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

aco/gfx11: apply clamp/omod to vinterp

Browse Source 
Totals from 2504 (3.27% of 76572) affected shaders:
MaxWaves: 74098 -> 74106 (+0.01%)
Instrs: 1829278 -> 1823427 (-0.32%); split: -0.32%, +0.00%
CodeSize: 9775908 -> 9759308 (-0.17%); split: -0.18%, +0.01%
Latency: 13494107 -> 13485390 (-0.06%); split: -0.10%, +0.04%
InvThroughput: 2052428 -> 2048724 (-0.18%); split: -0.18%, +0.00%
VClause: 26637 -> 26640 (+0.01%); split: -0.04%, +0.05%
SClause: 62027 -> 61988 (-0.06%); split: -0.14%, +0.08%
Copies: 73776 -> 73815 (+0.05%); split: -0.07%, +0.12%
PreVGPRs: 84403 -> 84397 (-0.01%)

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25220>
This commit is contained in:
Georg Lehmann
2023-09-14 13:01:09 +02:00
committed by Marge Bot
parent 7196b5318d
commit 5e9fad48bf
3 changed files with 61 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/amd/compiler/tests/test_optimizer.cpp
+35
View File
@@ -2192,3 +2192,38 @@ BEGIN_TEST(optimize.neg_mul_opsel)
finish_opt_test();
END_TEST
BEGIN_TEST(optimize.vinterp_inreg_output_modifiers)
//>> v1: %a, v1: %b, v1: %c = p_startpgm
if (!setup_cs("v1 v1 v1", GFX11))
return;
//! v1: %res0 = v_interp_p2_f32_inreg %a, %b, %c clamp
//! p_unit_test 0, %res0
Temp tmp = bld.vinterp_inreg(aco_opcode::v_interp_p2_f32_inreg, bld.def(v1), inputs[0],
inputs[1], inputs[2]);
writeout(0, fsat(tmp));
//! v1: %res1 = v_fma_f32 %b, %a, %c *2 quad_perm:[2,2,2,2] fi
//! p_unit_test 1, %res1
tmp = bld.vinterp_inreg(aco_opcode::v_interp_p2_f32_inreg, bld.def(v1), inputs[1], inputs[0],
inputs[2]);
tmp = bld.vop2(aco_opcode::v_mul_f32, bld.def(v1), Operand::c32(0x40000000u), tmp);
writeout(1, tmp);
//! v2b: %res2 = v_interp_p2_f16_f32_inreg %a, %b, %c clamp
//! p_unit_test 2, %res2
tmp = bld.vinterp_inreg(aco_opcode::v_interp_p2_f16_f32_inreg, bld.def(v2b), inputs[0],
inputs[1], inputs[2]);
writeout(2, fsat(tmp));
//! v2b: %tmp3 = v_interp_p2_f16_f32_inreg %b, %a, %c
//! v2b: %res3 = v_mul_f16 2.0, %tmp3
//! p_unit_test 3, %res3
tmp = bld.vinterp_inreg(aco_opcode::v_interp_p2_f16_f32_inreg, bld.def(v2b), inputs[1],
inputs[0], inputs[2]);
tmp = bld.vop2(aco_opcode::v_mul_f16, bld.def(v2b), Operand::c16(0x4000u), tmp);
writeout(3, tmp);
finish_opt_test();
END_TEST