radeonsi: merge and simplify tbuffer_store functions

Reviewed-by: Dave Airlie <airlied@redhat.com>

diff --git a/src/amd/common/ac_llvm_build.c b/src/amd/common/ac_llvm_build.c
index 34085bb7e5331aed4ac91bd1e5a10b61441dcc23..cc1eaf1385c280a18be4eb0e21ac65cd6f677f59 100644 (file)
--- a/src/amd/common/ac_llvm_build.c
+++ b/src/amd/common/ac_llvm_build.c
@@ -543,21 +543,25 @@ ac_build_indexed_load_const(struct ac_llvm_context *ctx,
  * or v4i32 (num_channels=3,4).
  */
 void
-ac_build_tbuffer_store(struct ac_llvm_context *ctx,
-                      LLVMValueRef rsrc,
-                      LLVMValueRef vdata,
-                      unsigned num_channels,
-                      LLVMValueRef vaddr,
-                      LLVMValueRef soffset,
-                      unsigned inst_offset,
-                      unsigned dfmt,
-                      unsigned nfmt,
-                      unsigned offen,
-                      unsigned idxen,
-                      unsigned glc,
-                      unsigned slc,
-                      unsigned tfe)
+ac_build_buffer_store_dword(struct ac_llvm_context *ctx,
+                           LLVMValueRef rsrc,
+                           LLVMValueRef vdata,
+                           unsigned num_channels,
+                           LLVMValueRef vaddr,
+                           LLVMValueRef soffset,
+                           unsigned inst_offset,
+                           bool offen,
+                           bool glc,
+                           bool slc)
 {
+       static unsigned dfmt[] = {
+               V_008F0C_BUF_DATA_FORMAT_32,
+               V_008F0C_BUF_DATA_FORMAT_32_32,
+               V_008F0C_BUF_DATA_FORMAT_32_32_32,
+               V_008F0C_BUF_DATA_FORMAT_32_32_32_32
+       };
+       assert(num_channels >= 1 && num_channels <= 4);
+
        LLVMValueRef args[] = {
                rsrc,
                vdata,
@@ -565,13 +569,13 @@ ac_build_tbuffer_store(struct ac_llvm_context *ctx,
                vaddr,
                soffset,
                LLVMConstInt(ctx->i32, inst_offset, 0),
-               LLVMConstInt(ctx->i32, dfmt, 0),
-               LLVMConstInt(ctx->i32, nfmt, 0),
+               LLVMConstInt(ctx->i32, dfmt[num_channels - 1], 0),
+               LLVMConstInt(ctx->i32, V_008F0C_BUF_NUM_FORMAT_UINT, 0),
                LLVMConstInt(ctx->i32, offen, 0),
-               LLVMConstInt(ctx->i32, idxen, 0),
+               LLVMConstInt(ctx->i32, 0, 0), /* idxen */
                LLVMConstInt(ctx->i32, glc, 0),
                LLVMConstInt(ctx->i32, slc, 0),
-               LLVMConstInt(ctx->i32, tfe, 0)
+               LLVMConstInt(ctx->i32, 0, 0), /* tfe*/
        };
 
        /* The instruction offset field has 12 bits */
@@ -588,28 +592,6 @@ ac_build_tbuffer_store(struct ac_llvm_context *ctx,
                               AC_FUNC_ATTR_LEGACY);
 }
 
-void
-ac_build_tbuffer_store_dwords(struct ac_llvm_context *ctx,
-                             LLVMValueRef rsrc,
-                             LLVMValueRef vdata,
-                             unsigned num_channels,
-                             LLVMValueRef vaddr,
-                             LLVMValueRef soffset,
-                             unsigned inst_offset)
-{
-       static unsigned dfmt[] = {
-               V_008F0C_BUF_DATA_FORMAT_32,
-               V_008F0C_BUF_DATA_FORMAT_32_32,
-               V_008F0C_BUF_DATA_FORMAT_32_32_32,
-               V_008F0C_BUF_DATA_FORMAT_32_32_32_32
-       };
-       assert(num_channels >= 1 && num_channels <= 4);
-
-       ac_build_tbuffer_store(ctx, rsrc, vdata, num_channels, vaddr, soffset,
-                              inst_offset, dfmt[num_channels - 1],
-                              V_008F0C_BUF_NUM_FORMAT_UINT, 1, 0, 1, 1, 0);
-}
-
 LLVMValueRef
 ac_build_buffer_load(struct ac_llvm_context *ctx,
                     LLVMValueRef rsrc,

diff --git a/src/amd/common/ac_llvm_build.h b/src/amd/common/ac_llvm_build.h
index e7773d734817904aed89728547b144a0d9c7af58..65a9a055a72978c0af8f218ae96e7f913b7dff47 100644 (file)
--- a/src/amd/common/ac_llvm_build.h
+++ b/src/amd/common/ac_llvm_build.h
@@ -122,30 +122,16 @@ ac_build_indexed_load_const(struct ac_llvm_context *ctx,
                            LLVMValueRef base_ptr, LLVMValueRef index);
 
 void
-ac_build_tbuffer_store_dwords(struct ac_llvm_context *ctx,
-                             LLVMValueRef rsrc,
-                             LLVMValueRef vdata,
-                             unsigned num_channels,
-                             LLVMValueRef vaddr,
-                             LLVMValueRef soffset,
-                             unsigned inst_offset);
-
-void
-ac_build_tbuffer_store(struct ac_llvm_context *ctx,
-                      LLVMValueRef rsrc,
-                      LLVMValueRef vdata,
-                      unsigned num_channels,
-                      LLVMValueRef vaddr,
-                      LLVMValueRef soffset,
-                      unsigned inst_offset,
-                      unsigned dfmt,
-                      unsigned nfmt,
-                      unsigned offen,
-                      unsigned idxen,
-                      unsigned glc,
-                      unsigned slc,
-                      unsigned tfe);
-
+ac_build_buffer_store_dword(struct ac_llvm_context *ctx,
+                           LLVMValueRef rsrc,
+                           LLVMValueRef vdata,
+                           unsigned num_channels,
+                           LLVMValueRef vaddr,
+                           LLVMValueRef soffset,
+                           unsigned inst_offset,
+                           bool offen,
+                           bool glc,
+                           bool slc);
 LLVMValueRef
 ac_build_buffer_load(struct ac_llvm_context *ctx,
                     LLVMValueRef rsrc,

diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c
index 47b95e40b2982d7f268e7149ae8255aefd4b50bf..d7c9d5968d95b68a869da0356fd1eb7dfd2cfa56 100644 (file)
--- a/src/amd/common/ac_nir_to_llvm.c
+++ b/src/amd/common/ac_nir_to_llvm.c
@@ -3156,12 +3156,10 @@ visit_emit_vertex(struct nir_to_llvm_context *ctx,
 
                        out_val = LLVMBuildBitCast(ctx->builder, out_val, ctx->i32, "");
 
-                       ac_build_tbuffer_store(&ctx->ac, ctx->gsvs_ring,
-                                              out_val, 1,
-                                              voffset, ctx->gs2vs_offset, 0,
-                                              V_008F0C_BUF_DATA_FORMAT_32,
-                                              V_008F0C_BUF_NUM_FORMAT_UINT,
-                                              1, 0, 1, 1, 0);
+                       ac_build_buffer_store_dword(&ctx->ac, ctx->gsvs_ring,
+                                                   out_val, 1,
+                                                   voffset, ctx->gs2vs_offset, 0,
+                                                   1, 1, 1);
                }
                idx += slot_inc;
        }
@@ -4672,14 +4670,12 @@ handle_es_outputs_post(struct nir_to_llvm_context *ctx)
                        LLVMValueRef out_val = LLVMBuildLoad(ctx->builder, out_ptr[j], "");
                        out_val = LLVMBuildBitCast(ctx->builder, out_val, ctx->i32, "");
 
-                       ac_build_tbuffer_store(&ctx->ac,
+                       ac_build_buffer_store_dword(&ctx->ac,
                                               ctx->esgs_ring,
                                               out_val, 1,
                                               LLVMGetUndef(ctx->i32), ctx->es2gs_offset,
                                               (4 * param_index + j + start) * 4,
-                                              V_008F0C_BUF_DATA_FORMAT_32,
-                                              V_008F0C_BUF_NUM_FORMAT_UINT,
-                                              0, 0, 1, 1, 0);
+                                              0, 1, 1);
                }
        }
        ctx->shader_info->vs.esgs_itemsize = (max_output_written + 1) * 16;

diff --git a/src/gallium/drivers/radeonsi/si_shader.c b/src/gallium/drivers/radeonsi/si_shader.c
index 75567af778eb6c708cb8c719636502422ee24f6d..11f2affbbb35516d9f75bb850e8fa8802709f667 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_shader.c
+++ b/src/gallium/drivers/radeonsi/si_shader.c
@@ -1047,17 +1047,17 @@ static void store_output_tcs(struct lp_build_tgsi_context *bld_base,
                values[chan_index] = value;
 
                if (inst->Dst[0].Register.WriteMask != 0xF && !is_tess_factor) {
-                       ac_build_tbuffer_store_dwords(&ctx->ac, buffer, value, 1,
-                                                     buf_addr, base,
-                                                     4 * chan_index);
+                       ac_build_buffer_store_dword(&ctx->ac, buffer, value, 1,
+                                                   buf_addr, base,
+                                                   4 * chan_index, 1, 1, 1);
                }
        }
 
        if (inst->Dst[0].Register.WriteMask == 0xF && !is_tess_factor) {
                LLVMValueRef value = lp_build_gather_values(bld_base->base.gallivm,
                                                            values, 4);
-               ac_build_tbuffer_store_dwords(&ctx->ac, buffer, value, 4, buf_addr,
-                                             base, 0);
+               ac_build_buffer_store_dword(&ctx->ac, buffer, value, 4, buf_addr,
+                                           base, 0, 1, 1, 1);
        }
 }
 
@@ -2083,11 +2083,11 @@ static void emit_streamout_output(struct si_shader_context *ctx,
                break;
        }
 
-       ac_build_tbuffer_store_dwords(&ctx->ac, so_buffers[buf_idx],
-                                     vdata, num_comps,
-                                     so_write_offsets[buf_idx],
-                                     LLVMConstInt(ctx->i32, 0, 0),
-                                     stream_out->dst_offset * 4);
+       ac_build_buffer_store_dword(&ctx->ac, so_buffers[buf_idx],
+                                   vdata, num_comps,
+                                   so_write_offsets[buf_idx],
+                                   LLVMConstInt(ctx->i32, 0, 0),
+                                   stream_out->dst_offset * 4, 1, 1, 1);
 }
 
 /**
@@ -2411,8 +2411,8 @@ static void si_copy_tcs_inputs(struct lp_build_tgsi_context *bld_base)
                LLVMValueRef value = lds_load(bld_base, TGSI_TYPE_SIGNED, ~0,
                                              lds_ptr);
 
-               ac_build_tbuffer_store_dwords(&ctx->ac, buffer, value, 4, buffer_addr,
-                                             buffer_offset, 0);
+               ac_build_buffer_store_dword(&ctx->ac, buffer, value, 4, buffer_addr,
+                                           buffer_offset, 0, 1, 1, 1);
        }
 }
 
@@ -2524,18 +2524,21 @@ static void si_write_tess_factors(struct lp_build_tgsi_context *bld_base,
                                  rel_patch_id, bld_base->uint_bld.zero, ""));
 
        /* Store the dynamic HS control word. */
-       ac_build_tbuffer_store_dwords(&ctx->ac, buffer,
-                                     lp_build_const_int32(gallivm, 0x80000000),
-                                     1, lp_build_const_int32(gallivm, 0), tf_base, 0);
+       ac_build_buffer_store_dword(&ctx->ac, buffer,
+                                   lp_build_const_int32(gallivm, 0x80000000),
+                                   1, lp_build_const_int32(gallivm, 0), tf_base,
+                                   0, 1, 1, 1);
 
        lp_build_endif(&inner_if_ctx);
 
        /* Store the tessellation factors. */
-       ac_build_tbuffer_store_dwords(&ctx->ac, buffer, vec0,
-                                     MIN2(stride, 4), byteoffset, tf_base, 4);
+       ac_build_buffer_store_dword(&ctx->ac, buffer, vec0,
+                                   MIN2(stride, 4), byteoffset, tf_base,
+                                   4, 1, 1, 1);
        if (vec1)
-               ac_build_tbuffer_store_dwords(&ctx->ac, buffer, vec1,
-                                             stride - 4, byteoffset, tf_base, 20);
+               ac_build_buffer_store_dword(&ctx->ac, buffer, vec1,
+                                           stride - 4, byteoffset, tf_base,
+                                           20, 1, 1, 1);
 
        /* Store the tess factors into the offchip buffer if TES reads them. */
        if (shader->key.part.tcs.epilog.tes_reads_tess_factors) {
@@ -2555,9 +2558,9 @@ static void si_write_tess_factors(struct lp_build_tgsi_context *bld_base,
                outer_vec = lp_build_gather_values(gallivm, outer,
                                                   util_next_power_of_two(outer_comps));
 
-               ac_build_tbuffer_store_dwords(&ctx->ac, buf, outer_vec,
-                                             outer_comps, tf_outer_offset,
-                                             base, 0);
+               ac_build_buffer_store_dword(&ctx->ac, buf, outer_vec,
+                                           outer_comps, tf_outer_offset,
+                                           base, 0, 1, 1, 1);
                if (inner_comps) {
                        param_inner = si_shader_io_get_unique_index(
                                              TGSI_SEMANTIC_TESSINNER, 0);
@@ -2566,9 +2569,9 @@ static void si_write_tess_factors(struct lp_build_tgsi_context *bld_base,
 
                        inner_vec = inner_comps == 1 ? inner[0] :
                                    lp_build_gather_values(gallivm, inner, inner_comps);
-                       ac_build_tbuffer_store_dwords(&ctx->ac, buf, inner_vec,
-                                                     inner_comps, tf_inner_offset,
-                                                     base, 0);
+                       ac_build_buffer_store_dword(&ctx->ac, buf, inner_vec,
+                                                   inner_comps, tf_inner_offset,
+                                                   base, 0, 1, 1, 1);
                }
        }
 
@@ -2688,14 +2691,12 @@ static void si_llvm_emit_es_epilogue(struct lp_build_tgsi_context *bld_base)
                        LLVMValueRef out_val = LLVMBuildLoad(gallivm->builder, out_ptr[chan], "");
                        out_val = LLVMBuildBitCast(gallivm->builder, out_val, ctx->i32, "");
 
-                       ac_build_tbuffer_store(&ctx->ac,
-                                              ctx->esgs_ring,
-                                              out_val, 1,
-                                              LLVMGetUndef(ctx->i32), soffset,
-                                              (4 * param_index + chan) * 4,
-                                              V_008F0C_BUF_DATA_FORMAT_32,
-                                              V_008F0C_BUF_NUM_FORMAT_UINT,
-                                              0, 0, 1, 1, 0);
+                       ac_build_buffer_store_dword(&ctx->ac,
+                                                   ctx->esgs_ring,
+                                                   out_val, 1,
+                                                   LLVMGetUndef(ctx->i32), soffset,
+                                                   (4 * param_index + chan) * 4,
+                                                   0, 1, 1);
                }
        }
 }
@@ -5056,13 +5057,11 @@ static void si_llvm_emit_vertex(
 
                        out_val = LLVMBuildBitCast(gallivm->builder, out_val, ctx->i32, "");
 
-                       ac_build_tbuffer_store(&ctx->ac,
-                                              ctx->gsvs_ring[stream],
-                                              out_val, 1,
-                                              voffset, soffset, 0,
-                                              V_008F0C_BUF_DATA_FORMAT_32,
-                                              V_008F0C_BUF_NUM_FORMAT_UINT,
-                                              1, 0, 1, 1, 0);
+                       ac_build_buffer_store_dword(&ctx->ac,
+                                                   ctx->gsvs_ring[stream],
+                                                   out_val, 1,
+                                                   voffset, soffset, 0,
+                                                   1, 1, 1);
                }
        }