diff options
author | Matt Turner <mattst88@gmail.com> | 2016-02-01 11:14:01 -0800 |
---|---|---|
committer | Matt Turner <mattst88@gmail.com> | 2016-02-04 09:30:58 -0800 |
commit | 8c7a42b3e880804e201ee9582a0a73763cdd2a01 (patch) | |
tree | bb4a2b5ce8a598df28d9ad84123e3b7ab2d8d771 /src/mesa/drivers/dri/i965/brw_fs_combine_constants.cpp | |
parent | 8ec24678ac0f459af4acb4b40a70d7802906e41a (diff) | |
download | external_mesa3d-8c7a42b3e880804e201ee9582a0a73763cdd2a01.zip external_mesa3d-8c7a42b3e880804e201ee9582a0a73763cdd2a01.tar.gz external_mesa3d-8c7a42b3e880804e201ee9582a0a73763cdd2a01.tar.bz2 |
i965/fs: Allocate single register at a time for constants.
No instruction counts changed, but:
total cycles in shared programs: 64834502 -> 64781530 (-0.08%)
cycles in affected programs: 16331544 -> 16278572 (-0.32%)
helped: 4757
HURT: 4288
GAINED: 66
LOST: 20
I remember trying this when I first wrote the pass, but it wasn't
helpful at the time.
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Diffstat (limited to 'src/mesa/drivers/dri/i965/brw_fs_combine_constants.cpp')
-rw-r--r-- | src/mesa/drivers/dri/i965/brw_fs_combine_constants.cpp | 6 |
1 files changed, 3 insertions, 3 deletions
diff --git a/src/mesa/drivers/dri/i965/brw_fs_combine_constants.cpp b/src/mesa/drivers/dri/i965/brw_fs_combine_constants.cpp index 994c699..d7a1456 100644 --- a/src/mesa/drivers/dri/i965/brw_fs_combine_constants.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs_combine_constants.cpp @@ -268,7 +268,7 @@ fs_visitor::opt_combine_constants() qsort(table.imm, table.len, sizeof(struct imm), compare); /* Insert MOVs to load the constant values into GRFs. */ - fs_reg reg(VGRF, alloc.allocate(dispatch_width / 8)); + fs_reg reg(VGRF, alloc.allocate(1)); reg.stride = 0; for (int i = 0; i < table.len; i++) { struct imm *imm = &table.imm[i]; @@ -284,8 +284,8 @@ fs_visitor::opt_combine_constants() imm->subreg_offset = reg.subreg_offset; reg.subreg_offset += sizeof(float); - if ((unsigned)reg.subreg_offset == dispatch_width * sizeof(float)) { - reg.nr = alloc.allocate(dispatch_width / 8); + if ((unsigned)reg.subreg_offset == 8 * sizeof(float)) { + reg.nr = alloc.allocate(1); reg.subreg_offset = 0; } } |