diff options
author | Rob Clark <[email protected]> | 2015-04-16 15:35:50 -0400 |
---|---|---|
committer | Rob Clark <[email protected]> | 2015-04-17 11:40:14 -0400 |
commit | 95e68adcd9f2589ae6d998328c72b84ffc49edc7 (patch) | |
tree | 6f22fc178e205c65a631abd4fc1e8a751a192466 /src/gallium/drivers/freedreno/ir3 | |
parent | efbf14e8936384ab1d243afbe3fa9bb0f40e3898 (diff) |
freedreno/ir3/nir: few little fixes
isaml needs to scale up coords based on LoD. Also fix bogus bary.f
varying # when there are non-bary frag shader inputs. And use sub.s of
a positive immediate rather than add.s of negative (since CP is better
about figuring out that those can be collapsed into the cat2 instr).
Signed-off-by: Rob Clark <[email protected]>
Diffstat (limited to 'src/gallium/drivers/freedreno/ir3')
-rw-r--r-- | src/gallium/drivers/freedreno/ir3/ir3_compiler_nir.c | 49 |
1 files changed, 28 insertions, 21 deletions
diff --git a/src/gallium/drivers/freedreno/ir3/ir3_compiler_nir.c b/src/gallium/drivers/freedreno/ir3/ir3_compiler_nir.c index 0b42cd70381..e5f6c2a5e83 100644 --- a/src/gallium/drivers/freedreno/ir3/ir3_compiler_nir.c +++ b/src/gallium/drivers/freedreno/ir3/ir3_compiler_nir.c @@ -566,13 +566,13 @@ create_frag_coord(struct ir3_compile *ctx, unsigned comp) * to subtract (integer) 8 and divide by 16 (right- * shift by 4) then convert to float: * - * add.s tmp, src, -8 + * sub.s tmp, src, 8 * shr.b tmp, tmp, 4 * mov.u32f32 dst, tmp * */ - instr = ir3_ADD_S(block, ctx->frag_coord[comp], 0, - create_immed(block, -8), 0); + instr = ir3_SUB_S(block, ctx->frag_coord[comp], 0, + create_immed(block, 8), 0); instr = ir3_SHR_B(block, instr, 0, create_immed(block, 4), 0); instr = ir3_COV(block, instr, TYPE_U32, TYPE_F32); @@ -1381,6 +1381,29 @@ emit_tex(struct ir3_compile *ctx, nir_tex_instr *tex) } } + switch (tex->op) { + case nir_texop_tex: opc = OPC_SAM; break; + case nir_texop_txb: opc = OPC_SAMB; break; + case nir_texop_txl: opc = OPC_SAML; break; + case nir_texop_txd: opc = OPC_SAMGQ; break; + case nir_texop_txf: opc = OPC_ISAML; break; + case nir_texop_txf_ms: + case nir_texop_txs: + case nir_texop_lod: + case nir_texop_tg4: + case nir_texop_query_levels: + compile_error(ctx, "Unhandled NIR tex type: %d\n", tex->op); + return; + } + + tex_info(tex, &flags, &coords); + + /* scale up integer coords for TXF based on the LOD */ + if (opc == OPC_ISAML) { + assert(has_lod); + for (i = 0; i < coords; i++) + coord[i] = ir3_SHL_B(b, coord[i], 0, lod, 0); + } /* * lay out the first argument in the proper order: * - actual coordinates first @@ -1392,8 +1415,6 @@ emit_tex(struct ir3_compile *ctx, nir_tex_instr *tex) * bias/lod go into the second arg */ - tex_info(tex, &flags, &coords); - /* insert tex coords: */ for (i = 0; i < coords; i++) src0[nsrc0++] = coord[i]; @@ -1450,21 +1471,6 @@ emit_tex(struct ir3_compile *ctx, nir_tex_instr *tex) src1[nsrc1++] = lod; } - switch (tex->op) { - case nir_texop_tex: opc = OPC_SAM; break; - case nir_texop_txb: opc = OPC_SAMB; break; - case nir_texop_txl: opc = OPC_SAML; break; - case nir_texop_txd: opc = OPC_SAMGQ; break; - case nir_texop_txf: opc = OPC_ISAML; break; - case nir_texop_txf_ms: - case nir_texop_txs: - case nir_texop_lod: - case nir_texop_tg4: - case nir_texop_query_levels: - compile_error(ctx, "Unhandled NIR tex type: %d\n", tex->op); - return; - } - switch (tex->dest_type) { case nir_type_invalid: case nir_type_float: @@ -1694,7 +1700,8 @@ setup_input(struct ir3_compile *ctx, nir_variable *in) so->inputs[n].bary = true; - instr = create_frag_input(ctx, idx, use_ldlv); + instr = create_frag_input(ctx, + so->inputs[n].inloc + i - 8, use_ldlv); } } else { instr = create_input(ctx->block, NULL, idx); |