ac/nir: set TRUNC_COORD=0 for nir_texop_tg4

Fixes black squares in Assassin's Creed: Valhalla and rendering of
FidelityFX-CACAO demo.

shader-db (sienna cichlid):
Totals:
SGPRS: 2977068 -> 2977220 (0.01 %)
VGPRS: 1929624 -> 1929616 (-0.00 %)
Spilled SGPRs: 5769 -> 5769 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Private memory VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 61423092 -> 61424672 (0.00 %) bytes
Max Waves: 895765 -> 895766 (0.00 %)

Totals from affected shaders:
SGPRS: 9520 -> 9672 (1.60 %)
VGPRS: 7464 -> 7456 (-0.11 %)
Spilled SGPRs: 0 -> 0 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Private memory VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 527432 -> 529012 (0.30 %) bytes
Max Waves: 1819 -> 1820 (0.05 %)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Fixes: 58f25098a0 ("radv: Use TRUNC_COORD on samplers")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10036>
(cherry picked from commit 6a7b89c89d)
This commit is contained in:
Rhys Perry 2021-04-05 14:26:33 +01:00 committed by Eric Engestrom
parent ef12d4ebb4
commit 70e8b11f37
2 changed files with 8 additions and 1 deletions

View file

@ -751,7 +751,7 @@
"description": "ac/nir: set TRUNC_COORD=0 for nir_texop_tg4",
"nominated": true,
"nomination_type": 1,
"resolution": 0,
"resolution": 1,
"master_sha": null,
"because_sha": "58f25098a0dc4f4976dadacdc4e7a9db42ec0c50"
},

View file

@ -4625,6 +4625,13 @@ static void visit_tex(struct ac_nir_context *ctx, nir_tex_instr *instr)
}
}
/* Set TRUNC_COORD=0 for textureGather(). */
if (instr->op == nir_texop_tg4) {
LLVMValueRef dword0 = LLVMBuildExtractElement(ctx->ac.builder, args.sampler, ctx->ac.i32_0, "");
dword0 = LLVMBuildAnd(ctx->ac.builder, dword0, LLVMConstInt(ctx->ac.i32, C_008F30_TRUNC_COORD, 0), "");
args.sampler = LLVMBuildInsertElement(ctx->ac.builder, args.sampler, dword0, ctx->ac.i32_0, "");
}
assert(instr->dest.is_ssa);
args.d16 = instr->dest.ssa.bit_size == 16;
args.tfe = instr->is_sparse;