going to write to the
- * X or Y components of the destination vector.
- */
- if (likely(inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_XY)) {
- r = tgsi_setup_trig(ctx);
- if (r)
- return r;
- }
-
- /* dst.x = COS */
- if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_X) {
- if (ctx->bc->chip_class == CAYMAN) {
- for (i = 0 ; i < 3; i++) {
- memset(&alu, 0, sizeof(struct r600_bytecode_alu));
- alu.op = ALU_OP1_COS;
- tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
-
- if (i == 0)
- alu.dst.write = 1;
- else
- alu.dst.write = 0;
- alu.src[0].sel = ctx->temp_reg;
- alu.src[0].chan = 0;
- if (i == 2)
- alu.last = 1;
- r = r600_bytecode_add_alu(ctx->bc, &alu);
- if (r)
- return r;
- }
- } else {
- memset(&alu, 0, sizeof(struct r600_bytecode_alu));
- alu.op = ALU_OP1_COS;
- tgsi_dst(ctx, &inst->Dst[0], 0, &alu.dst);
-
- alu.src[0].sel = ctx->temp_reg;
- alu.src[0].chan = 0;
- alu.last = 1;
- r = r600_bytecode_add_alu(ctx->bc, &alu);
- if (r)
- return r;
- }
- }
-
- /* dst.y = SIN */
- if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Y) {
- if (ctx->bc->chip_class == CAYMAN) {
- for (i = 0 ; i < 3; i++) {
- memset(&alu, 0, sizeof(struct r600_bytecode_alu));
- alu.op = ALU_OP1_SIN;
- tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
- if (i == 1)
- alu.dst.write = 1;
- else
- alu.dst.write = 0;
- alu.src[0].sel = ctx->temp_reg;
- alu.src[0].chan = 0;
- if (i == 2)
- alu.last = 1;
- r = r600_bytecode_add_alu(ctx->bc, &alu);
- if (r)
- return r;
- }
- } else {
- memset(&alu, 0, sizeof(struct r600_bytecode_alu));
- alu.op = ALU_OP1_SIN;
- tgsi_dst(ctx, &inst->Dst[0], 1, &alu.dst);
-
- alu.src[0].sel = ctx->temp_reg;
- alu.src[0].chan = 0;
- alu.last = 1;
- r = r600_bytecode_add_alu(ctx->bc, &alu);
- if (r)
- return r;
- }
- }
-
- /* dst.z = 0.0; */
- if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Z) {
- memset(&alu, 0, sizeof(struct r600_bytecode_alu));
-
- alu.op = ALU_OP1_MOV;
-
- tgsi_dst(ctx, &inst->Dst[0], 2, &alu.dst);
-
- alu.src[0].sel = V_SQ_ALU_SRC_0;
- alu.src[0].chan = 0;
-
- alu.last = 1;
-
- r = r600_bytecode_add_alu(ctx->bc, &alu);
- if (r)
- return r;
- }
-
- /* dst.w = 1.0; */
- if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_W) {
- memset(&alu, 0, sizeof(struct r600_bytecode_alu));
-
- alu.op = ALU_OP1_MOV;
-
- tgsi_dst(ctx, &inst->Dst[0], 3, &alu.dst);
-
- alu.src[0].sel = V_SQ_ALU_SRC_1;
- alu.src[0].chan = 0;
-
- alu.last = 1;
-
- r = r600_bytecode_add_alu(ctx->bc, &alu);
- if (r)
- return r;
- }
-
- return 0;
-}
-
-static int tgsi_kill(struct r600_shader_ctx *ctx)
-{
- const struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
- struct r600_bytecode_alu alu;
- int i, r;
-
- for (i = 0; i < 4; i++) {
- memset(&alu, 0, sizeof(struct r600_bytecode_alu));
- alu.op = ctx->inst_info->op;
-
- alu.dst.chan = i;
-
- alu.src[0].sel = V_SQ_ALU_SRC_0;
-
- if (inst->Instruction.Opcode == TGSI_OPCODE_KILL) {
- alu.src[1].sel = V_SQ_ALU_SRC_1;
- alu.src[1].neg = 1;
- } else {
- r600_bytecode_src(&alu.src[1], &ctx->src[0], i);
- }
- if (i == 3) {
- alu.last = 1;
- }
- r = r600_bytecode_add_alu(ctx->bc, &alu);
+ for (i = 0; i < 4; i++) {
+ memset(&alu, 0, sizeof(struct r600_bytecode_alu));
+ alu.op = ctx->inst_info->op;
+
+ alu.dst.chan = i;
+
+ alu.src[0].sel = V_SQ_ALU_SRC_0;
+
+ if (inst->Instruction.Opcode == TGSI_OPCODE_KILL) {
+ alu.src[1].sel = V_SQ_ALU_SRC_1;
+ alu.src[1].neg = 1;
+ } else {
+ r600_bytecode_src(&alu.src[1], &ctx->src[0], i);
+ }
+ if (i == 3) {
+ alu.last = 1;
+ }
+ r = r600_bytecode_add_alu(ctx->bc, &alu);
if (r)
return r;
}
{
int chan;
int sel;
- int i;
+ unsigned i;
if (ctx->bc->chip_class == CAYMAN) {
for (i = 0; i < 3; i++) {
alu.last = 1;
} else
alu.dst.write = 0;
-
+
r = r600_bytecode_add_alu(ctx->bc, &alu);
if (r)
return r;
memset(&alu, 0, sizeof(struct r600_bytecode_alu));
alu.op = ALU_OP1_FLT_TO_UINT;
-
+
alu.dst.sel = tmp0;
alu.dst.chan = 0;
alu.dst.write = 1;
} else {
r600_bytecode_src(&alu.src[1], &ctx->src[1], i);
}
-
+
alu.last = 1;
if ((r = r600_bytecode_add_alu(ctx->bc, &alu)))
return r;
} else {
r600_bytecode_src(&alu.src[0], &ctx->src[1], i);
}
-
+
alu.src[1].sel = tmp0;
alu.src[1].chan = 2;
struct r600_bytecode_alu alu;
int r, i = 0, k, interp_gpr, interp_base_chan, tmp, lasti;
unsigned location;
- int input;
+ const int input = inst->Src[0].Register.Index + ctx->shader->nsys_inputs;
assert(inst->Src[0].Register.File == TGSI_FILE_INPUT);
- input = inst->Src[0].Register.Index;
-
/* Interpolators have been marked for use already by allocate_system_value_inputs */
if (inst->Instruction.Opcode == TGSI_OPCODE_INTERP_OFFSET ||
inst->Instruction.Opcode == TGSI_OPCODE_INTERP_SAMPLE) {
int i, j, r;
int lasti = tgsi_last_instruction(inst->Dst[0].Register.WriteMask);
int temp_regs[4];
+ unsigned op = ctx->inst_info->op;
+
+ if (op == ALU_OP3_MULADD_IEEE &&
+ ctx->info.properties[TGSI_PROPERTY_MUL_ZERO_WINS])
+ op = ALU_OP3_MULADD;
for (j = 0; j < inst->Instruction.NumSrcRegs; j++) {
temp_regs[j] = 0;
continue;
memset(&alu, 0, sizeof(struct r600_bytecode_alu));
- alu.op = ctx->inst_info->op;
+ alu.op = op;
for (j = 0; j < inst->Instruction.NumSrcRegs; j++) {
r = tgsi_make_src_for_op3(ctx, temp_regs[j], i, &alu.src[j], &ctx->src[j]);
if (r)
struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
struct r600_bytecode_alu alu;
int i, j, r;
+ unsigned op = ctx->inst_info->op;
+ if (op == ALU_OP2_DOT4_IEEE &&
+ ctx->info.properties[TGSI_PROPERTY_MUL_ZERO_WINS])
+ op = ALU_OP2_DOT4;
for (i = 0; i < 4; i++) {
memset(&alu, 0, sizeof(struct r600_bytecode_alu));
- alu.op = ctx->inst_info->op;
+ alu.op = op;
for (j = 0; j < inst->Instruction.NumSrcRegs; j++) {
r600_bytecode_src(&alu.src[j], &ctx->src[j], i);
}
alu.src[0].chan = alu.src[1].chan = 0;
}
break;
- case TGSI_OPCODE_DPH:
- if (i == 3) {
- alu.src[0].sel = V_SQ_ALU_SRC_1;
- alu.src[0].chan = 0;
- alu.src[0].neg = 0;
- }
- break;
default:
break;
}
inst->Src[index].Register.File != TGSI_FILE_INPUT &&
inst->Src[index].Register.File != TGSI_FILE_OUTPUT) ||
ctx->src[index].neg || ctx->src[index].abs ||
- (inst->Src[index].Register.File == TGSI_FILE_INPUT && ctx->type == TGSI_PROCESSOR_GEOMETRY);
+ (inst->Src[index].Register.File == TGSI_FILE_INPUT && ctx->type == PIPE_SHADER_GEOMETRY);
}
static inline unsigned tgsi_tex_get_src_gpr(struct r600_shader_ctx *ctx,
static int tgsi_tex(struct r600_shader_ctx *ctx)
{
- static float one_point_five = 1.5f;
struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
struct r600_bytecode_tex tex;
struct r600_bytecode_alu alu;
/* Texture fetch instructions can only use gprs as source.
* Also they cannot negate the source or take the absolute value */
- const boolean src_requires_loading = (inst->Instruction.Opcode != TGSI_OPCODE_TXQ_LZ &&
- inst->Instruction.Opcode != TGSI_OPCODE_TXQS &&
+ const boolean src_requires_loading = (inst->Instruction.Opcode != TGSI_OPCODE_TXQS &&
tgsi_tex_src_requires_loading(ctx, 0)) ||
read_compressed_msaa || txf_add_offsets;
boolean src_loaded = FALSE;
- unsigned sampler_src_reg = inst->Instruction.Opcode == TGSI_OPCODE_TXQ_LZ ? 0 : 1;
+ unsigned sampler_src_reg = 1;
int8_t offset_x = 0, offset_y = 0, offset_z = 0;
boolean has_txq_cube_array_z = false;
unsigned sampler_index_mode;
inst->Texture.Texture == TGSI_TEXTURE_CUBE_ARRAY ||
inst->Texture.Texture == TGSI_TEXTURE_SHADOWCUBE ||
inst->Texture.Texture == TGSI_TEXTURE_SHADOWCUBE_ARRAY) &&
- inst->Instruction.Opcode != TGSI_OPCODE_TXQ &&
- inst->Instruction.Opcode != TGSI_OPCODE_TXQ_LZ) {
+ inst->Instruction.Opcode != TGSI_OPCODE_TXQ) {
static const unsigned src0_swizzle[] = {2, 2, 0, 1};
static const unsigned src1_swizzle[] = {1, 0, 2, 2};
alu.src[2].sel = V_SQ_ALU_SRC_LITERAL;
alu.src[2].chan = 0;
- alu.src[2].value = *(uint32_t *)&one_point_five;
+ alu.src[2].value = u_bitcast_f2u(1.5f);
alu.dst.sel = ctx->temp_reg;
alu.dst.chan = 0;
alu.src[2].sel = V_SQ_ALU_SRC_LITERAL;
alu.src[2].chan = 0;
- alu.src[2].value = *(uint32_t *)&one_point_five;
+ alu.src[2].value = u_bitcast_f2u(1.5f);
alu.dst.sel = ctx->temp_reg;
alu.dst.chan = 1;
r = r600_bytecode_add_alu(ctx->bc, &alu);
if (r)
return r;
- /* write initial compare value into Z component
+ /* write initial compare value into Z component
- W src 0 for shadow cube
- X src 1 for shadow cube array */
if (inst->Texture.Texture == TGSI_TEXTURE_SHADOWCUBE ||
inst->Texture.Texture == TGSI_TEXTURE_SHADOWCUBE_ARRAY) {
if (ctx->bc->chip_class >= EVERGREEN) {
int mytmp = r600_get_temp(ctx);
- static const float eight = 8.0f;
memset(&alu, 0, sizeof(struct r600_bytecode_alu));
alu.op = ALU_OP1_MOV;
alu.src[0].sel = ctx->temp_reg;
r600_bytecode_src(&alu.src[0], &ctx->src[0], 3);
alu.src[1].sel = V_SQ_ALU_SRC_LITERAL;
alu.src[1].chan = 0;
- alu.src[1].value = *(uint32_t *)&eight;
+ alu.src[1].value = u_bitcast_f2u(8.0f);
alu.src[2].sel = mytmp;
alu.src[2].chan = 0;
alu.dst.sel = ctx->temp_reg;
r = r600_bytecode_add_alu(ctx->bc, &alu);
if (r)
return r;
-
+
r = r600_bytecode_add_tex(ctx->bc, &tex);
if (r)
return r;
/* does this shader want a num layers from TXQ for a cube array? */
if (has_txq_cube_array_z) {
int id = tgsi_tex_get_src_gpr(ctx, sampler_src_reg);
-
+
memset(&alu, 0, sizeof(struct r600_bytecode_alu));
alu.op = ALU_OP1_MOV;
}
- if (inst->Instruction.Opcode == TGSI_OPCODE_TXQ_LZ ||
- inst->Instruction.Opcode == TGSI_OPCODE_TXQS) {
+ if (inst->Instruction.Opcode == TGSI_OPCODE_TXQS) {
tex.src_sel_x = 4;
tex.src_sel_y = 4;
tex.src_sel_z = 4;
return 0;
}
+static int find_hw_atomic_counter(struct r600_shader_ctx *ctx,
+ struct tgsi_full_src_register *src)
+{
+ int i;
+
+ if (src->Register.Indirect) {
+ for (i = 0; i < ctx->shader->nhwatomic_ranges; i++) {
+ if (src->Indirect.ArrayID == ctx->shader->atomics[i].array_id)
+ return ctx->shader->atomics[i].hw_idx;
+ }
+ } else {
+ uint32_t index = src->Register.Index;
+ for (i = 0; i < ctx->shader->nhwatomic_ranges; i++) {
+ if (ctx->shader->atomics[i].buffer_id != src->Dimension.Index)
+ continue;
+ if (index > ctx->shader->atomics[i].end)
+ continue;
+ if (index < ctx->shader->atomics[i].start)
+ continue;
+ uint32_t offset = (index - ctx->shader->atomics[i].start);
+ return ctx->shader->atomics[i].hw_idx + offset;
+ }
+ }
+ assert(0);
+ return -1;
+}
+
+
+static int tgsi_load_gds(struct r600_shader_ctx *ctx)
+{
+ struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+ int r;
+ struct r600_bytecode_gds gds;
+ int uav_id = 0;
+ int uav_index_mode = 0;
+
+ uav_id = find_hw_atomic_counter(ctx, &inst->Src[0]);
+
+ if (inst->Src[0].Register.Indirect)
+ uav_index_mode = 2;
+
+ memset(&gds, 0, sizeof(struct r600_bytecode_gds));
+ gds.op = FETCH_OP_GDS_READ_RET;
+ gds.dst_gpr = ctx->file_offset[inst->Dst[0].Register.File] + inst->Dst[0].Register.Index;
+ gds.uav_id = uav_id;
+ gds.uav_index_mode = uav_index_mode;
+ gds.src_gpr = ctx->temp_reg;
+ gds.src_sel_x = 4;
+ gds.src_sel_y = 4;
+ gds.src_sel_z = 4;
+ gds.dst_sel_x = 0;
+ gds.dst_sel_y = 7;
+ gds.dst_sel_z = 7;
+ gds.dst_sel_w = 7;
+ gds.src_gpr2 = ctx->temp_reg;
+ gds.alloc_consume = 1;
+ r = r600_bytecode_add_gds(ctx->bc, &gds);
+ if (r)
+ return r;
+
+ ctx->bc->cf_last->vpm = 1;
+ return 0;
+}
+
+static int tgsi_load(struct r600_shader_ctx *ctx)
+{
+ struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+ if (inst->Src[0].Register.File == TGSI_FILE_HW_ATOMIC)
+ return tgsi_load_gds(ctx);
+ return 0;
+}
+
+static int get_gds_op(int opcode)
+{
+ switch (opcode) {
+ case TGSI_OPCODE_ATOMUADD:
+ return FETCH_OP_GDS_ADD_RET;
+ case TGSI_OPCODE_ATOMAND:
+ return FETCH_OP_GDS_AND_RET;
+ case TGSI_OPCODE_ATOMOR:
+ return FETCH_OP_GDS_OR_RET;
+ case TGSI_OPCODE_ATOMXOR:
+ return FETCH_OP_GDS_XOR_RET;
+ case TGSI_OPCODE_ATOMUMIN:
+ return FETCH_OP_GDS_MIN_UINT_RET;
+ case TGSI_OPCODE_ATOMUMAX:
+ return FETCH_OP_GDS_MAX_UINT_RET;
+ case TGSI_OPCODE_ATOMXCHG:
+ return FETCH_OP_GDS_XCHG_RET;
+ case TGSI_OPCODE_ATOMCAS:
+ return FETCH_OP_GDS_CMP_XCHG_RET;
+ default:
+ return -1;
+ }
+}
+
+static int tgsi_atomic_op_gds(struct r600_shader_ctx *ctx)
+{
+ struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+ struct r600_bytecode_gds gds;
+ struct r600_bytecode_alu alu;
+ int gds_op = get_gds_op(inst->Instruction.Opcode);
+ int r;
+ int uav_id = 0;
+ int uav_index_mode = 0;
+
+ if (gds_op == -1) {
+ fprintf(stderr, "unknown GDS op for opcode %d\n", inst->Instruction.Opcode);
+ return -1;
+ }
+
+ uav_id = find_hw_atomic_counter(ctx, &inst->Src[0]);
+
+ if (inst->Src[0].Register.Indirect)
+ uav_index_mode = 2;
+
+ if (inst->Src[2].Register.File == TGSI_FILE_IMMEDIATE) {
+ int value = (ctx->literals[4 * inst->Src[2].Register.Index + inst->Src[2].Register.SwizzleX]);
+ int abs_value = abs(value);
+ if (abs_value != value && gds_op == FETCH_OP_GDS_ADD_RET)
+ gds_op = FETCH_OP_GDS_SUB_RET;
+ memset(&alu, 0, sizeof(struct r600_bytecode_alu));
+ alu.op = ALU_OP1_MOV;
+ alu.dst.sel = ctx->temp_reg;
+ alu.dst.chan = 0;
+ alu.src[0].sel = V_SQ_ALU_SRC_LITERAL;
+ alu.src[0].value = abs_value;
+ alu.last = 1;
+ alu.dst.write = 1;
+ r = r600_bytecode_add_alu(ctx->bc, &alu);
+ if (r)
+ return r;
+ } else {
+ memset(&alu, 0, sizeof(struct r600_bytecode_alu));
+ alu.op = ALU_OP1_MOV;
+ alu.dst.sel = ctx->temp_reg;
+ alu.dst.chan = 0;
+ r600_bytecode_src(&alu.src[0], &ctx->src[2], 0);
+ alu.last = 1;
+ alu.dst.write = 1;
+ r = r600_bytecode_add_alu(ctx->bc, &alu);
+ if (r)
+ return r;
+ }
+
+ memset(&gds, 0, sizeof(struct r600_bytecode_gds));
+ gds.op = gds_op;
+ gds.dst_gpr = ctx->file_offset[inst->Dst[0].Register.File] + inst->Dst[0].Register.Index;
+ gds.uav_id = uav_id;
+ gds.uav_index_mode = uav_index_mode;
+ gds.src_gpr = ctx->temp_reg;
+ gds.src_gpr2 = ctx->temp_reg;
+ gds.src_sel_x = 4;
+ gds.src_sel_y = 0;
+ gds.src_sel_z = 4;
+ gds.dst_sel_x = 0;
+ gds.dst_sel_y = 7;
+ gds.dst_sel_z = 7;
+ gds.dst_sel_w = 7;
+ gds.alloc_consume = 1;
+ r = r600_bytecode_add_gds(ctx->bc, &gds);
+ if (r)
+ return r;
+ ctx->bc->cf_last->vpm = 1;
+ return 0;
+}
+
+static int tgsi_atomic_op(struct r600_shader_ctx *ctx)
+{
+ struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+ if (inst->Src[0].Register.File == TGSI_FILE_HW_ATOMIC)
+ return tgsi_atomic_op_gds(ctx);
+ return 0;
+}
+
static int tgsi_lrp(struct r600_shader_ctx *ctx)
{
struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
int i, r, j;
int lasti = tgsi_last_instruction(inst->Dst[0].Register.WriteMask);
int temp_regs[3];
+ unsigned op;
+
+ if (ctx->src[0].abs && ctx->src[0].neg) {
+ op = ALU_OP3_CNDE;
+ ctx->src[0].abs = 0;
+ ctx->src[0].neg = 0;
+ } else {
+ op = ALU_OP3_CNDGE;
+ }
for (j = 0; j < inst->Instruction.NumSrcRegs; j++) {
temp_regs[j] = 0;
continue;
memset(&alu, 0, sizeof(struct r600_bytecode_alu));
- alu.op = ALU_OP3_CNDGE;
+ alu.op = op;
r = tgsi_make_src_for_op3(ctx, temp_regs[0], i, &alu.src[0], &ctx->src[0]);
if (r)
return r;
return 0;
}
-static int tgsi_xpd(struct r600_shader_ctx *ctx)
-{
- struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
- static const unsigned int src0_swizzle[] = {2, 0, 1};
- static const unsigned int src1_swizzle[] = {1, 2, 0};
- struct r600_bytecode_alu alu;
- uint32_t use_temp = 0;
- int i, r;
-
- if (inst->Dst[0].Register.WriteMask != 0xf)
- use_temp = 1;
-
- for (i = 0; i < 4; i++) {
- memset(&alu, 0, sizeof(struct r600_bytecode_alu));
- alu.op = ALU_OP2_MUL;
- if (i < 3) {
- r600_bytecode_src(&alu.src[0], &ctx->src[0], src0_swizzle[i]);
- r600_bytecode_src(&alu.src[1], &ctx->src[1], src1_swizzle[i]);
- } else {
- alu.src[0].sel = V_SQ_ALU_SRC_0;
- alu.src[0].chan = i;
- alu.src[1].sel = V_SQ_ALU_SRC_0;
- alu.src[1].chan = i;
- }
-
- alu.dst.sel = ctx->temp_reg;
- alu.dst.chan = i;
- alu.dst.write = 1;
-
- if (i == 3)
- alu.last = 1;
- r = r600_bytecode_add_alu(ctx->bc, &alu);
- if (r)
- return r;
- }
-
- for (i = 0; i < 4; i++) {
- memset(&alu, 0, sizeof(struct r600_bytecode_alu));
- alu.op = ALU_OP3_MULADD;
-
- if (i < 3) {
- r600_bytecode_src(&alu.src[0], &ctx->src[0], src1_swizzle[i]);
- r600_bytecode_src(&alu.src[1], &ctx->src[1], src0_swizzle[i]);
- } else {
- alu.src[0].sel = V_SQ_ALU_SRC_0;
- alu.src[0].chan = i;
- alu.src[1].sel = V_SQ_ALU_SRC_0;
- alu.src[1].chan = i;
- }
-
- alu.src[2].sel = ctx->temp_reg;
- alu.src[2].neg = 1;
- alu.src[2].chan = i;
-
- if (use_temp)
- alu.dst.sel = ctx->temp_reg;
- else
- tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
- alu.dst.chan = i;
- alu.dst.write = 1;
- alu.is_op3 = 1;
- if (i == 3)
- alu.last = 1;
- r = r600_bytecode_add_alu(ctx->bc, &alu);
- if (r)
- return r;
- }
- if (use_temp)
- return tgsi_helper_copy(ctx, inst);
- return 0;
-}
-
static int tgsi_exp(struct r600_shader_ctx *ctx)
{
struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
struct r600_bytecode_alu alu;
int r;
- int i;
+ unsigned i;
/* result.x = 2^floor(src); */
if (inst->Dst[0].Register.WriteMask & 1) {
struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
struct r600_bytecode_alu alu;
int r;
- int i;
+ unsigned i;
/* result.x = floor(log2(|src|)); */
if (inst->Dst[0].Register.WriteMask & 1) {
alu.op = ALU_OP1_LOG_IEEE;
r600_bytecode_src(&alu.src[0], &ctx->src[0], 0);
r600_bytecode_src_set_abs(&alu.src[0]);
-
+
alu.dst.sel = ctx->temp_reg;
alu.dst.chan = i;
- if (i == 0)
+ if (i == 0)
alu.dst.write = 1;
if (i == 2)
alu.last = 1;
alu.op = ALU_OP1_LOG_IEEE;
r600_bytecode_src(&alu.src[0], &ctx->src[0], 0);
r600_bytecode_src_set_abs(&alu.src[0]);
-
+
alu.dst.sel = ctx->temp_reg;
alu.dst.chan = 0;
alu.dst.write = 1;
alu.dst.write = 1;
if (i == 2)
alu.last = 1;
-
+
r = r600_bytecode_add_alu(ctx->bc, &alu);
if (r)
- return r;
+ return r;
}
} else {
memset(&alu, 0, sizeof(struct r600_bytecode_alu));
alu.dst.write = 1;
if (i == 2)
alu.last = 1;
-
+
r = r600_bytecode_add_alu(ctx->bc, &alu);
if (r)
return r;
static void fc_pushlevel(struct r600_shader_ctx *ctx, int type)
{
- ctx->bc->fc_sp++;
+ assert(ctx->bc->fc_sp < ARRAY_SIZE(ctx->bc->fc_stack));
ctx->bc->fc_stack[ctx->bc->fc_sp].type = type;
ctx->bc->fc_stack[ctx->bc->fc_sp].start = ctx->bc->cf_last;
+ ctx->bc->fc_sp++;
}
static void fc_poplevel(struct r600_shader_ctx *ctx)
{
- struct r600_cf_stack_entry *sp = &ctx->bc->fc_stack[ctx->bc->fc_sp];
+ struct r600_cf_stack_entry *sp = &ctx->bc->fc_stack[ctx->bc->fc_sp - 1];
free(sp->mid);
sp->mid = NULL;
sp->num_mid = 0;
r600_bytecode_add_cfinst(ctx->bc, CF_OP_ELSE);
ctx->bc->cf_last->pop_count = 1;
- fc_set_mid(ctx, ctx->bc->fc_sp);
- ctx->bc->fc_stack[ctx->bc->fc_sp].start->cf_addr = ctx->bc->cf_last->id;
+ fc_set_mid(ctx, ctx->bc->fc_sp - 1);
+ ctx->bc->fc_stack[ctx->bc->fc_sp - 1].start->cf_addr = ctx->bc->cf_last->id;
return 0;
}
static int tgsi_endif(struct r600_shader_ctx *ctx)
{
pops(ctx, 1);
- if (ctx->bc->fc_stack[ctx->bc->fc_sp].type != FC_IF) {
+ if (ctx->bc->fc_stack[ctx->bc->fc_sp - 1].type != FC_IF) {
R600_ERR("if/endif unbalanced in shader\n");
return -1;
}
- if (ctx->bc->fc_stack[ctx->bc->fc_sp].mid == NULL) {
- ctx->bc->fc_stack[ctx->bc->fc_sp].start->cf_addr = ctx->bc->cf_last->id + 2;
- ctx->bc->fc_stack[ctx->bc->fc_sp].start->pop_count = 1;
+ if (ctx->bc->fc_stack[ctx->bc->fc_sp - 1].mid == NULL) {
+ ctx->bc->fc_stack[ctx->bc->fc_sp - 1].start->cf_addr = ctx->bc->cf_last->id + 2;
+ ctx->bc->fc_stack[ctx->bc->fc_sp - 1].start->pop_count = 1;
} else {
- ctx->bc->fc_stack[ctx->bc->fc_sp].mid[0]->cf_addr = ctx->bc->cf_last->id + 2;
+ ctx->bc->fc_stack[ctx->bc->fc_sp - 1].mid[0]->cf_addr = ctx->bc->cf_last->id + 2;
}
fc_poplevel(ctx);
static int tgsi_endloop(struct r600_shader_ctx *ctx)
{
- int i;
+ unsigned i;
r600_bytecode_add_cfinst(ctx->bc, CF_OP_LOOP_END);
- if (ctx->bc->fc_stack[ctx->bc->fc_sp].type != FC_LOOP) {
+ if (ctx->bc->fc_stack[ctx->bc->fc_sp - 1].type != FC_LOOP) {
R600_ERR("loop/endloop in shader code are not paired.\n");
return -EINVAL;
}
LOOP START point to CF after LOOP END
BRK/CONT point to LOOP END CF
*/
- ctx->bc->cf_last->cf_addr = ctx->bc->fc_stack[ctx->bc->fc_sp].start->id + 2;
+ ctx->bc->cf_last->cf_addr = ctx->bc->fc_stack[ctx->bc->fc_sp - 1].start->id + 2;
- ctx->bc->fc_stack[ctx->bc->fc_sp].start->cf_addr = ctx->bc->cf_last->id + 2;
+ ctx->bc->fc_stack[ctx->bc->fc_sp - 1].start->cf_addr = ctx->bc->cf_last->id + 2;
- for (i = 0; i < ctx->bc->fc_stack[ctx->bc->fc_sp].num_mid; i++) {
- ctx->bc->fc_stack[ctx->bc->fc_sp].mid[i]->cf_addr = ctx->bc->cf_last->id;
+ for (i = 0; i < ctx->bc->fc_stack[ctx->bc->fc_sp - 1].num_mid; i++) {
+ ctx->bc->fc_stack[ctx->bc->fc_sp - 1].mid[i]->cf_addr = ctx->bc->cf_last->id;
}
/* XXX add LOOPRET support */
fc_poplevel(ctx);
return 0;
}
-static int tgsi_loop_breakc(struct r600_shader_ctx *ctx)
-{
- int r;
- unsigned int fscp;
-
- for (fscp = ctx->bc->fc_sp; fscp > 0; fscp--)
- {
- if (FC_LOOP == ctx->bc->fc_stack[fscp].type)
- break;
- }
- if (fscp == 0) {
- R600_ERR("BREAKC not inside loop/endloop pair\n");
- return -EINVAL;
- }
-
- if (ctx->bc->chip_class == EVERGREEN &&
- ctx->bc->family != CHIP_CYPRESS &&
- ctx->bc->family != CHIP_JUNIPER) {
- /* HW bug: ALU_BREAK does not save the active mask correctly */
- r = tgsi_uif(ctx);
- if (r)
- return r;
-
- r = r600_bytecode_add_cfinst(ctx->bc, CF_OP_LOOP_BREAK);
- if (r)
- return r;
- fc_set_mid(ctx, fscp);
-
- return tgsi_endif(ctx);
- } else {
- r = emit_logic_pred(ctx, ALU_OP2_PRED_SETE_INT, CF_OP_ALU_BREAK);
- if (r)
- return r;
- fc_set_mid(ctx, fscp);
- }
-
- return 0;
-}
-
static int tgsi_loop_brk_cont(struct r600_shader_ctx *ctx)
{
unsigned int fscp;
for (fscp = ctx->bc->fc_sp; fscp > 0; fscp--)
{
- if (FC_LOOP == ctx->bc->fc_stack[fscp].type)
+ if (FC_LOOP == ctx->bc->fc_stack[fscp - 1].type)
break;
}
r600_bytecode_add_cfinst(ctx->bc, ctx->inst_info->op);
- fc_set_mid(ctx, fscp);
+ fc_set_mid(ctx, fscp - 1);
return 0;
}
alu.src[0].sel = ctx->temp_reg;
alu.src[0].chan = i;
-
+
r600_bytecode_src(&alu.src[1], &ctx->src[2], i);
if (i == lasti) {
alu.last = 1;
return 0;
}
+static int tgsi_pk2h(struct r600_shader_ctx *ctx)
+{
+ struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+ struct r600_bytecode_alu alu;
+ int r, i;
+ int lasti = tgsi_last_instruction(inst->Dst[0].Register.WriteMask);
+
+ /* temp.xy = f32_to_f16(src) */
+ memset(&alu, 0, sizeof(struct r600_bytecode_alu));
+ alu.op = ALU_OP1_FLT32_TO_FLT16;
+ alu.dst.chan = 0;
+ alu.dst.sel = ctx->temp_reg;
+ alu.dst.write = 1;
+ r600_bytecode_src(&alu.src[0], &ctx->src[0], 0);
+ r = r600_bytecode_add_alu(ctx->bc, &alu);
+ if (r)
+ return r;
+ alu.dst.chan = 1;
+ r600_bytecode_src(&alu.src[0], &ctx->src[0], 1);
+ alu.last = 1;
+ r = r600_bytecode_add_alu(ctx->bc, &alu);
+ if (r)
+ return r;
+
+ /* dst.x = temp.y * 0x10000 + temp.x */
+ for (i = 0; i < lasti + 1; i++) {
+ if (!(inst->Dst[0].Register.WriteMask & (1 << i)))
+ continue;
+
+ memset(&alu, 0, sizeof(struct r600_bytecode_alu));
+ alu.op = ALU_OP3_MULADD_UINT24;
+ alu.is_op3 = 1;
+ tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
+ alu.last = i == lasti;
+ alu.src[0].sel = ctx->temp_reg;
+ alu.src[0].chan = 1;
+ alu.src[1].sel = V_SQ_ALU_SRC_LITERAL;
+ alu.src[1].value = 0x10000;
+ alu.src[2].sel = ctx->temp_reg;
+ alu.src[2].chan = 0;
+ r = r600_bytecode_add_alu(ctx->bc, &alu);
+ if (r)
+ return r;
+ }
+
+ return 0;
+}
+
+static int tgsi_up2h(struct r600_shader_ctx *ctx)
+{
+ struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+ struct r600_bytecode_alu alu;
+ int r, i;
+ int lasti = tgsi_last_instruction(inst->Dst[0].Register.WriteMask);
+
+ /* temp.x = src.x */
+ /* note: no need to mask out the high bits */
+ memset(&alu, 0, sizeof(struct r600_bytecode_alu));
+ alu.op = ALU_OP1_MOV;
+ alu.dst.chan = 0;
+ alu.dst.sel = ctx->temp_reg;
+ alu.dst.write = 1;
+ r600_bytecode_src(&alu.src[0], &ctx->src[0], 0);
+ r = r600_bytecode_add_alu(ctx->bc, &alu);
+ if (r)
+ return r;
+
+ /* temp.y = src.x >> 16 */
+ memset(&alu, 0, sizeof(struct r600_bytecode_alu));
+ alu.op = ALU_OP2_LSHR_INT;
+ alu.dst.chan = 1;
+ alu.dst.sel = ctx->temp_reg;
+ alu.dst.write = 1;
+ r600_bytecode_src(&alu.src[0], &ctx->src[0], 0);
+ alu.src[1].sel = V_SQ_ALU_SRC_LITERAL;
+ alu.src[1].value = 16;
+ alu.last = 1;
+ r = r600_bytecode_add_alu(ctx->bc, &alu);
+ if (r)
+ return r;
+
+ /* dst.wz = dst.xy = f16_to_f32(temp.xy) */
+ for (i = 0; i < lasti + 1; i++) {
+ if (!(inst->Dst[0].Register.WriteMask & (1 << i)))
+ continue;
+ memset(&alu, 0, sizeof(struct r600_bytecode_alu));
+ tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
+ alu.op = ALU_OP1_FLT16_TO_FLT32;
+ alu.src[0].sel = ctx->temp_reg;
+ alu.src[0].chan = i % 2;
+ alu.last = i == lasti;
+ r = r600_bytecode_add_alu(ctx->bc, &alu);
+ if (r)
+ return r;
+ }
+
+ return 0;
+}
+
static const struct r600_shader_tgsi_instruction r600_shader_tgsi_instruction[] = {
[TGSI_OPCODE_ARL] = { ALU_OP0_NOP, tgsi_r600_arl},
[TGSI_OPCODE_MOV] = { ALU_OP1_MOV, tgsi_op2},
[TGSI_OPCODE_RSQ] = { ALU_OP0_NOP, tgsi_rsq},
[TGSI_OPCODE_EXP] = { ALU_OP0_NOP, tgsi_exp},
[TGSI_OPCODE_LOG] = { ALU_OP0_NOP, tgsi_log},
- [TGSI_OPCODE_MUL] = { ALU_OP2_MUL, tgsi_op2},
+ [TGSI_OPCODE_MUL] = { ALU_OP2_MUL_IEEE, tgsi_op2},
[TGSI_OPCODE_ADD] = { ALU_OP2_ADD, tgsi_op2},
- [TGSI_OPCODE_DP3] = { ALU_OP2_DOT4, tgsi_dp},
- [TGSI_OPCODE_DP4] = { ALU_OP2_DOT4, tgsi_dp},
+ [TGSI_OPCODE_DP3] = { ALU_OP2_DOT4_IEEE, tgsi_dp},
+ [TGSI_OPCODE_DP4] = { ALU_OP2_DOT4_IEEE, tgsi_dp},
[TGSI_OPCODE_DST] = { ALU_OP0_NOP, tgsi_opdst},
[TGSI_OPCODE_MIN] = { ALU_OP2_MIN, tgsi_op2},
[TGSI_OPCODE_MAX] = { ALU_OP2_MAX, tgsi_op2},
[TGSI_OPCODE_SLT] = { ALU_OP2_SETGT, tgsi_op2_swap},
[TGSI_OPCODE_SGE] = { ALU_OP2_SETGE, tgsi_op2},
- [TGSI_OPCODE_MAD] = { ALU_OP3_MULADD, tgsi_op3},
- [TGSI_OPCODE_SUB] = { ALU_OP2_ADD, tgsi_op2},
+ [TGSI_OPCODE_MAD] = { ALU_OP3_MULADD_IEEE, tgsi_op3},
[TGSI_OPCODE_LRP] = { ALU_OP0_NOP, tgsi_lrp},
[TGSI_OPCODE_FMA] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_SQRT] = { ALU_OP1_SQRT_IEEE, tgsi_trans_srcx_replicate},
- [TGSI_OPCODE_DP2A] = { ALU_OP0_NOP, tgsi_unsupported},
+ [21] = { ALU_OP0_NOP, tgsi_unsupported},
[22] = { ALU_OP0_NOP, tgsi_unsupported},
[23] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_FRC] = { ALU_OP1_FRACT, tgsi_op2},
- [TGSI_OPCODE_CLAMP] = { ALU_OP0_NOP, tgsi_unsupported},
+ [25] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_FLR] = { ALU_OP1_FLOOR, tgsi_op2},
[TGSI_OPCODE_ROUND] = { ALU_OP1_RNDNE, tgsi_op2},
[TGSI_OPCODE_EX2] = { ALU_OP1_EXP_IEEE, tgsi_trans_srcx_replicate},
[TGSI_OPCODE_LG2] = { ALU_OP1_LOG_IEEE, tgsi_trans_srcx_replicate},
[TGSI_OPCODE_POW] = { ALU_OP0_NOP, tgsi_pow},
- [TGSI_OPCODE_XPD] = { ALU_OP0_NOP, tgsi_xpd},
+ [31] = { ALU_OP0_NOP, tgsi_unsupported},
[32] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ABS] = { ALU_OP1_MOV, tgsi_op2},
+ [33] = { ALU_OP0_NOP, tgsi_unsupported},
[34] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_DPH] = { ALU_OP2_DOT4, tgsi_dp},
+ [35] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_COS] = { ALU_OP1_COS, tgsi_trig},
[TGSI_OPCODE_DDX] = { FETCH_OP_GET_GRADIENTS_H, tgsi_tex},
[TGSI_OPCODE_DDY] = { FETCH_OP_GET_GRADIENTS_V, tgsi_tex},
[TGSI_OPCODE_RET] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_SSG] = { ALU_OP0_NOP, tgsi_ssg},
[TGSI_OPCODE_CMP] = { ALU_OP0_NOP, tgsi_cmp},
- [TGSI_OPCODE_SCS] = { ALU_OP0_NOP, tgsi_scs},
+ [67] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_TXB] = { FETCH_OP_SAMPLE_LB, tgsi_tex},
[69] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_DIV] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_DP2] = { ALU_OP2_DOT4, tgsi_dp},
+ [TGSI_OPCODE_DP2] = { ALU_OP2_DOT4_IEEE, tgsi_dp},
[TGSI_OPCODE_TXL] = { FETCH_OP_SAMPLE_L, tgsi_tex},
[TGSI_OPCODE_BRK] = { CF_OP_LOOP_BREAK, tgsi_loop_brk_cont},
[TGSI_OPCODE_IF] = { ALU_OP0_NOP, tgsi_if},
[TGSI_OPCODE_ENDIF] = { ALU_OP0_NOP, tgsi_endif},
[TGSI_OPCODE_DDX_FINE] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_DDY_FINE] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_PUSHA] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_POPA] = { ALU_OP0_NOP, tgsi_unsupported},
+ [81] = { ALU_OP0_NOP, tgsi_unsupported},
+ [82] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_CEIL] = { ALU_OP1_CEIL, tgsi_op2},
[TGSI_OPCODE_I2F] = { ALU_OP1_INT_TO_FLT, tgsi_op2_trans},
[TGSI_OPCODE_NOT] = { ALU_OP1_NOT_INT, tgsi_op2},
[TGSI_OPCODE_OR] = { ALU_OP2_OR_INT, tgsi_op2},
[TGSI_OPCODE_MOD] = { ALU_OP0_NOP, tgsi_imod},
[TGSI_OPCODE_XOR] = { ALU_OP2_XOR_INT, tgsi_op2},
- [TGSI_OPCODE_SAD] = { ALU_OP0_NOP, tgsi_unsupported},
+ [93] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_TXF] = { FETCH_OP_LD, tgsi_tex},
[TGSI_OPCODE_TXQ] = { FETCH_OP_GET_TEXTURE_RESINFO, tgsi_tex},
[TGSI_OPCODE_CONT] = { CF_OP_LOOP_CONTINUE, tgsi_loop_brk_cont},
[TGSI_OPCODE_BGNSUB] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_ENDLOOP] = { ALU_OP0_NOP, tgsi_endloop},
[TGSI_OPCODE_ENDSUB] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_TXQ_LZ] = { FETCH_OP_GET_TEXTURE_RESINFO, tgsi_tex},
+ [103] = { FETCH_OP_GET_TEXTURE_RESINFO, tgsi_tex},
[TGSI_OPCODE_TXQS] = { FETCH_OP_GET_NUMBER_OF_SAMPLES, tgsi_tex},
- [105] = { ALU_OP0_NOP, tgsi_unsupported},
+ [TGSI_OPCODE_RESQ] = { ALU_OP0_NOP, tgsi_unsupported},
[106] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_NOP] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_FSEQ] = { ALU_OP2_SETE_DX10, tgsi_op2},
[TGSI_OPCODE_FSGE] = { ALU_OP2_SETGE_DX10, tgsi_op2},
[TGSI_OPCODE_FSLT] = { ALU_OP2_SETGT_DX10, tgsi_op2_swap},
[TGSI_OPCODE_FSNE] = { ALU_OP2_SETNE_DX10, tgsi_op2_swap},
- [112] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_CALLNZ] = { ALU_OP0_NOP, tgsi_unsupported},
+ [TGSI_OPCODE_MEMBAR] = { ALU_OP0_NOP, tgsi_unsupported},
+ [113] = { ALU_OP0_NOP, tgsi_unsupported},
[114] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_BREAKC] = { ALU_OP0_NOP, tgsi_loop_breakc},
+ [115] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_KILL_IF] = { ALU_OP2_KILLGT, tgsi_kill}, /* conditional kill */
[TGSI_OPCODE_END] = { ALU_OP0_NOP, tgsi_end}, /* aka HALT */
- [118] = { ALU_OP0_NOP, tgsi_unsupported},
+ [TGSI_OPCODE_DFMA] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_F2I] = { ALU_OP1_FLT_TO_INT, tgsi_op2_trans},
[TGSI_OPCODE_IDIV] = { ALU_OP0_NOP, tgsi_idiv},
[TGSI_OPCODE_IMAX] = { ALU_OP2_MAX_INT, tgsi_op2},
[TGSI_OPCODE_ISSG] = { 0, tgsi_issg},
[TGSI_OPCODE_LOAD] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_STORE] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_MFENCE] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_LFENCE] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_SFENCE] = { ALU_OP0_NOP, tgsi_unsupported},
+ [163] = { ALU_OP0_NOP, tgsi_unsupported},
+ [164] = { ALU_OP0_NOP, tgsi_unsupported},
+ [165] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_BARRIER] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_ATOMUADD] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_ATOMXCHG] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_RSQ] = { ALU_OP1_RECIPSQRT_IEEE, tgsi_rsq},
[TGSI_OPCODE_EXP] = { ALU_OP0_NOP, tgsi_exp},
[TGSI_OPCODE_LOG] = { ALU_OP0_NOP, tgsi_log},
- [TGSI_OPCODE_MUL] = { ALU_OP2_MUL, tgsi_op2},
+ [TGSI_OPCODE_MUL] = { ALU_OP2_MUL_IEEE, tgsi_op2},
[TGSI_OPCODE_ADD] = { ALU_OP2_ADD, tgsi_op2},
- [TGSI_OPCODE_DP3] = { ALU_OP2_DOT4, tgsi_dp},
- [TGSI_OPCODE_DP4] = { ALU_OP2_DOT4, tgsi_dp},
+ [TGSI_OPCODE_DP3] = { ALU_OP2_DOT4_IEEE, tgsi_dp},
+ [TGSI_OPCODE_DP4] = { ALU_OP2_DOT4_IEEE, tgsi_dp},
[TGSI_OPCODE_DST] = { ALU_OP0_NOP, tgsi_opdst},
[TGSI_OPCODE_MIN] = { ALU_OP2_MIN, tgsi_op2},
[TGSI_OPCODE_MAX] = { ALU_OP2_MAX, tgsi_op2},
[TGSI_OPCODE_SLT] = { ALU_OP2_SETGT, tgsi_op2_swap},
[TGSI_OPCODE_SGE] = { ALU_OP2_SETGE, tgsi_op2},
- [TGSI_OPCODE_MAD] = { ALU_OP3_MULADD, tgsi_op3},
- [TGSI_OPCODE_SUB] = { ALU_OP2_ADD, tgsi_op2},
+ [TGSI_OPCODE_MAD] = { ALU_OP3_MULADD_IEEE, tgsi_op3},
[TGSI_OPCODE_LRP] = { ALU_OP0_NOP, tgsi_lrp},
- [TGSI_OPCODE_FMA] = { ALU_OP0_NOP, tgsi_unsupported},
+ [TGSI_OPCODE_FMA] = { ALU_OP3_FMA, tgsi_op3},
[TGSI_OPCODE_SQRT] = { ALU_OP1_SQRT_IEEE, tgsi_trans_srcx_replicate},
- [TGSI_OPCODE_DP2A] = { ALU_OP0_NOP, tgsi_unsupported},
+ [21] = { ALU_OP0_NOP, tgsi_unsupported},
[22] = { ALU_OP0_NOP, tgsi_unsupported},
[23] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_FRC] = { ALU_OP1_FRACT, tgsi_op2},
- [TGSI_OPCODE_CLAMP] = { ALU_OP0_NOP, tgsi_unsupported},
+ [25] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_FLR] = { ALU_OP1_FLOOR, tgsi_op2},
[TGSI_OPCODE_ROUND] = { ALU_OP1_RNDNE, tgsi_op2},
[TGSI_OPCODE_EX2] = { ALU_OP1_EXP_IEEE, tgsi_trans_srcx_replicate},
[TGSI_OPCODE_LG2] = { ALU_OP1_LOG_IEEE, tgsi_trans_srcx_replicate},
[TGSI_OPCODE_POW] = { ALU_OP0_NOP, tgsi_pow},
- [TGSI_OPCODE_XPD] = { ALU_OP0_NOP, tgsi_xpd},
+ [31] = { ALU_OP0_NOP, tgsi_unsupported},
[32] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ABS] = { ALU_OP1_MOV, tgsi_op2},
+ [33] = { ALU_OP0_NOP, tgsi_unsupported},
[34] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_DPH] = { ALU_OP2_DOT4, tgsi_dp},
+ [35] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_COS] = { ALU_OP1_COS, tgsi_trig},
[TGSI_OPCODE_DDX] = { FETCH_OP_GET_GRADIENTS_H, tgsi_tex},
[TGSI_OPCODE_DDY] = { FETCH_OP_GET_GRADIENTS_V, tgsi_tex},
[TGSI_OPCODE_KILL] = { ALU_OP2_KILLGT, tgsi_kill}, /* unconditional kill */
- [TGSI_OPCODE_PK2H] = { ALU_OP0_NOP, tgsi_unsupported},
+ [TGSI_OPCODE_PK2H] = { ALU_OP0_NOP, tgsi_pk2h},
[TGSI_OPCODE_PK2US] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_PK4B] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_PK4UB] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_TEX] = { FETCH_OP_SAMPLE, tgsi_tex},
[TGSI_OPCODE_TXD] = { FETCH_OP_SAMPLE_G, tgsi_tex},
[TGSI_OPCODE_TXP] = { FETCH_OP_SAMPLE, tgsi_tex},
- [TGSI_OPCODE_UP2H] = { ALU_OP0_NOP, tgsi_unsupported},
+ [TGSI_OPCODE_UP2H] = { ALU_OP0_NOP, tgsi_up2h},
[TGSI_OPCODE_UP2US] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_UP4B] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_UP4UB] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_RET] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_SSG] = { ALU_OP0_NOP, tgsi_ssg},
[TGSI_OPCODE_CMP] = { ALU_OP0_NOP, tgsi_cmp},
- [TGSI_OPCODE_SCS] = { ALU_OP0_NOP, tgsi_scs},
+ [67] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_TXB] = { FETCH_OP_SAMPLE_LB, tgsi_tex},
[69] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_DIV] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_DP2] = { ALU_OP2_DOT4, tgsi_dp},
+ [TGSI_OPCODE_DP2] = { ALU_OP2_DOT4_IEEE, tgsi_dp},
[TGSI_OPCODE_TXL] = { FETCH_OP_SAMPLE_L, tgsi_tex},
[TGSI_OPCODE_BRK] = { CF_OP_LOOP_BREAK, tgsi_loop_brk_cont},
[TGSI_OPCODE_IF] = { ALU_OP0_NOP, tgsi_if},
[TGSI_OPCODE_ENDIF] = { ALU_OP0_NOP, tgsi_endif},
[TGSI_OPCODE_DDX_FINE] = { FETCH_OP_GET_GRADIENTS_H, tgsi_tex},
[TGSI_OPCODE_DDY_FINE] = { FETCH_OP_GET_GRADIENTS_V, tgsi_tex},
- [TGSI_OPCODE_PUSHA] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_POPA] = { ALU_OP0_NOP, tgsi_unsupported},
+ [82] = { ALU_OP0_NOP, tgsi_unsupported},
+ [83] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_CEIL] = { ALU_OP1_CEIL, tgsi_op2},
[TGSI_OPCODE_I2F] = { ALU_OP1_INT_TO_FLT, tgsi_op2_trans},
[TGSI_OPCODE_NOT] = { ALU_OP1_NOT_INT, tgsi_op2},
[TGSI_OPCODE_OR] = { ALU_OP2_OR_INT, tgsi_op2},
[TGSI_OPCODE_MOD] = { ALU_OP0_NOP, tgsi_imod},
[TGSI_OPCODE_XOR] = { ALU_OP2_XOR_INT, tgsi_op2},
- [TGSI_OPCODE_SAD] = { ALU_OP0_NOP, tgsi_unsupported},
+ [93] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_TXF] = { FETCH_OP_LD, tgsi_tex},
[TGSI_OPCODE_TXQ] = { FETCH_OP_GET_TEXTURE_RESINFO, tgsi_tex},
[TGSI_OPCODE_CONT] = { CF_OP_LOOP_CONTINUE, tgsi_loop_brk_cont},
[TGSI_OPCODE_BGNSUB] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_ENDLOOP] = { ALU_OP0_NOP, tgsi_endloop},
[TGSI_OPCODE_ENDSUB] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_TXQ_LZ] = { FETCH_OP_GET_TEXTURE_RESINFO, tgsi_tex},
+ [103] = { FETCH_OP_GET_TEXTURE_RESINFO, tgsi_tex},
[TGSI_OPCODE_TXQS] = { FETCH_OP_GET_NUMBER_OF_SAMPLES, tgsi_tex},
- [105] = { ALU_OP0_NOP, tgsi_unsupported},
+ [TGSI_OPCODE_RESQ] = { ALU_OP0_NOP, tgsi_unsupported},
[106] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_NOP] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_FSEQ] = { ALU_OP2_SETE_DX10, tgsi_op2},
[TGSI_OPCODE_FSGE] = { ALU_OP2_SETGE_DX10, tgsi_op2},
[TGSI_OPCODE_FSLT] = { ALU_OP2_SETGT_DX10, tgsi_op2_swap},
[TGSI_OPCODE_FSNE] = { ALU_OP2_SETNE_DX10, tgsi_op2_swap},
- [112] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_CALLNZ] = { ALU_OP0_NOP, tgsi_unsupported},
+ [TGSI_OPCODE_MEMBAR] = { ALU_OP0_NOP, tgsi_unsupported},
+ [113] = { ALU_OP0_NOP, tgsi_unsupported},
[114] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_BREAKC] = { ALU_OP0_NOP, tgsi_unsupported},
+ [115] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_KILL_IF] = { ALU_OP2_KILLGT, tgsi_kill}, /* conditional kill */
[TGSI_OPCODE_END] = { ALU_OP0_NOP, tgsi_end}, /* aka HALT */
- [118] = { ALU_OP0_NOP, tgsi_unsupported},
+ /* Refer below for TGSI_OPCODE_DFMA */
[TGSI_OPCODE_F2I] = { ALU_OP1_FLT_TO_INT, tgsi_f2i},
[TGSI_OPCODE_IDIV] = { ALU_OP0_NOP, tgsi_idiv},
[TGSI_OPCODE_IMAX] = { ALU_OP2_MAX_INT, tgsi_op2},
[TGSI_OPCODE_UCMP] = { ALU_OP0_NOP, tgsi_ucmp},
[TGSI_OPCODE_IABS] = { 0, tgsi_iabs},
[TGSI_OPCODE_ISSG] = { 0, tgsi_issg},
- [TGSI_OPCODE_LOAD] = { ALU_OP0_NOP, tgsi_unsupported},
+ [TGSI_OPCODE_LOAD] = { ALU_OP0_NOP, tgsi_load},
[TGSI_OPCODE_STORE] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_MFENCE] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_LFENCE] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_SFENCE] = { ALU_OP0_NOP, tgsi_unsupported},
+ [163] = { ALU_OP0_NOP, tgsi_unsupported},
+ [164] = { ALU_OP0_NOP, tgsi_unsupported},
+ [165] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_BARRIER] = { ALU_OP0_GROUP_BARRIER, tgsi_barrier},
- [TGSI_OPCODE_ATOMUADD] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMXCHG] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMCAS] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMAND] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMOR] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMXOR] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMUMIN] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMUMAX] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMIMIN] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMIMAX] = { ALU_OP0_NOP, tgsi_unsupported},
+ [TGSI_OPCODE_ATOMUADD] = { V_RAT_INST_ADD_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMXCHG] = { V_RAT_INST_XCHG_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMCAS] = { V_RAT_INST_CMPXCHG_INT_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMAND] = { V_RAT_INST_AND_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMOR] = { V_RAT_INST_OR_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMXOR] = { V_RAT_INST_XOR_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMUMIN] = { V_RAT_INST_MIN_UINT_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMUMAX] = { V_RAT_INST_MAX_UINT_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMIMIN] = { V_RAT_INST_MIN_INT_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMIMAX] = { V_RAT_INST_MAX_INT_RTN, tgsi_atomic_op},
[TGSI_OPCODE_TEX2] = { FETCH_OP_SAMPLE, tgsi_tex},
[TGSI_OPCODE_TXB2] = { FETCH_OP_SAMPLE_LB, tgsi_tex},
[TGSI_OPCODE_TXL2] = { FETCH_OP_SAMPLE_L, tgsi_tex},
[TGSI_OPCODE_DNEG] = { ALU_OP2_ADD_64, tgsi_dneg},
[TGSI_OPCODE_DADD] = { ALU_OP2_ADD_64, tgsi_op2_64},
[TGSI_OPCODE_DMUL] = { ALU_OP2_MUL_64, cayman_mul_double_instr},
+ [TGSI_OPCODE_DDIV] = { 0, cayman_ddiv_instr },
[TGSI_OPCODE_DMAX] = { ALU_OP2_MAX_64, tgsi_op2_64},
[TGSI_OPCODE_DMIN] = { ALU_OP2_MIN_64, tgsi_op2_64},
[TGSI_OPCODE_DSLT] = { ALU_OP2_SETGT_64, tgsi_op2_64_single_dest_s},
[TGSI_OPCODE_DRCP] = { ALU_OP2_RECIP_64, cayman_emit_double_instr},
[TGSI_OPCODE_DSQRT] = { ALU_OP2_SQRT_64, cayman_emit_double_instr},
[TGSI_OPCODE_DMAD] = { ALU_OP3_FMA_64, tgsi_op3_64},
+ [TGSI_OPCODE_DFMA] = { ALU_OP3_FMA_64, tgsi_op3_64},
[TGSI_OPCODE_DFRAC] = { ALU_OP1_FRACT_64, tgsi_op2_64},
[TGSI_OPCODE_DLDEXP] = { ALU_OP2_LDEXP_64, tgsi_op2_64},
[TGSI_OPCODE_DFRACEXP] = { ALU_OP1_FREXP_64, tgsi_dfracexp},
[TGSI_OPCODE_RSQ] = { ALU_OP1_RECIPSQRT_IEEE, cayman_emit_float_instr},
[TGSI_OPCODE_EXP] = { ALU_OP0_NOP, tgsi_exp},
[TGSI_OPCODE_LOG] = { ALU_OP0_NOP, tgsi_log},
- [TGSI_OPCODE_MUL] = { ALU_OP2_MUL, tgsi_op2},
+ [TGSI_OPCODE_MUL] = { ALU_OP2_MUL_IEEE, tgsi_op2},
[TGSI_OPCODE_ADD] = { ALU_OP2_ADD, tgsi_op2},
- [TGSI_OPCODE_DP3] = { ALU_OP2_DOT4, tgsi_dp},
- [TGSI_OPCODE_DP4] = { ALU_OP2_DOT4, tgsi_dp},
+ [TGSI_OPCODE_DP3] = { ALU_OP2_DOT4_IEEE, tgsi_dp},
+ [TGSI_OPCODE_DP4] = { ALU_OP2_DOT4_IEEE, tgsi_dp},
[TGSI_OPCODE_DST] = { ALU_OP0_NOP, tgsi_opdst},
[TGSI_OPCODE_MIN] = { ALU_OP2_MIN, tgsi_op2},
[TGSI_OPCODE_MAX] = { ALU_OP2_MAX, tgsi_op2},
[TGSI_OPCODE_SLT] = { ALU_OP2_SETGT, tgsi_op2_swap},
[TGSI_OPCODE_SGE] = { ALU_OP2_SETGE, tgsi_op2},
- [TGSI_OPCODE_MAD] = { ALU_OP3_MULADD, tgsi_op3},
- [TGSI_OPCODE_SUB] = { ALU_OP2_ADD, tgsi_op2},
+ [TGSI_OPCODE_MAD] = { ALU_OP3_MULADD_IEEE, tgsi_op3},
[TGSI_OPCODE_LRP] = { ALU_OP0_NOP, tgsi_lrp},
- [TGSI_OPCODE_FMA] = { ALU_OP0_NOP, tgsi_unsupported},
+ [TGSI_OPCODE_FMA] = { ALU_OP3_FMA, tgsi_op3},
[TGSI_OPCODE_SQRT] = { ALU_OP1_SQRT_IEEE, cayman_emit_float_instr},
- [TGSI_OPCODE_DP2A] = { ALU_OP0_NOP, tgsi_unsupported},
+ [21] = { ALU_OP0_NOP, tgsi_unsupported},
[22] = { ALU_OP0_NOP, tgsi_unsupported},
[23] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_FRC] = { ALU_OP1_FRACT, tgsi_op2},
- [TGSI_OPCODE_CLAMP] = { ALU_OP0_NOP, tgsi_unsupported},
+ [25] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_FLR] = { ALU_OP1_FLOOR, tgsi_op2},
[TGSI_OPCODE_ROUND] = { ALU_OP1_RNDNE, tgsi_op2},
[TGSI_OPCODE_EX2] = { ALU_OP1_EXP_IEEE, cayman_emit_float_instr},
[TGSI_OPCODE_LG2] = { ALU_OP1_LOG_IEEE, cayman_emit_float_instr},
[TGSI_OPCODE_POW] = { ALU_OP0_NOP, cayman_pow},
- [TGSI_OPCODE_XPD] = { ALU_OP0_NOP, tgsi_xpd},
+ [31] = { ALU_OP0_NOP, tgsi_unsupported},
[32] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ABS] = { ALU_OP1_MOV, tgsi_op2},
+ [33] = { ALU_OP0_NOP, tgsi_unsupported},
[34] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_DPH] = { ALU_OP2_DOT4, tgsi_dp},
+ [35] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_COS] = { ALU_OP1_COS, cayman_trig},
[TGSI_OPCODE_DDX] = { FETCH_OP_GET_GRADIENTS_H, tgsi_tex},
[TGSI_OPCODE_DDY] = { FETCH_OP_GET_GRADIENTS_V, tgsi_tex},
[TGSI_OPCODE_KILL] = { ALU_OP2_KILLGT, tgsi_kill}, /* unconditional kill */
- [TGSI_OPCODE_PK2H] = { ALU_OP0_NOP, tgsi_unsupported},
+ [TGSI_OPCODE_PK2H] = { ALU_OP0_NOP, tgsi_pk2h},
[TGSI_OPCODE_PK2US] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_PK4B] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_PK4UB] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_TEX] = { FETCH_OP_SAMPLE, tgsi_tex},
[TGSI_OPCODE_TXD] = { FETCH_OP_SAMPLE_G, tgsi_tex},
[TGSI_OPCODE_TXP] = { FETCH_OP_SAMPLE, tgsi_tex},
- [TGSI_OPCODE_UP2H] = { ALU_OP0_NOP, tgsi_unsupported},
+ [TGSI_OPCODE_UP2H] = { ALU_OP0_NOP, tgsi_up2h},
[TGSI_OPCODE_UP2US] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_UP4B] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_UP4UB] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_RET] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_SSG] = { ALU_OP0_NOP, tgsi_ssg},
[TGSI_OPCODE_CMP] = { ALU_OP0_NOP, tgsi_cmp},
- [TGSI_OPCODE_SCS] = { ALU_OP0_NOP, tgsi_scs},
+ [67] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_TXB] = { FETCH_OP_SAMPLE_LB, tgsi_tex},
[69] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_DIV] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_DP2] = { ALU_OP2_DOT4, tgsi_dp},
+ [TGSI_OPCODE_DP2] = { ALU_OP2_DOT4_IEEE, tgsi_dp},
[TGSI_OPCODE_TXL] = { FETCH_OP_SAMPLE_L, tgsi_tex},
[TGSI_OPCODE_BRK] = { CF_OP_LOOP_BREAK, tgsi_loop_brk_cont},
[TGSI_OPCODE_IF] = { ALU_OP0_NOP, tgsi_if},
[TGSI_OPCODE_ENDIF] = { ALU_OP0_NOP, tgsi_endif},
[TGSI_OPCODE_DDX_FINE] = { FETCH_OP_GET_GRADIENTS_H, tgsi_tex},
[TGSI_OPCODE_DDY_FINE] = { FETCH_OP_GET_GRADIENTS_V, tgsi_tex},
- [TGSI_OPCODE_PUSHA] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_POPA] = { ALU_OP0_NOP, tgsi_unsupported},
+ [82] = { ALU_OP0_NOP, tgsi_unsupported},
+ [83] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_CEIL] = { ALU_OP1_CEIL, tgsi_op2},
[TGSI_OPCODE_I2F] = { ALU_OP1_INT_TO_FLT, tgsi_op2},
[TGSI_OPCODE_NOT] = { ALU_OP1_NOT_INT, tgsi_op2},
[TGSI_OPCODE_OR] = { ALU_OP2_OR_INT, tgsi_op2},
[TGSI_OPCODE_MOD] = { ALU_OP0_NOP, tgsi_imod},
[TGSI_OPCODE_XOR] = { ALU_OP2_XOR_INT, tgsi_op2},
- [TGSI_OPCODE_SAD] = { ALU_OP0_NOP, tgsi_unsupported},
+ [93] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_TXF] = { FETCH_OP_LD, tgsi_tex},
[TGSI_OPCODE_TXQ] = { FETCH_OP_GET_TEXTURE_RESINFO, tgsi_tex},
[TGSI_OPCODE_CONT] = { CF_OP_LOOP_CONTINUE, tgsi_loop_brk_cont},
[TGSI_OPCODE_BGNSUB] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_ENDLOOP] = { ALU_OP0_NOP, tgsi_endloop},
[TGSI_OPCODE_ENDSUB] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_TXQ_LZ] = { FETCH_OP_GET_TEXTURE_RESINFO, tgsi_tex},
+ [103] = { FETCH_OP_GET_TEXTURE_RESINFO, tgsi_tex},
[TGSI_OPCODE_TXQS] = { FETCH_OP_GET_NUMBER_OF_SAMPLES, tgsi_tex},
- [105] = { ALU_OP0_NOP, tgsi_unsupported},
+ [TGSI_OPCODE_RESQ] = { ALU_OP0_NOP, tgsi_unsupported},
[106] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_NOP] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_FSEQ] = { ALU_OP2_SETE_DX10, tgsi_op2},
[TGSI_OPCODE_FSGE] = { ALU_OP2_SETGE_DX10, tgsi_op2},
[TGSI_OPCODE_FSLT] = { ALU_OP2_SETGT_DX10, tgsi_op2_swap},
[TGSI_OPCODE_FSNE] = { ALU_OP2_SETNE_DX10, tgsi_op2_swap},
- [112] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_CALLNZ] = { ALU_OP0_NOP, tgsi_unsupported},
+ [TGSI_OPCODE_MEMBAR] = { ALU_OP0_NOP, tgsi_unsupported},
+ [113] = { ALU_OP0_NOP, tgsi_unsupported},
[114] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_BREAKC] = { ALU_OP0_NOP, tgsi_unsupported},
+ [115] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_KILL_IF] = { ALU_OP2_KILLGT, tgsi_kill}, /* conditional kill */
[TGSI_OPCODE_END] = { ALU_OP0_NOP, tgsi_end}, /* aka HALT */
- [118] = { ALU_OP0_NOP, tgsi_unsupported},
+ /* Refer below for TGSI_OPCODE_DFMA */
[TGSI_OPCODE_F2I] = { ALU_OP1_FLT_TO_INT, tgsi_op2},
[TGSI_OPCODE_IDIV] = { ALU_OP0_NOP, tgsi_idiv},
[TGSI_OPCODE_IMAX] = { ALU_OP2_MAX_INT, tgsi_op2},
[TGSI_OPCODE_UCMP] = { ALU_OP0_NOP, tgsi_ucmp},
[TGSI_OPCODE_IABS] = { 0, tgsi_iabs},
[TGSI_OPCODE_ISSG] = { 0, tgsi_issg},
- [TGSI_OPCODE_LOAD] = { ALU_OP0_NOP, tgsi_unsupported},
+ [TGSI_OPCODE_LOAD] = { ALU_OP0_NOP, tgsi_load},
[TGSI_OPCODE_STORE] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_MFENCE] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_LFENCE] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_SFENCE] = { ALU_OP0_NOP, tgsi_unsupported},
+ [163] = { ALU_OP0_NOP, tgsi_unsupported},
+ [164] = { ALU_OP0_NOP, tgsi_unsupported},
+ [165] = { ALU_OP0_NOP, tgsi_unsupported},
[TGSI_OPCODE_BARRIER] = { ALU_OP0_GROUP_BARRIER, tgsi_barrier},
- [TGSI_OPCODE_ATOMUADD] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMXCHG] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMCAS] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMAND] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMOR] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMXOR] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMUMIN] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMUMAX] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMIMIN] = { ALU_OP0_NOP, tgsi_unsupported},
- [TGSI_OPCODE_ATOMIMAX] = { ALU_OP0_NOP, tgsi_unsupported},
+ [TGSI_OPCODE_ATOMUADD] = { V_RAT_INST_ADD_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMXCHG] = { V_RAT_INST_XCHG_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMCAS] = { V_RAT_INST_CMPXCHG_INT_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMAND] = { V_RAT_INST_AND_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMOR] = { V_RAT_INST_OR_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMXOR] = { V_RAT_INST_XOR_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMUMIN] = { V_RAT_INST_MIN_UINT_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMUMAX] = { V_RAT_INST_MAX_UINT_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMIMIN] = { V_RAT_INST_MIN_INT_RTN, tgsi_atomic_op},
+ [TGSI_OPCODE_ATOMIMAX] = { V_RAT_INST_MAX_INT_RTN, tgsi_atomic_op},
[TGSI_OPCODE_TEX2] = { FETCH_OP_SAMPLE, tgsi_tex},
[TGSI_OPCODE_TXB2] = { FETCH_OP_SAMPLE_LB, tgsi_tex},
[TGSI_OPCODE_TXL2] = { FETCH_OP_SAMPLE_L, tgsi_tex},
[TGSI_OPCODE_DNEG] = { ALU_OP2_ADD_64, tgsi_dneg},
[TGSI_OPCODE_DADD] = { ALU_OP2_ADD_64, tgsi_op2_64},
[TGSI_OPCODE_DMUL] = { ALU_OP2_MUL_64, cayman_mul_double_instr},
+ [TGSI_OPCODE_DDIV] = { 0, cayman_ddiv_instr },
[TGSI_OPCODE_DMAX] = { ALU_OP2_MAX_64, tgsi_op2_64},
[TGSI_OPCODE_DMIN] = { ALU_OP2_MIN_64, tgsi_op2_64},
[TGSI_OPCODE_DSLT] = { ALU_OP2_SETGT_64, tgsi_op2_64_single_dest_s},
[TGSI_OPCODE_DRCP] = { ALU_OP2_RECIP_64, cayman_emit_double_instr},
[TGSI_OPCODE_DSQRT] = { ALU_OP2_SQRT_64, cayman_emit_double_instr},
[TGSI_OPCODE_DMAD] = { ALU_OP3_FMA_64, tgsi_op3_64},
+ [TGSI_OPCODE_DFMA] = { ALU_OP3_FMA_64, tgsi_op3_64},
[TGSI_OPCODE_DFRAC] = { ALU_OP1_FRACT_64, tgsi_op2_64},
[TGSI_OPCODE_DLDEXP] = { ALU_OP2_LDEXP_64, tgsi_op2_64},
[TGSI_OPCODE_DFRACEXP] = { ALU_OP1_FREXP_64, tgsi_dfracexp},
projects / mesa.git / blobdiff