Perform more optimization in supercard platform (#107)

arm9/source/patches/platform/supercard/SuperCardCommon.h

@@ -6,8 +6,8 @@ DEFINE_SECTION_SYMBOLS(scsd_change_mode);
 DEFINE_SECTION_SYMBOLS(scsd_common);
 
 extern "C" void sccmn_changeMode();
-
 extern "C" void sccmn_sdSendClock10();
+
 extern "C" void sccmn_sdio4BitCrc16();
 
 class SuperCardChangeModePatchCode : public PatchCode
@@ -20,6 +20,11 @@ public:
     {
         return GetAddressAtTarget((void*)sccmn_changeMode);
     }
+
+    const void* GetSdSendClock10Function() const
+    {
+        return GetAddressAtTarget((void*)sccmn_sdSendClock10);
+    }
 };
 
 class SuperCardCommonPatchCode : public PatchCode
@@ -28,11 +33,6 @@ public:
     explicit SuperCardCommonPatchCode(PatchHeap& patchHeap)
         : PatchCode(SECTION_START(scsd_common), SECTION_SIZE(scsd_common), patchHeap) { }
 
-    const void* GetSdSendClock10Function() const
-    {
-        return GetAddressAtTarget((void*)sccmn_sdSendClock10);
-    }
-
     const void* GetCrc16ChecksumFunction() const
     {
         return GetAddressAtTarget((void*)sccmn_sdio4BitCrc16);

arm9/source/patches/platform/supercard/SuperCardCommon.s

@@ -8,6 +8,7 @@
 
 @ void sc_change_mode(uint16_t mode);
 BEGIN_THUMB_FUNCTION sccmn_changeMode
+	push {r0-r3,lr}
     ldr r2,= 0x09FFFFFE
     ldr r3,= 0xA55A
     strh r3, [r2]
@@ -22,21 +23,23 @@ BEGIN_THUMB_FUNCTION sccmn_changeMode
     ldr r2, =sd_resetaddr
     strh r2, [r2]
 1:
-    mov pc, lr
+    pop {r0-r3,pc}
 
-.section "scsd_common", "ax"
-@ this function will trash r4 but leave r0 and r1 untouched
 @ void SDSendClock10(void)
 BEGIN_THUMB_FUNCTION sccmn_sdSendClock10
+	@ here we get lr from the caller
+	push {r0-r3}
     movs r3, #0x10
     @ loads reg_scsd_cmd
     movs r2, #0x98
     lsls r2, r2, #20
 1:
-    ldrh r4, [r2]
+    ldrh r0, [r2]
     subs r3, r3, #1
     bne 1b
-    mov pc, lr
+    pop {r0-r3,pc}
+
+.section "scsd_common", "ax"
 
 @ static uint64_t inline calSingleCRC16(uint64_t crc, uint32_t data_in){
 @ 	// Shift out 8 bits for each line
@@ -73,8 +76,10 @@ BEGIN_THUMB_FUNCTION sccmn_sdSendClock10
 
 @ uint64_t sdio_crc16_4bit_checksum(void*)
 @ r0 and r1 are left untouched, the result is returned in r2-r3 instead
+@ a "continuation function" is provided in r5, and we have to jump to it
 BEGIN_THUMB_FUNCTION sccmn_sdio4BitCrc16
-    push {r0,r1,r4-r5,lr}
+	push {r5,lr}
+    push {r0,r1,r4-r5}
     movs r4, #0 @ r4 = crc_lo
     movs r5, #0 @ r5 = crc_hi
     movs r1, #128

arm9/source/patches/platform/supercard/sclite/SuperCardLiteImpl.h

@@ -83,7 +83,7 @@ public:
         INTERWORK_LABEL(sclite_sdCommandAndDropResponse6, readInterwork)
             = (u32)superCardSdCommandAndDropLitePatchCode->GetSdCommandAndDropResponse6Function();
         INTERWORK_LABEL(sclite_readData, readInterwork) = (u32)superCardReadDataLitePatchCode->GetReadDataLiteFunction();
-        INTERWORK_LABEL(sccmn_sdSendClock10, readInterwork) = (u32)superCardCommonPatchCode->GetSdSendClock10Function();
+        INTERWORK_LABEL(sccmn_sdSendClock10, readInterwork) = (u32)superCardChangeModePatchCode->GetSdSendClock10Function();
     }
 
     const ReadSectorsFunc GetReadSectorsFunction() const override
@@ -104,7 +104,7 @@ public:
     {
         INTERWORK_LABEL(sclite_writeData, writeInterwork) = (u32)superCardWriteDataLitePatchCode->GetWriteDataLiteFunction();
         INTERWORK_LABEL(sccmn_sdio4BitCrc16, writeInterwork) = (u32)superCardCommonPatchCode->GetCrc16ChecksumFunction();
-        INTERWORK_LABEL(sccmn_sdSendClock10, writeInterwork) = (u32)superCardCommonPatchCode->GetSdSendClock10Function();
+        INTERWORK_LABEL(sccmn_sdSendClock10, writeInterwork) = (u32)superCardChangeModePatchCode->GetSdSendClock10Function();
         INTERWORK_LABEL(sccmn_changeMode, writeInterwork) = (u32)superCardChangeModePatchCode->GetScChangeModeFunction();
         INTERWORK_LABEL(sclite_sdCommandAndDropResponse6, writeInterwork)
             = (u32)superCardSdCommandAndDropLitePatchCode->GetSdCommandAndDropResponse6Function();

arm9/source/patches/platform/supercard/sclite/SuperCardLiteImpl.s

@@ -40,12 +40,19 @@
 
 .section "sclite_sd_command_drop", "ax"
 @ void sclite_sdCommandAndDropResponse6(uint32_t dummy, uint32_t argument, SD_COMMANDS command)
-@ command is passed in r2
+@ command is passed in r2, in r6 is the pointer to sccmn_sdSendClock10
 BEGIN_THUMB_FUNCTION sclite_sdCommandAndDropResponse6
+	push {lr}
+	cmp r2, #0x52
+	@ if command is READ_MULTIPLE_BLOCK, don't send extra clock cycles
+	beq 1f
+	@ we need to call sccmn_sdSendClock10 after the command is sent, push sccmn_sdSendClock10 and the lr to the stack
+	push {r6}
+1:
     @ among the pushed registers there are the command and
     @ argument one, which are then used at the loop to
     @ send the sd command
-    push {r1-r7,lr}
+    push {r1-r7}
 
     @ loads reg_scsd_cmd
     movs r7, #0x98
@@ -100,8 +107,11 @@ SDCommand_drop_resp:
     subs r6, r6, #1
     bne SDCommand_drop_resp
 
-    @ restore stack space
+	@ we pop from the stack r6 as pc, which corresponds to sccmn_sdSendClock10 or lr, if it is sccmn_sdSendClock10 that function
+	@ is called, and that function doesn't push a new lr but pops a preexisting one from the stack, which in our case is the
+	@ lr provided to this function
     pop {r1-r7,pc}
+
 .balign 4
 .pool
 
@@ -110,7 +120,7 @@ SDCommand_drop_resp:
 BEGIN_THUMB_FUNCTION sclite_writeSector
     push {r1,r2,r4-r7,lr}
 
-    @ load EXMEMCNT register
+    @ load EXMEMCNT register into r7
     LOAD_EXMEMCNT
 
     @ r1 for now holds the sector
@@ -120,80 +130,96 @@ sclite_writeSectorSdhcLabel:
     lsls r1, r0, #9
     @ movs r1, r0
 
+	@ r4 sccmn_changeMode
+	@ r5 sclite_sdCommandAndDropResponse6
+	@ r6 sccmn_sdSendClock10
+
+	adr r3, sccmn_changeMode_writeInterworkLite_address
+	ldmia r3!, {r4-r6}
+
     @ enable sd access
-    @ this function won't touch r1
+    @ this function won't touch anything
     movs r0, #3
-    CALL sccmn_changeMode writeInterwork
+    @ call sccmn_changeMode
+	bl interwork_r4
 
     @ WRITE_MULTIPLE_BLOCK
+	@ puts the value in r2
     SD_COMMAND_ARGUMENT #25
     @ 2nd parameter is in r1 from above
 
-    CALL sclite_sdCommandAndDropResponse6 writeInterwork
+    @ call sclite_sdCommandAndDropResponse6
+	bl interwork_r5
 
-    @ this function will trash r4 but leave r0 and r1 untouched
+	@ load the rest of the functions
-    CALL sccmn_sdSendClock10 writeInterwork
+	@ r4 sccmn_sdio4BitCrc16
+	@ r5 sclite_writeData
+	ldmia r3!, {r4-r5}
 
-    @ loads the saved r0 and r2 (readnum)
+    @ loads the saved r1 (buff) and r2 (writenum)
     @ into r0 and r1
     pop {r0,r1}
+
 write_sector_loop:
-    @ all the functions called in this loop don't change the value of r0 and r1
+    @ all the functions called in this loop don't change the value of any register
     @ except sclite_writeData, which will increase r0 by 512
     @ sccmn_sdio4BitCrc16 will write the checksum to r2-r3
-    CALL sccmn_sdio4BitCrc16 writeInterwork
+    @ call sccmn_sdio4BitCrc16, this function will then tail call to r5 (sclite_writeData)
-
+	bl interwork_r4
-    @ first argument is buffer
-    @ regs r2 and r3 hold the crc
-    CALL sclite_writeData writeInterwork
-
-    @ this function will trash r4 but leave r0 and r1 untouched
-    CALL sccmn_sdSendClock10 writeInterwork
 
     subs r1, #1
     bne write_sector_loop
 
+	adr r3, sccmn_changeMode_writeInterworkLite_address
+	@ r4 sccmn_changeMode
+	@ r5 sclite_sdCommandAndDropResponse6
+	ldmia r3!, {r4-r5}
+
     @ STOP_TRANSMISSION
+	@ puts the value in r2
     SD_COMMAND_ARGUMENT #12
     @ 2nd parameter is passed in r1
     @ and from the loop above r1 is already 0
 
-    CALL sclite_sdCommandAndDropResponse6 writeInterwork
+	@ call sclite_sdCommandAndDropResponse6
-
+	bl interwork_r5
-    @ this function will trash r4 but leave r0 and r1 untouched
-    CALL sccmn_sdSendClock10 writeInterwork
 
     @ loads sd_dataadd
     movs r0, #0x90
     lsls r0, r0, #20
 
-    @ while(*r1 &0x100) == 0
+    @ while(*r0 &0x100) == 0
 beginwhile_WriteSector:
     ldrh r1, [r0]
     lsrs r1, #9
     bcc beginwhile_WriteSector
 
     movs r0, #1
-    CALL sccmn_changeMode writeInterwork
+    @ call sccmn_changeMode writeInterwork
+	bl interwork_r4
 
     @ restore EXMEMCNT register
     RESTORE_EXMEMCNT
 
     pop {r4-r7,pc}
-INTERWORK writeInterwork
+interwork_r4:
-INTERWORK_FUNCTION sclite_writeData writeInterwork
+	bx r4
-INTERWORK_FUNCTION sccmn_sdio4BitCrc16 writeInterwork
+interwork_r5:
-INTERWORK_FUNCTION sccmn_sdSendClock10 writeInterwork
+	bx r5
-INTERWORK_FUNCTION sccmn_changeMode writeInterwork
-INTERWORK_FUNCTION sclite_sdCommandAndDropResponse6 writeInterwork
-
 .balign 4
 .pool
+INTERWORK_FUNCTION sccmn_changeMode writeInterwork
+INTERWORK_FUNCTION sclite_sdCommandAndDropResponse6 writeInterwork
+INTERWORK_FUNCTION sccmn_sdSendClock10 writeInterwork
+INTERWORK_FUNCTION sccmn_sdio4BitCrc16 writeInterwork
+INTERWORK_FUNCTION sclite_writeData writeInterwork
 
 .section "sclite_read_sector", "ax"
 @ bool sclite_readSector(uint32_t sector, uint8_t *buff, uint32_t readnum)
 BEGIN_THUMB_FUNCTION sclite_readSector
     push {r1,r2-r7,lr}
+
+    @ load EXMEMCNT register into r7
     LOAD_EXMEMCNT
 
     @ r1 for now holds the sector
@@ -204,21 +230,22 @@ sclite_readSectorSdhcLabel:
     @ movs r1, r0
 
     @ enable sd access
-    @ this function won't touch r1
+    @ this function won't touch anything
     movs r0, #3
     CALL sccmn_changeMode readInterwork
 
     @ READ_MULTIPLE_BLOCK
+	@ puts the value in r2
     SD_COMMAND_ARGUMENT #18
     @ 2nd parameter is in r1 from above
 
     CALL sclite_sdCommandAndDropResponse6 readInterwork
 
-    @ loads the saved r0 and r2 (writenum)
+    @ loads the saved r1 (buff) and r2 (readnum)
     @ into r0 and r1
     pop {r0,r1}
 read_sector_loop:
-    @ all the functions called in this loop don't change the value of r0 or r1
+    @ all the functions called in this loop don't change the value of any register
     @ except sclite_readData, which will increase r0 by 512 automatically
     CALL sclite_readData readInterwork
 
@@ -226,14 +253,15 @@ read_sector_loop:
     bne read_sector_loop
 
     @ STOP_TRANSMISSION
+	@ puts the value in r2
     SD_COMMAND_ARGUMENT #12
     @ 2nd parameter is passed in r1
     @ and from the loop above r1 is already 0
-
+	LOAD_INTERWORK_FUNCTION sccmn_sdSendClock10 readInterwork r6
     CALL sclite_sdCommandAndDropResponse6 readInterwork
 
-    @ this function will trash r4 but leave r0 and r1 untouched
+    @ this function won't touch anything
-    CALL sccmn_sdSendClock10 readInterwork
+    @ CALL sccmn_sdSendClock10 readInterwork
 
 
     movs r0, #1
@@ -312,15 +340,17 @@ sclite_readData_loop:
 
 .section "sclite_write_data", "ax"
 @ void sclite_writeData(void*& buffer, int value_to_keep, int crc_buff1, int crc_buff2)
-@ this function updates r0, leaves r1 untouched and thrashes r4
+@ this function updates r0 and leaves everything else untouched
+@ in r6 is the pointer to sccmn_sdSendClock10
 BEGIN_THUMB_FUNCTION sclite_writeData
+	@ push sccmn_sdSendClock10 and the current lr reg to the stack, so that at the end we can
+	@ call in sequence sccmn_sdSendClock10 and then return
+	push {r6}
+    push {r1-r7}
     @ loads sd_dataadd
     movs r4, #0x90
     lsls r4, r4, #20
 
-    @ save sd_dataadd constant above for later
-    push {r1-r7,lr}
-
 waitOnWriteFalse_WriteData:
     ldrh r1, [r4]
     lsrs r1, #9
@@ -356,9 +386,12 @@ sclite_writeData_loop:
 
     @ r1 holds a value that has not to be changed
     @ r2-r3 hold the crc value to be written
-    @ r4 holds the value sd_dataadd saved at the start
+    pop {r1-r3}
-    pop {r1-r4}
+
     stmia r7!, {r2-r3}
+    @ loads sd_dataadd
+    movs r4, #0x90
+    lsls r4, r4, #20
 
     @ write end bit
     movs r3, #0xff
@@ -374,7 +407,11 @@ waitOnWriteTrue_WriteData:
     str r3, [r4]
     str r3, [r4]
 
-    pop {r5-r7,pc}
+
+	@ we pop from the stack r6 as pc, which corresponds to sccmn_sdSendClock10 so we jump to it
+	@ sccmn_sdSendClock10 doesn't push a new lr but pops a preexisting one from the stack, which in our case is the
+	@ lr provided to this function
+    pop {r4-r7,pc}
 
 .balign 4
 .pool

arm9/source/patches/platform/supercard/scsd/SuperCardSDImpl.h

@@ -81,7 +81,7 @@ public:
     {
         INTERWORK_LABEL(scsd_writeData, writeInterwork) = (u32)superCardWriteDataPatchCode->GetWriteDataFunction();
         INTERWORK_LABEL(sccmn_sdio4BitCrc16, writeInterwork) = (u32)superCardCommonPatchCode->GetCrc16ChecksumFunction();
-        INTERWORK_LABEL(sccmn_sdSendClock10, writeInterwork) = (u32)superCardCommonPatchCode->GetSdSendClock10Function();
+        INTERWORK_LABEL(sccmn_sdSendClock10, writeInterwork) = (u32)superCardChangeModePatchCode->GetSdSendClock10Function();
         INTERWORK_LABEL(sccmn_changeMode, writeInterwork) = (u32)superCardChangeModePatchCode->GetScChangeModeFunction();
         INTERWORK_LABEL(scsd_sdCommandAndDropResponse6, writeInterwork)
             = (u32)superCardSdCommandAndDropPatchCode->GetSdCommandAndDropResponse6Function();
@@ -107,7 +107,7 @@ public:
         INTERWORK_LABEL(scsd_sdCommandAndDropResponse6, readInterwork)
             = (u32)superCardSdCommandAndDropPatchCode->GetSdCommandAndDropResponse6Function();
         INTERWORK_LABEL(scsd_readData, readInterwork) = (u32)superCardReadDataPatchCode->GetReadDataFunction();
-        INTERWORK_LABEL(sccmn_sdSendClock10, readInterwork) = (u32)superCardCommonPatchCode->GetSdSendClock10Function();
+        INTERWORK_LABEL(sccmn_sdSendClock10, readInterwork) = (u32)superCardChangeModePatchCode->GetSdSendClock10Function();
     }
 
     const ReadSectorsFunc GetReadSectorsFunction() const override

arm9/source/patches/platform/supercard/scsd/SuperCardSDImpl.s

@@ -23,9 +23,9 @@
 
 .macro LOAD_FAST_EXMEMCNT
     @ loads EXMEMCNT register address
-    movs r6, #0x18
+    movs r0, #0x18
     @ waitstate 2,1 and arm9 slot2 access
-    strb r6, [r7, #4]
+    strb r0, [r7, #4]
 .endm
 
 .macro RESTORE_EXMEMCNT
@@ -43,11 +43,18 @@
 @ void scsd_sdCommandAndDropResponse6(uint32_t dummy, uint32_t argument, SD_COMMANDS command)
 @ argument is passed in r1
 BEGIN_THUMB_FUNCTION scsd_sdCommandAndDropResponse6
+	push {lr}
+	cmp r2, #0x52
+	@ if command is READ_MULTIPLE_BLOCK, don't send extra clock cycles
+	beq 1f
+	@ we need to call sccmn_sdSendClock10 after the command is sent, push sccmn_sdSendClock10 and the lr to the stack
+	push {r6}
+1:
     @ among the pushed registers there are the command and
     @ argument one, which are then used by the crc7 function
     @ and in the loop at the bottom
     @ also allocate an extra 4 bytes for the SD_CRC7 function to put the crc byte
-    push {r0-r7,lr}
+    push {r0-r7}
 
     @ pass the buffer incremented by 4, so that the crc function
     @ can use a proper indexing method
@@ -99,7 +106,9 @@ SDCommand_drop_resp:
     subs r6, r6, #1
     bne SDCommand_drop_resp
 
-    @ restore stack space
+	@ we pop from the stack r6 as pc, which corresponds to sccmn_sdSendClock10 or lr, if it is sccmn_sdSendClock10 that function
+	@ is called, and that function doesn't push a new lr but pops a preexisting one from the stack, which in our case is the
+	@ lr provided to this function
     pop {r0-r7,pc}
 
 @ inline uint8_t CRC7_one(uint8_t crcIn, uint8_t data) {
@@ -169,7 +178,7 @@ skip_xor:
 BEGIN_THUMB_FUNCTION scsd_writeSector
     push {r1,r2,r4-r7,lr}
 
-    @ load EXMEMCNT register
+    @ load EXMEMCNT register into r7
     LOAD_SLOW_EXMEMCNT
 
     @ r1 for now holds the sector
@@ -179,23 +188,34 @@ scsd_writeSectorSdhcLabel:
     lsls r1, r0, #9
     @ movs r1, r0
 
+	@ r4 sccmn_changeMode
+	@ r5 scsd_sdCommandAndDropResponse6
+	@ r6 sccmn_sdSendClock10
+
+	adr r3, sccmn_changeMode_writeInterwork_address
+	ldmia r3!, {r4-r6}
+
     @ enable sd access
-    @ this function won't touch r1
+    @ this function won't touch anything
     movs r0, #3
-    bl scsd_sccmn_changeMode_call
+    @ call sccmn_changeMode
+	bl interwork_r4
 
     @ WRITE_MULTIPLE_BLOCK
     SD_COMMAND_ARGUMENT #25
     @ 2nd parameter is in r1 from above
 
-    bl scsd_sdCommandAndDropResponse6_call
+    @ call scsd_sdCommandAndDropResponse6
-
+	bl interwork_r5
-    @ this function will trash r4 but leave r0 and r1 untouched
-    bl scsd_sccmn_sdSendClock10_call
 
     LOAD_FAST_EXMEMCNT
 
-    @ loads the saved r0 and r2 (readnum)
+	@ load the rest of the functions
+	@ r4 sccmn_sdio4BitCrc16
+	@ r5 scsd_writeData
+	ldmia r3!, {r4-r5}
+
+    @ loads the saved r1 (buff) and r2 (writenum)
     @ into r0 and r1
     pop {r0,r1}
 
@@ -203,18 +223,17 @@ write_sector_loop:
     @ all the functions called in this loop don't change the value of r0 and r1
     @ except scsd_writeData, which will increase r0 by 512
     @ sccmn_sdio4BitCrc16 will write the checksum to r2-r3
-    CALL sccmn_sdio4BitCrc16 writeInterwork
+    @ call sccmn_sdio4BitCrc16, this function will then tail call to r5 (scsd_writeData)
-
+	bl interwork_r4
-    @ first argument is buffer
-    @ regs r2 and r3 hold the crc
-    CALL scsd_writeData writeInterwork
-
-    @ this function will trash r4 but leave r0 and r1 untouched
-    bl scsd_sccmn_sdSendClock10_call
 
     subs r1, #1
     bne write_sector_loop
 
+	adr r3, sccmn_changeMode_writeInterwork_address
+	@ r4 sccmn_changeMode
+	@ r5 scsd_sdCommandAndDropResponse6
+	ldmia r3!, {r4-r5}
+
     RELOAD_SLOW_EXMEMCNT
 
     @ STOP_TRANSMISSION
@@ -222,39 +241,38 @@ write_sector_loop:
     @ 2nd parameter is passed in r1
     @ and from the loop above r1 is already 0
 
-    bl scsd_sdCommandAndDropResponse6_call
+	@ call scsd_sdCommandAndDropResponse6
+	bl interwork_r5
 
-    @ this function will trash r4 but leave r0 and r1 untouched
+    @ loads sd_dataadd
-    bl scsd_sccmn_sdSendClock10_call
+    movs r0, #0x90
+    lsls r0, r0, #20
 
-    @ SD_DATAADD is loaded by scsd_writeData into r5
+    @ while(*r0 &0x100) == 0
-    @ while(*r5 &0x100) == 0
 beginwhile_WriteSector:
-    ldrh r1, [r5]
+    ldrh r1, [r0]
     lsrs r1, #9
     bcc beginwhile_WriteSector
 
     movs r0, #1
-    bl scsd_sccmn_changeMode_call
+    @ call sccmn_changeMode writeInterwork
+	bl interwork_r4
 
     @ restore EXMEMCNT register
     RESTORE_EXMEMCNT
 
     pop {r4-r7,pc}
-scsd_sccmn_sdSendClock10_call:
+interwork_r4:
-	LOAD_INTERWORK_FUNCTION sccmn_sdSendClock10 writeInterwork r4
 	bx r4
-scsd_sdCommandAndDropResponse6_call:
+interwork_r5:
-	LOAD_INTERWORK_FUNCTION scsd_sdCommandAndDropResponse6 writeInterwork r4
+	bx r5
-	bx r4
+.balign 4
-scsd_sccmn_changeMode_call:
+.pool
-	LOAD_INTERWORK_FUNCTION sccmn_changeMode writeInterwork r4
-INTERWORK writeInterwork
-INTERWORK_FUNCTION scsd_writeData writeInterwork
-INTERWORK_FUNCTION sccmn_sdio4BitCrc16 writeInterwork
-INTERWORK_FUNCTION sccmn_sdSendClock10 writeInterwork
 INTERWORK_FUNCTION sccmn_changeMode writeInterwork
 INTERWORK_FUNCTION scsd_sdCommandAndDropResponse6 writeInterwork
+INTERWORK_FUNCTION sccmn_sdSendClock10 writeInterwork
+INTERWORK_FUNCTION sccmn_sdio4BitCrc16 writeInterwork
+INTERWORK_FUNCTION scsd_writeData writeInterwork
 
 
 .section "scsd_read_sector", "ax"
@@ -283,7 +301,7 @@ scsd_readSectorSdhcLabel:
 
     LOAD_FAST_EXMEMCNT
 
-    @ loads the saved r0 and r2 (writenum)
+    @ loads the saved r1 (buff) and r2 (readnum)
     @ into r0 and r1
     pop {r0,r1}
 read_sector_loop:
@@ -300,13 +318,9 @@ read_sector_loop:
     SD_COMMAND_ARGUMENT #12
     @ 2nd parameter is passed in r1
     @ and from the loop above r1 is already 0
-
+	LOAD_INTERWORK_FUNCTION sccmn_sdSendClock10 readInterwork r6
     CALL scsd_sdCommandAndDropResponse6 readInterwork
 
-    @ this function will trash r4 but leave r0 and r1 untouched
-    CALL sccmn_sdSendClock10 readInterwork
-
-
     movs r0, #1
     CALL sccmn_changeMode readInterwork
 
@@ -357,14 +371,15 @@ write32_loop:
 
 @ void scsd_writeData(void* buffer, int value_to_keep, int crc_buff1, int crc_buff2)
 BEGIN_THUMB_FUNCTION scsd_writeData
+	@ push sccmn_sdSendClock10 and the current lr reg to the stack, so that at the end we can
+	@ call in sequence sccmn_sdSendClock10 and then return
+	push {r6}
+    push {r1-r7}
 
-    @ loads SD_DATAADD, do it before the push so that
+    @ loads SD_DATAADD
-	@ it becomes available to the callee afterwards
     movs r5, #0x90
     lsls r5, r5, #20
 
-    push {r1-r7,lr}
-
     @ while(*r5 &0x100) == 0
 scsd_writeData_waitOnWriteFalse:
     ldrh r6, [r5]
@@ -384,6 +399,7 @@ scsd_writeData_waitOnWriteFalse:
     lsls r1, r1, #2
     @ no need for special handling because those 2 functions will be in the same block
     bl SCSD_writeBuffer32
+	@ save incremented r0 register
     push {r0}
 
     movs r1, #8
@@ -403,6 +419,11 @@ scsd_writeData_waitOnWriteTrue:
     lsrs r6, #9
     bcs scsd_writeData_waitOnWriteTrue
 
+
+	@ we pop from the stack r6 as pc, which corresponds to sccmn_sdSendClock10 so we jump to it
+	@ sccmn_sdSendClock10 doesn't push a new lr but pops a preexisting one from the stack, which in our case is the
+	@ lr provided to this function
+	@ also we load in r0 the value stored above
     pop {r0-r7,pc}
 
 .balign 4