Commit 1583f49e authored by Ian Romanick's avatar Ian Romanick

i965/vec4: Allow CSE on subset VF constant loads

v2: Rewrite the code that generates the VF mask.  Suggested by Ken.

No changes on other platforms.

Haswell, Ivy Bridge, and Sandy Bridge had similar results. (Haswell shown)
total instructions in shared programs: 13059891 -> 13059884 (<.01%)
instructions in affected programs: 431 -> 424 (-1.62%)
helped: 7
HURT: 0
helped stats (abs) min: 1 max: 1 x̄: 1.00 x̃: 1
helped stats (rel) min: 1.19% max: 5.26% x̄: 2.05% x̃: 1.49%
95% mean confidence interval for instructions value: -1.00 -1.00
95% mean confidence interval for instructions %-change: -3.39% -0.71%
Instructions are helped.

total cycles in shared programs: 409260032 -> 409260018 (<.01%)
cycles in affected programs: 4228 -> 4214 (-0.33%)
helped: 7
HURT: 0
helped stats (abs) min: 2 max: 2 x̄: 2.00 x̃: 2
helped stats (rel) min: 0.28% max: 2.04% x̄: 0.54% x̃: 0.28%
95% mean confidence interval for cycles value: -2.00 -2.00
95% mean confidence interval for cycles %-change: -1.15% 0.07%

Inconclusive result (%-change mean confidence interval includes 0).
Signed-off-by: default avatarIan Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke's avatarKenneth Graunke <kenneth@whitecape.org>
parent 360899d4
......@@ -104,6 +104,25 @@ operands_match(const vec4_instruction *a, const vec4_instruction *b)
return xs[0].equals(ys[0]) &&
((xs[1].equals(ys[1]) && xs[2].equals(ys[2])) ||
(xs[2].equals(ys[1]) && xs[1].equals(ys[2])));
} else if (a->opcode == BRW_OPCODE_MOV &&
xs[0].file == IMM &&
xs[0].type == BRW_REGISTER_TYPE_VF) {
src_reg tmp_x = xs[0];
src_reg tmp_y = ys[0];
/* Smash out the values that are not part of the writemask. Otherwise
* the equals operator will fail due to mismatches in unused components.
*/
const unsigned ab_writemask = a->dst.writemask & b->dst.writemask;
const uint32_t mask = ((ab_writemask & WRITEMASK_X) ? 0x000000ff : 0) |
((ab_writemask & WRITEMASK_Y) ? 0x0000ff00 : 0) |
((ab_writemask & WRITEMASK_Z) ? 0x00ff0000 : 0) |
((ab_writemask & WRITEMASK_W) ? 0xff000000 : 0);
tmp_x.ud &= mask;
tmp_y.ud &= mask;
return tmp_x.equals(tmp_y);
} else if (!a->is_commutative()) {
return xs[0].equals(ys[0]) && xs[1].equals(ys[1]) && xs[2].equals(ys[2]);
} else {
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment