AlexIndustrial/mesa
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

vc4: Pack VPM attr contents according to just the size of the attribute.

Browse Source 
total instructions in shared programs: 40960 -> 39753 (-2.95%)
instructions in affected programs:     20871 -> 19664 (-5.78%)
This commit is contained in:
Eric Anholt
2015-01-09 12:56:34 -08:00
parent 72cb6619cb
commit 92a0b0bd70
3 changed files with 9 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/gallium/drivers/vc4/vc4_draw.c
+5 -2
View File
@@ -197,6 +197,7 @@ vc4_draw_vbo(struct pipe_context *pctx, const struct pipe_draw_info *info)
cl_u32(&vc4->shader_rec, 0); /* UBO offset written by kernel */
uint32_t max_index = 0xffff;
uint32_t vpm_offset = 0;
for (int i = 0; i < vtx->num_elements; i++) {
struct pipe_vertex_element *elem = &vtx->pipe[i];
struct pipe_vertex_buffer *vb =
@@ -210,8 +211,10 @@ vc4_draw_vbo(struct pipe_context *pctx, const struct pipe_draw_info *info)
cl_reloc(vc4, &vc4->shader_rec, rsc->bo, offset);
cl_u8(&vc4->shader_rec, elem_size - 1);
cl_u8(&vc4->shader_rec, vb->stride);
cl_u8(&vc4->shader_rec, i * 16); /* VS VPM offset */
cl_u8(&vc4->shader_rec, i * 16); /* CS VPM offset */
cl_u8(&vc4->shader_rec, vpm_offset); /* VS VPM offset */
cl_u8(&vc4->shader_rec, vpm_offset); /* CS VPM offset */
vpm_offset += align(elem_size, 4) / 4;
if (vb->stride > 0) {
max_index = MIN2(max_index,
src/gallium/drivers/vc4/vc4_program.c
+3 -9
View File
@@ -1076,17 +1076,11 @@ static void
emit_vertex_input(struct vc4_compile *c, int attr)
{
enum pipe_format format = c->vs_key->attr_formats[attr];
uint32_t attr_size = util_format_get_blocksize(format);
struct qreg vpm_reads[4];
/* Right now, we're setting the VPM offsets to be 16 bytes wide every
* time, so we always read 4 32-bit VPM entries.
*/
for (int i = 0; i < 4; i++) {
vpm_reads[i] = qir_get_temp(c);
qir_emit(c, qir_inst(QOP_VPM_READ,
vpm_reads[i],
c->undef,
c->undef));
for (int i = 0; i < align(attr_size, 4) / 4; i++) {
vpm_reads[i] = qir_VPM_READ(c);
c->num_inputs++;
}
src/gallium/drivers/vc4/vc4_qir.h
+1
View File
@@ -495,6 +495,7 @@ QIR_ALU0(FRAG_W)
QIR_ALU0(FRAG_REV_FLAG)
QIR_ALU0(TEX_RESULT)
QIR_ALU0(TLB_COLOR_READ)
QIR_ALU0(VPM_READ)
QIR_NODST_1(TLB_Z_WRITE)
QIR_NODST_1(TLB_DISCARD_SETUP)
QIR_NODST_1(TLB_STENCIL_SETUP)