April 2024 - Wine-commits - list.winehq.org

Alexandre Julliard : widl: Output register parameter assignments on ARM platforms.
by Alexandre Julliard 16 Apr '24

16 Apr '24

Module: wine Branch: master Commit: 235f76594e1aec4b37014ff2e37b5a487b1df33e URL: https://gitlab.winehq.org/wine/wine/-/commit/235f76594e1aec4b37014ff2e37b5a… Author: Alexandre Julliard <julliard(a)winehq.org> Date: Tue Apr 16 18:09:52 2024 +0200 widl: Output register parameter assignments on ARM platforms. --- tools/widl/typegen.c | 177 ++++++++++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 169 insertions(+), 8 deletions(-) diff --git a/tools/widl/typegen.c b/tools/widl/typegen.c index 063080bd9de..6f362bc1d8d 100644 --- a/tools/widl/typegen.c +++ b/tools/widl/typegen.c @@ -1372,6 +1372,129 @@ int is_interpreted_func( const type_t *iface, const var_t *func ) return interpreted_mode; } +/* replace consecutive params code by a repeat sequence: 0x9d code<1> repeat_count<2> */ +static unsigned int compress_params_array( unsigned char *params, unsigned int count ) +{ + unsigned int i, j; + + for (i = 0; i + 4 <= count; i++) + { + for (j = 1; i + j < count; j++) if (params[i + j] != params[i]) break; + if (j < 4) continue; + params[i] = 0x9d; + params[i + 2] = j & 0xff; + params[i + 3] = j >> 8; + memmove( params + i + 4, params + i + j, count - (i + j) ); + count -= j - 4; + i += 3; + } + return count; +} + +/* fill the parameters array for the procedure extra data on ARM platforms */ +static unsigned int fill_params_array( const type_t *iface, const var_t *func, + unsigned char *params, unsigned int count ) +{ + unsigned int reg_count = 0, float_count = 0, double_count = 0, stack_pos = 0, offset = 0; + var_list_t *args = type_function_get_args( func->declspec.type ); + enum type_basic_type type; + unsigned int size, pos, align; + var_t *var; + + memset( params, 0x9f /* padding */, count ); + + if (is_object( iface )) + { + params[0] = 0x80 + reg_count++; + offset += pointer_size; + } + + if (args) LIST_FOR_EACH_ENTRY( var, args, var_t, entry ) + { + type = TYPE_BASIC_LONG; + if (type_get_type( var->declspec.type ) == TYPE_BASIC) + type = type_basic_get_type( var->declspec.type ); + + size = get_stack_size( var, &align, NULL ); + offset = ROUND_SIZE( offset, align ); + pos = offset / pointer_size; + + if (target.cpu == CPU_ARM64) + { + switch (type) + { + case TYPE_BASIC_FLOAT: + case TYPE_BASIC_DOUBLE: + if (double_count >= 8) break; + params[pos] = 0x88 + double_count++; + offset += size; + continue; + + default: + reg_count = ROUND_SIZE( reg_count, align / pointer_size ); + if (reg_count > 8 - size / pointer_size) break; + while (size) + { + params[pos++] = 0x80 + reg_count++; + offset += pointer_size; + size -= pointer_size; + } + continue; + } + } + else /* CPU_ARM */ + { + switch (type) + { + case TYPE_BASIC_FLOAT: + if (!(float_count % 2)) float_count = max( float_count, double_count * 2 ); + if (float_count >= 16) + { + stack_pos = ROUND_SIZE( stack_pos, align ); + params[pos] = 0x100 - (offset - stack_pos) / pointer_size; + stack_pos += size; + } + else + { + params[pos] = 0x84 + float_count++; + } + offset += size; + continue; + + case TYPE_BASIC_DOUBLE: + double_count = max( double_count, (float_count + 1) / 2 ); + if (double_count >= 8) break; + params[pos] = 0x84 + 2 * double_count; + params[pos + 1] = 0x84 + 2 * double_count + 1; + double_count++; + offset += size; + continue; + + default: + reg_count = ROUND_SIZE( reg_count, align / pointer_size ); + if (reg_count <= 4 - size / pointer_size || !stack_pos) + { + while (size && reg_count < 4) + { + params[pos++] = 0x80 + reg_count++; + offset += pointer_size; + size -= pointer_size; + } + } + break; + } + } + + stack_pos = ROUND_SIZE( stack_pos, align ); + memset( params + pos, 0x100 - (offset - stack_pos) / pointer_size, size / pointer_size ); + stack_pos += size; + offset += size; + } + + while (count && params[count - 1] == 0x9f) count--; + return count; +} + static void write_proc_func_interp( FILE *file, int indent, const type_t *iface, const var_t *func, unsigned int *offset, unsigned short num_proc ) @@ -1390,6 +1513,7 @@ static void write_proc_func_interp( FILE *file, int indent, const type_t *iface, unsigned int stack_size = 0; unsigned int stack_offset = 0; unsigned int stack_align; + unsigned int extra_size = 0; unsigned short param_num = 0; unsigned short handle_stack_offset = 0; unsigned short handle_param_num = 0; @@ -1477,16 +1601,21 @@ static void write_proc_func_interp( FILE *file, int indent, const type_t *iface, print_file( file, indent, "NdrFcShort(0x%x),\t/* server buffer = %u */\n", size, size ); print_file( file, indent, "0x%02x,\n", oi2_flags ); print_file( file, indent, "0x%02x,\t/* %u params */\n", nb_args, nb_args ); - print_file( file, indent, "0x%02x,\n", pointer_size == 8 ? 10 : 8 ); - print_file( file, indent, "0x%02x,\n", ext_flags ); - print_file( file, indent, "NdrFcShort(0x0),\n" ); /* server corr hint */ - print_file( file, indent, "NdrFcShort(0x0),\n" ); /* client corr hint */ - print_file( file, indent, "NdrFcShort(0x0),\n" ); /* FIXME: notify index */ - *offset += 14; - if (pointer_size == 8) + *offset += 6; + extra_size = 8; + + switch (target.cpu) + { + case CPU_x86_64: { unsigned short pos = 0, fpu_mask = 0; + extra_size += 2; + print_file( file, indent, "0x%02x,\n", extra_size ); + print_file( file, indent, "0x%02x,\n", ext_flags ); + print_file( file, indent, "NdrFcShort(0x0),\n" ); /* server corr hint */ + print_file( file, indent, "NdrFcShort(0x0),\n" ); /* client corr hint */ + print_file( file, indent, "NdrFcShort(0x0),\n" ); /* FIXME: notify index */ if (is_object( iface )) pos += 2; if (args) LIST_FOR_EACH_ENTRY( var, args, var_t, entry ) { @@ -1503,8 +1632,40 @@ static void write_proc_func_interp( FILE *file, int indent, const type_t *iface, if (pos >= 16) break; } print_file( file, indent, "NdrFcShort(0x%x),\n", fpu_mask ); /* floating point mask */ - *offset += 2; + break; + } + case CPU_ARM: + case CPU_ARM64: + { + unsigned int i, len, count = stack_size / pointer_size; + unsigned char *params = xmalloc( count ); + + count = fill_params_array( iface, func, params, count ); + len = compress_params_array( params, count ); + + extra_size += 3 + len + !(len % 2); + print_file( file, indent, "0x%02x,\n", extra_size ); + print_file( file, indent, "0x%02x,\n", ext_flags ); + print_file( file, indent, "NdrFcShort(0x0),\n" ); /* server corr hint */ + print_file( file, indent, "NdrFcShort(0x0),\n" ); /* client corr hint */ + print_file( file, indent, "NdrFcShort(0x0),\n" ); /* FIXME: notify index */ + print_file( file, indent, "NdrFcShort(0x%02x),\n", count ); + print_file( file, indent, "0x%02x,\n", len ); + for (i = 0; i < len; i++) print_file( file, indent, "0x%02x,\n", params[i] ); + if (!(len % 2)) print_file( file, indent, "0x00,\n" ); + free( params ); + break; + } + case CPU_i386: + default: + print_file( file, indent, "0x%02x,\n", extra_size ); + print_file( file, indent, "0x%02x,\n", ext_flags ); + print_file( file, indent, "NdrFcShort(0x0),\n" ); /* server corr hint */ + print_file( file, indent, "NdrFcShort(0x0),\n" ); /* client corr hint */ + print_file( file, indent, "NdrFcShort(0x0),\n" ); /* FIXME: notify index */ + break; } + *offset += extra_size; /* emit argument data */ if (args) LIST_FOR_EACH_ENTRY( var, args, var_t, entry )

1 0

Alexandre Julliard : widl: Clear RobustEarly flag also for data structure conformance.
by Alexandre Julliard 16 Apr '24

16 Apr '24

Module: wine Branch: master Commit: 57b145e468e15af71236907b72cd471f08219acd URL: https://gitlab.winehq.org/wine/wine/-/commit/57b145e468e15af71236907b72cd47… Author: Alexandre Julliard <julliard(a)winehq.org> Date: Tue Apr 16 21:26:20 2024 +0200 widl: Clear RobustEarly flag also for data structure conformance. --- tools/widl/typegen.c | 1 + 1 file changed, 1 insertion(+) diff --git a/tools/widl/typegen.c b/tools/widl/typegen.c index 6f362bc1d8d..d164317a666 100644 --- a/tools/widl/typegen.c +++ b/tools/widl/typegen.c @@ -1987,6 +1987,7 @@ static unsigned int write_conf_or_var_desc(FILE *file, const type_t *cont_type, break; } offset += size; + if (offset > baseoff) robust_flags &= ~RobustEarly; } }

1 0

Alexandre Julliard : rpcrt4/tests: Add some more parameter passing tests.
by Alexandre Julliard 16 Apr '24

16 Apr '24

Module: wine Branch: master Commit: 4fc1591fc6f78e5e2bb8ecd9ad0f89ffee6a8084 URL: https://gitlab.winehq.org/wine/wine/-/commit/4fc1591fc6f78e5e2bb8ecd9ad0f89… Author: Alexandre Julliard <julliard(a)winehq.org> Date: Mon Apr 15 11:43:16 2024 +0200 rpcrt4/tests: Add some more parameter passing tests. --- dlls/rpcrt4/tests/server.c | 131 +++++++++++++++++++++++++++++++++++++++++++ dlls/rpcrt4/tests/server.idl | 37 ++++++++++++ 2 files changed, 168 insertions(+) diff --git a/dlls/rpcrt4/tests/server.c b/dlls/rpcrt4/tests/server.c index f92ed401a69..baf00aee744 100644 --- a/dlls/rpcrt4/tests/server.c +++ b/dlls/rpcrt4/tests/server.c @@ -68,6 +68,20 @@ static hyper (__cdecl *sum_hyper)(hyper x, hyper y); static int (__cdecl *sum_hyper_int)(hyper x, hyper y); static int (__cdecl *sum_char_hyper)(signed char x, hyper y); static void (__cdecl *square_out)(int x, int *y); +static int (__cdecl *sum_chars)(int a, chars_t x, chars_t y); +static int (__cdecl *sum_ints)(int a, ints_t x, ints_t y); +static int (__cdecl *sum_flts)(int a, flts_t x, flts_t y); +static int (__cdecl *sum_dbls)(int a, dbls_t x, dbls_t y); +static int (__cdecl *sum_iiff)(int x0, int y0, float x1, float y1); +static int (__cdecl *sum_ifif)(int x0, float y0, int x1, float y1); +static int (__cdecl *sum_iidd)(int x0, int y0, double x1, double y1); +static int (__cdecl *sum_idid)(int x0, double y0, int x1, double y1); +static int (__cdecl *sum_ififififififif)(int x0, float y0, int x1, float y1, int x2, float y2, int x3, float y3, int x4, float y4, int x5, float y5, int x6, float y6); +static int (__cdecl *sum_ididididididid)(int x0, double y0, int x1, double y1, int x2, double y2, int x3, double y3, int x4, double y4, int x5, double y5, int x6, double y6); +static int (__cdecl *sum_idfdifdfifdfidifdidf)(int x0, double y0, float z0, double y1, int x1, float z1, double y2, float z2, int x2, float z3, double y3, float z4, int x3, double y4, int x4, float z5, double y5, int x5, double y6, float z6); +static int (__cdecl *sum_ididididididididididid)(int x0, double y0, int x1, double y1, int x2, double y2, int x3, double y3, int x4, double y4, int x5, double y5, int x6, double y6, int x7, double y7, int x8, double y8); +static int (__cdecl *sum_iidiidiidiidiidiidiidiidiidiidiid)(int x0, int x1, double y0, int x2, int x3, double y1, int x4, int x5, double y2, int x6, int x7, double y3, int x8, int x9, double y4, int x10, int x11, double y5, int x12, int x13, double y6, int x14, int x15, double y7, int x16, int x17, double y8); +static int (__cdecl *sum_iyiy)(int x0, hyper y0, int x1, hyper y1); static void (__cdecl *square_ref)(int *x); static int (__cdecl *str_length)(const char *s); static int (__cdecl *str_t_length)(str_t s); @@ -162,6 +176,20 @@ static void (__cdecl *test_I_RpcBindingInqLocalClientPID)(unsigned int protseq, X(sum_hyper) \ X(sum_hyper_int) \ X(sum_char_hyper) \ + X(sum_chars) \ + X(sum_ints) \ + X(sum_flts) \ + X(sum_dbls) \ + X(sum_iiff) \ + X(sum_ifif) \ + X(sum_iidd) \ + X(sum_idid) \ + X(sum_ififififififif) \ + X(sum_ididididididid) \ + X(sum_idfdifdfifdfidifdidf) \ + X(sum_ididididididididididid) \ + X(sum_iidiidiidiidiidiidiidiidiidiidiid) \ + X(sum_iyiy) \ X(square_out) \ X(square_ref) \ X(str_length) \ @@ -338,6 +366,76 @@ int __cdecl s_sum_char_hyper(signed char x, hyper y) return x + y; } +int __cdecl s_sum_chars(int a, chars_t x, chars_t y) +{ + return a * (x.a + x.b + x.c + x.d + x.e + y.a + y.b + y.c + y.d + y.e); +} + +int __cdecl s_sum_ints(int a, ints_t x, ints_t y) +{ + return a * (x.i + x.j + y.i + y.j); +} + +int __cdecl s_sum_flts(int a, flts_t x, flts_t y) +{ + return a * (x.i + x.f + y.i + y.f); +} + +int __cdecl s_sum_dbls(int a, dbls_t x, dbls_t y) +{ + return a * (x.i + x.d + y.i + y.d); +} + +int __cdecl s_sum_iiff(int x0, int y0, float x1, float y1) +{ + return x0 + y0 + x1 + y1; +} + +int __cdecl s_sum_ifif(int x0, float y0, int x1, float y1) +{ + return x0 + y0 + x1 + y1; +} + +int __cdecl s_sum_iidd(int x0, int y0, double x1, double y1) +{ + return x0 + y0 + x1 + y1; +} + +int __cdecl s_sum_idid(int x0, double y0, int x1, double y1) +{ + return x0 + y0 + x1 + y1; +} + +int __cdecl s_sum_ififififififif(int x0, float y0, int x1, float y1, int x2, float y2, int x3, float y3, int x4, float y4, int x5, float y5, int x6, float y6) +{ + return x0 + y0 + x1 + y1 + x2 + y2 + x3 + y3 + x4 + y4 + x5 + y5 + x6 + y6; +} + +int __cdecl s_sum_ididididididid(int x0, double y0, int x1, double y1, int x2, double y2, int x3, double y3, int x4, double y4, int x5, double y5, int x6, double y6) +{ + return x0 + y0 + x1 + y1 + x2 + y2 + x3 + y3 + x4 + y4 + x5 + y5 + x6 + y6; +} + +int __cdecl s_sum_idfdifdfifdfidifdidf(int x0, double y0, float z0, double y1, int x1, float z1, double y2, float z2, int x2, float z3, double y3, float z4, int x3, double y4, int x4, float z5, double y5, int x5, double y6, float z6) +{ + return x0 + y0 + z0 + x1 + y1 + z1 + x2 + y2 + z2 + x3 + y3 + z3 + x4 + y4 + z4 + x5 + y5 + z5 + y6 + z6; +} + +int __cdecl s_sum_ididididididididididid(int x0, double y0, int x1, double y1, int x2, double y2, int x3, double y3, int x4, double y4, int x5, double y5, int x6, double y6, int x7, double y7, int x8, double y8) +{ + return x0 + y0 + x1 + y1 + x2 + y2 + x3 + y3 + x4 + y4 + x5 + y5 + x6 + y6 + x7 + y7 + x8 + y8; +} + +int __cdecl s_sum_iidiidiidiidiidiidiidiidiidiidiid(int x0, int x1, double y0, int x2, int x3, double y1, int x4, int x5, double y2, int x6, int x7, double y3, int x8, int x9, double y4, int x10, int x11, double y5, int x12, int x13, double y6, int x14, int x15, double y7, int x16, int x17, double y8) +{ + return x0 + x1 + y0 + x2 + x3 + y1 + x4 + x5 + y2 + x6 + x7 + y3 + x8 + x9 + y4 + x10 + x11 + y5 + x12 + x13 + y6 + x14 + x15 + y7 + x16 + x17 + y8; +} + +int __cdecl s_sum_iyiy(int x0, hyper y0, int x1, hyper y1) +{ + return x0 + y0 + x1 + y1; +} + void __cdecl s_square_out(int x, int *y) { *y = s_square(x); @@ -1271,6 +1369,10 @@ basic_tests(void) str_t str; se_t se; renum_t re; + chars_t cs1 = { 2, 4, 6, 8, 10 }, cs2 = { -1, -2, -3, -4, -5 }; + ints_t is1 = { 2, 4 }, is2 = { 5, 8 }; + flts_t fs1 = { 3, 5.5 }, fs2 = { 8, 2.3 }; + dbls_t ds1 = { 9, -2.5 }, ds2 = { 2, -4.2 }; ok(int_return() == INT_CODE, "RPC int_return\n"); @@ -1291,6 +1393,35 @@ basic_tests(void) ok(x == 0x12120000, "RPC hyper_int got 0x%x\n", x); x = sum_char_hyper( 12, ((hyper)0x42424242 << 32) | 0x33334444 ); ok(x == 0x33334450, "RPC char_hyper got 0x%x\n", x); + x = sum_chars( 4, cs1, cs2 ); + ok(x == 60, "got %d\n", x); + x = sum_ints( 2, is1, is2 ); + ok(x == 38, "got %d\n", x); + x = sum_flts( 3, fs1, fs2 ); + ok(x == 56, "got %d\n", x); + x = sum_dbls( 7, ds1, ds2 ); + ok(x == 30, "got %d\n", x); + x = sum_iiff( 12, 23, 3.4, 4.7 ); + ok(x == 43, "got %d\n", x); + x = sum_ifif( 12, 11.2, 23, 34.5 ); + ok(x == 80, "got %d\n", x); + x = sum_iidd( 6, 5, 4.3, 2.1 ); + ok(x == 17, "got %d\n", x); + x = sum_idid( 55, 44.55, 33, 22.44 ); + ok(x == 154, "got %d\n", x); + x = sum_ififififififif( 1, 2.1, 3, 4.2, 5, 6.3, 7, 8.4, 9, 10.5, 11, 12.2, 13, 14.9 ); + ok(x == 107, "got %d\n", x); + x = sum_ididididididid( -1, -2.1, -3, -4, -5, -6.7, -8, -9.1, -11, -12.3, -13, -14.2, -15, -16.3 ); + ok(x == -120, "got %d\n", x); + x = sum_idfdifdfifdfidifdidf( 2, 1.2, 2.1, 2.3, 4, 3.2, 4.5, 5.5, -3, -2.2, -4.4, -5.5, 6, 6.4, -3, 8.1, 9.2, 7, -10.3, 12.4 ); + ok(x == 45, "got %d\n", x); + x = sum_ididididididididididid( 2, 3, 4, 5, 6, 7, 8, 9, -1, -2, -3, -4, -5, -6, -7, -8, -9, 5 ); + ok(x == 4, "got %d\n", x); + x = sum_iidiidiidiidiidiidiidiidiidiidiid( 11, 22, 33.1, 44, 55, 66.1, 77, 88, 99.1, 111, 222, 333.1, 444, + 555, 666.1, 777, 888, 999.1, 1, 2, 3.1, 4, 5, 6.1, 7, 8, 9.1 ); + ok(x == 5535, "got %d\n", x); + x = sum_iyiy( 1, 1234567890, 2, -1234567800 ); + ok(x == 93, "got %d\n", x); x = 0; square_out(11, &x); diff --git a/dlls/rpcrt4/tests/server.idl b/dlls/rpcrt4/tests/server.idl index 8de5dce5d4e..d8886ee2dd2 100644 --- a/dlls/rpcrt4/tests/server.idl +++ b/dlls/rpcrt4/tests/server.idl @@ -95,6 +95,29 @@ cpp_quote("#ifndef SKIP_TYPE_DECLS") int s; } sun_t; + + typedef struct + { + signed char a, b, c, d, e; + } chars_t; + + typedef struct + { + int i; + hyper j; + } ints_t; + + typedef struct + { + int i; + float f; + } flts_t; + + typedef struct + { + int i; + double d; + } dbls_t; cpp_quote("#endif") int int_return(void); @@ -107,6 +130,20 @@ cpp_quote("#endif") hyper sum_hyper(hyper x, hyper y); int sum_hyper_int(hyper x, hyper y); int sum_char_hyper(signed char x, hyper y); + int sum_ints(int a, ints_t x, ints_t y); + int sum_chars(int a, chars_t x, chars_t y); + int sum_flts(int a, flts_t x, flts_t y); + int sum_dbls(int a, dbls_t x, dbls_t y); + int sum_iiff(int x0, int y0, float x1, float y1); + int sum_ifif(int x0, float y0, int x1, float y1); + int sum_iidd(int x0, int y0, double x1, double y1); + int sum_idid(int x0, double y0, int x1, double y1); + int sum_ififififififif(int x0, float y0, int x1, float y1, int x2, float y2, int x3, float y3, int x4, float y4, int x5, float y5, int x6, float y6); + int sum_ididididididid(int x0, double y0, int x1, double y1, int x2, double y2, int x3, double y3, int x4, double y4, int x5, double y5, int x6, double y6); + int sum_idfdifdfifdfidifdidf(int x0, double y0, float z0, double y1, int x1, float z1, double y2, float z2, int x2, float z3, double y3, float z4, int x3, double y4, int x4, float z5, double y5, int x5, double y6, float z6); + int sum_ididididididididididid(int x0, double y0, int x1, double y1, int x2, double y2, int x3, double y3, int x4, double y4, int x5, double y5, int x6, double y6, int x7, double y7, int x8, double y8); + int sum_iidiidiidiidiidiidiidiidiidiidiid(int x0, int x1, double y0, int x2, int x3, double y1, int x4, int x5, double y2, int x6, int x7, double y3, int x8, int x9, double y4, int x10, int x11, double y5, int x12, int x13, double y6, int x14, int x15, double y7, int x16, int x17, double y8); + int sum_iyiy(int x0, hyper y0, int x1, hyper y1); void square_out(int x, [out] int *y); void square_ref([in, out] int *x); int str_length([string] const char *s);

1 0

Henri Verbeet : tests: Test creating compute pipeline states from shaders with embedded root signatures.
by Alexandre Julliard 15 Apr '24

15 Apr '24

Module: vkd3d Branch: master Commit: 18986ddb5005293e224d9a2b6022ea7b3297cf62 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/18986ddb5005293e224d9a2b6022e… Author: Henri Verbeet <hverbeet(a)codeweavers.com> Date: Thu Apr 11 20:04:28 2024 +0200 tests: Test creating compute pipeline states from shaders with embedded root signatures. --- tests/d3d12.c | 28 ++++++++++++++++++++++++++++ 1 file changed, 28 insertions(+) diff --git a/tests/d3d12.c b/tests/d3d12.c index aaea5e31..e414ca36 100644 --- a/tests/d3d12.c +++ b/tests/d3d12.c @@ -3045,6 +3045,22 @@ static void test_create_compute_pipeline_state(void) ULONG refcount; HRESULT hr; + static const DWORD cs_with_rs[] = + { +#if 0 + [RootSignature("")] + [numthreads(1, 1, 1)] + void main() + { + } +#endif + 0x43425844, 0x215835dd, 0xdcf65f2e, 0x076d1ec0, 0xb1664d2b, 0x00000001, 0x00000098, 0x00000004, + 0x00000030, 0x00000040, 0x00000050, 0x00000078, 0x4e475349, 0x00000008, 0x00000000, 0x00000008, + 0x4e47534f, 0x00000008, 0x00000000, 0x00000008, 0x58454853, 0x00000020, 0x00050050, 0x00000008, + 0x0100086a, 0x0400009b, 0x00000001, 0x00000001, 0x00000001, 0x0100003e, 0x30535452, 0x00000018, + 0x00000001, 0x00000000, 0x00000018, 0x00000000, 0x00000018, 0x00000000, + }; + static const char shader_code[] = "[numthreads(1, 1, 1)]\n" "void main() { }\n"; @@ -3111,6 +3127,18 @@ static void test_create_compute_pipeline_state(void) refcount = ID3D12RootSignature_Release(root_signature); ok(!refcount, "ID3D12RootSignature has %u references left.\n", (unsigned int)refcount); + pipeline_state_desc.pRootSignature = NULL; + hr = ID3D12Device_CreateComputePipelineState(device, &pipeline_state_desc, + &IID_ID3D12PipelineState, (void **)&pipeline_state); + ok(hr == E_INVALIDARG, "Got hr %#x.\n", hr); + + pipeline_state_desc.CS = shader_bytecode(cs_with_rs, sizeof(cs_with_rs)); + hr = ID3D12Device_CreateComputePipelineState(device, &pipeline_state_desc, + &IID_ID3D12PipelineState, (void **)&pipeline_state); + todo ok(hr == S_OK, "Got hr %#x.\n", hr); + if (SUCCEEDED(hr)) + ID3D12PipelineState_Release(pipeline_state); + refcount = ID3D12Device_Release(device); ok(!refcount, "ID3D12Device has %u references left.\n", (unsigned int)refcount);

1 0

Henri Verbeet : vkd3d: Implement creating compute pipeline states from shaders with embedded root signatures.
by Alexandre Julliard 15 Apr '24

15 Apr '24

Module: vkd3d Branch: master Commit: 7d6f0f2592a8aedf749c2dff36ea330e9ccb49d1 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/7d6f0f2592a8aedf749c2dff36ea3… Author: Henri Verbeet <hverbeet(a)codeweavers.com> Date: Tue Sep 5 15:39:06 2023 +0200 vkd3d: Implement creating compute pipeline states from shaders with embedded root signatures. --- libs/vkd3d/state.c | 28 +++++++++++++++++++++++++--- libs/vkd3d/vkd3d_private.h | 1 + tests/d3d12.c | 5 ++--- 3 files changed, 28 insertions(+), 6 deletions(-) diff --git a/libs/vkd3d/state.c b/libs/vkd3d/state.c index 08cc110e..b8328216 100644 --- a/libs/vkd3d/state.c +++ b/libs/vkd3d/state.c @@ -2045,6 +2045,9 @@ static ULONG STDMETHODCALLTYPE d3d12_pipeline_state_Release(ID3D12PipelineState d3d12_pipeline_uav_counter_state_cleanup(&state->uav_counters, device); + if (state->implicit_root_signature) + d3d12_root_signature_Release(state->implicit_root_signature); + vkd3d_free(state); d3d12_device_release(device); @@ -2413,8 +2416,8 @@ static HRESULT d3d12_pipeline_state_init_compute(struct d3d12_pipeline_state *st const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs; struct vkd3d_shader_interface_info shader_interface; struct vkd3d_shader_descriptor_offset_info offset_info; - const struct d3d12_root_signature *root_signature; struct vkd3d_shader_spirv_target_info target_info; + struct d3d12_root_signature *root_signature; VkPipelineLayout vk_pipeline_layout; HRESULT hr; @@ -2425,13 +2428,27 @@ static HRESULT d3d12_pipeline_state_init_compute(struct d3d12_pipeline_state *st if (!(root_signature = unsafe_impl_from_ID3D12RootSignature(desc->root_signature))) { - WARN("Root signature is NULL.\n"); - return E_INVALIDARG; + TRACE("Root signature is NULL, looking for an embedded signature.\n"); + if (FAILED(hr = d3d12_root_signature_create(device, + desc->cs.pShaderBytecode, desc->cs.BytecodeLength, &root_signature))) + { + WARN("Failed to find an embedded root signature, hr %s.\n", debugstr_hresult(hr)); + return hr; + } + state->implicit_root_signature = &root_signature->ID3D12RootSignature_iface; + } + else + { + state->implicit_root_signature = NULL; } if (FAILED(hr = d3d12_pipeline_state_find_and_init_uav_counters(state, device, root_signature, &desc->cs, VK_SHADER_STAGE_COMPUTE_BIT))) + { + if (state->implicit_root_signature) + d3d12_root_signature_Release(state->implicit_root_signature); return hr; + } memset(&target_info, 0, sizeof(target_info)); target_info.type = VKD3D_SHADER_STRUCTURE_TYPE_SPIRV_TARGET_INFO; @@ -2476,6 +2493,8 @@ static HRESULT d3d12_pipeline_state_init_compute(struct d3d12_pipeline_state *st { WARN("Failed to create Vulkan compute pipeline, hr %s.\n", debugstr_hresult(hr)); d3d12_pipeline_uav_counter_state_cleanup(&state->uav_counters, device); + if (state->implicit_root_signature) + d3d12_root_signature_Release(state->implicit_root_signature); return hr; } @@ -2483,6 +2502,8 @@ static HRESULT d3d12_pipeline_state_init_compute(struct d3d12_pipeline_state *st { VK_CALL(vkDestroyPipeline(device->vk_device, state->u.compute.vk_pipeline, NULL)); d3d12_pipeline_uav_counter_state_cleanup(&state->uav_counters, device); + if (state->implicit_root_signature) + d3d12_root_signature_Release(state->implicit_root_signature); return hr; } @@ -3484,6 +3505,7 @@ static HRESULT d3d12_pipeline_state_init_graphics(struct d3d12_pipeline_state *s goto fail; state->vk_bind_point = VK_PIPELINE_BIND_POINT_GRAPHICS; + state->implicit_root_signature = NULL; d3d12_device_add_ref(state->device = device); return S_OK; diff --git a/libs/vkd3d/vkd3d_private.h b/libs/vkd3d/vkd3d_private.h index 91ffa6e9..8de57a33 100644 --- a/libs/vkd3d/vkd3d_private.h +++ b/libs/vkd3d/vkd3d_private.h @@ -1213,6 +1213,7 @@ struct d3d12_pipeline_state struct d3d12_pipeline_uav_counter_state uav_counters; + ID3D12RootSignature *implicit_root_signature; struct d3d12_device *device; struct vkd3d_private_store private_store; diff --git a/tests/d3d12.c b/tests/d3d12.c index e414ca36..31f57c40 100644 --- a/tests/d3d12.c +++ b/tests/d3d12.c @@ -3135,9 +3135,8 @@ static void test_create_compute_pipeline_state(void) pipeline_state_desc.CS = shader_bytecode(cs_with_rs, sizeof(cs_with_rs)); hr = ID3D12Device_CreateComputePipelineState(device, &pipeline_state_desc, &IID_ID3D12PipelineState, (void **)&pipeline_state); - todo ok(hr == S_OK, "Got hr %#x.\n", hr); - if (SUCCEEDED(hr)) - ID3D12PipelineState_Release(pipeline_state); + ok(hr == S_OK, "Got hr %#x.\n", hr); + ID3D12PipelineState_Release(pipeline_state); refcount = ID3D12Device_Release(device); ok(!refcount, "ID3D12Device has %u references left.\n", (unsigned int)refcount);

1 0

Giovanni Mascellani : vkd3d-shader/ir: Prepare for changing the destination list in vsir_cfg_optimize_recurse().
by Alexandre Julliard 15 Apr '24

15 Apr '24

Module: vkd3d Branch: master Commit: a2c6b5450b3f6e21c60c9280df5e23694c91f6a5 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/a2c6b5450b3f6e21c60c9280df5e2… Author: Giovanni Mascellani <gmascellani(a)codeweavers.com> Date: Thu Mar 14 15:04:29 2024 +0100 vkd3d-shader/ir: Prepare for changing the destination list in vsir_cfg_optimize_recurse(). --- libs/vkd3d-shader/ir.c | 33 ++++++++++++++++----------------- 1 file changed, 16 insertions(+), 17 deletions(-) diff --git a/libs/vkd3d-shader/ir.c b/libs/vkd3d-shader/ir.c index e0928ff7..da23e228 100644 --- a/libs/vkd3d-shader/ir.c +++ b/libs/vkd3d-shader/ir.c @@ -4307,19 +4307,21 @@ static enum vkd3d_result vsir_cfg_append_loop(struct vsir_cfg *cfg, static enum vkd3d_result vsir_cfg_optimize_recurse(struct vsir_cfg *cfg, struct vsir_cfg_structure_list *list) { - struct vsir_cfg_structure_list new_list = {0}; + struct vsir_cfg_structure_list old_list = *list, *new_list = list; enum vkd3d_result ret; size_t i; - for (i = 0; i < list->count; ++i) + memset(new_list, 0, sizeof(*new_list)); + + for (i = 0; i < old_list.count; ++i) { - struct vsir_cfg_structure *loop = &list->structures[i]; + struct vsir_cfg_structure *loop = &old_list.structures[i]; struct vsir_cfg_structure_list *loop_body; if (loop->type != STRUCTURE_TYPE_LOOP) { - if ((ret = vsir_cfg_structure_list_append_from_region(&new_list, loop, 1)) < 0) - goto fail; + if ((ret = vsir_cfg_structure_list_append_from_region(new_list, loop, 1)) < 0) + goto out; memset(loop, 0, sizeof(*loop)); continue; } @@ -4328,8 +4330,8 @@ static enum vkd3d_result vsir_cfg_optimize_recurse(struct vsir_cfg *cfg, struct if (loop_body->count == 0) { - if ((ret = vsir_cfg_structure_list_append_from_region(&new_list, loop, 1)) < 0) - goto fail; + if ((ret = vsir_cfg_structure_list_append_from_region(new_list, loop, 1)) < 0) + goto out; memset(loop, 0, sizeof(*loop)); continue; } @@ -4337,22 +4339,19 @@ static enum vkd3d_result vsir_cfg_optimize_recurse(struct vsir_cfg *cfg, struct vsir_cfg_remove_trailing_continue(cfg, loop_body, loop->u.loop.idx); if ((ret = vsir_cfg_optimize_recurse(cfg, loop_body)) < 0) - goto fail; + goto out; if ((ret = vsir_cfg_synthesize_selections(cfg, loop_body)) < 0) - goto fail; + goto out; - if ((ret = vsir_cfg_append_loop(cfg, &new_list, loop)) < 0) - goto fail; + if ((ret = vsir_cfg_append_loop(cfg, new_list, loop)) < 0) + goto out; } - vsir_cfg_structure_list_cleanup(list); - *list = new_list; + ret = VKD3D_OK; - return VKD3D_OK; - -fail: - vsir_cfg_structure_list_cleanup(list); +out: + vsir_cfg_structure_list_cleanup(&old_list); return ret; }

1 0

Giovanni Mascellani : vkd3d-shader/ir: Append code to the non-breaking branch if there is breaking one.
by Alexandre Julliard 15 Apr '24

15 Apr '24

Module: vkd3d Branch: master Commit: be7eec75dcc769ebd809418e804e6bda4f4a9f71 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/be7eec75dcc769ebd809418e804e6… Author: Giovanni Mascellani <gmascellani(a)codeweavers.com> Date: Thu Mar 14 15:33:19 2024 +0100 vkd3d-shader/ir: Append code to the non-breaking branch if there is breaking one. --- libs/vkd3d-shader/ir.c | 18 +++++++++++++++++- 1 file changed, 17 insertions(+), 1 deletion(-) diff --git a/libs/vkd3d-shader/ir.c b/libs/vkd3d-shader/ir.c index da23e228..a9588279 100644 --- a/libs/vkd3d-shader/ir.c +++ b/libs/vkd3d-shader/ir.c @@ -4315,7 +4315,7 @@ static enum vkd3d_result vsir_cfg_optimize_recurse(struct vsir_cfg *cfg, struct for (i = 0; i < old_list.count; ++i) { - struct vsir_cfg_structure *loop = &old_list.structures[i]; + struct vsir_cfg_structure *loop = &old_list.structures[i], *selection; struct vsir_cfg_structure_list *loop_body; if (loop->type != STRUCTURE_TYPE_LOOP) @@ -4346,6 +4346,22 @@ static enum vkd3d_result vsir_cfg_optimize_recurse(struct vsir_cfg *cfg, struct if ((ret = vsir_cfg_append_loop(cfg, new_list, loop)) < 0) goto out; + + /* If the last pushed instruction is a selection and one of the branches terminates with a + * `break', start pushing to the other branch, in the hope of eventually push a `break' + * there too and be able to remove a loop. */ + if (new_list->count == 0) + continue; + + selection = &new_list->structures[new_list->count - 1]; + + if (selection->type == STRUCTURE_TYPE_SELECTION) + { + if (vsir_cfg_get_trailing_break(&selection->u.selection.if_body)) + new_list = &selection->u.selection.else_body; + else if (vsir_cfg_get_trailing_break(&selection->u.selection.else_body)) + new_list = &selection->u.selection.if_body; + } } ret = VKD3D_OK;

1 0

Giovanni Mascellani : vkd3d-shader/ir: Move breaks out of selections again after having optimized loops.
by Alexandre Julliard 15 Apr '24

15 Apr '24

Module: vkd3d Branch: master Commit: eb15f97083e03975bc27eb72f1f5c159058df577 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/eb15f97083e03975bc27eb72f1f5c… Author: Giovanni Mascellani <gmascellani(a)codeweavers.com> Date: Thu Mar 14 16:26:52 2024 +0100 vkd3d-shader/ir: Move breaks out of selections again after having optimized loops. --- libs/vkd3d-shader/ir.c | 21 ++++++++++++++++++++- 1 file changed, 20 insertions(+), 1 deletion(-) diff --git a/libs/vkd3d-shader/ir.c b/libs/vkd3d-shader/ir.c index a9588279..6d428e22 100644 --- a/libs/vkd3d-shader/ir.c +++ b/libs/vkd3d-shader/ir.c @@ -4200,6 +4200,25 @@ static enum vkd3d_result vsir_cfg_move_breaks_out_of_selections(struct vsir_cfg return VKD3D_OK; } +static enum vkd3d_result vsir_cfg_move_breaks_out_of_selections_recursively(struct vsir_cfg *cfg, + struct vsir_cfg_structure_list *list) +{ + struct vsir_cfg_structure *trailing; + + if (list->count == 0) + return VKD3D_OK; + + trailing = &list->structures[list->count - 1]; + + if (trailing->type != STRUCTURE_TYPE_SELECTION) + return VKD3D_OK; + + vsir_cfg_move_breaks_out_of_selections_recursively(cfg, &trailing->u.selection.if_body); + vsir_cfg_move_breaks_out_of_selections_recursively(cfg, &trailing->u.selection.else_body); + + return vsir_cfg_move_breaks_out_of_selections(cfg, list); +} + static enum vkd3d_result vsir_cfg_synthesize_selections(struct vsir_cfg *cfg, struct vsir_cfg_structure_list *list) { @@ -4364,7 +4383,7 @@ static enum vkd3d_result vsir_cfg_optimize_recurse(struct vsir_cfg *cfg, struct } } - ret = VKD3D_OK; + ret = vsir_cfg_move_breaks_out_of_selections_recursively(cfg, list); out: vsir_cfg_structure_list_cleanup(&old_list);

1 0

Giovanni Mascellani : vkd3d-shader/ir: Split program structurization.
by Alexandre Julliard 15 Apr '24

15 Apr '24

Module: vkd3d Branch: master Commit: d891ac5017807f404f449061f54ca3441f8c6f2c URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/d891ac5017807f404f449061f54ca… Author: Giovanni Mascellani <gmascellani(a)codeweavers.com> Date: Wed Apr 10 15:21:52 2024 +0200 vkd3d-shader/ir: Split program structurization. --- libs/vkd3d-shader/ir.c | 76 +++++++++++++++++++++++--------------------------- 1 file changed, 35 insertions(+), 41 deletions(-) diff --git a/libs/vkd3d-shader/ir.c b/libs/vkd3d-shader/ir.c index 6d428e22..e7e66662 100644 --- a/libs/vkd3d-shader/ir.c +++ b/libs/vkd3d-shader/ir.c @@ -4677,6 +4677,40 @@ fail: return ret; } +static enum vkd3d_result vsir_program_structurize(struct vsir_program *program, + struct vkd3d_shader_message_context *message_context) +{ + enum vkd3d_result ret; + struct vsir_cfg cfg; + + if ((ret = vsir_cfg_init(&cfg, program, message_context)) < 0) + return ret; + + vsir_cfg_compute_dominators(&cfg); + + if ((ret = vsir_cfg_compute_loops(&cfg)) < 0) + goto out; + + if ((ret = vsir_cfg_sort_nodes(&cfg)) < 0) + goto out; + + if ((ret = vsir_cfg_generate_synthetic_loop_intervals(&cfg)) < 0) + goto out; + + if ((ret = vsir_cfg_build_structured_program(&cfg)) < 0) + goto out; + + if ((ret = vsir_cfg_optimize(&cfg)) < 0) + goto out; + + ret = vsir_cfg_emit_structured_program(&cfg); + +out: + vsir_cfg_cleanup(&cfg); + + return ret; +} + static void register_map_undominated_use(struct vkd3d_shader_register *reg, struct ssas_to_temps_alloc *alloc, struct vsir_block *block, struct vsir_block **origin_blocks) { @@ -5654,48 +5688,8 @@ enum vkd3d_result vsir_program_normalise(struct vsir_program *program, uint64_t if ((result = lower_switch_to_if_ladder(program)) < 0) return result; - if ((result = vsir_cfg_init(&cfg, program, message_context)) < 0) - return result; - - vsir_cfg_compute_dominators(&cfg); - - if ((result = vsir_cfg_compute_loops(&cfg)) < 0) - { - vsir_cfg_cleanup(&cfg); + if ((result = vsir_program_structurize(program, message_context)) < 0) return result; - } - - if ((result = vsir_cfg_sort_nodes(&cfg)) < 0) - { - vsir_cfg_cleanup(&cfg); - return result; - } - - if ((result = vsir_cfg_generate_synthetic_loop_intervals(&cfg)) < 0) - { - vsir_cfg_cleanup(&cfg); - return result; - } - - if ((result = vsir_cfg_build_structured_program(&cfg)) < 0) - { - vsir_cfg_cleanup(&cfg); - return result; - } - - if ((result = vsir_cfg_optimize(&cfg)) < 0) - { - vsir_cfg_cleanup(&cfg); - return result; - } - - if ((result = vsir_cfg_emit_structured_program(&cfg)) < 0) - { - vsir_cfg_cleanup(&cfg); - return result; - } - - vsir_cfg_cleanup(&cfg); if ((result = vsir_program_flatten_control_flow_constructs(program, message_context)) < 0) return result;

1 0

Giovanni Mascellani : vkd3d-shader/ir: Split undominated SSA materialization.
by Alexandre Julliard 15 Apr '24

15 Apr '24

Module: vkd3d Branch: master Commit: 27801e56129f3288382b3227b623f664a6f91a1e URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/27801e56129f3288382b3227b623f… Author: Giovanni Mascellani <gmascellani(a)codeweavers.com> Date: Wed Apr 10 15:26:50 2024 +0200 vkd3d-shader/ir: Split undominated SSA materialization. --- libs/vkd3d-shader/ir.c | 30 +++++++++++++++++++----------- 1 file changed, 19 insertions(+), 11 deletions(-) diff --git a/libs/vkd3d-shader/ir.c b/libs/vkd3d-shader/ir.c index e7e66662..b022d75c 100644 --- a/libs/vkd3d-shader/ir.c +++ b/libs/vkd3d-shader/ir.c @@ -4803,6 +4803,24 @@ done: return VKD3D_OK; } +static enum vkd3d_result vsir_program_materialize_undominated_ssas_to_temps(struct vsir_program *program, + struct vkd3d_shader_message_context *message_context) +{ + enum vkd3d_result ret; + struct vsir_cfg cfg; + + if ((ret = vsir_cfg_init(&cfg, program, message_context)) < 0) + return ret; + + vsir_cfg_compute_dominators(&cfg); + + ret = vsir_cfg_materialize_undominated_ssas_to_temps(&cfg); + + vsir_cfg_cleanup(&cfg); + + return ret; +} + struct validation_context { struct vkd3d_shader_message_context *message_context; @@ -5680,8 +5698,6 @@ enum vkd3d_result vsir_program_normalise(struct vsir_program *program, uint64_t if (program->shader_version.major >= 6) { - struct vsir_cfg cfg; - if ((result = vsir_program_materialise_phi_ssas_to_temps(program)) < 0) return result; @@ -5694,15 +5710,7 @@ enum vkd3d_result vsir_program_normalise(struct vsir_program *program, uint64_t if ((result = vsir_program_flatten_control_flow_constructs(program, message_context)) < 0) return result; - if ((result = vsir_cfg_init(&cfg, program, message_context)) < 0) - return result; - vsir_cfg_compute_dominators(&cfg); - - result = vsir_cfg_materialize_undominated_ssas_to_temps(&cfg); - - vsir_cfg_cleanup(&cfg); - - if (result < 0) + if ((result = vsir_program_materialize_undominated_ssas_to_temps(program, message_context)) < 0) return result; } else

1 0