From: Marek Olšák <marek.ol...@amd.com>
I'll need indexed loads without the meta data flag for tessellation later.
Also rename load_const to buffer_load_const to distinguish it from indexed
const loads.
---
src/gallium/drivers/radeonsi/si_shader.c | 68 +++++++++++++++++---------------
1 file changed, 36 insertions(+), 32 deletions(-)
diff --git a/src/gallium/drivers/radeonsi/si_shader.c
b/src/gallium/drivers/radeonsi/si_shader.c
index cd78f66..0373f34 100644
--- a/src/gallium/drivers/radeonsi/si_shader.c
+++ b/src/gallium/drivers/radeonsi/si_shader.c
@@ -181,6 +181,20 @@ static int get_param_index(unsigned semantic_name,
unsigned index,
return -1;
}
+static LLVMValueRef build_indexed_load(struct si_shader_context *si_shader_ctx,
+ LLVMValueRef base_ptr, LLVMValueRef
index)
+{
+ struct lp_build_tgsi_context *bld_base =
&si_shader_ctx->radeon_bld.soa.bld_base;
+ struct gallivm_state *gallivm = bld_base->base.gallivm;
+ LLVMValueRef indices[2], pointer;
+
+ indices[0] = bld_base->uint_bld.zero;
+ indices[1] = index;
+
+ pointer = LLVMBuildGEP(gallivm->builder, base_ptr, indices, 2, "");
+ return LLVMBuildLoad(gallivm->builder, pointer, "");
+}
+
/**
* Build an LLVM bytecode indexed load using LLVMBuildGEP + LLVMBuildLoad
*
@@ -193,21 +207,11 @@ static int get_param_index(unsigned semantic_name,
unsigned index,
* into a dword offset automatically.
*
*/
-static LLVMValueRef build_indexed_load(
+static LLVMValueRef build_indexed_load_const(
struct si_shader_context * si_shader_ctx,
- LLVMValueRef base_ptr,
- LLVMValueRef offset)
+ LLVMValueRef base_ptr, LLVMValueRef index)
{
- struct lp_build_context * base =
&si_shader_ctx->radeon_bld.soa.bld_base.base;
-
- LLVMValueRef indices[2] = {
- LLVMConstInt(LLVMInt64TypeInContext(base->gallivm->context), 0,
false),
- offset
- };
- LLVMValueRef computed_ptr = LLVMBuildGEP(
- base->gallivm->builder, base_ptr, indices, 2, "");
-
- LLVMValueRef result = LLVMBuildLoad(base->gallivm->builder,
computed_ptr, "");
+ LLVMValueRef result = build_indexed_load(si_shader_ctx, base_ptr,
index);
LLVMSetMetadata(result, 1, si_shader_ctx->const_md);
return result;
}
@@ -259,7 +263,7 @@ static void declare_input_vs(
t_offset = lp_build_const_int32(gallivm, input_index);
- t_list = build_indexed_load(si_shader_ctx, t_list_ptr, t_offset);
+ t_list = build_indexed_load_const(si_shader_ctx, t_list_ptr, t_offset);
/* Build the attribute offset */
attribute_offset = lp_build_const_int32(gallivm, 0);
@@ -354,7 +358,7 @@ static LLVMValueRef fetch_input_gs(
/* Load the ESGS ring resource descriptor */
t_list_ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn,
SI_PARAM_RW_BUFFERS);
- t_list = build_indexed_load(si_shader_ctx, t_list_ptr,
+ t_list = build_indexed_load_const(si_shader_ctx, t_list_ptr,
lp_build_const_int32(gallivm,
SI_RING_ESGS));
args[0] = t_list;
@@ -562,8 +566,8 @@ static LLVMValueRef get_sample_id(struct
radeon_llvm_context *radeon_bld)
return value;
}
-static LLVMValueRef load_const(LLVMBuilderRef builder, LLVMValueRef resource,
- LLVMValueRef offset, LLVMTypeRef return_type)
+static LLVMValueRef buffer_load_const(LLVMBuilderRef builder, LLVMValueRef
resource,
+ LLVMValueRef offset, LLVMTypeRef
return_type)
{
LLVMValueRef args[2] = {resource, offset};
@@ -602,15 +606,15 @@ static void declare_system_value(
LLVMBuilderRef builder = gallivm->builder;
LLVMValueRef desc =
LLVMGetParam(si_shader_ctx->radeon_bld.main_fn, SI_PARAM_CONST);
LLVMValueRef buf_index = lp_build_const_int32(gallivm,
SI_DRIVER_STATE_CONST_BUF);
- LLVMValueRef resource = build_indexed_load(si_shader_ctx, desc,
buf_index);
+ LLVMValueRef resource = build_indexed_load_const(si_shader_ctx,
desc, buf_index);
/* offset = sample_id * 8 (8 = 2 floats containing
samplepos.xy) */
LLVMValueRef offset0 = lp_build_mul_imm(uint_bld,
get_sample_id(radeon_bld), 8);
LLVMValueRef offset1 = LLVMBuildAdd(builder, offset0,
lp_build_const_int32(gallivm, 4), "");
LLVMValueRef pos[4] = {
- load_const(builder, resource, offset0,
radeon_bld->soa.bld_base.base.elem_type),
- load_const(builder, resource, offset1,
radeon_bld->soa.bld_base.base.elem_type),
+ buffer_load_const(builder, resource, offset0,
radeon_bld->soa.bld_base.base.elem_type),
+ buffer_load_const(builder, resource, offset1,
radeon_bld->soa.bld_base.base.elem_type),
lp_build_const_float(gallivm, 0),
lp_build_const_float(gallivm, 0)
};
@@ -661,7 +665,7 @@ static LLVMValueRef fetch_constant(
addr = lp_build_add(&bld_base->uint_bld, addr,
lp_build_const_int32(base->gallivm, idx * 4));
- result = load_const(base->gallivm->builder,
si_shader_ctx->const_resource[buf],
+ result = buffer_load_const(base->gallivm->builder,
si_shader_ctx->const_resource[buf],
addr, base->elem_type);
return bitcast(bld_base, type, result);
@@ -812,7 +816,7 @@ static void si_llvm_emit_clipvertex(struct
lp_build_tgsi_context * bld_base,
LLVMValueRef base_elt;
LLVMValueRef ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn,
SI_PARAM_CONST);
LLVMValueRef constbuf_index = lp_build_const_int32(base->gallivm,
SI_DRIVER_STATE_CONST_BUF);
- LLVMValueRef const_resource = build_indexed_load(si_shader_ctx, ptr,
constbuf_index);
+ LLVMValueRef const_resource = build_indexed_load_const(si_shader_ctx,
ptr, constbuf_index);
for (reg_index = 0; reg_index < 2; reg_index ++) {
LLVMValueRef *args = pos[2 + reg_index];
@@ -830,7 +834,7 @@ static void si_llvm_emit_clipvertex(struct
lp_build_tgsi_context * bld_base,
args[1] = lp_build_const_int32(base->gallivm,
((reg_index * 4
+ chan) * 4 +
const_chan) *
4);
- base_elt = load_const(base->gallivm->builder,
const_resource,
+ base_elt =
buffer_load_const(base->gallivm->builder, const_resource,
args[1], base->elem_type);
args[5 + chan] =
lp_build_add(base, args[5 + chan],
@@ -1242,7 +1246,7 @@ static void si_llvm_emit_es_epilogue(struct
lp_build_tgsi_context * bld_base)
/* Load the ESGS ring resource descriptor */
t_list_ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn,
SI_PARAM_RW_BUFFERS);
- t_list = build_indexed_load(si_shader_ctx, t_list_ptr,
+ t_list = build_indexed_load_const(si_shader_ctx, t_list_ptr,
lp_build_const_int32(gallivm,
SI_RING_ESGS));
for (i = 0; i < info->num_outputs; i++) {
@@ -2153,7 +2157,7 @@ static void si_llvm_emit_vertex(
/* Load the GSVS ring resource descriptor */
t_list_ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn,
SI_PARAM_RW_BUFFERS);
- t_list = build_indexed_load(si_shader_ctx, t_list_ptr,
+ t_list = build_indexed_load_const(si_shader_ctx, t_list_ptr,
lp_build_const_int32(gallivm,
SI_RING_GSVS));
/* Write vertex attribute values to GSVS ring */
@@ -2408,12 +2412,12 @@ static void preload_constants(struct si_shader_context
*si_shader_ctx)
/* Load the resource descriptor */
si_shader_ctx->const_resource[buf] =
- build_indexed_load(si_shader_ctx, ptr,
lp_build_const_int32(gallivm, buf));
+ build_indexed_load_const(si_shader_ctx, ptr,
lp_build_const_int32(gallivm, buf));
/* Load the constants, we rely on the code sinking to do the
rest */
for (i = 0; i < num_const * 4; ++i) {
si_shader_ctx->constants[buf][i] =
- load_const(gallivm->builder,
+ buffer_load_const(gallivm->builder,
si_shader_ctx->const_resource[buf],
lp_build_const_int32(gallivm, i * 4),
bld_base->base.elem_type);
@@ -2442,17 +2446,17 @@ static void preload_samplers(struct si_shader_context
*si_shader_ctx)
for (i = 0; i < num_samplers; ++i) {
/* Resource */
offset = lp_build_const_int32(gallivm, i);
- si_shader_ctx->resources[i] = build_indexed_load(si_shader_ctx,
res_ptr, offset);
+ si_shader_ctx->resources[i] =
build_indexed_load_const(si_shader_ctx, res_ptr, offset);
/* Sampler */
offset = lp_build_const_int32(gallivm, i);
- si_shader_ctx->samplers[i] = build_indexed_load(si_shader_ctx,
samp_ptr, offset);
+ si_shader_ctx->samplers[i] =
build_indexed_load_const(si_shader_ctx, samp_ptr, offset);
/* FMASK resource */
if (info->is_msaa_sampler[i]) {
offset = lp_build_const_int32(gallivm,
SI_FMASK_TEX_OFFSET + i);
si_shader_ctx->resources[SI_FMASK_TEX_OFFSET + i] =
- build_indexed_load(si_shader_ctx, res_ptr,
offset);
+ build_indexed_load_const(si_shader_ctx,
res_ptr, offset);
}
}
}
@@ -2477,7 +2481,7 @@ static void preload_streamout_buffers(struct
si_shader_context *si_shader_ctx)
LLVMValueRef offset = lp_build_const_int32(gallivm,
SI_SO_BUF_OFFSET + i);
- si_shader_ctx->so_buffers[i] =
build_indexed_load(si_shader_ctx, buf_ptr, offset);
+ si_shader_ctx->so_buffers[i] =
build_indexed_load_const(si_shader_ctx, buf_ptr, offset);
}
}
}
@@ -2599,7 +2603,7 @@ static int si_generate_gs_copy_shader(struct si_screen
*sscreen,
/* Load the GSVS ring resource descriptor */
t_list_ptr = LLVMGetParam(si_shader_ctx->radeon_bld.main_fn,
SI_PARAM_RW_BUFFERS);
- t_list = build_indexed_load(si_shader_ctx, t_list_ptr,
+ t_list = build_indexed_load_const(si_shader_ctx, t_list_ptr,
lp_build_const_int32(gallivm,
SI_RING_GSVS));
args[0] = t_list;
--
1.9.1
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
