r600g: Use LOOP_START_DX10 for loops

LOOP_START_DX10 ignores the LOOP_CONFIG* registers, so it is not limited
to 4096 iterations like the other LOOP_* instructions.  Compute shaders
need to use this instruction, and since we aren't optimizing loops with
the LOOP_CONFIG* registers for pixel and vertex shaders, it seems like
we should just use it for everything.

Reviewed-by: Marek Olšák <[email protected]>

3 files changed, 11 insertions, 2 deletions

diff --git a/src/gallium/drivers/r600/eg_asm.c b/src/gallium/drivers/r600/eg_asm.c
index d2c1679796a..00ac4a8c25b 100644
--- a/src/gallium/drivers/r600/eg_asm.c
+++ b/src/gallium/drivers/r600/eg_asm.c
@@ -121,6 +121,7 @@ int eg_bytecode_cf_build(struct r600_bytecode *bc, struct r600_bytecode_cf *cf)
 	case EG_V_SQ_CF_WORD1_SQ_CF_INST_ELSE:
 	case EG_V_SQ_CF_WORD1_SQ_CF_INST_POP:
 	case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_NO_AL:
+	case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_DX10:
 	case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_END:
 	case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_CONTINUE:
 	case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_BREAK:
diff --git a/src/gallium/drivers/r600/r600_asm.c b/src/gallium/drivers/r600/r600_asm.c
index 648e8b6ed5d..58350c6f75d 100644
--- a/src/gallium/drivers/r600/r600_asm.c
+++ b/src/gallium/drivers/r600/r600_asm.c
@@ -1878,6 +1878,7 @@ static int r600_bytecode_cf_build(struct r600_bytecode *bc, struct r600_bytecode
 	case V_SQ_CF_WORD1_SQ_CF_INST_ELSE:
 	case V_SQ_CF_WORD1_SQ_CF_INST_POP:
 	case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_NO_AL:
+	case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_DX10:
 	case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_END:
 	case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_CONTINUE:
 	case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_BREAK:
@@ -1952,6 +1953,7 @@ int r600_bytecode_build(struct r600_bytecode *bc)
 			case EG_V_SQ_CF_WORD1_SQ_CF_INST_ELSE:
 			case EG_V_SQ_CF_WORD1_SQ_CF_INST_POP:
 			case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_NO_AL:
+			case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_DX10:
 			case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_END:
 			case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_CONTINUE:
 			case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_BREAK:
@@ -1986,7 +1988,7 @@ int r600_bytecode_build(struct r600_bytecode *bc)
 			case V_SQ_CF_WORD1_SQ_CF_INST_JUMP:
 			case V_SQ_CF_WORD1_SQ_CF_INST_ELSE:
 			case V_SQ_CF_WORD1_SQ_CF_INST_POP:
-			case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_NO_AL:
+			case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_DX10:
 			case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_END:
 			case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_CONTINUE:
 			case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_BREAK:
@@ -2089,6 +2091,7 @@ int r600_bytecode_build(struct r600_bytecode *bc)
 			case EG_V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_MEM_STREAM3_BUF1:
 			case EG_V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_MEM_STREAM3_BUF2:
 			case EG_V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_MEM_STREAM3_BUF3:
+			case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_DX10:
 			case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_NO_AL:
 			case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_END:
 			case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_CONTINUE:
@@ -2172,6 +2175,7 @@ int r600_bytecode_build(struct r600_bytecode *bc)
 			case V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_MEM_STREAM2:
 			case V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_MEM_STREAM3:
 			case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_NO_AL:
+			case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_DX10:
 			case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_END:
 			case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_CONTINUE:
 			case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_BREAK:
@@ -2360,6 +2364,7 @@ void r600_bytecode_dump(struct r600_bytecode *bc)
 			case EG_V_SQ_CF_WORD1_SQ_CF_INST_ELSE:
 			case EG_V_SQ_CF_WORD1_SQ_CF_INST_POP:
 			case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_NO_AL:
+			case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_DX10:
 			case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_END:
 			case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_CONTINUE:
 			case EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_BREAK:
@@ -2454,6 +2459,7 @@ void r600_bytecode_dump(struct r600_bytecode *bc)
 			case V_SQ_CF_WORD1_SQ_CF_INST_ELSE:
 			case V_SQ_CF_WORD1_SQ_CF_INST_POP:
 			case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_NO_AL:
+			case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_DX10:
 			case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_END:
 			case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_CONTINUE:
 			case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_BREAK:
diff --git a/src/gallium/drivers/r600/r600_shader.c b/src/gallium/drivers/r600/r600_shader.c
index 3e746e5e2e2..8288c040a63 100644
--- a/src/gallium/drivers/r600/r600_shader.c
+++ b/src/gallium/drivers/r600/r600_shader.c
@@ -5101,7 +5101,9 @@ static int tgsi_endif(struct r600_shader_ctx *ctx)
 
 static int tgsi_bgnloop(struct r600_shader_ctx *ctx)
 {
-	r600_bytecode_add_cfinst(ctx->bc, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_NO_AL));
+	/* LOOP_START_DX10 ignores the LOOP_CONFIG* registers, so it is not
+	 * limited to 4096 iterations, like the other LOOP_* instructions. */
+	r600_bytecode_add_cfinst(ctx->bc, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_DX10));
 
 	fc_pushlevel(ctx, FC_LOOP);