[Mesa-dev] [PATCH 1/2] r600: correct texture offset for array index lookup
Gert Wollny
gert.wollny at collabora.com
Fri Jun 22 09:37:39 UTC 2018
For texture array lookup the slice index is evaluated according to
idx = floor(z + 0.5)
This patch implements the first part by adding 0.5 to the according
texture coordinate when appropriate.
Signed-off-by: Gert Wollny <gert.wollny at collabora.com>
---
src/gallium/drivers/r600/r600_shader.c | 64 +++++++++++++++++++++++++++++++++-
1 file changed, 63 insertions(+), 1 deletion(-)
diff --git a/src/gallium/drivers/r600/r600_shader.c b/src/gallium/drivers/r600/r600_shader.c
index c466a48262..a0c5e3fe39 100644
--- a/src/gallium/drivers/r600/r600_shader.c
+++ b/src/gallium/drivers/r600/r600_shader.c
@@ -7456,6 +7456,8 @@ static int tgsi_tex(struct r600_shader_ctx *ctx)
int8_t offset_x = 0, offset_y = 0, offset_z = 0;
boolean has_txq_cube_array_z = false;
unsigned sampler_index_mode;
+ int *array_index_offset = NULL;
+ int array_index_offset_channel = 2; /* default channel for array offsets */
if (inst->Instruction.Opcode == TGSI_OPCODE_TXQ &&
((inst->Texture.Texture == TGSI_TEXTURE_CUBE_ARRAY ||
@@ -8411,18 +8413,78 @@ static int tgsi_tex(struct r600_shader_ctx *ctx)
opcode == FETCH_OP_SAMPLE_C_LB) {
/* the array index is read from Y */
tex.coord_type_y = 0;
+ array_index_offset = &tex.offset_y;
+ array_index_offset_channel = 1;
} else {
/* the array index is read from Z */
tex.coord_type_z = 0;
tex.src_sel_z = tex.src_sel_y;
+ array_index_offset = &tex.offset_z;
}
} else if (inst->Texture.Texture == TGSI_TEXTURE_2D_ARRAY ||
inst->Texture.Texture == TGSI_TEXTURE_SHADOW2D_ARRAY ||
((inst->Texture.Texture == TGSI_TEXTURE_CUBE_ARRAY ||
inst->Texture.Texture == TGSI_TEXTURE_SHADOWCUBE_ARRAY) &&
- (ctx->bc->chip_class >= EVERGREEN)))
+ (ctx->bc->chip_class >= EVERGREEN))) {
/* the array index is read from Z */
tex.coord_type_z = 0;
+ array_index_offset = &tex.offset_z;
+ }
+
+ /* We have array access and the the coordinates are not int */
+ if (array_index_offset && opcode != FETCH_OP_LD) {
+ /* The z-offset range is not yet exhausted and we actually use the
+ * offset registers. The offsets are stored as 5-bit S3.1 floating point
+ * values with a range [0,31) = [-8, 7.5].
+ */
+ if (*array_index_offset < 31 &&
+ opcode != FETCH_OP_GATHER4_C_O &&
+ opcode != FETCH_OP_GATHER4_O) {
+ *array_index_offset += 1;
+ } else {
+ /* offset registers are ignord or already at the limit, add the offset
+ * to the lookup coordinate.
+ */
+ if (!tex.src_rel) {
+ memset(&alu, 0, sizeof(struct r600_bytecode_alu));
+ alu.op =ALU_OP2_ADD;
+ alu.src[0].sel = tex.src_gpr;
+ alu.src[0].chan = array_index_offset_channel;
+ alu.src[1].sel = V_SQ_ALU_SRC_0_5;
+ alu.dst.sel = tex.src_gpr;
+ alu.dst.chan = array_index_offset_channel;
+ alu.dst.write = 1;
+ alu.last = 1;
+ r = r600_bytecode_add_alu(ctx->bc, &alu);
+ if (r)
+ return r;
+ } else {
+ /* We can not read and write to a relative register in the same instruction, so
+ * create a temp and move everything over.
+ */
+ int ofs_tmp = r600_get_temp(ctx);
+ for (int i = 0; i < 4; ++i) {
+ memset(&alu, 0, sizeof(struct r600_bytecode_alu));
+ if (i == array_index_offset_channel) {
+ alu.op = ALU_OP2_ADD;
+ alu.src[1].sel = V_SQ_ALU_SRC_0_5;
+ } else
+ alu.op = ALU_OP1_MOV;
+ alu.src[0].sel = tex.src_gpr;
+ alu.src[0].rel = tex.src_rel;
+ alu.src[0].chan = i;
+ alu.dst.sel = ofs_tmp;
+ alu.dst.chan = i;
+ alu.dst.write = 1;
+ alu.last = (i == 3);
+ r = r600_bytecode_add_alu(ctx->bc, &alu);
+ if (r)
+ return r;
+ }
+ tex.src_gpr = ofs_tmp;
+ }
+ }
+ }
/* mask unused source components */
if (opcode == FETCH_OP_SAMPLE || opcode == FETCH_OP_GATHER4) {
--
2.16.4
More information about the mesa-dev
mailing list