r300: simplify rc_find_free_temporary

Back when we had a stupid register allocator we did a lot of tricks to
optimize the register usage. The old version of rc_find_free_temporary
did a full program search each time it was called to find out what
registers and channels are actually used and than used that info to give
us the first free register to use.

Now that we have a proper register allocator both for vertex and
fragment shaders, this is no longer needed. Just scan the program when
called for the first time to find the first unused temporary index and
than increment by one everytime. Regalloc can sort it out later.

No change in shader-db confirms this assumption is sound.

Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Reviewed-by: Filip Gawin <filip@gawin.net>
Tested-by: Filip Gawin <filip@gawin.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19853>

src/gallium/drivers/r300/compiler/radeon_compiler.c

@@ -45,6 +45,7 @@ void rc_init(struct radeon_compiler * c, const struct rc_regalloc_state *rs)
 	c->Program.Instructions.Next = &c->Program.Instructions;
 	c->Program.Instructions.U.I.Opcode = RC_OPCODE_ILLEGAL_OPCODE;
 	c->regalloc_state = rs;
+	c->max_temp_index = -1;
 }
 
 void rc_destroy(struct radeon_compiler * c)

src/gallium/drivers/r300/compiler/radeon_compiler.h

@@ -62,6 +62,8 @@ struct radeon_compiler {
 	int max_alu_insts;
 	unsigned max_tex_insts;
 
+	int max_temp_index;
+
 	/* Whether to remove unused constants and empty holes in constant space. */
 	unsigned remove_unused_constants:1;
 

src/gallium/drivers/r300/compiler/radeon_program.c

@@ -71,98 +71,28 @@ void rc_local_transform(
 	}
 }
 
-struct get_used_temporaries_data {
-	unsigned char * Used;
-	unsigned int UsedLength;
-};
-
-static void get_used_temporaries_cb(
-	void * userdata,
-	struct rc_instruction * inst,
-	rc_register_file file,
-	unsigned int index,
-	unsigned int mask)
-{
-	struct get_used_temporaries_data * d = userdata;
-
-	if (file != RC_FILE_TEMPORARY)
-		return;
-
-	if (index >= d->UsedLength)
-		return;
-
-	d->Used[index] |= mask;
-}
-
-/**
- * This function fills in the parameter 'used' with a writemask that
- * represent which components of each temporary register are used by the
- * program.  This is meant to be combined with rc_find_free_temporary_list as a
- * more efficient version of rc_find_free_temporary.
- * @param used The function does not initialize this parameter.
- */
-void rc_get_used_temporaries(
-	struct radeon_compiler * c,
-	unsigned char * used,
-	unsigned int used_length)
-{
-	struct rc_instruction * inst;
-	struct get_used_temporaries_data d;
-	d.Used = used;
-	d.UsedLength = used_length;
-
-	for(inst = c->Program.Instructions.Next;
-			inst != &c->Program.Instructions; inst = inst->Next) {
-
-		rc_for_all_reads_mask(inst, get_used_temporaries_cb, &d);
-		rc_for_all_writes_mask(inst, get_used_temporaries_cb, &d);
-	}
-}
-
-/* Search a list of used temporaries for a free one
- * \sa rc_get_used_temporaries
- * @note If this functions finds a free temporary, it will mark it as used
- * in the used temporary list (param 'used')
- * @param used list of used temporaries
- * @param used_length number of items in param 'used'
- * @param mask which components must be free in the temporary index that is
- * returned.
- * @return -1 If there are no more free temporaries, otherwise the index of
- * a temporary register where the components specified in param 'mask' are
- * not being used.
- */
-int rc_find_free_temporary_list(
-	struct radeon_compiler * c,
-	unsigned char * used,
-	unsigned int used_length,
-	unsigned int mask)
-{
-	int i;
-	for(i = 0; i < used_length; i++) {
-		if ((~used[i] & mask) == mask) {
-			used[i] |= mask;
-			return i;
-		}
-	}
-	return -1;
-}
-
 unsigned int rc_find_free_temporary(struct radeon_compiler * c)
 {
-	unsigned char used[RC_REGISTER_MAX_INDEX];
-	int free;
+	/* Find the largest used temp index when called for the first time. */
+	if (c->max_temp_index == -1) {
+		for (struct rc_instruction * inst = c->Program.Instructions.Next;
+			inst != &c->Program.Instructions; inst = inst->Next) {
+			const struct rc_opcode_info * opcode =
+				rc_get_opcode_info(inst->U.I.Opcode);
+			if (opcode->HasDstReg &&
+				inst->U.I.DstReg.File == RC_FILE_TEMPORARY &&
+				inst->U.I.WriteALUResult == RC_ALURESULT_NONE &&
+				inst->U.I.DstReg.Index > c->max_temp_index)
+				c->max_temp_index = inst->U.I.DstReg.Index;
+		}
+	}
 
-	memset(used, 0, sizeof(used));
-
-	rc_get_used_temporaries(c, used, RC_REGISTER_MAX_INDEX);
-
-	free = rc_find_free_temporary_list(c, used, RC_REGISTER_MAX_INDEX,
-								RC_MASK_XYZW);
-	if (free < 0) {
+	c->max_temp_index++;
+	if (c->max_temp_index > RC_REGISTER_MAX_INDEX) {
 		rc_error(c, "Ran out of temporary registers\n");
 		return 0;
 	}
-	return free;
+	return c->max_temp_index;
 }
 
 

src/gallium/drivers/r300/compiler/radeon_rename_regs.c

@@ -48,9 +48,7 @@
  */
 void rc_rename_regs(struct radeon_compiler *c, void *user)
 {
-	unsigned int used_length;
 	struct rc_instruction * inst;
-	unsigned char * used;
 	struct rc_list * variables;
 	struct rc_list * var_ptr;
 
@@ -62,11 +60,6 @@ void rc_rename_regs(struct radeon_compiler *c, void *user)
 			return;
 	}
 
-	used_length = MIN2(2 * rc_recompute_ips(c), RC_REGISTER_MAX_INDEX);
-	used = memory_pool_malloc(&c->Pool, sizeof(unsigned char) * used_length);
-	memset(used, 0, sizeof(unsigned char) * used_length);
-
-	rc_get_used_temporaries(c, used, used_length);
 	variables = rc_get_variables(c);
 
 	for (var_ptr = variables; var_ptr; var_ptr = var_ptr->Next) {
@@ -78,8 +71,7 @@ void rc_rename_regs(struct radeon_compiler *c, void *user)
 			continue;
 		}
 
-		new_index = rc_find_free_temporary_list(c, used, used_length,
-						RC_MASK_XYZW);
+		new_index = rc_find_free_temporary(c);
 		if (new_index < 0) {
 			rc_error(c, "Ran out of temporary registers\n");
 			return;