#define RADEON_LLVM_MAX_INPUTS (VARYING_SLOT_VAR31 + 1)
#define RADEON_LLVM_MAX_OUTPUTS (VARYING_SLOT_VAR31 + 1)
-enum desc_type {
- DESC_IMAGE,
- DESC_FMASK,
- DESC_SAMPLER,
- DESC_BUFFER,
-};
-
struct nir_to_llvm_context;
struct ac_nir_context {
return container_of(abi, ctx, abi);
}
-static LLVMValueRef get_sampler_desc(struct nir_to_llvm_context *ctx,
+static LLVMValueRef get_sampler_desc(struct ac_nir_context *ctx,
const nir_deref_var *deref,
- enum desc_type desc_type);
+ enum ac_descriptor_type desc_type,
+ bool image, bool write);
+
static unsigned radeon_llvm_reg_index_soa(unsigned index, unsigned chan)
{
return (index * 4) + chan;
LLVMConstInt(ac->i32, index, false), "");
}
-static LLVMValueRef trim_vector(struct nir_to_llvm_context *ctx,
+static LLVMValueRef trim_vector(struct ac_llvm_context *ctx,
LLVMValueRef value, unsigned count)
{
unsigned num_components = get_llvm_num_components(value);
}
static LLVMValueRef
-get_buffer_size(struct nir_to_llvm_context *ctx, LLVMValueRef descriptor, bool in_elements)
+get_buffer_size(struct ac_nir_context *ctx, LLVMValueRef descriptor, bool in_elements)
{
LLVMValueRef size =
- LLVMBuildExtractElement(ctx->builder, descriptor,
- LLVMConstInt(ctx->i32, 2, false), "");
+ LLVMBuildExtractElement(ctx->ac.builder, descriptor,
+ LLVMConstInt(ctx->ac.i32, 2, false), "");
/* VI only */
- if (ctx->options->chip_class >= VI && in_elements) {
+ if (ctx->abi->chip_class >= VI && in_elements) {
/* On VI, the descriptor contains the size in bytes,
* but TXQ must return the size in elements.
* The stride is always non-zero for resources using TXQ.
*/
LLVMValueRef stride =
- LLVMBuildExtractElement(ctx->builder, descriptor,
- LLVMConstInt(ctx->i32, 1, false), "");
- stride = LLVMBuildLShr(ctx->builder, stride,
- LLVMConstInt(ctx->i32, 16, false), "");
- stride = LLVMBuildAnd(ctx->builder, stride,
- LLVMConstInt(ctx->i32, 0x3fff, false), "");
+ LLVMBuildExtractElement(ctx->ac.builder, descriptor,
+ LLVMConstInt(ctx->ac.i32, 1, false), "");
+ stride = LLVMBuildLShr(ctx->ac.builder, stride,
+ LLVMConstInt(ctx->ac.i32, 16, false), "");
+ stride = LLVMBuildAnd(ctx->ac.builder, stride,
+ LLVMConstInt(ctx->ac.i32, 0x3fff, false), "");
- size = LLVMBuildUDiv(ctx->builder, size, stride, "");
+ size = LLVMBuildUDiv(ctx->ac.builder, size, stride, "");
}
return size;
}
strcpy(buf, "i32");
}
-static LLVMValueRef radv_lower_gather4_integer(struct nir_to_llvm_context *ctx,
+static LLVMValueRef radv_lower_gather4_integer(struct ac_llvm_context *ctx,
struct ac_image_args *args,
const nir_tex_instr *instr)
{
txq_args.da = instr->is_array || instr->sampler_dim == GLSL_SAMPLER_DIM_CUBE;
txq_args.opcode = ac_image_get_resinfo;
txq_args.dmask = 0xf;
- txq_args.addr = ctx->i32zero;
+ txq_args.addr = ctx->i32_0;
txq_args.resource = args->resource;
- LLVMValueRef size = ac_build_image_opcode(&ctx->ac, &txq_args);
+ LLVMValueRef size = ac_build_image_opcode(ctx, &txq_args);
for (c = 0; c < 2; c++) {
half_texel[c] = LLVMBuildExtractElement(ctx->builder, size,
LLVMConstInt(ctx->i32, c, false), "");
half_texel[c] = LLVMBuildUIToFP(ctx->builder, half_texel[c], ctx->f32, "");
- half_texel[c] = ac_build_fdiv(&ctx->ac, ctx->f32one, half_texel[c]);
+ half_texel[c] = ac_build_fdiv(ctx, ctx->f32_1, half_texel[c]);
half_texel[c] = LLVMBuildFMul(ctx->builder, half_texel[c],
LLVMConstReal(ctx->f32, -0.5), "");
}
/* workaround 8/8/8/8 uint/sint cube gather bug */
/* first detect it then change to a scaled read and f2i */
- tmp = LLVMBuildExtractElement(ctx->builder, args->resource, ctx->i32one, "");
+ tmp = LLVMBuildExtractElement(ctx->builder, args->resource, ctx->i32_1, "");
tmp2 = tmp;
/* extract the DATA_FORMAT */
- tmp = ac_build_bfe(&ctx->ac, tmp, LLVMConstInt(ctx->i32, 20, false),
+ tmp = ac_build_bfe(ctx, tmp, LLVMConstInt(ctx->i32, 20, false),
LLVMConstInt(ctx->i32, 6, false), false);
/* is the DATA_FORMAT == 8_8_8_8 */
tmp2 = LLVMBuildAnd(ctx->builder, tmp2, LLVMConstInt(ctx->i32, C_008F14_NUM_FORMAT_GFX6, false), "");
tmp2 = LLVMBuildOr(ctx->builder, tmp2, tmp, "");
- args->resource = LLVMBuildInsertElement(ctx->builder, args->resource, tmp2, ctx->i32one, "");
+ args->resource = LLVMBuildInsertElement(ctx->builder, args->resource, tmp2, ctx->i32_1, "");
/* don't modify the coordinates for this case */
coord = LLVMBuildSelect(ctx->builder, compare_cube_wa, orig_coords, coord, "");
}
args->addr = coord;
- result = ac_build_image_opcode(&ctx->ac, args);
+ result = ac_build_image_opcode(ctx, args);
if (instr->sampler_dim == GLSL_SAMPLER_DIM_CUBE) {
LLVMValueRef tmp, tmp2;
return result;
}
-static LLVMValueRef build_tex_intrinsic(struct nir_to_llvm_context *ctx,
+static LLVMValueRef build_tex_intrinsic(struct ac_nir_context *ctx,
const nir_tex_instr *instr,
bool lod_is_zero,
struct ac_image_args *args)
return ac_build_buffer_load_format(&ctx->ac,
args->resource,
args->addr,
- LLVMConstInt(ctx->i32, 0, false),
+ LLVMConstInt(ctx->ac.i32, 0, false),
true);
}
if (instr->op == nir_texop_tg4) {
enum glsl_base_type stype = glsl_get_sampler_result_type(instr->texture->var->type);
if (stype == GLSL_TYPE_UINT || stype == GLSL_TYPE_INT) {
- return radv_lower_gather4_integer(ctx, args, instr);
+ return radv_lower_gather4_integer(&ctx->ac, args, instr);
}
}
return ac_build_image_opcode(&ctx->ac, args);
return LLVMBuildLoad(ctx->builder, ptr, "");
}
-static LLVMValueRef visit_get_buffer_size(struct nir_to_llvm_context *ctx,
+static LLVMValueRef visit_get_buffer_size(struct ac_nir_context *ctx,
const nir_intrinsic_instr *instr)
{
- LLVMValueRef desc = get_src(ctx->nir, instr->src[0]);
+ LLVMValueRef desc = get_src(ctx, instr->src[0]);
return get_buffer_size(ctx, desc, false);
}
data_type = LLVMVectorType(ctx->f32, components_32bit);
base_data = to_float(&ctx->ac, src_data);
- base_data = trim_vector(ctx, base_data, instr->num_components);
+ base_data = trim_vector(&ctx->ac, base_data, instr->num_components);
base_data = LLVMBuildBitCast(ctx->builder, base_data,
data_type, "");
base_offset = get_src(ctx->nir, instr->src[2]); /* voffset */
result = ac_build_buffer_load(&ctx->ac, ctx->hs_ring_tess_offchip, instr->num_components, NULL,
buf_addr, ctx->oc_lds, is_compact ? (4 * const_index) : 0, 1, 0, true, false);
- result = trim_vector(ctx, result, instr->num_components);
+ result = trim_vector(&ctx->ac, result, instr->num_components);
result = LLVMBuildBitCast(ctx->builder, result, get_def_type(ctx->nir, &instr->dest.ssa), "");
return result;
}
* The sample index should be adjusted as follows:
* sample_index = (fmask >> (sample_index * 4)) & 0xF;
*/
-static LLVMValueRef adjust_sample_index_using_fmask(struct nir_to_llvm_context *ctx,
+static LLVMValueRef adjust_sample_index_using_fmask(struct ac_llvm_context *ctx,
LLVMValueRef coord_x, LLVMValueRef coord_y,
LLVMValueRef coord_z,
LLVMValueRef sample_index,
args.da = coord_z ? true : false;
args.resource = fmask_desc_ptr;
args.dmask = 0xf;
- args.addr = ac_build_gather_values(&ctx->ac, fmask_load_address, coord_z ? 4 : 2);
+ args.addr = ac_build_gather_values(ctx, fmask_load_address, coord_z ? 4 : 2);
- res = ac_build_image_opcode(&ctx->ac, &args);
+ res = ac_build_image_opcode(ctx, &args);
- res = to_integer(&ctx->ac, res);
+ res = to_integer(ctx, res);
LLVMValueRef four = LLVMConstInt(ctx->i32, 4, false);
LLVMValueRef F = LLVMConstInt(ctx->i32, 0xf, false);
LLVMValueRef fmask = LLVMBuildExtractElement(ctx->builder,
res,
- ctx->i32zero, "");
+ ctx->i32_0, "");
LLVMValueRef sample_index4 =
LLVMBuildMul(ctx->builder, sample_index, four, "");
LLVMValueRef fmask_word1 =
LLVMBuildExtractElement(ctx->builder, fmask_desc,
- ctx->i32one, "");
+ ctx->i32_1, "");
LLVMValueRef word1_is_nonzero =
LLVMBuildICmp(ctx->builder, LLVMIntNE,
- fmask_word1, ctx->i32zero, "");
+ fmask_word1, ctx->i32_0, "");
/* Replace the MSAA sample index. */
sample_index =
return sample_index;
}
-static LLVMValueRef get_image_coords(struct nir_to_llvm_context *ctx,
+static LLVMValueRef get_image_coords(struct ac_nir_context *ctx,
const nir_intrinsic_instr *instr)
{
const struct glsl_type *type = instr->variables[0]->var->type;
if(instr->variables[0]->deref.child)
type = instr->variables[0]->deref.child->type;
- LLVMValueRef src0 = get_src(ctx->nir, instr->src[0]);
+ LLVMValueRef src0 = get_src(ctx, instr->src[0]);
LLVMValueRef coords[4];
LLVMValueRef masks[] = {
- LLVMConstInt(ctx->i32, 0, false), LLVMConstInt(ctx->i32, 1, false),
- LLVMConstInt(ctx->i32, 2, false), LLVMConstInt(ctx->i32, 3, false),
+ LLVMConstInt(ctx->ac.i32, 0, false), LLVMConstInt(ctx->ac.i32, 1, false),
+ LLVMConstInt(ctx->ac.i32, 2, false), LLVMConstInt(ctx->ac.i32, 3, false),
};
LLVMValueRef res;
- LLVMValueRef sample_index = llvm_extract_elem(&ctx->ac, get_src(ctx->nir, instr->src[1]), 0);
+ LLVMValueRef sample_index = llvm_extract_elem(&ctx->ac, get_src(ctx, instr->src[1]), 0);
int count;
enum glsl_sampler_dim dim = glsl_get_sampler_dim(type);
LLVMValueRef fmask_load_address[3];
int chan;
- fmask_load_address[0] = LLVMBuildExtractElement(ctx->builder, src0, masks[0], "");
- fmask_load_address[1] = LLVMBuildExtractElement(ctx->builder, src0, masks[1], "");
+ fmask_load_address[0] = LLVMBuildExtractElement(ctx->ac.builder, src0, masks[0], "");
+ fmask_load_address[1] = LLVMBuildExtractElement(ctx->ac.builder, src0, masks[1], "");
if (glsl_sampler_type_is_array(type))
- fmask_load_address[2] = LLVMBuildExtractElement(ctx->builder, src0, masks[2], "");
+ fmask_load_address[2] = LLVMBuildExtractElement(ctx->ac.builder, src0, masks[2], "");
else
fmask_load_address[2] = NULL;
if (add_frag_pos) {
for (chan = 0; chan < 2; ++chan)
- fmask_load_address[chan] = LLVMBuildAdd(ctx->builder, fmask_load_address[chan], LLVMBuildFPToUI(ctx->builder, ctx->frag_pos[chan], ctx->i32, ""), "");
+ fmask_load_address[chan] = LLVMBuildAdd(ctx->ac.builder, fmask_load_address[chan], LLVMBuildFPToUI(ctx->ac.builder, ctx->nctx->frag_pos[chan], ctx->ac.i32, ""), "");
}
- sample_index = adjust_sample_index_using_fmask(ctx,
+ sample_index = adjust_sample_index_using_fmask(&ctx->ac,
fmask_load_address[0],
fmask_load_address[1],
fmask_load_address[2],
sample_index,
- get_sampler_desc(ctx, instr->variables[0], DESC_FMASK));
+ get_sampler_desc(ctx, instr->variables[0], AC_DESC_FMASK, true, false));
}
if (count == 1) {
if (instr->src[0].ssa->num_components)
- res = LLVMBuildExtractElement(ctx->builder, src0, masks[0], "");
+ res = LLVMBuildExtractElement(ctx->ac.builder, src0, masks[0], "");
else
res = src0;
} else {
if (is_ms)
count--;
for (chan = 0; chan < count; ++chan) {
- coords[chan] = LLVMBuildExtractElement(ctx->builder, src0, masks[chan], "");
+ coords[chan] = LLVMBuildExtractElement(ctx->ac.builder, src0, masks[chan], "");
}
if (add_frag_pos) {
for (chan = 0; chan < count; ++chan)
- coords[chan] = LLVMBuildAdd(ctx->builder, coords[chan], LLVMBuildFPToUI(ctx->builder, ctx->frag_pos[chan], ctx->i32, ""), "");
+ coords[chan] = LLVMBuildAdd(ctx->ac.builder, coords[chan], LLVMBuildFPToUI(ctx->ac.builder, ctx->nctx->frag_pos[chan], ctx->ac.i32, ""), "");
}
if (is_ms) {
coords[count] = sample_index;
}
if (count == 3) {
- coords[3] = LLVMGetUndef(ctx->i32);
+ coords[3] = LLVMGetUndef(ctx->ac.i32);
count = 4;
}
res = ac_build_gather_values(&ctx->ac, coords, count);
return res;
}
-static LLVMValueRef visit_image_load(struct nir_to_llvm_context *ctx,
+static LLVMValueRef visit_image_load(struct ac_nir_context *ctx,
const nir_intrinsic_instr *instr)
{
LLVMValueRef params[7];
char intrinsic_name[64];
const nir_variable *var = instr->variables[0]->var;
const struct glsl_type *type = var->type;
+ LLVMValueRef i1false = LLVMConstInt(ctx->ac.i1, 0, false);
+ LLVMValueRef i1true = LLVMConstInt(ctx->ac.i1, 1, false);
+
if(instr->variables[0]->deref.child)
type = instr->variables[0]->deref.child->type;
type = glsl_without_array(type);
if (glsl_get_sampler_dim(type) == GLSL_SAMPLER_DIM_BUF) {
- params[0] = get_sampler_desc(ctx, instr->variables[0], DESC_BUFFER);
- params[1] = LLVMBuildExtractElement(ctx->builder, get_src(ctx->nir, instr->src[0]),
- LLVMConstInt(ctx->i32, 0, false), ""); /* vindex */
- params[2] = LLVMConstInt(ctx->i32, 0, false); /* voffset */
- params[3] = ctx->i1false; /* glc */
- params[4] = ctx->i1false; /* slc */
- res = ac_build_intrinsic(&ctx->ac, "llvm.amdgcn.buffer.load.format.v4f32", ctx->v4f32,
+ params[0] = get_sampler_desc(ctx, instr->variables[0], AC_DESC_BUFFER, true, false);
+ params[1] = LLVMBuildExtractElement(ctx->ac.builder, get_src(ctx, instr->src[0]),
+ ctx->ac.i32_0, ""); /* vindex */
+ params[2] = ctx->ac.i32_0; /* voffset */
+ params[3] = i1false; /* glc */
+ params[4] = i1false; /* slc */
+ res = ac_build_intrinsic(&ctx->ac, "llvm.amdgcn.buffer.load.format.v4f32", ctx->ac.v4f32,
params, 5, 0);
- res = trim_vector(ctx, res, instr->dest.ssa.num_components);
+ res = trim_vector(&ctx->ac, res, instr->dest.ssa.num_components);
res = to_integer(&ctx->ac, res);
} else {
bool is_da = glsl_sampler_type_is_array(type) ||
glsl_get_sampler_dim(type) == GLSL_SAMPLER_DIM_CUBE;
- LLVMValueRef da = is_da ? ctx->i1true : ctx->i1false;
- LLVMValueRef glc = ctx->i1false;
- LLVMValueRef slc = ctx->i1false;
+ LLVMValueRef da = is_da ? i1true : i1false;
+ LLVMValueRef glc = i1false;
+ LLVMValueRef slc = i1false;
params[0] = get_image_coords(ctx, instr);
- params[1] = get_sampler_desc(ctx, instr->variables[0], DESC_IMAGE);
- params[2] = LLVMConstInt(ctx->i32, 15, false); /* dmask */
+ params[1] = get_sampler_desc(ctx, instr->variables[0], AC_DESC_IMAGE, true, false);
+ params[2] = LLVMConstInt(ctx->ac.i32, 15, false); /* dmask */
if (HAVE_LLVM <= 0x0309) {
- params[3] = ctx->i1false; /* r128 */
+ params[3] = i1false; /* r128 */
params[4] = da;
params[5] = glc;
params[6] = slc;
} else {
- LLVMValueRef lwe = ctx->i1false;
+ LLVMValueRef lwe = i1false;
params[3] = glc;
params[4] = slc;
params[5] = lwe;
}
ac_get_image_intr_name("llvm.amdgcn.image.load",
- ctx->v4f32, /* vdata */
+ ctx->ac.v4f32, /* vdata */
LLVMTypeOf(params[0]), /* coords */
LLVMTypeOf(params[1]), /* rsrc */
intrinsic_name, sizeof(intrinsic_name));
- res = ac_build_intrinsic(&ctx->ac, intrinsic_name, ctx->v4f32,
+ res = ac_build_intrinsic(&ctx->ac, intrinsic_name, ctx->ac.v4f32,
params, 7, AC_FUNC_ATTR_READONLY);
}
return to_integer(&ctx->ac, res);
}
-static void visit_image_store(struct nir_to_llvm_context *ctx,
+static void visit_image_store(struct ac_nir_context *ctx,
nir_intrinsic_instr *instr)
{
LLVMValueRef params[8];
char intrinsic_name[64];
const nir_variable *var = instr->variables[0]->var;
const struct glsl_type *type = glsl_without_array(var->type);
- LLVMValueRef glc = ctx->i1false;
- bool force_glc = ctx->options->chip_class == SI;
+ LLVMValueRef i1false = LLVMConstInt(ctx->ac.i1, 0, false);
+ LLVMValueRef i1true = LLVMConstInt(ctx->ac.i1, 1, false);
+ LLVMValueRef glc = i1false;
+ bool force_glc = ctx->abi->chip_class == SI;
if (force_glc)
- glc = ctx->i1true;
- if (ctx->stage == MESA_SHADER_FRAGMENT)
- ctx->shader_info->fs.writes_memory = true;
+ glc = i1true;
if (glsl_get_sampler_dim(type) == GLSL_SAMPLER_DIM_BUF) {
- params[0] = to_float(&ctx->ac, get_src(ctx->nir, instr->src[2])); /* data */
- params[1] = get_sampler_desc(ctx, instr->variables[0], DESC_BUFFER);
- params[2] = LLVMBuildExtractElement(ctx->builder, get_src(ctx->nir, instr->src[0]),
- LLVMConstInt(ctx->i32, 0, false), ""); /* vindex */
- params[3] = LLVMConstInt(ctx->i32, 0, false); /* voffset */
+ params[0] = to_float(&ctx->ac, get_src(ctx, instr->src[2])); /* data */
+ params[1] = get_sampler_desc(ctx, instr->variables[0], AC_DESC_BUFFER, true, true);
+ params[2] = LLVMBuildExtractElement(ctx->ac.builder, get_src(ctx, instr->src[0]),
+ ctx->ac.i32_0, ""); /* vindex */
+ params[3] = ctx->ac.i32_0; /* voffset */
params[4] = glc; /* glc */
- params[5] = ctx->i1false; /* slc */
- ac_build_intrinsic(&ctx->ac, "llvm.amdgcn.buffer.store.format.v4f32", ctx->voidt,
+ params[5] = i1false; /* slc */
+ ac_build_intrinsic(&ctx->ac, "llvm.amdgcn.buffer.store.format.v4f32", ctx->ac.voidt,
params, 6, 0);
} else {
bool is_da = glsl_sampler_type_is_array(type) ||
glsl_get_sampler_dim(type) == GLSL_SAMPLER_DIM_CUBE;
- LLVMValueRef da = is_da ? ctx->i1true : ctx->i1false;
- LLVMValueRef slc = ctx->i1false;
+ LLVMValueRef da = is_da ? i1true : i1false;
+ LLVMValueRef slc = i1false;
- params[0] = to_float(&ctx->ac, get_src(ctx->nir, instr->src[2]));
+ params[0] = to_float(&ctx->ac, get_src(ctx, instr->src[2]));
params[1] = get_image_coords(ctx, instr); /* coords */
- params[2] = get_sampler_desc(ctx, instr->variables[0], DESC_IMAGE);
- params[3] = LLVMConstInt(ctx->i32, 15, false); /* dmask */
+ params[2] = get_sampler_desc(ctx, instr->variables[0], AC_DESC_IMAGE, true, true);
+ params[3] = LLVMConstInt(ctx->ac.i32, 15, false); /* dmask */
if (HAVE_LLVM <= 0x0309) {
- params[4] = ctx->i1false; /* r128 */
+ params[4] = i1false; /* r128 */
params[5] = da;
params[6] = glc;
params[7] = slc;
} else {
- LLVMValueRef lwe = ctx->i1false;
+ LLVMValueRef lwe = i1false;
params[4] = glc;
params[5] = slc;
params[6] = lwe;
LLVMTypeOf(params[2]), /* rsrc */
intrinsic_name, sizeof(intrinsic_name));
- ac_build_intrinsic(&ctx->ac, intrinsic_name, ctx->voidt,
+ ac_build_intrinsic(&ctx->ac, intrinsic_name, ctx->ac.voidt,
params, 8, 0);
}
}
-static LLVMValueRef visit_image_atomic(struct nir_to_llvm_context *ctx,
+static LLVMValueRef visit_image_atomic(struct ac_nir_context *ctx,
const nir_intrinsic_instr *instr)
{
LLVMValueRef params[6];
const char *atomic_name;
char intrinsic_name[41];
const struct glsl_type *type = glsl_without_array(var->type);
+ LLVMValueRef i1false = LLVMConstInt(ctx->ac.i1, 0, false);
+ LLVMValueRef i1true = LLVMConstInt(ctx->ac.i1, 1, false);
MAYBE_UNUSED int length;
- if (ctx->stage == MESA_SHADER_FRAGMENT)
- ctx->shader_info->fs.writes_memory = true;
-
switch (instr->intrinsic) {
case nir_intrinsic_image_atomic_add:
atomic_name = "add";
}
if (instr->intrinsic == nir_intrinsic_image_atomic_comp_swap)
- params[param_count++] = get_src(ctx->nir, instr->src[3]);
- params[param_count++] = get_src(ctx->nir, instr->src[2]);
+ params[param_count++] = get_src(ctx, instr->src[3]);
+ params[param_count++] = get_src(ctx, instr->src[2]);
if (glsl_get_sampler_dim(type) == GLSL_SAMPLER_DIM_BUF) {
- params[param_count++] = get_sampler_desc(ctx, instr->variables[0], DESC_BUFFER);
- params[param_count++] = LLVMBuildExtractElement(ctx->builder, get_src(ctx->nir, instr->src[0]),
- LLVMConstInt(ctx->i32, 0, false), ""); /* vindex */
- params[param_count++] = ctx->i32zero; /* voffset */
- params[param_count++] = ctx->i1false; /* slc */
+ params[param_count++] = get_sampler_desc(ctx, instr->variables[0], AC_DESC_BUFFER,
+ true, true);
+ params[param_count++] = LLVMBuildExtractElement(ctx->ac.builder, get_src(ctx, instr->src[0]),
+ ctx->ac.i32_0, ""); /* vindex */
+ params[param_count++] = ctx->ac.i32_0; /* voffset */
+ params[param_count++] = i1false; /* slc */
length = snprintf(intrinsic_name, sizeof(intrinsic_name),
"llvm.amdgcn.buffer.atomic.%s", atomic_name);
glsl_get_sampler_dim(type) == GLSL_SAMPLER_DIM_CUBE;
LLVMValueRef coords = params[param_count++] = get_image_coords(ctx, instr);
- params[param_count++] = get_sampler_desc(ctx, instr->variables[0], DESC_IMAGE);
- params[param_count++] = ctx->i1false; /* r128 */
- params[param_count++] = da ? ctx->i1true : ctx->i1false; /* da */
- params[param_count++] = ctx->i1false; /* slc */
+ params[param_count++] = get_sampler_desc(ctx, instr->variables[0], AC_DESC_IMAGE,
+ true, true);
+ params[param_count++] = i1false; /* r128 */
+ params[param_count++] = da ? i1true : i1false; /* da */
+ params[param_count++] = i1false; /* slc */
build_int_type_name(LLVMTypeOf(coords),
coords_type, sizeof(coords_type));
}
assert(length < sizeof(intrinsic_name));
- return ac_build_intrinsic(&ctx->ac, intrinsic_name, ctx->i32, params, param_count, 0);
+ return ac_build_intrinsic(&ctx->ac, intrinsic_name, ctx->ac.i32, params, param_count, 0);
}
-static LLVMValueRef visit_image_size(struct nir_to_llvm_context *ctx,
+static LLVMValueRef visit_image_size(struct ac_nir_context *ctx,
const nir_intrinsic_instr *instr)
{
LLVMValueRef res;
type = instr->variables[0]->deref.child->type;
if (glsl_get_sampler_dim(type) == GLSL_SAMPLER_DIM_BUF)
- return get_buffer_size(ctx, get_sampler_desc(ctx, instr->variables[0], DESC_BUFFER), true);
+ return get_buffer_size(ctx,
+ get_sampler_desc(ctx, instr->variables[0],
+ AC_DESC_BUFFER, true, false), true);
struct ac_image_args args = { 0 };
args.da = da;
args.dmask = 0xf;
- args.resource = get_sampler_desc(ctx, instr->variables[0], DESC_IMAGE);
+ args.resource = get_sampler_desc(ctx, instr->variables[0], AC_DESC_IMAGE, true, false);
args.opcode = ac_image_get_resinfo;
- args.addr = ctx->i32zero;
+ args.addr = ctx->ac.i32_0;
res = ac_build_image_opcode(&ctx->ac, &args);
if (glsl_get_sampler_dim(type) == GLSL_SAMPLER_DIM_CUBE &&
glsl_sampler_type_is_array(type)) {
- LLVMValueRef two = LLVMConstInt(ctx->i32, 2, false);
- LLVMValueRef six = LLVMConstInt(ctx->i32, 6, false);
- LLVMValueRef z = LLVMBuildExtractElement(ctx->builder, res, two, "");
- z = LLVMBuildSDiv(ctx->builder, z, six, "");
- res = LLVMBuildInsertElement(ctx->builder, res, z, two, "");
+ LLVMValueRef two = LLVMConstInt(ctx->ac.i32, 2, false);
+ LLVMValueRef six = LLVMConstInt(ctx->ac.i32, 6, false);
+ LLVMValueRef z = LLVMBuildExtractElement(ctx->ac.builder, res, two, "");
+ z = LLVMBuildSDiv(ctx->ac.builder, z, six, "");
+ res = LLVMBuildInsertElement(ctx->ac.builder, res, z, two, "");
}
return res;
}
const nir_intrinsic_instr *instr)
{
LLVMValueRef cond;
- ctx->shader_info->fs.can_discard = true;
cond = LLVMBuildICmp(ctx->builder, LLVMIntNE,
get_src(ctx->nir, instr->src[0]),
break;
case nir_intrinsic_load_instance_id:
result = ctx->abi->instance_id;
- ctx->nctx->shader_info->vs.vgpr_comp_cnt = MAX2(3,
- ctx->nctx->shader_info->vs.vgpr_comp_cnt);
break;
case nir_intrinsic_load_num_work_groups:
result = ctx->nctx->num_work_groups;
result = visit_load_ubo_buffer(ctx, instr);
break;
case nir_intrinsic_get_buffer_size:
- result = visit_get_buffer_size(ctx->nctx, instr);
+ result = visit_get_buffer_size(ctx, instr);
break;
case nir_intrinsic_load_var:
result = visit_load_var(ctx, instr);
visit_store_var(ctx, instr);
break;
case nir_intrinsic_image_load:
- result = visit_image_load(ctx->nctx, instr);
+ result = visit_image_load(ctx, instr);
break;
case nir_intrinsic_image_store:
- visit_image_store(ctx->nctx, instr);
+ visit_image_store(ctx, instr);
break;
case nir_intrinsic_image_atomic_add:
case nir_intrinsic_image_atomic_min:
case nir_intrinsic_image_atomic_xor:
case nir_intrinsic_image_atomic_exchange:
case nir_intrinsic_image_atomic_comp_swap:
- result = visit_image_atomic(ctx->nctx, instr);
+ result = visit_image_atomic(ctx, instr);
break;
case nir_intrinsic_image_size:
- result = visit_image_size(ctx->nctx, instr);
+ result = visit_image_size(ctx, instr);
break;
case nir_intrinsic_discard:
- ctx->nctx->shader_info->fs.can_discard = true;
ac_build_intrinsic(&ctx->ac, "llvm.AMDGPU.kilp",
LLVMVoidTypeInContext(ctx->ac.context),
NULL, 0, AC_FUNC_ATTR_LEGACY);
}
}
-static LLVMValueRef get_sampler_desc(struct nir_to_llvm_context *ctx,
- const nir_deref_var *deref,
- enum desc_type desc_type)
+static LLVMValueRef radv_get_sampler_desc(struct ac_shader_abi *abi,
+ unsigned descriptor_set,
+ unsigned base_index,
+ unsigned constant_index,
+ LLVMValueRef index,
+ enum ac_descriptor_type desc_type,
+ bool image, bool write)
{
- unsigned desc_set = deref->var->data.descriptor_set;
- LLVMValueRef list = ctx->descriptor_sets[desc_set];
- struct radv_descriptor_set_layout *layout = ctx->options->layout->set[desc_set].layout;
- struct radv_descriptor_set_binding_layout *binding = layout->binding + deref->var->data.binding;
+ struct nir_to_llvm_context *ctx = nir_to_llvm_context_from_abi(abi);
+ LLVMValueRef list = ctx->descriptor_sets[descriptor_set];
+ struct radv_descriptor_set_layout *layout = ctx->options->layout->set[descriptor_set].layout;
+ struct radv_descriptor_set_binding_layout *binding = layout->binding + base_index;
unsigned offset = binding->offset;
unsigned stride = binding->size;
unsigned type_size;
LLVMBuilderRef builder = ctx->builder;
LLVMTypeRef type;
- LLVMValueRef index = NULL;
- unsigned constant_index = 0;
- assert(deref->var->data.binding < layout->binding_count);
+ assert(base_index < layout->binding_count);
+
+ if (write && ctx->stage == MESA_SHADER_FRAGMENT)
+ ctx->shader_info->fs.writes_memory = true;
switch (desc_type) {
- case DESC_IMAGE:
+ case AC_DESC_IMAGE:
type = ctx->v8i32;
type_size = 32;
break;
- case DESC_FMASK:
+ case AC_DESC_FMASK:
type = ctx->v8i32;
offset += 32;
type_size = 32;
break;
- case DESC_SAMPLER:
+ case AC_DESC_SAMPLER:
type = ctx->v4i32;
if (binding->type == VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER)
offset += 64;
type_size = 16;
break;
- case DESC_BUFFER:
+ case AC_DESC_BUFFER:
type = ctx->v4i32;
type_size = 16;
break;
unreachable("invalid desc_type\n");
}
- if (deref->deref.child) {
- const nir_deref_array *child =
- (const nir_deref_array *)deref->deref.child;
-
- assert(child->deref_array_type != nir_deref_array_type_wildcard);
- offset += child->base_offset * stride;
- if (child->deref_array_type == nir_deref_array_type_indirect) {
- index = get_src(ctx->nir, child->indirect);
- }
+ offset += constant_index * stride;
- constant_index = child->base_offset;
- }
- if (desc_type == DESC_SAMPLER && binding->immutable_samplers_offset &&
+ if (desc_type == AC_DESC_SAMPLER && binding->immutable_samplers_offset &&
(!index || binding->immutable_samplers_equal)) {
if (binding->immutable_samplers_equal)
constant_index = 0;
return ac_build_indexed_load_const(&ctx->ac, list, index);
}
-static void set_tex_fetch_args(struct nir_to_llvm_context *ctx,
+static LLVMValueRef get_sampler_desc(struct ac_nir_context *ctx,
+ const nir_deref_var *deref,
+ enum ac_descriptor_type desc_type,
+ bool image, bool write)
+{
+ LLVMValueRef index = NULL;
+ unsigned constant_index = 0;
+ const nir_deref *tail = &deref->deref;
+
+ while (tail->child) {
+ const nir_deref_array *child = nir_deref_as_array(tail->child);
+ unsigned array_size = glsl_get_aoa_size(tail->child->type);
+
+ if (!array_size)
+ array_size = 1;
+
+ assert(child->deref_array_type != nir_deref_array_type_wildcard);
+
+ if (child->deref_array_type == nir_deref_array_type_indirect) {
+ LLVMValueRef indirect = get_src(ctx, child->indirect);
+
+ indirect = LLVMBuildMul(ctx->ac.builder, indirect,
+ LLVMConstInt(ctx->ac.i32, array_size, false), "");
+
+ if (!index)
+ index = indirect;
+ else
+ index = LLVMBuildAdd(ctx->ac.builder, index, indirect, "");
+ }
+
+ constant_index += child->base_offset * array_size;
+
+ tail = &child->deref;
+ }
+
+ return ctx->abi->load_sampler_desc(ctx->abi,
+ deref->var->data.descriptor_set,
+ deref->var->data.binding,
+ constant_index, index,
+ desc_type, image, write);
+}
+
+static void set_tex_fetch_args(struct ac_llvm_context *ctx,
struct ac_image_args *args,
const nir_tex_instr *instr,
nir_texop op,
param[count++] = LLVMGetUndef(ctx->i32);
if (count > 1)
- args->addr = ac_build_gather_values(&ctx->ac, param, count);
+ args->addr = ac_build_gather_values(ctx, param, count);
else
args->addr = param[0];
* VI:
* The ANISO_OVERRIDE sampler field enables this fix in TA.
*/
-static LLVMValueRef sici_fix_sampler_aniso(struct nir_to_llvm_context *ctx,
+static LLVMValueRef sici_fix_sampler_aniso(struct ac_nir_context *ctx,
LLVMValueRef res, LLVMValueRef samp)
{
- LLVMBuilderRef builder = ctx->builder;
+ LLVMBuilderRef builder = ctx->ac.builder;
LLVMValueRef img7, samp0;
- if (ctx->options->chip_class >= VI)
+ if (ctx->abi->chip_class >= VI)
return samp;
img7 = LLVMBuildExtractElement(builder, res,
- LLVMConstInt(ctx->i32, 7, 0), "");
+ LLVMConstInt(ctx->ac.i32, 7, 0), "");
samp0 = LLVMBuildExtractElement(builder, samp,
- LLVMConstInt(ctx->i32, 0, 0), "");
+ LLVMConstInt(ctx->ac.i32, 0, 0), "");
samp0 = LLVMBuildAnd(builder, samp0, img7, "");
return LLVMBuildInsertElement(builder, samp, samp0,
- LLVMConstInt(ctx->i32, 0, 0), "");
+ LLVMConstInt(ctx->ac.i32, 0, 0), "");
}
-static void tex_fetch_ptrs(struct nir_to_llvm_context *ctx,
+static void tex_fetch_ptrs(struct ac_nir_context *ctx,
nir_tex_instr *instr,
LLVMValueRef *res_ptr, LLVMValueRef *samp_ptr,
LLVMValueRef *fmask_ptr)
{
if (instr->sampler_dim == GLSL_SAMPLER_DIM_BUF)
- *res_ptr = get_sampler_desc(ctx, instr->texture, DESC_BUFFER);
+ *res_ptr = get_sampler_desc(ctx, instr->texture, AC_DESC_BUFFER, false, false);
else
- *res_ptr = get_sampler_desc(ctx, instr->texture, DESC_IMAGE);
+ *res_ptr = get_sampler_desc(ctx, instr->texture, AC_DESC_IMAGE, false, false);
if (samp_ptr) {
if (instr->sampler)
- *samp_ptr = get_sampler_desc(ctx, instr->sampler, DESC_SAMPLER);
+ *samp_ptr = get_sampler_desc(ctx, instr->sampler, AC_DESC_SAMPLER, false, false);
else
- *samp_ptr = get_sampler_desc(ctx, instr->texture, DESC_SAMPLER);
+ *samp_ptr = get_sampler_desc(ctx, instr->texture, AC_DESC_SAMPLER, false, false);
if (instr->sampler_dim < GLSL_SAMPLER_DIM_RECT)
*samp_ptr = sici_fix_sampler_aniso(ctx, *res_ptr, *samp_ptr);
}
if (fmask_ptr && !instr->sampler && (instr->op == nir_texop_txf_ms ||
instr->op == nir_texop_samples_identical))
- *fmask_ptr = get_sampler_desc(ctx, instr->texture, DESC_FMASK);
+ *fmask_ptr = get_sampler_desc(ctx, instr->texture, AC_DESC_FMASK, false, false);
}
-static LLVMValueRef apply_round_slice(struct nir_to_llvm_context *ctx,
+static LLVMValueRef apply_round_slice(struct ac_llvm_context *ctx,
LLVMValueRef coord)
{
- coord = to_float(&ctx->ac, coord);
- coord = ac_build_intrinsic(&ctx->ac, "llvm.rint.f32", ctx->f32, &coord, 1, 0);
- coord = to_integer(&ctx->ac, coord);
+ coord = to_float(ctx, coord);
+ coord = ac_build_intrinsic(ctx, "llvm.rint.f32", ctx->f32, &coord, 1, 0);
+ coord = to_integer(ctx, coord);
return coord;
}
-static void visit_tex(struct nir_to_llvm_context *ctx, nir_tex_instr *instr)
+static void visit_tex(struct ac_nir_context *ctx, nir_tex_instr *instr)
{
LLVMValueRef result = NULL;
struct ac_image_args args = { 0 };
unsigned chan, count = 0;
unsigned const_src = 0, num_deriv_comp = 0;
bool lod_is_zero = false;
+
tex_fetch_ptrs(ctx, instr, &res_ptr, &samp_ptr, &fmask_ptr);
for (unsigned i = 0; i < instr->num_srcs; i++) {
switch (instr->src[i].src_type) {
case nir_tex_src_coord:
- coord = get_src(ctx->nir, instr->src[i].src);
+ coord = get_src(ctx, instr->src[i].src);
break;
case nir_tex_src_projector:
break;
case nir_tex_src_comparator:
- comparator = get_src(ctx->nir, instr->src[i].src);
+ comparator = get_src(ctx, instr->src[i].src);
break;
case nir_tex_src_offset:
- offsets = get_src(ctx->nir, instr->src[i].src);
+ offsets = get_src(ctx, instr->src[i].src);
const_src = i;
break;
case nir_tex_src_bias:
- bias = get_src(ctx->nir, instr->src[i].src);
+ bias = get_src(ctx, instr->src[i].src);
break;
case nir_tex_src_lod: {
nir_const_value *val = nir_src_as_const_value(instr->src[i].src);
if (val && val->i32[0] == 0)
lod_is_zero = true;
- lod = get_src(ctx->nir, instr->src[i].src);
+ lod = get_src(ctx, instr->src[i].src);
break;
}
case nir_tex_src_ms_index:
- sample_index = get_src(ctx->nir, instr->src[i].src);
+ sample_index = get_src(ctx, instr->src[i].src);
break;
case nir_tex_src_ms_mcs:
break;
case nir_tex_src_ddx:
- ddx = get_src(ctx->nir, instr->src[i].src);
+ ddx = get_src(ctx, instr->src[i].src);
num_deriv_comp = instr->src[i].src.ssa->num_components;
break;
case nir_tex_src_ddy:
- ddy = get_src(ctx->nir, instr->src[i].src);
+ ddy = get_src(ctx, instr->src[i].src);
break;
case nir_tex_src_texture_offset:
case nir_tex_src_sampler_offset:
if (instr->op == nir_texop_texture_samples) {
LLVMValueRef res, samples, is_msaa;
- res = LLVMBuildBitCast(ctx->builder, res_ptr, ctx->v8i32, "");
- samples = LLVMBuildExtractElement(ctx->builder, res,
- LLVMConstInt(ctx->i32, 3, false), "");
- is_msaa = LLVMBuildLShr(ctx->builder, samples,
- LLVMConstInt(ctx->i32, 28, false), "");
- is_msaa = LLVMBuildAnd(ctx->builder, is_msaa,
- LLVMConstInt(ctx->i32, 0xe, false), "");
- is_msaa = LLVMBuildICmp(ctx->builder, LLVMIntEQ, is_msaa,
- LLVMConstInt(ctx->i32, 0xe, false), "");
-
- samples = LLVMBuildLShr(ctx->builder, samples,
- LLVMConstInt(ctx->i32, 16, false), "");
- samples = LLVMBuildAnd(ctx->builder, samples,
- LLVMConstInt(ctx->i32, 0xf, false), "");
- samples = LLVMBuildShl(ctx->builder, ctx->i32one,
+ res = LLVMBuildBitCast(ctx->ac.builder, res_ptr, ctx->ac.v8i32, "");
+ samples = LLVMBuildExtractElement(ctx->ac.builder, res,
+ LLVMConstInt(ctx->ac.i32, 3, false), "");
+ is_msaa = LLVMBuildLShr(ctx->ac.builder, samples,
+ LLVMConstInt(ctx->ac.i32, 28, false), "");
+ is_msaa = LLVMBuildAnd(ctx->ac.builder, is_msaa,
+ LLVMConstInt(ctx->ac.i32, 0xe, false), "");
+ is_msaa = LLVMBuildICmp(ctx->ac.builder, LLVMIntEQ, is_msaa,
+ LLVMConstInt(ctx->ac.i32, 0xe, false), "");
+
+ samples = LLVMBuildLShr(ctx->ac.builder, samples,
+ LLVMConstInt(ctx->ac.i32, 16, false), "");
+ samples = LLVMBuildAnd(ctx->ac.builder, samples,
+ LLVMConstInt(ctx->ac.i32, 0xf, false), "");
+ samples = LLVMBuildShl(ctx->ac.builder, ctx->ac.i32_1,
samples, "");
- samples = LLVMBuildSelect(ctx->builder, is_msaa, samples,
- ctx->i32one, "");
+ samples = LLVMBuildSelect(ctx->ac.builder, is_msaa, samples,
+ ctx->ac.i32_1, "");
result = samples;
goto write_result;
}
if (offsets && instr->op != nir_texop_txf) {
LLVMValueRef offset[3], pack;
for (chan = 0; chan < 3; ++chan)
- offset[chan] = ctx->i32zero;
+ offset[chan] = ctx->ac.i32_0;
args.offset = true;
for (chan = 0; chan < get_llvm_num_components(offsets); chan++) {
offset[chan] = llvm_extract_elem(&ctx->ac, offsets, chan);
- offset[chan] = LLVMBuildAnd(ctx->builder, offset[chan],
- LLVMConstInt(ctx->i32, 0x3f, false), "");
+ offset[chan] = LLVMBuildAnd(ctx->ac.builder, offset[chan],
+ LLVMConstInt(ctx->ac.i32, 0x3f, false), "");
if (chan)
- offset[chan] = LLVMBuildShl(ctx->builder, offset[chan],
- LLVMConstInt(ctx->i32, chan * 8, false), "");
+ offset[chan] = LLVMBuildShl(ctx->ac.builder, offset[chan],
+ LLVMConstInt(ctx->ac.i32, chan * 8, false), "");
}
- pack = LLVMBuildOr(ctx->builder, offset[0], offset[1], "");
- pack = LLVMBuildOr(ctx->builder, pack, offset[2], "");
+ pack = LLVMBuildOr(ctx->ac.builder, offset[0], offset[1], "");
+ pack = LLVMBuildOr(ctx->ac.builder, pack, offset[2], "");
address[count++] = pack;
}
if (instr->sampler_dim == GLSL_SAMPLER_DIM_CUBE && coord) {
if (instr->is_array && instr->op != nir_texop_lod)
- coords[3] = apply_round_slice(ctx, coords[3]);
+ coords[3] = apply_round_slice(&ctx->ac, coords[3]);
for (chan = 0; chan < instr->coord_components; chan++)
coords[chan] = to_float(&ctx->ac, coords[chan]);
if (instr->coord_components == 3)
- coords[3] = LLVMGetUndef(ctx->f32);
+ coords[3] = LLVMGetUndef(ctx->ac.f32);
ac_prepare_cube_coords(&ctx->ac,
instr->op == nir_texop_txd, instr->is_array,
coords, derivs);
address[count++] = coords[0];
if (instr->coord_components > 1) {
if (instr->sampler_dim == GLSL_SAMPLER_DIM_1D && instr->is_array && instr->op != nir_texop_txf) {
- coords[1] = apply_round_slice(ctx, coords[1]);
+ coords[1] = apply_round_slice(&ctx->ac, coords[1]);
}
address[count++] = coords[1];
}
if (instr->sampler_dim != GLSL_SAMPLER_DIM_3D &&
instr->sampler_dim != GLSL_SAMPLER_DIM_CUBE &&
instr->op != nir_texop_txf) {
- coords[2] = apply_round_slice(ctx, coords[2]);
+ coords[2] = apply_round_slice(&ctx->ac, coords[2]);
}
address[count++] = coords[2];
}
if (lod)
address[count++] = lod;
else
- address[count++] = ctx->i32zero;
+ address[count++] = ctx->ac.i32_0;
}
for (chan = 0; chan < count; chan++) {
- address[chan] = LLVMBuildBitCast(ctx->builder,
- address[chan], ctx->i32, "");
+ address[chan] = LLVMBuildBitCast(ctx->ac.builder,
+ address[chan], ctx->ac.i32, "");
}
if (instr->op == nir_texop_samples_identical) {
memcpy(txf_address, address, sizeof(txf_address));
if (!instr->is_array)
- txf_address[2] = ctx->i32zero;
- txf_address[3] = ctx->i32zero;
+ txf_address[2] = ctx->ac.i32_0;
+ txf_address[3] = ctx->ac.i32_0;
- set_tex_fetch_args(ctx, &txf_args, instr, nir_texop_txf,
+ set_tex_fetch_args(&ctx->ac, &txf_args, instr, nir_texop_txf,
fmask_ptr, NULL,
txf_address, txf_count, 0xf);
result = build_tex_intrinsic(ctx, instr, false, &txf_args);
- result = LLVMBuildExtractElement(ctx->builder, result, ctx->i32zero, "");
- result = emit_int_cmp(&ctx->ac, LLVMIntEQ, result, ctx->i32zero);
+ result = LLVMBuildExtractElement(ctx->ac.builder, result, ctx->ac.i32_0, "");
+ result = emit_int_cmp(&ctx->ac, LLVMIntEQ, result, ctx->ac.i32_0);
goto write_result;
}
if (instr->sampler_dim == GLSL_SAMPLER_DIM_MS &&
instr->op != nir_texop_txs) {
unsigned sample_chan = instr->is_array ? 3 : 2;
- address[sample_chan] = adjust_sample_index_using_fmask(ctx,
+ address[sample_chan] = adjust_sample_index_using_fmask(&ctx->ac,
address[0],
address[1],
instr->is_array ? address[2] : NULL,
assert(const_offset);
num_offsets = MIN2(num_offsets, instr->coord_components);
if (num_offsets > 2)
- address[2] = LLVMBuildAdd(ctx->builder,
- address[2], LLVMConstInt(ctx->i32, const_offset->i32[2], false), "");
+ address[2] = LLVMBuildAdd(ctx->ac.builder,
+ address[2], LLVMConstInt(ctx->ac.i32, const_offset->i32[2], false), "");
if (num_offsets > 1)
- address[1] = LLVMBuildAdd(ctx->builder,
- address[1], LLVMConstInt(ctx->i32, const_offset->i32[1], false), "");
- address[0] = LLVMBuildAdd(ctx->builder,
- address[0], LLVMConstInt(ctx->i32, const_offset->i32[0], false), "");
+ address[1] = LLVMBuildAdd(ctx->ac.builder,
+ address[1], LLVMConstInt(ctx->ac.i32, const_offset->i32[1], false), "");
+ address[0] = LLVMBuildAdd(ctx->ac.builder,
+ address[0], LLVMConstInt(ctx->ac.i32, const_offset->i32[0], false), "");
}
else
dmask = 1 << instr->component;
}
- set_tex_fetch_args(ctx, &args, instr, instr->op,
+ set_tex_fetch_args(&ctx->ac, &args, instr, instr->op,
res_ptr, samp_ptr, address, count, dmask);
result = build_tex_intrinsic(ctx, instr, lod_is_zero, &args);
if (instr->op == nir_texop_query_levels)
- result = LLVMBuildExtractElement(ctx->builder, result, LLVMConstInt(ctx->i32, 3, false), "");
- else if (instr->is_shadow && instr->op != nir_texop_txs && instr->op != nir_texop_lod && instr->op != nir_texop_tg4)
- result = LLVMBuildExtractElement(ctx->builder, result, ctx->i32zero, "");
+ result = LLVMBuildExtractElement(ctx->ac.builder, result, LLVMConstInt(ctx->ac.i32, 3, false), "");
+ else if (instr->is_shadow && instr->is_new_style_shadow &&
+ instr->op != nir_texop_txs && instr->op != nir_texop_lod &&
+ instr->op != nir_texop_tg4)
+ result = LLVMBuildExtractElement(ctx->ac.builder, result, ctx->ac.i32_0, "");
else if (instr->op == nir_texop_txs &&
instr->sampler_dim == GLSL_SAMPLER_DIM_CUBE &&
instr->is_array) {
- LLVMValueRef two = LLVMConstInt(ctx->i32, 2, false);
- LLVMValueRef six = LLVMConstInt(ctx->i32, 6, false);
- LLVMValueRef z = LLVMBuildExtractElement(ctx->builder, result, two, "");
- z = LLVMBuildSDiv(ctx->builder, z, six, "");
- result = LLVMBuildInsertElement(ctx->builder, result, z, two, "");
+ LLVMValueRef two = LLVMConstInt(ctx->ac.i32, 2, false);
+ LLVMValueRef six = LLVMConstInt(ctx->ac.i32, 6, false);
+ LLVMValueRef z = LLVMBuildExtractElement(ctx->ac.builder, result, two, "");
+ z = LLVMBuildSDiv(ctx->ac.builder, z, six, "");
+ result = LLVMBuildInsertElement(ctx->ac.builder, result, z, two, "");
} else if (instr->dest.ssa.num_components != 4)
- result = trim_vector(ctx, result, instr->dest.ssa.num_components);
+ result = trim_vector(&ctx->ac, result, instr->dest.ssa.num_components);
write_result:
if (result) {
assert(instr->dest.is_ssa);
result = to_integer(&ctx->ac, result);
- _mesa_hash_table_insert(ctx->nir->defs, &instr->dest.ssa, result);
+ _mesa_hash_table_insert(ctx->defs, &instr->dest.ssa, result);
}
}
visit_intrinsic(ctx, nir_instr_as_intrinsic(instr));
break;
case nir_instr_type_tex:
- visit_tex(ctx->nctx, nir_instr_as_tex(instr));
+ visit_tex(ctx, nir_instr_as_tex(instr));
break;
case nir_instr_type_phi:
visit_phi(ctx, nir_instr_as_phi(instr));
memset(shader_info, 0, sizeof(*shader_info));
ac_nir_shader_info_pass(nir, options, &shader_info->info);
-
+
LLVMSetTarget(ctx.module, options->supports_spill ? "amdgcn-mesa-mesa3d" : "amdgcn--");
LLVMTargetDataRef data_layout = LLVMCreateTargetDataLayout(tm);
ctx.gs_max_out_vertices = nir->info.gs.vertices_out;
} else if (nir->stage == MESA_SHADER_TESS_EVAL) {
ctx.tes_primitive_mode = nir->info.tess.primitive_mode;
+ } else if (nir->stage == MESA_SHADER_VERTEX) {
+ if (shader_info->info.vs.needs_instance_id) {
+ ctx.shader_info->vs.vgpr_comp_cnt =
+ MAX2(3, ctx.shader_info->vs.vgpr_comp_cnt);
+ }
+ } else if (nir->stage == MESA_SHADER_FRAGMENT) {
+ shader_info->fs.can_discard = nir->info.fs.uses_discard;
}
ac_setup_rings(&ctx);
if (nir->stage == MESA_SHADER_FRAGMENT)
handle_fs_inputs_pre(&ctx, nir);
+ ctx.abi.chip_class = options->chip_class;
ctx.abi.inputs = &ctx.inputs[0];
ctx.abi.emit_outputs = handle_shader_outputs_post;
+ ctx.abi.load_sampler_desc = radv_get_sampler_desc;
nir_foreach_variable(variable, &nir->outputs)
scan_shader_output_decl(&ctx, variable);