r600/ssbo: refactor out buffer coord calcs and use for atomic path.

The atomic rat path has a bug in the ssbo path, refactor out the
address calcs from the load/store paths and reuse to fix the bug
in the buffer rat atomic path.

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

diff --git a/src/gallium/drivers/r600/r600_shader.c b/src/gallium/drivers/r600/r600_shader.c
index 02e8354baa29c06a8b7cb917236a0da071fa624d..7768f503c0ccd1077c452db8156b1a4f164a61b6 100644 (file)
--- a/src/gallium/drivers/r600/r600_shader.c
+++ b/src/gallium/drivers/r600/r600_shader.c
@@ -7952,22 +7952,13 @@ static int load_index_src(struct r600_shader_ctx *ctx, int src_index, int *idx_g
        return 0;
 }
 
-static int tgsi_load_buffer(struct r600_shader_ctx *ctx)
+static int load_buffer_coord(struct r600_shader_ctx *ctx, int src_idx,
+                            int temp_reg)
 {
        struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
-       /* have to work out the offset into the RAT immediate return buffer */
-       struct r600_bytecode_vtx vtx;
-       struct r600_bytecode_cf *cf;
        int r;
-       int temp_reg = r600_get_temp(ctx);
-       unsigned rat_index_mode;
-       unsigned base;
-
-       rat_index_mode = inst->Src[0].Indirect.Index == 2 ? 2 : 0; // CF_INDEX_1 : CF_INDEX_NONE
-       base = R600_IMAGE_REAL_RESOURCE_OFFSET + ctx->info.file_count[TGSI_FILE_IMAGE];
-
-       if (inst->Src[1].Register.File == TGSI_FILE_IMMEDIATE) {
-               int value = (ctx->literals[4 * inst->Src[1].Register.Index + inst->Src[1].Register.SwizzleX]);
+       if (inst->Src[src_idx].Register.File == TGSI_FILE_IMMEDIATE) {
+               int value = (ctx->literals[4 * inst->Src[src_idx].Register.Index + inst->Src[src_idx].Register.SwizzleX]);
                r = single_alu_op2(ctx, ALU_OP1_MOV,
                                   temp_reg, 0,
                                   V_SQ_ALU_SRC_LITERAL, value >> 2,
@@ -7978,7 +7969,7 @@ static int tgsi_load_buffer(struct r600_shader_ctx *ctx)
                struct r600_bytecode_alu alu;
                memset(&alu, 0, sizeof(struct r600_bytecode_alu));
                alu.op = ALU_OP2_LSHR_INT;
-               r600_bytecode_src(&alu.src[0], &ctx->src[1], 0);
+               r600_bytecode_src(&alu.src[0], &ctx->src[src_idx], 0);
                alu.src[1].sel = V_SQ_ALU_SRC_LITERAL;
                alu.src[1].value = 2;
                alu.dst.sel = temp_reg;
@@ -7988,6 +7979,26 @@ static int tgsi_load_buffer(struct r600_shader_ctx *ctx)
                if (r)
                        return r;
        }
+       return 0;
+}
+
+static int tgsi_load_buffer(struct r600_shader_ctx *ctx)
+{
+       struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+       /* have to work out the offset into the RAT immediate return buffer */
+       struct r600_bytecode_vtx vtx;
+       struct r600_bytecode_cf *cf;
+       int r;
+       int temp_reg = r600_get_temp(ctx);
+       unsigned rat_index_mode;
+       unsigned base;
+
+       rat_index_mode = inst->Src[0].Indirect.Index == 2 ? 2 : 0; // CF_INDEX_1 : CF_INDEX_NONE
+       base = R600_IMAGE_REAL_RESOURCE_OFFSET + ctx->info.file_count[TGSI_FILE_IMAGE];
+
+       r = load_buffer_coord(ctx, 1, temp_reg);
+       if (r)
+               return r;
        ctx->bc->cf_last->barrier = 1;
        memset(&vtx, 0, sizeof(struct r600_bytecode_vtx));
        vtx.op = FETCH_OP_VFETCH;
@@ -8125,22 +8136,9 @@ static int tgsi_store_buffer_rat(struct r600_shader_ctx *ctx)
        int lasti;
        int temp_reg = r600_get_temp(ctx), treg2 = r600_get_temp(ctx);
 
-       if (inst->Src[0].Register.File == TGSI_FILE_IMMEDIATE) {
-               int value = (ctx->literals[4 * inst->Src[0].Register.Index + inst->Src[0].Register.SwizzleX]);
-               r = single_alu_op2(ctx, ALU_OP1_MOV,
-                                  treg2, 0,
-                                  V_SQ_ALU_SRC_LITERAL, value >> 2,
-                                  0, 0);
-               if (r)
-                       return r;
-       } else {
-               r = single_alu_op2(ctx, ALU_OP2_LSHR_INT,
-                                  treg2, 0,
-                                  ctx->src[0].sel, ctx->src[0].swizzle[0],
-                                  V_SQ_ALU_SRC_LITERAL, 2);
-               if (r)
-                       return r;
-       }
+       r = load_buffer_coord(ctx, 0, treg2);
+       if (r)
+               return r;
 
        rat_index_mode = inst->Dst[0].Indirect.Index == 2 ? 2 : 0; // CF_INDEX_1 : CF_INDEX_NONE
        if (rat_index_mode)
@@ -8294,14 +8292,19 @@ static int tgsi_atomic_op_rat(struct r600_shader_ctx *ctx)
         if (inst->Src[0].Register.File == TGSI_FILE_BUFFER) {
                immed_base += ctx->info.file_count[TGSI_FILE_IMAGE];
                rat_base += ctx->info.file_count[TGSI_FILE_IMAGE];
+
+               r = load_buffer_coord(ctx, 1, ctx->temp_reg);
+               if (r)
+                       return r;
+               idx_gpr = ctx->temp_reg;
+       } else {
+               r = load_index_src(ctx, 1, &idx_gpr);
+               if (r)
+                       return r;
        }
 
        rat_index_mode = inst->Src[0].Indirect.Index == 2 ? 2 : 0; // CF_INDEX_1 : CF_INDEX_NONE
 
-       r = load_index_src(ctx, 1, &idx_gpr);
-       if (r)
-               return r;
-
        if (ctx->inst_info->op == V_RAT_INST_CMPXCHG_INT_RTN) {
                memset(&alu, 0, sizeof(struct r600_bytecode_alu));
                alu.op = ALU_OP1_MOV;