Dram data storage: fast_get()/_put()

Kconfig.sof

@@ -130,6 +130,17 @@ config COLD_STORE_EXECUTE_DRAM
 	  option to enable this feature to save SRAM and to speed up SRAM
 	  copying of performance-critical data and code.
 
+config FAST_GET
+	bool "Enable simple refcounting dram data copier"
+	default n
+	help
+	  Enable simple refcounting DRAM data copier for copying processing
+	  module data from DRAM to SRAM when the data is needed and freeing
+	  the SRAM when the data is not needed anymore. If multiple module
+	  instances need the same chunk the same copy is used with reference
+	  counting. Source is src/lib/fast-get.c. The option should be selected
+	  by the modules using it.
+
 rsource "src/Kconfig"
 
 # See zephyr/modules/Kconfig

scripts/llext_link_helper.py

@@ -76,6 +76,7 @@ def main():
 	executable = []
 	writable = []
 	readonly = []
+	readonly_dram = []
 
 	text_found = False
 
@@ -101,10 +102,7 @@ def main():
 		if (s_flags & (SH_FLAGS.SHF_ALLOC | SH_FLAGS.SHF_EXECINSTR) ==
                     SH_FLAGS.SHF_ALLOC | SH_FLAGS.SHF_EXECINSTR and
 		    s_type == 'SHT_PROGBITS'):
-			# An executable section, currently only a single .text is supported.
-			# In general additional executable sections are possible, e.g.
-			# .init. In the future support for arbitrary such sections can be
-			# added, similar to writable and read-only data below.
+			# An executable section.
 			if s_name == '.text':
 				text_found = True
 				text_addr = max_alignment(text_addr, 0x1000, s_alignment)
@@ -123,7 +121,10 @@ def main():
 
 		if s_type == 'SHT_PROGBITS' and s_flags & SH_FLAGS.SHF_ALLOC:
 			# .rodata or other read-only sections
-			readonly.append(section)
+			if s_name == '.coldrodata':
+				readonly_dram.append(section)
+			else:
+				readonly.append(section)
 
 	if not text_found:
 		raise RuntimeError('No .text section found in the object file')
@@ -136,24 +137,37 @@ def main():
 	# run at arbitrary memory locations. One of the use-cases is running
 	# parts of the module directly in DRAM - sacrificing performance but
 	# saving scarce SRAM. We achieve this by placing non-performance
-	# critical functions in a .cold ELF section. When compiling and linking
-	# such functions, an additional .cold.literal section is automatically
-	# created. Note, that for some reason the compiler also marks that
-	# section as executable.
+	# critical functions in a .cold ELF section, read-only data in a
+	# .coldrodata ELF section, etc. When compiling and linking such
+	# functions, an additional .cold.literal section is automatically
+	# created. Note, that for some reason the compiler also marks .cold as
+	# executable.
 	# This script links those sections at address 0. We could hard-code
 	# section names, but so far we choose to only link .text the "original"
-	# way and all other executable sections we link at 0.
-	exe_addr = 0
+	# way and all other executable sections we link at 0. For data sections
+	# we accept only the .coldrodata name for now.
+
+	dram_addr = 0
 
 	for section in executable:
 		s_alignment = section.header['sh_addralign']
 		s_name = section.name
 
-		exe_addr = align_up(exe_addr, s_alignment)
+		dram_addr = align_up(dram_addr, s_alignment)
+
+		command.append(f'-Wl,--section-start={s_name}=0x{dram_addr:x}')
+
+		dram_addr += section.header['sh_size']
+
+	for section in readonly_dram:
+		s_alignment = section.header['sh_addralign']
+		s_name = section.name
+
+		dram_addr = align_up(dram_addr, s_alignment)
 
-		command.append(f'-Wl,--section-start={s_name}=0x{exe_addr:x}')
+		command.append(f'-Wl,--section-start={s_name}=0x{dram_addr:x}')
 
-		exe_addr += section.header['sh_size']
+		dram_addr += section.header['sh_size']
 
 	start_addr = align_up(text_addr + text_size, 0x1000)
 

src/audio/src/coef/src_ipc4_int32_10_21_2500_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_10_21_2500_5000_fir[480] = {
+__cold_rodata static const int32_t src_int32_10_21_2500_5000_fir[480] = {
 	176197,
 	283398,
 	-489527,

src/audio/src/coef/src_ipc4_int32_10_21_3455_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_10_21_3455_5000_fir[640] = {
+__cold_rodata static const int32_t src_int32_10_21_3455_5000_fir[640] = {
 	110400,
 	517669,
 	162088,

src/audio/src/coef/src_ipc4_int32_10_21_4535_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_10_21_4535_5000_fir[2320] = {
+__cold_rodata static const int32_t src_int32_10_21_4535_5000_fir[2320] = {
 	26554,
 	22041,
 	-35569,

src/audio/src/coef/src_ipc4_int32_10_9_4535_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_10_9_4535_5000_fir[1080] = {
+__cold_rodata static const int32_t src_int32_10_9_4535_5000_fir[1080] = {
 	-35695,
 	60551,
 	-91611,

src/audio/src/coef/src_ipc4_int32_16_21_4319_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_16_21_4319_5000_fir[1472] = {
+__cold_rodata static const int32_t src_int32_16_21_4319_5000_fir[1472] = {
 	69743,
 	-28255,
 	-123867,

src/audio/src/coef/src_ipc4_int32_16_21_4535_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_16_21_4535_5000_fir[2048] = {
+__cold_rodata static const int32_t src_int32_16_21_4535_5000_fir[2048] = {
 	66387,
 	-68365,
 	-7975,

src/audio/src/coef/src_ipc4_int32_16_7_4082_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_16_7_4082_5000_fir[896] = {
+__cold_rodata static const int32_t src_int32_16_7_4082_5000_fir[896] = {
 	-71000,
 	181977,
 	-339747,

src/audio/src/coef/src_ipc4_int32_1_2_2268_5000.h

@@ -7,7 +7,7 @@
 /** \cond GENERATED_BY_TOOLS_TUNE_SRC */
 #include <stdint.h>
 
-static const int32_t src_int32_1_2_2268_5000_fir[36] = {
+__cold_rodata static const int32_t src_int32_1_2_2268_5000_fir[36] = {
 	1065827,
 	-37924,
 	-4976218,

src/audio/src/coef/src_ipc4_int32_1_2_2500_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_1_2_2500_5000_fir[40] = {
+__cold_rodata static const int32_t src_int32_1_2_2500_5000_fir[40] = {
 	-879692,
 	460291,
 	4237437,

src/audio/src/coef/src_ipc4_int32_1_2_2721_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_1_2_2721_5000_fir[44] = {
+__cold_rodata static const int32_t src_int32_1_2_2721_5000_fir[44] = {
 	776925,
 	-535235,
 	-3522824,

src/audio/src/coef/src_ipc4_int32_1_2_3401_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_1_2_3401_5000_fir[60] = {
+__cold_rodata static const int32_t src_int32_1_2_3401_5000_fir[60] = {
 	483288,
 	-83413,
 	-1522435,

src/audio/src/coef/src_ipc4_int32_1_2_3887_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_1_2_3887_5000_fir[84] = {
+__cold_rodata static const int32_t src_int32_1_2_3887_5000_fir[84] = {
 	488732,
 	-7737,
 	-1049640,

src/audio/src/coef/src_ipc4_int32_1_2_4535_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_1_2_4535_5000_fir[192] = {
+__cold_rodata static const int32_t src_int32_1_2_4535_5000_fir[192] = {
 	-215107,
 	-43725,
 	301513,

src/audio/src/coef/src_ipc4_int32_1_3_2268_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_1_3_2268_5000_fir[52] = {
+__cold_rodata static const int32_t src_int32_1_3_2268_5000_fir[52] = {
 	856478,
 	-618891,
 	-4156030,

src/audio/src/coef/src_ipc4_int32_1_3_4535_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_1_3_4535_5000_fir[260] = {
+__cold_rodata static const int32_t src_int32_1_3_4535_5000_fir[260] = {
 	-76785,
 	87265,
 	208768,

src/audio/src/coef/src_ipc4_int32_1_4_1512_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_1_4_1512_5000_fir[52] = {
+__cold_rodata static const int32_t src_int32_1_4_1512_5000_fir[52] = {
 	740488,
 	-111252,
 	-2633435,

src/audio/src/coef/src_ipc4_int32_1_4_2268_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_1_4_2268_5000_fir[60] = {
+__cold_rodata static const int32_t src_int32_1_4_2268_5000_fir[60] = {
 	-1265010,
 	-1300023,
 	42822,

src/audio/src/coef/src_ipc4_int32_1_4_4535_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_1_4_4535_5000_fir[332] = {
+__cold_rodata static const int32_t src_int32_1_4_4535_5000_fir[332] = {
 	-246503,
 	-173077,
 	24381,

src/audio/src/coef/src_ipc4_int32_1_6_1134_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_1_6_1134_5000_fir[68] = {
+__cold_rodata static const int32_t src_int32_1_6_1134_5000_fir[68] = {
 	-2393808,
 	-3445469,
 	-3907601,

src/audio/src/coef/src_ipc4_int32_20_21_1250_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_20_21_1250_5000_fir[320] = {
+__cold_rodata static const int32_t src_int32_20_21_1250_5000_fir[320] = {
 	134710,
 	1035873,
 	-9174506,

src/audio/src/coef/src_ipc4_int32_20_21_2500_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_20_21_2500_5000_fir[560] = {
+__cold_rodata static const int32_t src_int32_20_21_2500_5000_fir[560] = {
 	-5217,
 	-381011,
 	1487277,

src/audio/src/coef/src_ipc4_int32_20_21_3125_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_20_21_3125_5000_fir[640] = {
+__cold_rodata static const int32_t src_int32_20_21_3125_5000_fir[640] = {
 	81905,
 	-392860,
 	605436,

src/audio/src/coef/src_ipc4_int32_20_21_4167_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_20_21_4167_5000_fir[1200] = {
+__cold_rodata static const int32_t src_int32_20_21_4167_5000_fir[1200] = {
 	54594,
 	-184742,
 	399571,

src/audio/src/coef/src_ipc4_int32_20_21_4535_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_20_21_4535_5000_fir[2080] = {
+__cold_rodata static const int32_t src_int32_20_21_4535_5000_fir[2080] = {
 	-39854,
 	85177,
 	-146854,

src/audio/src/coef/src_ipc4_int32_20_7_2976_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_20_7_2976_5000_fir[560] = {
+__cold_rodata static const int32_t src_int32_20_7_2976_5000_fir[560] = {
 	-33375,
 	393038,
 	-1331352,

src/audio/src/coef/src_ipc4_int32_21_10_2500_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_21_10_2500_5000_fir[504] = {
+__cold_rodata static const int32_t src_int32_21_10_2500_5000_fir[504] = {
 	7704,
 	568453,
 	-2657822,

src/audio/src/coef/src_ipc4_int32_21_10_3455_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_21_10_3455_5000_fir[756] = {
+__cold_rodata static const int32_t src_int32_21_10_3455_5000_fir[756] = {
 	-37222,
 	240677,
 	-698458,

src/audio/src/coef/src_ipc4_int32_21_10_4535_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_21_10_4535_5000_fir[2520] = {
+__cold_rodata static const int32_t src_int32_21_10_4535_5000_fir[2520] = {
 	-9118,
 	18975,
 	-33943,

src/audio/src/coef/src_ipc4_int32_21_16_4319_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_21_16_4319_5000_fir[1596] = {
+__cold_rodata static const int32_t src_int32_21_16_4319_5000_fir[1596] = {
 	-50136,
 	105527,
 	-181643,

src/audio/src/coef/src_ipc4_int32_21_16_4535_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_21_16_4535_5000_fir[2436] = {
+__cold_rodata static const int32_t src_int32_21_16_4535_5000_fir[2436] = {
 	-15208,
 	29050,
 	-48628,

src/audio/src/coef/src_ipc4_int32_21_20_1250_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_21_20_1250_5000_fir[420] = {
+__cold_rodata static const int32_t src_int32_21_20_1250_5000_fir[420] = {
 	94664,
 	-1133291,
 	202774,

src/audio/src/coef/src_ipc4_int32_21_20_2500_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_21_20_2500_5000_fir[504] = {
+__cold_rodata static const int32_t src_int32_21_20_2500_5000_fir[504] = {
 	8116,
 	587965,
 	-2725209,

src/audio/src/coef/src_ipc4_int32_21_20_3125_5000.h

@@ -8,7 +8,7 @@
 
 #include <stdint.h>
 
-static const int32_t src_int32_21_20_3125_5000_fir[672] = {
+__cold_rodata static const int32_t src_int32_21_20_3125_5000_fir[672] = {
 	4928,
 	190845,
 	-929522,