fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-07 07:08:04 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 4

Revert "i965/fs: Use a stride of 1 and byte offsets for UBOs"

Browse source 
This reverts commit 0ae22b3ebd.

See the previous reverts.
This commit is contained in:
Emil Velikov 2015-12-19 00:18:54 +00:00
parent 0452dcd92d
commit eff2eea145
3 changed files with 16 additions and 13 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

16
src/mesa/drivers/dri/i965/brw_fs.cpp
View file

@ -185,7 +185,7 @@ fs_visitor::VARYING_PULL_CONSTANT_LOAD(const fs_builder &bld,
* the redundant ones. * the redundant ones.
*/ */
fs_reg vec4_offset = vgrf(glsl_type::int_type); fs_reg vec4_offset = vgrf(glsl_type::int_type);
bld.ADD(vec4_offset, varying_offset, fs_reg(const_offset & ~0xf)); bld.ADD(vec4_offset, varying_offset, fs_reg(const_offset & ~3));
int scale = 1; int scale = 1;
if (devinfo->gen == 4 && bld.dispatch_width() == 8) { if (devinfo->gen == 4 && bld.dispatch_width() == 8) {
@ -217,7 +217,7 @@ fs_visitor::VARYING_PULL_CONSTANT_LOAD(const fs_builder &bld,
inst->mlen = 1 + bld.dispatch_width() / 8; inst->mlen = 1 + bld.dispatch_width() / 8;
} }
bld.MOV(dst, offset(vec4_result, bld, ((const_offset & 0xf) / 4) * scale)); bld.MOV(dst, offset(vec4_result, bld, (const_offset & 3) * scale));
} }
/** /**
@ -1939,12 +1939,10 @@ fs_visitor::demote_pull_constants()
/* Generate a pull load into dst. */ /* Generate a pull load into dst. */
if (inst->src[i].reladdr) { if (inst->src[i].reladdr) {
fs_reg indirect = ibld.vgrf(BRW_REGISTER_TYPE_D);
ibld.MUL(indirect, *inst->src[i].reladdr, brw_imm_d(4));
VARYING_PULL_CONSTANT_LOAD(ibld, dst, VARYING_PULL_CONSTANT_LOAD(ibld, dst,
surf_index, surf_index,
indirect, *inst->src[i].reladdr,
pull_index * 4); pull_index);
inst->src[i].reladdr = NULL; inst->src[i].reladdr = NULL;
inst->src[i].stride = 1; inst->src[i].stride = 1;
} else { } else {
@ -2980,11 +2978,13 @@ fs_visitor::lower_uniform_pull_constant_loads()
continue; continue;
if (devinfo->gen >= 7) { if (devinfo->gen >= 7) {
/* The offset arg is a vec4-aligned immediate byte offset. */ /* The offset arg before was a vec4-aligned byte offset. We need to
* turn it into a dword offset.
*/
fs_reg const_offset_reg = inst->src[1]; fs_reg const_offset_reg = inst->src[1];
assert(const_offset_reg.file == IMM && assert(const_offset_reg.file == IMM &&
const_offset_reg.type == BRW_REGISTER_TYPE_UD); const_offset_reg.type == BRW_REGISTER_TYPE_UD);
assert(const_offset_reg.fixed_hw_reg.dw1.ud % 16 == 0); const_offset_reg.fixed_hw_reg.dw1.ud /= 4;
fs_reg payload, offset; fs_reg payload, offset;
if (devinfo->gen >= 9) { if (devinfo->gen >= 9) {

11
src/mesa/drivers/dri/i965/brw_fs_nir.cpp
View file

@ -1588,13 +1588,16 @@ fs_visitor::nir_emit_intrinsic(const fs_builder &bld, nir_intrinsic_instr *instr
} }
if (has_indirect) { if (has_indirect) {
fs_reg base_offset = retype(get_nir_src(instr->src[1]), /* Turn the byte offset into a dword offset. */
BRW_REGISTER_TYPE_D); fs_reg base_offset = vgrf(glsl_type::int_type);
bld.SHR(base_offset, retype(get_nir_src(instr->src[1]),
BRW_REGISTER_TYPE_D),
fs_reg(2));
unsigned vec4_offset = instr->const_index[0]; unsigned vec4_offset = instr->const_index[0] / 4;
for (int i = 0; i < instr->num_components; i++) for (int i = 0; i < instr->num_components; i++)
VARYING_PULL_CONSTANT_LOAD(bld, offset(dest, bld, i), surf_index, VARYING_PULL_CONSTANT_LOAD(bld, offset(dest, bld, i), surf_index,
base_offset, vec4_offset + i * 4); base_offset, vec4_offset + i);
} else { } else {
fs_reg packed_consts = vgrf(glsl_type::float_type); fs_reg packed_consts = vgrf(glsl_type::float_type);
packed_consts.type = dest.type; packed_consts.type = dest.type;

2
src/mesa/drivers/dri/i965/brw_wm_surface_state.c
View file

@ -403,7 +403,7 @@ brw_create_constant_surface(struct brw_context *brw,
uint32_t *out_offset, uint32_t *out_offset,
bool dword_pitch) bool dword_pitch)
{ {
uint32_t stride = dword_pitch ? 1 : 16; uint32_t stride = dword_pitch ? 4 : 16;
uint32_t elements = ALIGN(size, stride) / stride; uint32_t elements = ALIGN(size, stride) / stride;
brw->vtbl.emit_buffer_surface_state(brw, out_offset, bo, offset, brw->vtbl.emit_buffer_surface_state(brw, out_offset, bo, offset,