From: Nicolai Hähnle <nicolai.haeh...@amd.com>
---
src/gallium/auxiliary/gallivm/lp_bld_tgsi.c | 2 +-
src/gallium/auxiliary/gallivm/lp_bld_tgsi.h | 1 +
src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c | 15 +++++++--------
src/gallium/drivers/radeonsi/si_shader.c | 14 ++++++++------
src/gallium/drivers/radeonsi/si_shader_internal.h | 1 +
src/gallium/drivers/radeonsi/si_shader_tgsi_setup.c | 14 +++++++++-----
6 files changed, 27 insertions(+), 20 deletions(-)
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_tgsi.c
b/src/gallium/auxiliary/gallivm/lp_bld_tgsi.c
index e450092a82c..66f752989ab 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_tgsi.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_tgsi.c
@@ -306,21 +306,21 @@ lp_build_tgsi_inst_llvm(
if (info->output_mode == TGSI_OUTPUT_REPLICATE && bld_base->soa) {
val = emit_data.output[0];
memset(emit_data.output, 0, sizeof(emit_data.output));
TGSI_FOR_EACH_DST0_ENABLED_CHANNEL(inst, chan_index) {
emit_data.output[chan_index] = val;
}
}
}
if (info->num_dst > 0 && info->opcode != TGSI_OPCODE_STORE) {
- bld_base->emit_store(bld_base, inst, info, emit_data.output);
+ bld_base->emit_store(bld_base, inst, info, 0, emit_data.output);
}
return TRUE;
}
LLVMValueRef
lp_build_emit_fetch_src(
struct lp_build_tgsi_context *bld_base,
const struct tgsi_full_src_register *reg,
enum tgsi_opcode_type stype,
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_tgsi.h
b/src/gallium/auxiliary/gallivm/lp_bld_tgsi.h
index eb632b700ab..eeeea507810 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_tgsi.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld_tgsi.h
@@ -363,20 +363,21 @@ struct lp_build_tgsi_context
LLVMValueRef, unsigned, unsigned, unsigned, unsigned);
void (*emit_debug)(struct lp_build_tgsi_context *,
const struct tgsi_full_instruction *,
const struct tgsi_opcode_info *);
void (*emit_store)(struct lp_build_tgsi_context *,
const struct tgsi_full_instruction *,
const struct tgsi_opcode_info *,
+ unsigned index,
LLVMValueRef dst[4]);
void (*emit_declaration)(struct lp_build_tgsi_context *,
const struct tgsi_full_declaration *decl);
void (*emit_immediate)(struct lp_build_tgsi_context *,
const struct tgsi_full_immediate *imm);
/* Allow the user to store data in this structure rather than passing it
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
b/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
index f16c579f38d..45110e8b9fe 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
@@ -1906,33 +1906,32 @@ emit_debug(
lp_build_print_value(gallivm, " mask = ",
bld->exec_mask.exec_mask);
}
}
}
static void
emit_store(
struct lp_build_tgsi_context * bld_base,
const struct tgsi_full_instruction * inst,
const struct tgsi_opcode_info * info,
+ unsigned index,
LLVMValueRef dst[4])
{
- unsigned chan_index;
enum tgsi_opcode_type dtype =
tgsi_opcode_infer_dst_type(inst->Instruction.Opcode);
- if(info->num_dst) {
- TGSI_FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
-
- if (tgsi_type_is_64bit(dtype) && (chan_index == 1 || chan_index == 3))
- continue;
- emit_store_chan(bld_base, inst, 0, chan_index, dst[chan_index]);
- }
+ unsigned writemask = inst->Dst[index].Register.WriteMask;
+ while (writemask) {
+ unsigned chan_index = u_bit_scan(&writemask);
+ if (tgsi_type_is_64bit(dtype) && (chan_index == 1 || chan_index == 3))
+ continue;
+ emit_store_chan(bld_base, inst, index, chan_index, dst[chan_index]);
}
}
static unsigned
tgsi_to_pipe_tex_target(unsigned tgsi_target)
{
switch (tgsi_target) {
case TGSI_TEXTURE_BUFFER:
return PIPE_BUFFER;
case TGSI_TEXTURE_1D:
diff --git a/src/gallium/drivers/radeonsi/si_shader.c
b/src/gallium/drivers/radeonsi/si_shader.c
index c4e7f225a8f..1a1a70e23ea 100644
--- a/src/gallium/drivers/radeonsi/si_shader.c
+++ b/src/gallium/drivers/radeonsi/si_shader.c
@@ -1132,39 +1132,40 @@ static LLVMValueRef fetch_input_tes(
base = LLVMGetParam(ctx->main_fn, ctx->param_tcs_offchip_offset);
addr = get_tcs_tes_buffer_address_from_reg(ctx, NULL, reg);
return buffer_load(bld_base, type, swizzle, buffer, base, addr, true);
}
static void store_output_tcs(struct lp_build_tgsi_context *bld_base,
const struct tgsi_full_instruction *inst,
const struct tgsi_opcode_info *info,
+ unsigned index,
LLVMValueRef dst[4])
{
struct si_shader_context *ctx = si_shader_context(bld_base);
struct gallivm_state *gallivm = &ctx->gallivm;
- const struct tgsi_full_dst_register *reg = &inst->Dst[0];
+ const struct tgsi_full_dst_register *reg = &inst->Dst[index];
const struct tgsi_shader_info *sh_info = &ctx->shader->selector->info;
unsigned chan_index;
LLVMValueRef dw_addr, stride;
LLVMValueRef buffer, base, buf_addr;
LLVMValueRef values[4];
bool skip_lds_store;
bool is_tess_factor = false, is_tess_inner = false;
/* Only handle per-patch and per-vertex outputs here.
* Vectors will be lowered to scalars and this function will be called
again.
*/
if (reg->Register.File != TGSI_FILE_OUTPUT ||
(dst[0] && LLVMGetTypeKind(LLVMTypeOf(dst[0])) ==
LLVMVectorTypeKind)) {
- si_llvm_emit_store(bld_base, inst, info, dst);
+ si_llvm_emit_store(bld_base, inst, info, index, dst);
return;
}
if (reg->Register.Dimension) {
stride = get_tcs_out_vertex_dw_stride(ctx);
dw_addr = get_tcs_out_current_patch_offset(ctx);
dw_addr = get_dw_address(ctx, reg, NULL, stride, dw_addr);
skip_lds_store = !sh_info->reads_pervertex_outputs;
} else {
dw_addr = get_tcs_out_current_patch_data_offset(ctx);
@@ -1184,54 +1185,55 @@ static void store_output_tcs(struct
lp_build_tgsi_context *bld_base,
is_tess_inner = name == TGSI_SEMANTIC_TESSINNER;
}
}
}
buffer = desc_from_addr_base64k(ctx,
ctx->param_tcs_offchip_addr_base64k);
base = LLVMGetParam(ctx->main_fn, ctx->param_tcs_offchip_offset);
buf_addr = get_tcs_tes_buffer_address_from_reg(ctx, reg, NULL);
-
- TGSI_FOR_EACH_DST0_ENABLED_CHANNEL(inst, chan_index) {
+ uint32_t writemask = reg->Register.WriteMask;
+ while (writemask) {
+ chan_index = u_bit_scan(&writemask);
LLVMValueRef value = dst[chan_index];
if (inst->Instruction.Saturate)
value = ac_build_clamp(&ctx->ac, value);
/* Skip LDS stores if there is no LDS read of this output. */
if (!skip_lds_store)
lds_store(bld_base, chan_index, dw_addr, value);
value = LLVMBuildBitCast(gallivm->builder, value, ctx->i32, "");
values[chan_index] = value;
- if (inst->Dst[0].Register.WriteMask != 0xF && !is_tess_factor) {
+ if (reg->Register.WriteMask != 0xF && !is_tess_factor) {
ac_build_buffer_store_dword(&ctx->ac, buffer, value, 1,
buf_addr, base,
4 * chan_index, 1, 0, true,
false);
}
/* Write tess factors into VGPRs for the epilog. */
if (is_tess_factor &&
ctx->shader->selector->tcs_info.tessfactors_are_def_in_all_invocs) {
if (!is_tess_inner) {
LLVMBuildStore(gallivm->builder, value, /*
outer */
ctx->invoc0_tess_factors[chan_index]);
} else if (chan_index < 2) {
LLVMBuildStore(gallivm->builder, value, /*
inner */
ctx->invoc0_tess_factors[4 +
chan_index]);
}
}
}
- if (inst->Dst[0].Register.WriteMask == 0xF && !is_tess_factor) {
+ if (reg->Register.WriteMask == 0xF && !is_tess_factor) {
LLVMValueRef value = lp_build_gather_values(gallivm,
values, 4);
ac_build_buffer_store_dword(&ctx->ac, buffer, value, 4,
buf_addr,
base, 0, 1, 0, true, false);
}
}
static LLVMValueRef fetch_input_gs(
struct lp_build_tgsi_context *bld_base,
const struct tgsi_full_src_register *reg,
diff --git a/src/gallium/drivers/radeonsi/si_shader_internal.h
b/src/gallium/drivers/radeonsi/si_shader_internal.h
index 023f9a6a093..141dd34be3d 100644
--- a/src/gallium/drivers/radeonsi/si_shader_internal.h
+++ b/src/gallium/drivers/radeonsi/si_shader_internal.h
@@ -282,20 +282,21 @@ LLVMValueRef si_llvm_emit_fetch_64bit(struct
lp_build_tgsi_context *bld_base,
LLVMValueRef ptr2);
LLVMValueRef si_llvm_emit_fetch(struct lp_build_tgsi_context *bld_base,
const struct tgsi_full_src_register *reg,
enum tgsi_opcode_type type,
unsigned swizzle);
void si_llvm_emit_store(struct lp_build_tgsi_context *bld_base,
const struct tgsi_full_instruction *inst,
const struct tgsi_opcode_info *info,
+ unsigned index,
LLVMValueRef dst[4]);
/* Combine these with & instead of |. */
#define NOOP_WAITCNT 0xf7f
#define LGKM_CNT 0x07f
#define VM_CNT 0xf70
void si_emit_waitcnt(struct si_shader_context *ctx, unsigned simm16);
LLVMValueRef si_get_indirect_index(struct si_shader_context *ctx,
diff --git a/src/gallium/drivers/radeonsi/si_shader_tgsi_setup.c
b/src/gallium/drivers/radeonsi/si_shader_tgsi_setup.c
index 231f16f049d..67172729bb6 100644
--- a/src/gallium/drivers/radeonsi/si_shader_tgsi_setup.c
+++ b/src/gallium/drivers/radeonsi/si_shader_tgsi_setup.c
@@ -874,48 +874,52 @@ static void emit_declaration(struct lp_build_tgsi_context
*bld_base,
break;
default:
break;
}
}
void si_llvm_emit_store(struct lp_build_tgsi_context *bld_base,
const struct tgsi_full_instruction *inst,
const struct tgsi_opcode_info *info,
+ unsigned index,
LLVMValueRef dst[4])
{
struct si_shader_context *ctx = si_shader_context(bld_base);
struct gallivm_state *gallivm = &ctx->gallivm;
- const struct tgsi_full_dst_register *reg = &inst->Dst[0];
+ const struct tgsi_full_dst_register *reg = &inst->Dst[index];
LLVMBuilderRef builder = ctx->gallivm.builder;
LLVMValueRef temp_ptr, temp_ptr2 = NULL;
- unsigned chan, chan_index;
bool is_vec_store = false;
enum tgsi_opcode_type dtype =
tgsi_opcode_infer_dst_type(inst->Instruction.Opcode);
if (dst[0]) {
LLVMTypeKind k = LLVMGetTypeKind(LLVMTypeOf(dst[0]));
is_vec_store = (k == LLVMVectorTypeKind);
}
if (is_vec_store) {
LLVMValueRef values[4] = {};
- TGSI_FOR_EACH_DST0_ENABLED_CHANNEL(inst, chan) {
+ uint32_t writemask = reg->Register.WriteMask;
+ while (writemask) {
+ unsigned chan = u_bit_scan(&writemask);
LLVMValueRef index = LLVMConstInt(ctx->i32, chan, 0);
values[chan] =
LLVMBuildExtractElement(gallivm->builder,
dst[0], index, "");
}
- bld_base->emit_store(bld_base, inst, info, values);
+ bld_base->emit_store(bld_base, inst, info, index, values);
return;
}
- TGSI_FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
+ uint32_t writemask = reg->Register.WriteMask;
+ while (writemask) {
+ unsigned chan_index = u_bit_scan(&writemask);
LLVMValueRef value = dst[chan_index];
if (tgsi_type_is_64bit(dtype) && (chan_index == 1 || chan_index
== 3))
continue;
if (inst->Instruction.Saturate)
value = ac_build_clamp(&ctx->ac, value);
if (reg->Register.File == TGSI_FILE_ADDRESS) {
temp_ptr = ctx->addrs[reg->Register.Index][chan_index];
LLVMBuildStore(builder, value, temp_ptr);
--
2.11.0
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
