summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
Diffstat
-rw-r--r--src/mesa/tnl/t_vtx_x86_gcc.S200
1 files changed, 100 insertions, 100 deletions
diff --git a/src/mesa/tnl/t_vtx_x86_gcc.S b/src/mesa/tnl/t_vtx_x86_gcc.S
index 1d8682d5e9..bad87d3ee9 100644
--- a/src/mesa/tnl/t_vtx_x86_gcc.S
+++ b/src/mesa/tnl/t_vtx_x86_gcc.S
@@ -49,7 +49,7 @@ _##x:
// Someone who knew a lot about this sort of thing would use this
// macro to note current offsets, etc in a special region of the
-// object file & just make everything work out neat. I don't know
+// object file & just make everything work out neat. I do not know
// enough to do that...
#define SUBST( x ) (0x10101010 + x)
@@ -67,27 +67,27 @@ GLOBL ( _tnl_x86_Vertex1fv )
movl 4(%esp), %ecx
push %edi
push %esi
- movl SUBST(0), %edi // 0x0 --> tnl->vtx.vbptr
- movl (%ecx), %edx // load v[0]
- movl %edx, (%edi) // tnl->vtx.vbptr[0] = v[0]
- addl $4, %edi // tnl->vtx.vbptr += 1
- movl $SUBST(1), %ecx // 0x1 --> (tnl->vtx.vertex_size - 1)
- movl $SUBST(2), %esi // 0x2 --> (tnl->vtx.vertex + 1)
+ movl SUBST(0), %edi # 0x0 --> tnl->vtx.vbptr
+ movl (%ecx), %edx # load v[0]
+ movl %edx, (%edi) # tnl->vtx.vbptr[0] = v[0]
+ addl $4, %edi # tnl->vtx.vbptr += 1
+ movl $SUBST(1), %ecx # 0x1 --> (tnl->vtx.vertex_size - 1)
+ movl $SUBST(2), %esi # 0x2 --> (tnl->vtx.vertex + 1)
repz
movsl %ds:(%esi), %es:(%edi)
- movl %edi, SUBST(0) // 0x0 --> tnl->vtx.vbptr
- movl SUBST(3), %edx // 0x3 --> counter
+ movl %edi, SUBST(0) # 0x0 --> tnl->vtx.vbptr
+ movl SUBST(3), %edx # 0x3 --> counter
pop %esi
pop %edi
- dec %edx // counter--
- movl %edx, SUBST(3) // 0x3 --> counter
- jne .0 // if (counter != 0) return
- pushl $SUBST(4) // 0x4 --> ctx
- .byte 0xe8 // call ...
- .long SUBST(5) // ... _tnl_wrap_filled_vertex(ctx)
+ dec %edx # counter--
+ movl %edx, SUBST(3) # 0x3 --> counter
+ jne .0 # if (counter != 0) return
+ pushl $SUBST(4) # 0x4 --> ctx
+ .byte 0xe8 # call ...
+ .long SUBST(5) # ... _tnl_wrap_filled_vertex(ctx)
pop %eax
.0:
- ret // return
+ ret # return
GLOBL ( _tnl_x86_Vertex1fv_end )
@@ -96,29 +96,29 @@ GLOBL ( _tnl_x86_Vertex2fv )
movl 4(%esp), %ecx
push %edi
push %esi
- movl SUBST(0), %edi // load tnl->vtx.vbptr
- movl (%ecx), %edx // load v[0]
- movl 4(%ecx), %eax // load v[1]
- movl %edx, (%edi) // tnl->vtx.vbptr[0] = v[0]
- movl %eax, 4(%edi) // tnl->vtx.vbptr[1] = v[1]
- addl $8, %edi // tnl->vtx.vbptr += 2
- movl $SUBST(1), %ecx // vertex_size - 2
- movl $SUBST(2), %esi // tnl->vtx.vertex + 2
+ movl SUBST(0), %edi # load tnl->vtx.vbptr
+ movl (%ecx), %edx # load v[0]
+ movl 4(%ecx), %eax # load v[1]
+ movl %edx, (%edi) # tnl->vtx.vbptr[0] = v[0]
+ movl %eax, 4(%edi) # tnl->vtx.vbptr[1] = v[1]
+ addl $8, %edi # tnl->vtx.vbptr += 2
+ movl $SUBST(1), %ecx # vertex_size - 2
+ movl $SUBST(2), %esi # tnl->vtx.vertex + 2
repz
movsl %ds:(%esi), %es:(%edi)
- movl %edi, SUBST(0) // save tnl->vtx.vbptr
- movl SUBST(3), %edx // load counter
+ movl %edi, SUBST(0) # save tnl->vtx.vbptr
+ movl SUBST(3), %edx # load counter
pop %esi
pop %edi
- dec %edx // counter--
- movl %edx, SUBST(3) // save counter
- jne .1 // if (counter != 0) return
- pushl $SUBST(4) // load ctx
- .byte 0xe8 // call ...
- .long SUBST(5) // ... _tnl_wrap_filled_vertex(ctx)
+ dec %edx # counter--
+ movl %edx, SUBST(3) # save counter
+ jne .1 # if (counter != 0) return
+ pushl $SUBST(4) # load ctx
+ .byte 0xe8 # call ...
+ .long SUBST(5) # ... _tnl_wrap_filled_vertex(ctx)
pop %eax
.1:
- ret // return
+ ret # return
GLOBL ( _tnl_x86_Vertex2fv_end )
.align 4
@@ -126,31 +126,31 @@ GLOBL ( _tnl_x86_Vertex3fv )
movl 4(%esp), %ecx
push %edi
push %esi
- movl SUBST(0), %edi // load tnl->vtx.vbptr
- movl (%ecx), %edx // load v[0]
- movl 4(%ecx), %eax // load v[1]
- movl 8(%ecx), %esi // load v[2]
- movl %edx, (%edi) // tnl->vtx.vbptr[0] = v[0]
- movl %eax, 4(%edi) // tnl->vtx.vbptr[1] = v[1]
- movl %esi, 8(%edi) // tnl->vtx.vbptr[2] = v[2]
- addl $12, %edi // tnl->vtx.vbptr += 3
- movl $SUBST(1), %ecx // vertex_size - 3
- movl $SUBST(2), %esi // tnl->vtx.vertex + 3
+ movl SUBST(0), %edi # load tnl->vtx.vbptr
+ movl (%ecx), %edx # load v[0]
+ movl 4(%ecx), %eax # load v[1]
+ movl 8(%ecx), %esi # load v[2]
+ movl %edx, (%edi) # tnl->vtx.vbptr[0] = v[0]
+ movl %eax, 4(%edi) # tnl->vtx.vbptr[1] = v[1]
+ movl %esi, 8(%edi) # tnl->vtx.vbptr[2] = v[2]
+ addl $12, %edi # tnl->vtx.vbptr += 3
+ movl $SUBST(1), %ecx # vertex_size - 3
+ movl $SUBST(2), %esi # tnl->vtx.vertex + 3
repz
movsl %ds:(%esi), %es:(%edi)
- movl %edi, SUBST(0) // save tnl->vtx.vbptr
- movl SUBST(3), %edx // load counter
+ movl %edi, SUBST(0) # save tnl->vtx.vbptr
+ movl SUBST(3), %edx # load counter
pop %esi
pop %edi
- dec %edx // counter--
- movl %edx, SUBST(3) // save counter
- jne .2 // if (counter != 0) return
- pushl $SUBST(4) // load ctx
- .byte 0xe8 // call ...
- .long SUBST(5) // ... _tnl_wrap_filled_vertex(ctx)
+ dec %edx # counter--
+ movl %edx, SUBST(3) # save counter
+ jne .2 # if (counter != 0) return
+ pushl $SUBST(4) # load ctx
+ .byte 0xe8 # call ...
+ .long SUBST(5) # ... _tnl_wrap_filled_vertex(ctx)
pop %eax
.2:
- ret // return
+ ret # return
GLOBL ( _tnl_x86_Vertex3fv_end )
@@ -159,33 +159,33 @@ GLOBL ( _tnl_x86_Vertex4fv )
movl 4(%esp), %ecx
push %edi
push %esi
- movl SUBST(0), %edi // load tnl->vtx.vbptr
- movl (%ecx), %edx // load v[0]
- movl 4(%ecx), %eax // load v[1]
- movl 8(%ecx), %esi // load v[2]
- movl 12(%ecx), %ecx // load v[3]
- movl %edx, (%edi) // tnl->vtx.vbptr[0] = v[0]
- movl %eax, 4(%edi) // tnl->vtx.vbptr[1] = v[1]
- movl %esi, 8(%edi) // tnl->vtx.vbptr[2] = v[2]
- movl %ecx, 12(%edi) // tnl->vtx.vbptr[3] = v[3]
- addl $16, %edi // tnl->vtx.vbptr += 4
- movl $SUBST(1), %ecx // vertex_size - 4
- movl $SUBST(2), %esi // tnl->vtx.vertex + 3
+ movl SUBST(0), %edi # load tnl->vtx.vbptr
+ movl (%ecx), %edx # load v[0]
+ movl 4(%ecx), %eax # load v[1]
+ movl 8(%ecx), %esi # load v[2]
+ movl 12(%ecx), %ecx # load v[3]
+ movl %edx, (%edi) # tnl->vtx.vbptr[0] = v[0]
+ movl %eax, 4(%edi) # tnl->vtx.vbptr[1] = v[1]
+ movl %esi, 8(%edi) # tnl->vtx.vbptr[2] = v[2]
+ movl %ecx, 12(%edi) # tnl->vtx.vbptr[3] = v[3]
+ addl $16, %edi # tnl->vtx.vbptr += 4
+ movl $SUBST(1), %ecx # vertex_size - 4
+ movl $SUBST(2), %esi # tnl->vtx.vertex + 3
repz
movsl %ds:(%esi), %es:(%edi)
- movl %edi, SUBST(0) // save tnl->vtx.vbptr
- movl SUBST(3), %edx // load counter
+ movl %edi, SUBST(0) # save tnl->vtx.vbptr
+ movl SUBST(3), %edx # load counter
pop %esi
pop %edi
- dec %edx // counter--
- movl %edx, SUBST(3) // save counter
- jne .3 // if (counter != 0) return
- pushl $SUBST(4) // load ctx
- .byte 0xe8 // call ...
- .long SUBST(5) // ... _tnl_wrap_filled_vertex(ctx)
+ dec %edx # counter--
+ movl %edx, SUBST(3) # save counter
+ jne .3 # if (counter != 0) return
+ pushl $SUBST(4) # load ctx
+ .byte 0xe8 # call ...
+ .long SUBST(5) # ... _tnl_wrap_filled_vertex(ctx)
pop %eax
.3:
- ret // return
+ ret # return
GLOBL ( _tnl_x86_Vertex4fv_end )
@@ -243,13 +243,13 @@ GLOBL ( _tnl_x86_Attribute4fv_end )
GLOBL( _tnl_x86_choose_fv)
- subl $12, %esp // gcc does 16 byte alignment of stack frames?
- movl $SUBST(0), (%esp) // arg 0 - attrib
- movl $SUBST(1), 4(%esp) // arg 1 - N
- .byte 0xe8 // call ...
- .long SUBST(2) // ... do_choose
- add $12, %esp // tear down stack frame
- jmp *%eax // jump to new func
+ subl $12, %esp # gcc does 16 byte alignment of stack frames?
+ movl $SUBST(0), (%esp) # arg 0 - attrib
+ movl $SUBST(1), 4(%esp) # arg 1 - N
+ .byte 0xe8 # call ...
+ .long SUBST(2) # ... do_choose
+ add $12, %esp # tear down stack frame
+ jmp *%eax # jump to new func
GLOBL ( _tnl_x86_choose_fv_end )
@@ -261,7 +261,7 @@ GLOBL ( _tnl_x86_choose_fv_end )
// In the 1st level dispatch functions, switch to a different
// calling convention -- (const GLfloat *v) in %ecx.
//
-// As with regular (x86) dispatch, don't create a new stack frame -
+// As with regular (x86) dispatch, do not create a new stack frame -
// just let the 'ret' in the dispatched function return straight
// back to the original caller.
@@ -274,18 +274,18 @@ GLOBL ( _tnl_x86_choose_fv_end )
// Unfortunately, have to play with the stack in the non-fv case:
//
GLOBL( _tnl_x86_dispatch_attrf )
- subl $12, %esp // gcc does 16 byte alignment of stack frames?
- leal 16(%esp), %edx // address of first float on stack
- movl %edx, (%esp) // save as 'v'
- call *SUBST(0) // 0x0 --> tabfv[attr][n]
- addl $12, %esp // tear down frame
- ret // return
+ subl $12, %esp # gcc does 16 byte alignment of stack frames?
+ leal 16(%esp), %edx # address of first float on stack
+ movl %edx, (%esp) # save as 'v'
+ call *SUBST(0) # 0x0 --> tabfv[attr][n]
+ addl $12, %esp # tear down frame
+ ret # return
GLOBL( _tnl_x86_dispatch_attrf_end )
// The fv case is simpler:
//
GLOBL( _tnl_x86_dispatch_attrfv )
- jmp *SUBST(0) // 0x0 --> tabfv[attr][n]
+ jmp *SUBST(0) # 0x0 --> tabfv[attr][n]
GLOBL( _tnl_x86_dispatch_attrfv_end )
@@ -296,9 +296,9 @@ GLOBL( _tnl_x86_dispatch_attrfv_end )
// [dBorca]
// right, this would be the preferred approach, but gcc does not
// clean up the stack after each function call when optimizing (-fdefer-pop);
-// can it make assumptions about what's already on the stack? I dunno,
+// can it make assumptions about what is already on the stack? I dunno,
// but in this case, we can't mess with the caller's stack frame, and
-// we must use a model like `_x86_dispatch_attrfv' above. Caveat emptor!
+// we must use a model like '_x86_dispatch_attrfv' above. Caveat emptor!
// Also, will only need a maximum of four of each of these per context:
//
@@ -308,7 +308,7 @@ GLOBL( _tnl_x86_dispatch_multitexcoordf )
andl $7, %ecx
movl %edx, 4(%esp)
sall $4, %ecx
- jmp *SUBST(0)(%ecx) // 0x0 - tabfv[tex0][n]
+ jmp *SUBST(0)(%ecx) # 0x0 - tabfv[tex0][n]
GLOBL( _tnl_x86_dispatch_multitexcoordf_end )
GLOBL( _tnl_x86_dispatch_multitexcoordfv )
@@ -317,7 +317,7 @@ GLOBL( _tnl_x86_dispatch_multitexcoordfv )
andl $7, %ecx
movl %edx, 4(%esp)
sall $4, %ecx
- jmp *SUBST(0)(%ecx) // 0x0 - tabfv[tex0][n]
+ jmp *SUBST(0)(%ecx) # 0x0 - tabfv[tex0][n]
GLOBL( _tnl_x86_dispatch_multitexcoordfv_end )
// VertexAttrib: the address of the function pointer must be
@@ -326,23 +326,23 @@ GLOBL( _tnl_x86_dispatch_multitexcoordfv_end )
GLOBL( _tnl_x86_dispatch_vertexattribf )
movl 4(%esp), %eax
cmpl $16, %eax
- jb .8 // "cmovge" is not supported on all CPUs
+ jb .8 # "cmovge" is not supported on all CPUs
movl $16, %eax
.8:
- leal 8(%esp), %ecx // calculate 'v'
- movl %ecx, 4(%esp) // save in 1st arg slot
+ leal 8(%esp), %ecx # calculate 'v'
+ movl %ecx, 4(%esp) # save in 1st arg slot
sall $4, %eax
- jmp *SUBST(0)(%eax) // 0x0 - tabfv[0][n]
+ jmp *SUBST(0)(%eax) # 0x0 - tabfv[0][n]
GLOBL( _tnl_x86_dispatch_vertexattribf_end )
GLOBL( _tnl_x86_dispatch_vertexattribfv )
movl 4(%esp), %eax
cmpl $16, %eax
- jb .9 // "cmovge" is not supported on all CPUs
+ jb .9 # "cmovge" is not supported on all CPUs
movl $16, %eax
.9:
- movl 8(%esp), %ecx // load 'v'
- movl %ecx, 4(%esp) // save in 1st arg slot
+ movl 8(%esp), %ecx # load 'v'
+ movl %ecx, 4(%esp) # save in 1st arg slot
sall $4, %eax
- jmp *SUBST(0)(%eax) // 0x0 - tabfv[0][n]
+ jmp *SUBST(0)(%eax) # 0x0 - tabfv[0][n]
GLOBL( _tnl_x86_dispatch_vertexattribfv_end )
generated by cgit v1.2.3 (git 2.51.0) at 2025-10-31 20:42:04 +0000